Molecular community profiling reveals impacts of time, space, and disease status on the bacterial community associated with the Caribbean sponge Aplysina cauliformis

Authors

  • Julie B. Olson,

    Corresponding author
    1. Department of Biological Sciences, University of Alabama, Tuscaloosa, AL, USA
    • Correspondence: Julie B. Olson, Department of Biological Sciences, University of Alabama, 1325 SEC Bldg., Campus Box 870344, Tuscaloosa, AL 35487, USA. Tel.: 205 348 2633; fax: 205 348 1786; e-mail: jolson@as.ua.edu

    Search for more papers by this author
  • Robert W. Thacker,

    1. Department of Biology, University of Alabama at Birmingham, Birmingham, AL, USA
    Search for more papers by this author
  • Deborah J. Gochfeld

    1. National Center for Natural Products Research, University of Mississippi, University, MS, USA
    Search for more papers by this author

Abstract

Reports of marine sponge diseases have increased in recent years, but few etiologic agents have been identified. Aplysina red band syndrome (ARBS), a condition observed in the Caribbean sponge Aplysina cauliformis, is characterized by a rust-colored leading margin. Culture-independent methods (terminal restriction fragment length polymorphism and clone library analyses) were used to assess bacterial communities associated with healthy and ARBS-affected sponges from two locations over 2 years. Although the bacterial communities associated with healthy and ARBS-affected sponges were significantly different, the sponges maintained a core bacterial community across space, time, and health status. Ten terminal restriction fragments were shown to change significantly between sponge health conditions, with six increasing in abundance with disease and four decreasing. The prevalence of the photosymbiont Synechococcus spongiarum decreased with ARBS infection, suggesting a functional consequence of disease. After cultivating a red-pigmented Leptolyngbya strain from ARBS lesions, transmission studies were conducted to determine whether this organism was the ARBS pathogen. Despite significantly increased abundance of Leptolyngbya spp. in diseased sponges, signs of ARBS were not observed in healthy sponges following 24 days of contact with the cultured strain. Additional work with this model system is needed to increase our understanding of the dynamics of marine diseases.

Introduction

Sponges are integral components of coral reefs and provide many ecosystem services, including water filtration and nutrient cycling (Wilkinson, 1987; Diaz & Rützler, 2001; Lesser, 2005; Wulff, 2006). These sessile, filter-feeding organisms can dominate reef biomass in the Caribbean Sea (Diaz & Rützler, 2001; Wulff, 2006) and thus play critical roles in the ecology of these ecosystems. Although many marine sponges, termed ‘bacteriosponges’ or ‘high microbial abundance sponges’, support large and diverse microbial communities that are distinct from microorganisms found in ambient seawater (Santavy & Colwell, 1990; Hentschel et al., 2002; Webster et al., 2004, 2010), considerable debate surrounds the specificity and stability of sponge–microorganism interactions. While some studies demonstrated vertical transmission of bacterial associates (Enticknap et al., 2006; Sharp et al., 2007; Schmitt et al., 2008; Lee et al., 2009), recent high-throughput sequencing work (Webster et al., 2010; Taylor et al., 2013) suggested that many bacteria thought to be sponge specific are found in very low numbers in the ambient environment. These symbionts might therefore be acquired from the environment (horizontal transmission). Numerous sponge species appear to maintain a ‘core’ group of associated bacteria, with the remainder of the community being more variable (e.g. Taylor et al., 2007; Erwin et al., 2011; Montalvo & Hill, 2011; Schmitt et al., 2011, 2012), fitting a model of ‘leaky vertical transmission’ (Vrijenhoek, 2010; Thacker & Freeman, 2012).

Some sponges host cyanobacterial symbionts that provide nutrition to the sponge holobiont (e.g. Weisz et al., 2007; Erwin & Thacker, 2008; Freeman & Thacker, 2011). Healthy individuals of Aplysina cauliformis host large populations of the cyanobacterium Synechococcus spongiarum, which assimilate inorganic carbon and nitrogen, and then transfer these elements to the sponge (Freeman & Thacker, 2011). Synechococcus spongiarum is one of few sponge-associated microorganisms for which a function has been attributed; relatively little is known about the specific roles that most microorganisms play within sponge hosts.

Numerous diseases affecting coral reef species have been documented in recent decades, with the majority of reports describing conditions impacting corals. However, diseases affecting sponges have been reported with increasing frequency (Rützler, 1988; Olson et al., 2006; Webster, 2007; Wulff, 2007). Aplysina red band syndrome (ARBS) is a disease that affects the Caribbean rope sponge A. cauliformis, resulting in localized tissue necrosis, reduced growth, and other sublethal impacts (Olson et al., 2006; Gochfeld et al., 2012a). ARBS was first observed on a single patch reef in the Bahamas in 2004 and has since been documented on additional reefs in the Florida Keys, Bahamas, Belize, Cayman Islands, Panama, Curacao, and the U.S. Virgin Islands (Gochfeld et al., 2007; D.J. Gochfeld and J.B. Olson, unpublished data).

Rützler (1988) described a disease affecting the mangrove sponge Geodia papyracea and suggested overgrowth of its cyanobacterial symbiont as the cause. While A. cauliformis, the sponge affected by ARBS, is known to host S. spongiarum (Erwin & Thacker, 2008), little is known about the identity or functions of other members of its symbiotic microbial community, including other cyanobacteria, which might be harmful to the host sponge. When examined microscopically, the margins of ARBS lesions support a large community of cyanobacteria (Olson et al., 2006), including a red-colored Leptolyngbya species that was cultured in the laboratory. Because the coloration of this organism matches that of the leading margin of ARBS lesions, studies were initiated to determine whether it was the etiologic agent of ARBS. To test this hypothesis, both healthy and ARBS-affected sponges were collected from two sites (Carrie Bow Cay, Belize, and Lee Stocking Island, Bahamas) over 2 years (2008 and 2009). Culture-independent techniques [terminal restriction fragment length polymorphism (T-RFLP) analysis and clone library construction] were utilized to characterize and compare symbiotic bacterial communities associated with healthy and diseased sponges. T-RFLP methods have been shown to provide a robust index of community diversity with results that are well correlated with those obtained from clone libraries (Fierer & Jackson, 2006). Together, the information generated from these molecular community profiling approaches might provide a better understanding of the etiology and dynamics of a transmissible marine disease. Based on the limited recovery of microorganisms in cultivation approaches, molecular approaches have been shown to provide a more holistic representation of community composition. Care, however, should be taken in interpreting the resulting data, as these methods are also susceptible to biases (e.g. von Wintzingerode et al., 1997). To minimize bias and maximize information, multiple methods were combined (Rastogi & Sani, 2011). Finally, transmission experiments were performed using cultured cyanobacteria in an attempt to generate ARBS in healthy sponges.

Materials and methods

Study locations and collection methods

Field sites included patch reefs near the Perry Institute for Marine Science, Lee Stocking Island, Exuma Cays, Bahamas (23°46.31′N, 76°6.13′W), and the Smithsonian Field Station at Carrie Bow Cay, Belize (16°48.17′N, 88°4.93′W). Aplysina cauliformis was among the most prevalent sponge species at both locations. Visibly healthy and ARBS-affected individuals were selected haphazardly from small patch reefs (c. 25 m diameter) while SCUBA diving. Approximately 10-cm portions of a branch were removed from healthy individuals using scissors and placed into resealable plastic bags. For diseased individuals, affected branches were cut c. 5 cm below the lesion. Samples were kept in individual bags in seawater until arrival at the field stations; equal numbers of healthy and diseased sponges were collected on the same dive, and all samples from a location were collected within three consecutive days. Small wedges (c. 1 mm thick) were aseptically cut from each healthy branch or from the margin of the ARBS lesion and placed into sterile cryovials with 1.8 ml RNAlater® (Life Technologies). Five samples of healthy and ARBS-affected sponges were preserved at each location during the summers of 2008 and 2009. In 2009, seawater was collected in five 1-L Nalgene® bottles at both locations. Using 0.2-μm polycarbonate membrane filters, 250 mL per bottle was filtered. Filters were placed into 1.8 mL of RNAlater® in sterile cryovials. All samples were frozen at −20 °C until DNA extraction.

DNA extraction and PCR amplification for T-RFLP

Total DNA was extracted from sponge samples and from half of each filter using the FastDNA Spin Kit for Soil (MP Biochemicals), incubating samples overnight on a shaking platform at room temperature. DNA was eluted into 50 μL of elution buffer. As recommended by Marsh et al. (2000), two 16S rRNA gene primer pairs were utilized to better represent the entire bacterial community. The first pair, 8F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1392R (5′-ACGGGCGGTGTGTACA-3′), is considered universal and amplifies a large portion of the 16S ribosomal RNA gene, while the second pair, 359F (5′-GGGGAATYTTCCGCAATGGG-3′) and 1392R, includes a forward primer originally designed for cyanobacteria (Nübel et al., 1997), but which has been shown to amplify a much larger bacterial subset (Olson & Gao, 2013). Amplification mixtures contained 10 μL TaqMaster PCR enhancer (5 PRIME, Inc.), 5 μL reaction buffer, 2.5 μL, 25 mM magnesium solution, 25 pmol of each primer, each deoxynucleoside triphosphate at a concentration of 0.2 µM, 2 U Taq DNA polymerase, 1 μL of DNA diluted 1 : 10 in distilled water, and distilled water to a final volume of 50 μL. Reaction conditions were identical for both primer pairs, with a 5-min incubation at 85 °C followed by 30 cycles of 94 °C for 45 s, 55 °C for 1.0 min, and 72 °C for 1.5 min, and a final extension for 7 min at 72 °C. Reactions were verified on 1.5% agarose gels containing GelRed (Biotium, Inc.) at a constant voltage of 65 mV for 65 min. Gels were viewed under UV transillumination and photographed with a digital imaging system.

Terminal restriction fragment length polymorphism analysis

Triplicate PCRs were amplified for each sample using both primer pairs with a fluorescent S-hexachlorofluorescein (HEX) label on the forward primers. Products were cleaned individually using the QIAquick PCR Purification kit (Qiagen Inc.) and eluted in 30 μL of elution buffer. After combining the three eluates, DNA concentration was determined using a NanoDrop spectrophotometer (Thermo Fisher Scientific Inc.). Two restriction endonucleases were used individually for digestions, HaeIII and HhaI. In each reaction, 400 ng of DNA was added to 2 U enzyme, 5 μL of enzyme buffer, and distilled water to a final volume of 50 μL and incubated for 8 h at 37 °C. Enzymes were heat killed, and samples were ethanol-precipitated overnight. Pellets were resuspended in 10 μL of deionized formamide and 0.5 μL of 6-carboxytetramethylrhodamine (TAMRA) size standard prior to loading on an ABI 3100 genetic analyzer with a 50-cm capillary array (Applied Biosystems). Fragment lengths were determined using the Local Southern size-calling algorithm of the Genemapper analysis software (version 3.7; Applied Biosystems).

Data matrices were constructed as described in the study by Hodges & Olson (2009). Peaks were analyzed using T-REX (Culman et al., 2009), with standard deviation multipliers of 1.25–1.5 to filter noise and a clustering threshold of 1.0 to align terminal restriction fragments (T-RFs). Peak areas were normalized to total fluorescence units per sample; resulting data matrices were imported into primer 6.0 for analyses.

Clone library construction

Separate PCRs were performed using two randomly selected samples collected in Belize in 2008 (BZ-H2 and BZ-D2 from T-RFLP analyses, equivalent to BZ46H and BZ40D) and two from Bahamas in 2008 (LSI-H3 and LSI-D2, equivalent to LSI107H and LSI99D). DNA was extracted using the Wizard Genomic DNA Purification Kit (Promega). For each sample, PCR amplifications targeted bacteria [using primers 8F and 1392R and 106F (5′-CGGACGGGTGAGTAACGCGTGA-3′; Nübel et al., 1997) and 1392R]. Amplification mixtures and reaction conditions were described above. PCR products were gel-purified with the Wizard SV Gel Clean-Up System (Promega) and ligated into plasmids using the pGEM T-Easy Vector System (Promega). Eight clone libraries (four samples × two primer sets) were plated, randomly picked, and sequenced at the Washington University Genome Sequencing Center (Saint Louis, MO). From each library, 96 clones were sequenced, but many clones did not contain inserts or yielded poor-quality sequences.

Comparison of clone library and T-RFLP analyses

A database of T-RFs was created using 16S rRNA gene sequences from the clone libraries. In silico sequence digestions were performed using the restriction endonucleases HaeIII and HhaI. Five sequenced clones were also analyzed by T-RFLP to determine the impact of the HEX label on fragment migration during capillary electrophoresis. Based on these results, 2 base pairs were added to the predicted T-RFs. Predicted and observed T-RFs were compared to match individual T-RFs with specific clone sequences.

Statistical and phylogenetic analyses

Bray–Curtis similarity matrices were constructed using fourth-root transformations of the normalized T-RFLP peak areas. Two-way crossed and nested analyses of similarity (anosim) were used to test the effects of location, time, and health status among the T-RFLP profiles, while one-way similarity percentage (SIMPER) analyses were used to identify discriminating peaks (primer 6.0; Clarke, 1993). When multiple comparisons were made within anosim, Bonferroni corrections were applied to the significance level. Paired t-tests were used for comparisons between SIMPER values. To determine the significance of differences in normalized T-RF peak areas between healthy and diseased sponges, a power analysis was conducted using r, with a power level of 0.8, significance level of 0.05, sample size of 20, and the observed mean T-RF peak area and standard deviation within each group.

Aligner software (CodonCode Corp.) was used to assemble contigs. Clones with forward and reverse reads that formed unambiguous contigs were selected for data analyses along with single reads with an average PHRED score > 30. Geneious software (Biomatters Ltd) was used to conduct blastn searches (Altschul et al., 1990) against the GenBank database. Sequences were also compared with the SILVA database using the SINA aligner. Potential chimeric sequences in the clone libraries were removed using the UCHIME algorithm (Edgar et al., 2011), as implemented in Mothur (Schloss, 2009). Multiple sets of reference sequences were used for chimera detection (the Silva.gold alignment, a set that included the most closely related GenBank sequence for each clone sequence, and the clone sequences).

Sequences were aligned in mafft (Katoh et al., 2009) using the default settings. Phylogenetic trees were constructed using Bayesian criteria in MrBayes 3.1.2 (Ronquist & Huelsenbeck, 2003), with a general time-reversible (GTR) model of sequence evolution, a gamma distribution of variable substitution rates among sites, and a proportion of invariant sites. A minimum of 500 000 generations were sampled per phylogeny. Maximum-likelihood criteria were implemented in RAxML (Stamatakis et al., 2008), with a CAT-GTR model of sequence evolution.

Leptolyngbya sp. transmission experiment

Cyanobacteria from ARBS lesions were cultured on f/2-Si plates (Guillard & Ryther, 1962; Guillard, 1975). Growth was transferred to fresh f/2-Si plates to minimize bacterial contamination. Using cloning methods described above, the 16S rRNA gene was sequenced from a culture of a red-pigmented cyanobacterium; GenBank blast indicated that this culture is a Leptolyngbya sp. This culture was grown in liquid ASN-III medium (Rippka et al., 1979) supplemented with antibiotics (cycloheximide and nalidixic acid at 25 μg mL−1) to generate substantial biomass prior to being plated onto ASN-III agar plates (n = 10) with sterile gauze strips embedded in the agar surface. All cyanobacterial cultures were grown with a 12-h fluorescent light/12-h dark cycle.

Once substantial cyanobacterial growth was noted (after c. 14 days), the gauze strips were removed and placed into individual resealable plastic bags. Using SCUBA, divers placed the cyanobacterial growth on each gauze strip in contact with a healthy A. cauliformis sponge (n = 20), tying the ends of the strip around the sponge branch to maintain constant contact. Gauze strips embedded in sterile ASN-III agar plates were used as controls (n = 12). After 24 days, sponges were cut c. 3 cm below the gauze strip. Upon return to the laboratory, the gauze strips were removed, and the area of contact with either the agar or the Leptolyngbya culture was photographed and visually evaluated for evidence of ARBS transmission and lesion formation.

Results

Bacterial community analyses

The number of terminal restriction fragments (T-RFs) recovered ranged from 14 to 111 and 28 to 108 for amplifications with bacterial primers 8F-1392R for HaeIII and HhaI digestions, respectively, and from 24 to 95 (HaeIII) and 14 to 115 (HhaI) for reactions with primers 359F-1392R. For all four analyses, average T-RF richness was between 64 and 70 peaks. Two-way crossed anosim indicated that year (2008 or 2009), location (Bahamas or Belize), and condition (healthy, ARBS-affected, or ambient water) significantly influenced bacterial community structure (Table 1). Two-way nested anosim with condition as the primary factor also indicated statistically significant differences between locations and years (Table 1).

Table 1. Analyses of similarity (anosim) R values for T-RFLP data. Condition = healthy sponge, ARBS-affected sponge, or ambient water; Location = Bahamas or Belize; Year = 2008 or 2009 (P < 0.05). For two-way crossed analyses using condition, the reported value for each condition comparison is provided separately below the original comparison. A Bonferroni correction has been applied to the significance values for these comparisons (P < 0.017)
 8F HaeIII8F HhaI359F HaeIII359F HhaI
  1. ARBS, Aplysina red band syndrome.

Two-way crossed anosim
Condition × location0.445, P = 0.0010.510, P = 0.0010.557, P = 0.0010.544, P = 0.001
Healthy × ARBS0.081, P = 0.0170.135, P = 0.0110.176, P = 0.0020.121, P = 0.013
Healthy × water0.932, P = 0.0010.974, P = 0.0010.963, P = 0.0010.981, P = 0.001
ARBS × water0.704, P = 0.0010.789, P = 0.0010.947, P = 0.0010.993, P = 0.001
Condition × year0.395, P = 0.0010.518, P = 0.0010.517, P = 0.0010.464, P = 0.001
Healthy × ARBS0.092, P = 0.0110.155, P = 0.0040.245, P = 0.0010.076, P = 0.057
Healthy × water0.890, P = 0.0010.949, P = 0.0010.866, P = 0.0010.957, P = 0.001
ARBS × water0.653, P = 0.0010.955, P = 0.0010.851, P = 0.0010.947, P = 0.001
Location × year0.122, P = 0.0040.120, P = 0.0010.160, P = 0.0040.185, P = 0.002
Two-way nested anosim
Condition × location0.133, P = 0.0030.226, P = 0.0010.242, P = 0.0010.364, P = 0.001
Condition × year0.137, P = 0.0030.216, P = 0.0010.320, P = 0.0010.155, P = 0.004

The contribution of individual T-RFs to these differences was examined by SIMPER. One-way analyses of all four data sets revealed that the average dissimilarity between healthy and ARBS-affected sponge bacterial communities ranged from 35% to 46% (Table 2). Dissimilarities between healthy sponges and ambient water and between diseased sponges and ambient water were significantly greater (paired t-tests, P < 0.004), ranging from 51% to 67% and 52% to 66%, respectively, suggesting that A. cauliformis sponges maintain a core community of bacteria distinct from the water column, even when infected with ARBS. Dissimilarities between sampling locations (41–53%) and year of collection (41–53%) were also noted. In all SIMPER comparisons, differences between the bacterial communities were not driven by individual T-RFs, because the greatest contributor to the dissimilarity between groups accounted for < 3% of the total. Instead, these outcomes suggest that multiple taxa differ among communities in each comparison.

Table 2. One-way SIMPER average dissimilarity values
Comparisons8F HaeIII (%)8F HhaI (%)359F HaeIII (%)359F HhaI (%)
  1. ARBS, Aplysina red band syndrome; SIMPER, similarity percentage.

Healthy and ARBS45.445.534.939.7
Healthy and water67.462.351.066.3
ARBS and water66.462.752.466.1
Bahamas and Belize52.651.140.749.9
2008 and 200952.751.641.249.5

The 30 most dominant T-RFs (based on peak areas) from each primer set and endonuclease showed considerable variation between healthy and ARBS-affected sponges (Fig. 1; Supporting Information, Table S1). However, only one T-RF was detected solely in diseased sponges (T-RF 360 in HaeIII 359F-1392R data set); all other T-RFs were found in healthy and diseased individuals. Power analyses were used to determine whether changes in T-RF abundance were statistically unexpected. In total, 10 T-RFs were found to change significantly between sponge health condition (six increased in abundance with ARBS and four decreased; Fig. 1). Although interesting, these changes in T-RF abundance between sponge health statuses must also be examined in light of impacts of collection location, date of collection, and the presence of the T-RF in the ambient water column. Some T-RFs were specifically associated with A. cauliformis and were not present in water samples, while others were distributed abundantly in both sponges and seawater (Table S1).

Figure 1.

Percent difference from healthy sponges for the 30 most abundant T-RFs (in order from the most abundant T-RF) in healthy (n = 20) and ARBS-affected (n = 20) Aplysina cauliformis for each primer set and restriction enzyme. (a) 8F-1392R HaeIII, (b) 8F-1392R HhaI, (c) 359F-1392R HaeIII, (d) 359F-1392R HhaI. Asterisks (*) and black bars instead of gray denote significance at P < 0.05 between health conditions in a power analysis.

Phylogenetic and taxonomic diversity of bacteria in healthy and diseased A. cauliformis

After chimeric sequences were removed, 355 sequences were obtained from the eight clone libraries. The sequences of many clones were > 99% similar; thus, only 212 consensus sequences were submitted to GenBank (accession numbers KF286001KF286212). Comparison of clone sequences with those in GenBank and the SILVA database revealed the presence of eight bacterial phyla, including Acidobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, Chloroflexi, Planctomycetes, Proteobacteria, and Verrucomicrobia. In the 8F clone libraries, sequences related to Chloroflexi (17.0%), Synechococcus/Prochlorococcus (17.0%), and Leptolyngbya (16.5%) were the most abundant, representing half of the clones. In 359F libraries, Leptolyngbya (24.3%), Chloroflexi (22.8%), and Actinobacteria (13.2%) accounted for the most common phylogenetic affiliations. Healthy and diseased individuals showed different proportional representations of recovered phyla (Fig. 2).

Figure 2.

Representation of the various phyla within Aplysina cauliformis 8F (top) and 359F (bottom) clone libraries. The number of clones per library is shown under the sample name. BZ = Belize samples; LSI = Bahamas samples; D = ARBS-affected; H = healthy.

Based on GenBank comparisons, 64.3% of the most closely related sequences originated from bacteria reported from marine sponges. When coral-associated bacteria were included, 82.4% of the clones fell within this group. The remaining clones were closely related to organisms recovered from seawater, cuttlefish eggs, microbial mats, basalt, marine sediments, mangrove soil, ballast water, and members of culture collections.

Evaluation of combined clone library and T-RFLP data

In silico digestions of clone library sequences indicated that some clones could be matched to specific T-RFs, while other clone sequences yielded a T-RF that was either < 100 bp or > 500 bp and thus not detectable in our T-RFLP analyses. HaeIII generated fragments of a detectable length more often than that of HhaI and was therefore more useful for providing potential phylogenetic affiliations. As a result, the T-RFs that were found to be significantly different based on health status will be the focus of these comparisons. However, because many bacteria share significant sequence homology of the 16S rRNA gene, digestion products from multiple bacteria can yield the same T-RF, indicating that care should be taken in inferring taxonomic identities (Fierer & Jackson, 2006).

Sequences from the cyanobacterial photosymbiont Sspongiarum generated a T-RF of 290 bp for amplifications with the 8F primer and digestion with HaeIII. Notably, this T-RF also includes free-living Synechococcus and Prochlorococcus species. The 290-bp T-RF, which was detected in 47 of the 50 samples analyzed (40 sponge and 10 water), decreased significantly in ARBS-affected sponges. However, this T-RF was not detected in three ARBS-affected sponges from Belize (one in 2008, two in 2009). In the clone libraries, 15 sequences were most closely related to S. spongiarum clade A (Erwin & Thacker, 2008).

Clone sequences related to Leptolyngbya spp. and for the strain used in the transmission experiments yielded a T-RF of 292 bp for 8F HaeIII digestions. This T-RF, which also includes closely related cyanobacteria such as Phormidium spp., significantly increased in ARBS-affected sponges and was detected in 42 of the 50 samples, but not in two ARBS-affected sponges and six healthy sponges. Interestingly, 359F HaeIII digestions of Leptolyngbya (and Phormidium) spp. yielded a fragment of c. 361 bp, closely matching the 360-bp fragment found only in diseased sponges in the T-RFLP analyses. In the clone libraries, 61 sequences were most closely related to Leptolyngbya, forming a well-supported clade with Leptolyngbya spp. isolated from coral-associated black band disease (Fig. 3a).

Figure 3.

(a) Phylogeny of 16S rRNA gene sequences related to the genus Leptolyngbya recovered from healthy and diseased Aplysina cauliformis; these sequences match HaeIII 8F T-RF 292, which was more abundant in diseased individuals. (b) Phylogeny of 16S rRNA gene sequences related to HaeIII 8F T-RF 324 (Gammaproteobacteria) recovered significantly more often in diseased A. cauliformis sponges than in healthy conspecifics. (c) Phylogeny of 16S rRNA gene sequences related to HaeIII 8F T-RF 148 (Gammaproteobacteria) that was more abundant in healthy A. cauliformis individuals than in ARBS-affected sponges. No close relatives of this Gammaproteobacteria have been cultivated. (d) Phylogeny of 16S rRNA gene sequences related to HaeIII 8F T-RF 305 (Chloroflexi), which was more abundant in healthy A. cauliformis than in diseased conspecifics. Few Chloroflexi have been cultivated; thus, the closest relatives to our clones consist of clone sequences obtained from a variety of sponges and corals. Bold labels indicate sequences from the current study; all others indicate appropriate NCBI accession number. Tree topologies were constructed using Bayesian inference; posterior probability/maximum-likelihood values are shown at each node, with values < 0.50/50 denoted by an asterisk (*). Scale bars represent 0.02, 0.2, 0.04, and 0.2 substitutions per site, respectively.

A TR-F of 324 bp from HaeIII 8F-1392R digestions was significantly more prevalent in ARBS-affected samples, increasing 154% over healthy sponges. This T-RF matched two Gammaproteobacteria sequences in the clone libraries, grouping with sequences recovered from both healthy and diseased individuals of the sponge Ianthella basta (Luter et al., 2010; Fig. 3b). This clade is nested within a larger clade of symbionts reported from diverse marine invertebrates, which included an additional Gammaproteobacteria sequence from our clone libraries.

Three TR-Fs (148, 187, and 305 bp) were significantly more abundant in healthy sponges based on HaeIII 8F-1392R digestions. The 148-bp T-RF was detected in all healthy sponges, all but two diseased sponges, and in only one of 10 water samples, but was 61.4% less abundant in ARBS-affected individuals (Fig. 1a). This T-RF matched three Gammaproteobacteria clone sequences, grouping with a clade of sponge-specific symbionts recovered from the genera Plakortis, Ircinia, and Chondrilla (Hill et al., 2006; Taylor et al., 2007; Mohamed et al., 2010; Fig. 3c). Additional Gammaproteobacteria sequences in the clone libraries were not associated with this T-RF, but were closely related, forming a sponge-specific clade with symbionts recovered from the genera Ancorina, Svenzea, and Agelas (Taylor et al., 2007; Lee et al., 2009; Kamke et al., 2010). The 187-bp T-RF was detected in all 20 healthy sponge samples, 16 of the 20 ARBS-affected sponges, and five of the 10 water samples. This T-RF, which was 57% less abundant in diseased sponges, matched a single Gammaproteobacteria clone sequence (BZ46H8f_a09) that was most closely related to a clone recovered from the sponge Svenzea zeai (97.8% sequence homology). The 305-bp T-RF was detected in 36 of the 50 samples, but was 60.1% less abundant in ARBS-affected samples (Fig. 1). This T-RF was represented by a single clone (LSI107H8f_b07) whose sequence grouped with a diverse set of Chloroflexi sequences reported from sponges and corals, as well as five other sequences from the clone libraries (Fig. 3d). An additional 63 Chloroflexi sequences were in the clone libraries, but none matched the 305-bp T-RF.

In HaeIII digestions of 8F PCR products, six T-RFs were detected only in sponge samples (148, 222, 231, 259, 272, and 402 bp; Table S1). In silico digestions suggested taxonomic affiliations for three of the six sponge-specific T-RFs. T-RF 148 was most closely related to members of the Gammaproteobacteria commonly recovered from examinations of sponge-associated bacterial communities. Similarly, T-RF 272 was most closely related to a member of the Gammaproteobacteria recovered from an Ancorina sponge. Finally, T-RF 222 was most closely related to a member of the Bacteroidetes found associated with both corals and sponges.

Six T-RFs were also only detected in sponges from HaeIII 359F digestions (118, 146, 228, 321, 360, and 451 bp; Table S1); however, only T-RF 360 provided a potential identification (Leptolyngbya sp.) by in silico digestions. In the HhaI digestions, 7 (194, 196, 243, 270, 368, 423, and 452 bp) and 10 (135, 142, 144, 146, 196, 226, 228, 253, 351, and 376 bp) T-RFs appeared to be sponge specific for 8F and 359F primers, respectively. Tentative identifications of these T-RFs include Leptolyngbya (194 bp), Alphaproteobacteria (8F and 359F 196 bp), Deltaproteobacteria (368 bp), and Chloroflexi (146 bp). No match was found for the remaining T-RFs.

Leptolyngbya transmission experiment

Upon removal of the gauze strip, slight discoloration of the sponge tissue was detected in numerous experimental and control sponges, likely due to shading or abrasion of the sponge tissue by the gauze. None of the experimental or control sponges showed visible signs of ARBS lesions. Although cultivated from a lesion, the Leptolyngbya sp. isolate used in the transmission experiment was closely related to, but not the same as, clones recovered from additional ARBS lesions or associated with black band disease in corals (GenBank accession KF286167; Fig. 3a).

Discussion

Contrary to much of the published sponge microbiology literature (reviewed by Taylor et al., 2007; Webster & Taylor, 2011; Thacker & Freeman, 2012), we detected a significant variation in the composition of bacterial communities associated with A. cauliformis when comparing healthy and ARBS-affected sponges, sponges collected in different years, and sponges from different locations within the Caribbean. Although these communities are distinguishable, anosim and SIMPER results found considerable overlap of specific bacterial T-RFs, suggesting that A. cauliformis maintains a conserved ‘core’ group of bacterial associates, largely dominated by members of the Proteobacteria and Chloroflexi, supporting the recent findings of Schmitt et al. (2011), with other bacteria being transient members of the community. Similar findings have been reported for other sponge species (e.g. Taylor et al., 2007; Erwin et al., 2011; Montalvo & Hill, 2011; Schmitt et al., 2011, 2012; Fan et al., 2012), suggesting that this form of ‘leaky vertical transmission’ (Vrijenhoek, 2010) might be common among marine sponges that support high numbers of microbial associates. The core community likely plays a role in providing both nitrogen and carbon to the host sponge (Weisz et al., 2007; Erwin & Thacker, 2008; Fiore et al., 2010; Freeman et al., 2013), highlighting the functionality and potential specificity of these associations. Based on these results, however, we suggest that care should be taken in inferring stability in the composition of sponge-associated bacterial communities across space or time.

This study demonstrated that multiple, complementary approaches provide a more robust understanding of the diversity of sponge-associated bacterial communities, but that spatial and temporal variability must be considered. As evidenced by the inability to match numerous T-RFs to clone sequences, the detection of a T-RF does not provide taxonomic information (e.g. Lui et al., 1997; Marsh, 1999); instead, T-RFLP analyses are most useful for examining changes in community composition across treatments or variables. As the number of detectable T-RFs is generally < 100, this approach can often underestimate community diversity (Rastogi & Sani, 2011), especially as some sponges have now been shown to display bacterial richness in the range of 2500–3000 phylotypes (Webster et al., 2010; Schmitt et al., 2012). Clone libraries, which have provided much of our information regarding bacterial community composition, are also limited by cost, time, and required labor to effectively ‘capture’ the diversity present within a single sample (DeSantis et al., 2007). Thus, although the application of multiple methods is preferred for providing a more complete understanding of bacterial community composition, it was not unexpected that the data from these analyses did not perfectly overlap nor identify the entirety of the bacterial diversity present.

Currently, our understanding of sponge-associated bacterial communities includes both organisms that appear to be specific to sponge tissues [referred to as sponge-specific clusters (SC); e.g. Schmitt et al., 2012; Simister et al., 2012; Taylor et al., 2013] and those that are detectable within both sponge tissue and the ambient environment. In this study, a number of T-RFs were found only within the sponge samples, although recent work revealed the presence of a number of previously described SC at low concentrations in ambient environmental samples (Webster et al., 2010; Taylor et al., 2013). Because the taxonomic information from T-RFs is limited, it remains to be seen whether these organisms are truly sponge specific. However, the potential congruence with the clone library data suggests that six of the 10 putatively identified clones fall within phyla thought to contain SC (Alphaproteobacteria, Gammaproteobacteria, Cyanobacteria, and Chloroflexi; Taylor et al., 2013).

Reports of diseases affecting marine sponges appear to be increasing (Webster, 2007), but little is known about specific causes; only two sponge diseases have presumed etiologic agents (Rützler, 1988; Webster et al., 2002). Our knowledge has been limited by the complexity of the large, diverse, and changing microbial communities associated with many sponges. ARBS causes tissue necrosis, reductions in sponge growth, and changes in sponge physiology and biochemical composition (Gochfeld et al., 2012ab), but the absence of bacterial associates specific to ARBS-affected individuals, other than the presumptive Leptolyngbya sp. responsible for the 360-bp T-RF in HaeIII 359F-1392R digestions, makes identification of the etiologic agent(s) challenging. This is especially true as our transmission experiments using a cultivated Leptolyngbya strain failed to recreate disease in healthy individuals. Several possible explanations exist, including ARBS as a polymicrobial infection, similar to black band disease in hard corals, in which a consortium of organisms appears to be responsible for disease (e.g. Frias-Lopez et al., 2004; Richardson, 2004; Barneah et al., 2007; Sato et al., 2011; Glas et al., 2012); an opportunistic infection under altered homeostatic conditions by an organism regularly found as a part of the associated community; challenge with a nonpathogenic strain; or a viral or eukaryotic pathogen. Studies are underway to examine changes in the composition of associated bacterial communities during the process of infection with ARBS; these results might provide more insight into possible etiologic agents.

To better understand the effect of ARBS on its sponge host, we focused on the T-RFs that significantly changed in abundance between healthy and diseased sponges in power analyses, but we acknowledge that the etiologic agent(s) might not necessarily be numerically dominant. Some of these T-RFs were associated with bacteria detected in a variety of sponges, suggesting that they were likely common and nonpathogenic associates (Hill et al., 2006; Taylor et al., 2007; Lee et al., 2009; Kamke et al., 2010; Mohamed et al., 2010). However, one T-RF was closely related to a Gammaproteobacterium previously reported from both healthy and diseased Ianthella basta sponges (Luter et al., 2010). This bacterium might be an example of an organism that functions as an opportunistic pathogen under altered homeostatic conditions as a regular constituent of the associated community. Additional work with this organism is clearly needed before its potential role in disease can be determined. The increase in the abundance of the T-RF affiliated with Leptolyngbya spp. was intriguing, especially as the closest relative to our clones was a sequence recovered from a coral afflicted with black band disease (Voss et al., 2007). However, the prevalence of this T-RF, which also includes closely related cyanobacteria such as Phormidium spp., in the ambient water suggests that its presence in diseased individuals might simply reflect opportunistic colonization. Finally, the decrease in the abundance of the T-RF associated with the primary photosymbiont S. spongiarum in ARBS lesions suggests a functional consequence of disease. Although the etiologic agent of ARBS was not elucidated from these studies, we have identified a number of organisms that warrant further investigation.

The results of this study reveal the high diversity and variability of A. cauliformis-associated bacterial communities, contradicting a paradigm of general stability in bacterial communities for other sponges within different orders. However, the phyla detected within the associated bacterial community reflect what has been previously reported from a variety of sponge hosts (Hentschel et al., 2012). ARBS infection has significant effects on the bacterial community, but does not apparently introduce dominant new organisms. Interestingly, close relatives of all of the organisms whose abundance changed significantly between sponge conditions either have been previously associated with diseases of marine invertebrates or have been reported from healthy marine invertebrates. These data suggest that there might be bacterial species whose abundance above or below a threshold level is indicative of disease. Additional research is needed to tease apart the complexities of marine disease dynamics.

Acknowledgements

We thank M. Slattery, C. Easson, C. Freeman, J. Stevens, X. Gao, E. Hunkin, L. O'Donahue, and J. Weston for assistance. Funding was provided by collaborative NSF grants #0727833 to JBO, #0727996 to DJG, and #0726944 to RWT, and Smithsonian Marine Science Network grants. The Perry Institute of Marine Science (Bahamas) and Carrie Bow Cay Field Station (Belize) provided field support. All samples were collected under Bahamas Department of Marine Resources or Belize Fisheries Department permits.

Ancillary