Correspondence: Patrick M. Erwin, Center for Advanced Studies of Blanes (CEAB-CSIC), Accés Cala S. Francesc 14, 17300 Blanes, Girona, Spain. Tel.: +34 972 336101; fax: +34 972 337806; e-mail: email@example.com
Microbial symbionts form abundant and diverse components of marine sponge holobionts, yet the ecological and evolutionary factors that dictate their community structure are unresolved. Here, we characterized the bacterial symbiont communities of three sympatric host species in the genus Ircinia from the NW Mediterranean Sea, using electron microscopy and replicated 16S rRNA gene sequence clone libraries. All Ircinia host species harbored abundant and phylogenetically diverse symbiont consortia, comprised primarily of sequences related to other sponge-derived microorganisms. Community-level analyses of bacterial symbionts revealed host species-specific genetic differentiation and structuring of Ircinia-associated microbiota. Phylogenetic analyses of host sponges showed a close evolutionary relationship between Ircinia fasciculata and Ircinia variabilis, the two host species exhibiting more similar symbiont communities. In addition, several bacterial operational taxonomic units were shared between I. variabilis and Ircinia oros, the two host species inhabiting semi-sciophilous communities in more cryptic benthic habitats, and absent in I. fasciculata, which occurs in exposed, high-irradiance habitats. The generalist nature of individual symbionts and host-specific structure of entire communities suggest that: (1) a ‘specific mix of generalists’ framework applies to bacterial symbionts in Ircinia hosts and (2) factors specific to each host species contribute to the distinct symbiont mix observed in Ircinia hosts.
Sponges are sessile, filter-feeding invertebrates that inhabit diverse marine ecosystems and host remarkably abundant and diverse microbial symbiont populations (Taylor et al., 2007; Webster & Taylor, 2011), in some hosts accounting for up to 35% of sponge biomass (Vacelet, 1975) and consisting of hundreds to thousands of symbiont taxa (Webster et al., 2010; Lee et al., 2011). These diverse symbiont communities may enhance sponge holobiont metabolism through microbial processes, including photosynthesis (Erwin & Thacker, 2008), nitrification (López-Legentil et al., 2010), and sulfate reduction (Hoffmann et al., 2005), and can produce defensive secondary metabolites (Flatt et al., 2005) that decrease the susceptibility of host sponges to predation and fouling (Paul & Ritson-Williams, 2008). In turn, sponge-associated bacteria may benefit from the unique microenvironment within host tissues, potentially nourished by ammonia as the end product of animal metabolism and protected from open-ocean predation and UV exposure. Although empirical evidence for symbiont benefit is scarce (Taylor et al., 2007), the high biodiversity of sponge-associated microorganisms and their exclusivity to host sponges suggest that these niche habitats are fertile grounds for marine microorganisms.
The study of sponge microbiology has revealed striking trends in the distribution and specificity of microbial symbionts, yet little is known about the factors that structure these communities. Evidence for sponge-specific microbial lineages, or symbionts shared among unrelated sponge hosts from distant geographic regions (Hentschel et al., 2002; Olson & McCarthy, 2005; Hill et al., 2006), suggests a generalist distribution of symbionts and some degree of conformity in the sponge microbiota (Taylor et al., 2007). However, from a host perspective, the composition and structure of microbial symbiont communities have been recently reported as species-specific, despite the presence of sponge-specific clusters within these communities (Webster et al., 2010; Lee et al., 2011). Multiple factors may influence the composition and structure of microbial symbionts in sponges, including environmental factors such as temperature and nutrient levels (Mohamed et al., 2008b; Webster et al., 2008; Turque et al., 2010; Webster et al., 2011). In addition, the presence of distinct symbiont communities among unrelated sponge hosts from the same habitat (Lee et al., 2009b; Radwan et al., 2010; Webster et al., 2010; Lee et al., 2011) and similar symbiont communities among related sponges from different oceans (Montalvo & Hill, 2011) suggests that host-specific factors also play a role in structuring the sponge microbiota. Uncoupling the effects of environment and host on symbiont communities requires the study of related sponges from the same environment, a sampling design that to date has rarely been employed (Lee et al., 2009b) or limited to a specific portion of overall symbiont communities (Erpenbeck et al., 2002; Thacker & Starnes, 2003).
To investigate the structuring and specificity of bacterial symbionts in closely related high-microbial-abundance (HMA) host sponges, we studied three sympatric species (i.e. sponges with overlapping species distributions) in the genus Ircinia from the Mediterranean Sea. The genus Ircinia is a chemically diverse and symbiont-rich sponge taxon that exhibits high species richness, occurring in shallow to deepwater habitats of tropical and temperate marine environments. The bacterial diversity of Ircinia spp. is consistent with other HMA host sponges, composed largely of sponge-specific sequences from Acidobacteria, Actinobacteria, Chloroflexi, Nitrospira, Poribacteria, and Proteobacteria (Schmitt et al., 2007, 2008; Mohamed et al., 2008a, b, 2010; Webster et al., 2010; Yang et al., 2011). These studies have focused on the microbiota in Caribbean and Indo-Pacific host species, whereas the molecular diversity of microbial symbionts in Mediterranean Ircinia hosts has been addressed by a single study that focused specifically on cyanobacteria (Usher et al., 2004).
In this study, we characterized the bacterial communities in the common Mediterranean species, Ircinia fasciculata, I. variabilis, and Ircinia oros, using replicated clone libraries of 16S rRNA gene sequences and compared the richness, diversity, structure, and specificity of symbiont communities among these congeneric and sympatric hosts. Phylogenetic analyses were conducted to compare symbionts in Mediterranean Ircinia spp. with previously described sponge-associated bacteria, including sequences derived from Caribbean Ircinia spp. In addition, we resolved the phylogenetic relationships among the three host sponges using ribosomal and mitochondrial DNA markers, thus allowing for the determination of symbiont specificity within a well-defined host phylogenetic context. Herein, we use the terms ‘semi-sciophilous’ and ‘photophilic’ to refer to benthic communities commonly inhabited by Ircinia spp. Semi-sciophilous (‘shade-loving’) communities are found in low-irradiance and cryptic (i.e. less exposed) habitats, such as vertical walls and overhangs, while photophilic (‘light-loving’) communities occur in high-irradiance and exposed habitats, such as horizontal substrates.
Materials and methods
The marine sponges I. fasciculata (Pallas 1766; Fig. 1a), I. variabilis (Schmidt, 1862; Fig. 1b), and I. oros (Schmidt, 1864; Fig. 1c) were collected from shallow (< 20 m) littoral zones at three neighboring sites (< 12 km apart) along the Catalan Coast (Spain) in the northwestern Mediterranean Sea. Ircinia fasciculata colonies were sampled at Punta de S'Agulla (Blanes; 41°40′54.87″N, 2°49′00.01″E), I. variabilis at Mar Menuda (Tossa de Mar; 41°43′13.62″N, 2°56′26.90″E), and I. oros at Punta Santa Anna (Blanes; 41°40′21.48″N, 2°48′13.55″E) by SCUBA during three consecutive days in March 2010. At each site, ambient seawater samples (500 mL) were collected simultaneously and in close proximity (< 1 m) to sampled sponges. Sponge and seawater samples were transported in an insulated cooler to the laboratory (ca. 2 h transit time), where sponge samples were preserved in 100% ethanol and stored at −20 °C and seawater samples were concentrated on 0.2-μm filters and stored at −80 °C.
Transmission electron microscopy (TEM)
To visualize the bacterial diversity present in I. variabilis, I. fasciculata, and I. oros, small ectosomal and choanosomal tissue pieces (ca. 4 mm3) were dissected and fixed in a solution of 2.5% glutaraldehyde and 2% paraformaldehyde, buffered with filtered seawater. Samples were incubated in the fixative mixture overnight at 4 °C and subsequently rinsed with filtered seawater to remove fixative and then dehydrated in a graded ethanol series and embedded in Spurr resin at room temperature. A Reichert Ultracut microtome was used to produce ultrathin sections (ca. 60 nm) that were contrasted with uranyl acetate and lead citrate for ultrastructural observation (Reynolds, 1963). TEM observations were made at the Microscopy Unit of the Scientific and Technical Services of the University of Barcelona on a JEOL JEM-1010 (Tokyo, Japan) coupled with a Bioscan 972 camera (Gatan, Germany).
DNA extracts were prepared separately for three individuals of each host sponge species (including ectosomal and choanosomal tissue) and three samples of concentrated seawater (one from each collection site) using the DNeasy® Blood & Tissue kit (Qiagen®). Full-strength and 1 : 10 diluted DNA extracts were used as templates in PCR amplifications.
16S rRNA gene sequence clone libraries
The universal bacterial forward primer 8F (Reysenbach et al., 1994) and reverse primer 1509R (Martínez-Murcia et al., 1995) were used to amplify approximately 1500-bp fragments of bacterial 16S rRNA gene sequences from all sponge and seawater extracts. Total PCR volume was 50 μL, including 10 pmol of each primer, 10 nmol of each dNTP, 1× Reaction Buffer (Ecogen), and five units of BIOTAQ™ polymerase (Ecogen). Thermocycler reaction conditions were an initial denaturing time of 2 min at 94 °C, followed by 30 cycles of 1 min at 94 °C, 0.5 min at 50 °C, and 1.5 min at 72 °C, and a final extension time of 2 min at 72 °C. To minimize PCR amplification biases, a low annealing temperature and low cycle number were used and three separate reactions were conducted for each sample. PCR amplification products were gel-purified and cleaned using the QIAquick Gel Extraction kit (Qiagen®). Triplicate PCR products were combined and quantified using a Qubit™ fluorometer and Quant-iT™ dsDNA Assay kit (Invitrogen™). Purified PCR products (ca. 75 ng) were ligated into plasmids using the pGEM®-T Vector System (Promega).
Individual clones were PCR-screened using vector primers, and clones with approximately 1500-bp inserts were purified and sequenced at Macrogen, Inc. A single sequencing reaction was performed for all clones to recover the 5′-end of 16S rRNA gene sequences using the primer 800R (5′-TAC CAG GGT ATC TAA TCC-3′). Ambiguities on sequencing reaction ends were excluded by trimming sequences at the 5′-end to the highly conserved Escherichia coli site 54 and at the 3′-end to E. coli site 754, yielding sequences ranging from 613 to 725 bp (average length = 683 bp) that were used for diversity calculations and phylogenetic metrics. In addition, bidirectional sequencing reactions with vector primers were performed to recover near full-length 16S rRNA gene sequences (range = 1423–1523 bp; average = 1491 bp) of representative clones (total = 39) for phylogenetic analyses. Raw sequence data were processed in Geneious (Drummond et al., 2010), and low-quality sequencing reads were discarded. Sequences were screened for sequencing anomalies (e.g. chimeras) using Mallard (Ashelford et al., 2006) and a reference 16S rRNA gene sequence from E. coli (GenBank accession no. U00096). Putative sequence anomalies were subsequently confirmed or refuted using Pintail (Ashelford et al., 2005) and two related reference 16S rRNA gene sequences. All confirmed chimeras were removed from the dataset. Quality-checked sequences are archived in GenBank under accession nos. JN655200–JN655511.
Operational taxonomic unit (OTU) assignment and composition
Clone library sequences were ascribed to OTUs calculated at different sequence identity percentages (99%, 97%, 95%, 90%, 85%, and 80%) using the nearest neighbor algorithm, as implemented in the mothur software package (Schloss et al., 2009). The observed OTU richness (Sobs) for each bacterial community was compared across different OTU thresholds, calculated as total (combined sequence data by source) and average values (separated sequence data by samples). All subsequent OTU-based metrics were conducted using an OTU classification at 99% sequence identity. Representative sequences from each 99% OTU were analyzed using a nucleotide–nucleotide blast search (Altschul et al., 1990) to find the most closely related sequence, and using the Ribosomal Database Project II (Cole et al., 2007) sequence classifier to assess taxonomic affiliations.
Diversity, structure, and similarity of bacterial communities
The diversity of recovered bacterial communities in Ircinia spp. and ambient seawater was compared using multiple metrics for OTU richness, dominance, and evenness, calculated in the mothur software package. Richness calculations included observed species richness (99% OTUs), rarefaction analysis, and expected species richness using the Chao1 estimator (SChao1). The effect of increasing sequencing effort on OTU richness was estimated using the Boneh calculation (Boneh et al., 1998). Dominance metrics included Simpson's inverse index (1/D) and the Berger-Parker index (d), and evenness calculations included the Simpson's evenness measure (E1/D) and Smith & Wilson's evenness index (EVAR).
Genetic diversity of symbiont communities was compared among host species and seawater sources using nonparametric tests for homogeneity of molecular variance (homova), as implemented in the mothur software package, and an analysis of molecular variance (amova; Stewart & Excoffier, 1996). These tests provide OTU-independent assessments of genetic variation and differentiation within and among bacterial communities inhabiting Ircinia spp. and seawater. Distances were calculated for amova using the Tajima and Nei algorithm with α = 0.05. Using the arlequin software package, version 3.5 (Excoffier & Lischer, 2010), a hierarchical partitioning of genetic variation was assessed across different levels (among sources, among replicates within sources, and among sequences within replicates), and pairwise variation among sources was computed as FST, with statistical significance based on 1000 permutations. Distributions of unique lineages among bacterial communities were examined using a phylogenetic lineage-sorting test (P-test; Martin, 2002), also referred to as the parsimony test (Schloss, 2008), as implemented in the mothur software package.
To determine the distribution of bacterial OTUs within each community, OTU rank-abundance plots were constructed and compared to a fitted log-series and geometric series distributions by calculating the Kolmogorov–Smirnov test statistic (Dmax, α = 0.05 and 0.01). Community similarity among sources was calculated as Bray–Curtis similarity values and visualized in complete linkage similarity dendrograms using primer v6 (Plymouth Marine Laboratory, UK) computer software. Finally, the integral form of LIBSHUFF (∫-LIBSCHUFF; Schloss et al., 2004) was used to test pairwise differences in bacterial communities from each source, implemented in the mothur software package with significance values based on 100 000 randomizations and adjusted using Bonferroni corrections for multiple pairwise comparisons (Sokal & Rohlf, 1995).
Phylogenetic analysis of bacterial symbionts
Phylogenetic analyses of 16S rRNA gene sequences were conducted to determine the affiliations between sequences recovered from Ircinia spp. herein and previously characterized sponge symbionts. In particular, sequences from recent studies characterizing sponge-associated bacterial communities in Ircinia species from the Caribbean were targeted. Publicly available datasets for Ircinia felix (Schmitt et al., 2007, 2008) and Ircinia strobilina (Mohamed et al., 2008b; Yang et al., 2011) were retrieved from GenBank and grouped into 99% OTUs, following the methods employed for the clone libraries herein. Representative sequences from each OTU for I. felix (n =102) and I. strobilina (n =156), top matching sequences from blast searches (n =189), and near full-length 16S rRNA gene clones (n =39) and partial 16S rRNA gene clones (n =87) from this study were aligned to the greengenes reference database (DeSantis et al., 2006) using the mothur software package, with an outgroup sequence from Archaea (Haloarcula vallismortis, GenBank accession no. D50581). Maximum-likelihood (ML) phylogenetic trees were constructed in RAxML (Stamatakis et al., 2005) using the General Time Reversible model with a gamma distribution of variable substitution rates among sites (GTR + G). Data were resampled using 100 bootstrap replicates, and a thorough ML search was conducted to optimize the topology and recover the best-scoring tree. Owing to the variable length of 16S rRNA gene sequences being compared (422–1526 bp), a binary backbone constraint tree was constructed from long (> 1000 bp) sequences and used to restrict topology changes when introducing short (< 1000 bp) sequences into the phylogeny. This method allowed for: (1) accurate reconstruction of deeper nodes, based on the most informative sequences, and (2) precise placement of short 16S gene sequence fragments near terminal nodes, for comparative analysis with previous Ircinia sp. datasets (e.g. excised and sequenced denaturing gradient gel electrophoresis bands; Schmitt et al., 2007).
Molecular identification of host sponges
Sponge samples were identified to species using morphological observations, including gross morphology and fiber characteristics, and ribosomal and mitochondrial molecular markers. A segment of nuclear ribosomal DNA, corresponding to the 3′-end of the 5.8S subunit, the entire second internal transcribed spacer (ITS-2) region, and the 5′-end of the 28S subunit, was PCR-amplified following the method described by Erwin & Thacker (2007). A fragment of the mitochondrial gene cytochrome oxidase I (COI), corresponding to the standard barcoding partition (i.e. ‘Folmer’ partition) (Folmer et al., 1994; Herbert et al., 2003) and the I3-M11 partition (Erpenbeck et al., 2006), was PCR-amplified using a degenerated version of the universal barcoding forward primer dgLCO1490 (Meyer et al., 2005) and the reverse primer COX1-R1 (Rot et al., 2006). PCR amplicons were gel-purified and cleaned using the QIAquick Gel Extraction kit (Qiagen®) and ligated into plasmids using the pGEM®-T Vector System (Promega). Individual clones were PCR-screened using vector primers and then purified and bidirectionally sequenced at Macrogen, Inc. Consensus sequences were constructed by sponge individual (COI) or by clone (rRNA gene sequences), the latter to account for potential intragenomic variation in ITS-2 copies (Wörheide et al., 2004), and archived in GenBank under accession nos. JN655171–JN655199.
Consensus sequences for rRNA gene sequences were aligned using MAFFT (Katoh et al., 2002) with outgroup sequences from Smenospongia aurea (Dictyoceratida; Thorectidae) (Erwin & Thacker, 2007). Alignment of consensus COI sequences included the congeneric Caribbean species I. strobilina (Erpenbeck et al., 2009) and the outgroup species Hippospongia lachne (Dictyoceratida; Spongiidae) (Lavrov et al., 2008). Pairwise genetic distance matrices (uncorrected p-distance) were constructed using the software package mothur. For rRNA gene sequences, ML phylogenies were constructed using PHYML (Guidon & Gascuel, 2003) and the Hasegawa–Kishino–Yano model with a gamma distribution of variable substitution rates among sites (HKY + G), as suggested by FINDMODEL; data were resampled using 100 bootstrap replicates. Neighbor-joining trees were constructed using Geneious and the HKY model of nucleotide substitution; data were resampled using 1000 bootstrap replicates.
Transmission electron microscopy
Electron microscopy observations of host sponge tissue revealed a high density of bacterial cells (Fig. 1). Characteristic of HMA sponges, examined sections were comprised primarily of bacterial cells, with only occasional sponge cells (archaeocytes) and structural elements (spongin and collagen fibers). In I. fasciculata and I. variabilis, ectosomal (peripheral) tissue sections revealed dense populations of ‘Candidatus Synechococcus spongiarum’ identifiable by their characteristic spiral thylakoid membranes encompassing the perimeter of the cells. Synechococcus spongiarum cells dominated the ectosomal regions of host tissue and were observed to be actively dividing, exhibiting several stages of binary fission (Fig. 1d and e). In I. oros, no S. spongiarum symbionts or other cyanobacterial cells were observed in ectosomal tissue, rather a high density of heterotrophic bacteria occurred consisting of multiple bacterial morphotypes, some of which were also showing active cell division (Fig. 1g). Sections from deeper tissue regions (choanosome) of I. variabilis revealed the absence of S. spongiarum cells and the proliferation of heterotrophic bacteria cells, many with similar morphotypes to those observed in I. oros ectosomal tissue (Fig. 1f).
Bacterial OTUs and sampling coverage
Bacterial 16S rRNA gene sequences recovered from I. fasciculata (n =77), I. variabilis (n =80), I. oros (n =82), and ambient seawater (n =73) grouped into 124 OTUs, defined by 99% or greater sequence similarity. Grouping sequences according to lower identity thresholds reduced the number of OTUs within each source community; however, the same trend in comparative OTU richness among communities was observed (Supporting Information, Fig. S1). Coverage estimates revealed that sampled sponge-associated bacterial communities represented the majority of expected diversity (76.3% in I. variabilis, 80.5% in I. fasciculata, and 81.7% in I. oros), and doubling the sampling effort conducted herein was predicted to produce few (5–6) new OTUs. In contrast, recovered seawater bacterial communities represented less than half (43.8%) of the total expected diversity, and doubling the sampling effort was predicted to produce 12 new OTUs, indicating that more extensive sampling is required to fully characterize bacterioplankton diversity. Similarly, rarefaction analyses showed the sponge-associated bacterial communities beginning to reach OTU saturation, whereas seawater bacterial communities continued to steadily accumulate new OTUs (Fig. S2).
Composition of bacterial communities
Sponge-associated bacterial communities exhibited high diversity and were comprised of nine phyla, including representatives from four classes of Proteobacteria (Table 1). Sequences affiliated with Proteobacteria and Bacteroidetes were recovered from all three Ircinia host species and seawater, whereas Acidobacteria, Nitrospira, Chloroflexi, and Gemmatimonadetes were exclusively found in Ircinia-associated communities. Within the sponge-associated microbiota, Deltaproteobacteria-affiliated sequences comprised an abundant component of microbial communities in all three host species (> 15% total clones), although phylum-level differentiation of symbiont communities among hosts was apparent (Table 1). Ambient seawater bacteria represented 10 phyla and included four phyla not detected in sponge-associated communities (Table 1).
Table 1. Composition of bacterial communities in Ircinia spp. and ambient seawater
Percentage of total clones is shown by bacterial phyla (classes of Proteobacteria indicated with an asterisk). Values in parenthesis depict the number of 99% OTUs within each lineage.
The vast majority of 16S rRNA gene sequences recovered from I. fasciculata (87.0%), I. variabilis (85.0%), and I. oros (81.0%) matched most closely to other sponge-derived bacterial sequences, generally at high sequence identity levels (≥ 97%; Fig. 2). Sequences from Ircinia spp. not associated with sponge-derived clones were most commonly matched to sequences from marine sediment (n =19), corals (n =4), and seawater (n =3). Most sequences obtained from ambient seawater bacteria (98.6%) were closely related to sequence from other bacterioplankton sources, with nearly all sequences (90.3%) matching at very high sequence identity levels (≥ 99%; Fig. 2).
Diversity, structure, and similarity of bacterial communities
Seawater bacterioplankton communities were clearly differentiated from the sponge-associated bacterial communities, exhibiting higher OTU richness (observed and expected), lower dominance indices, and higher evenness indices (Table 2). Among the three host sponges, bacterial communities in I. variabilis and I. oros exhibited very similar richness and evenness values (Table 2). By comparison, the bacterial community in I. fasciculata exhibited lower OTU richness and a less even (more dominant) community structure; however, overlapping confidence intervals for index values were observed among all Ircinia hosts (Table 2), indicating similar symbiont diversity across the three host species. Genetic diversity analyses revealed similar trends to OTU-based metrics, with bacterial communities in seawater exhibiting significantly higher genetic diversity compared to those in Ircinia sponges (homova, P <0.005) and no significant pairwise differences in genetic diversity among sponge-associated communities (P >0.237; Table 3).
Table 2. Diversity metrics comparing the richness, dominance, and evenness of bacterial communities in Ircinia spp. and seawater
Lower and upper 95% confidence intervals are shown in parentheses where available.
Observed OTUs Sobs
Expected OTUs SChao1
Simpson Index 1/D
Smith & Wilson Evar
Simpson Index E1/D
Table 3. Pairwise statistical comparisons of genetic diversity and community structure of bacterial communities in Ircinia fasciculata (IF), Ircinia variabilis (IV), Ircinia oros (IO), and seawater (H2O)
Rank-abundance analyses revealed the presence of a few, dominant OTUs and numerous, rare OTUs within each clone library, consistent with a log-series distribution of bacterial OTUs (Dmax, P >0.05). The most dominant OTU in the seawater community accounted for 8.0% of the total community, and singleton OTUs (n =41) accounted for over 80% of all OTUs. In fact, only five OTUs (9.4%) were recovered more than twice in seawater clone libraries. Dominant OTUs in I. variabilis and I. oros accounted for 15% of each community, with the top three most abundant OTUs accounting for one-third of the total community. In I. fasciculata, a single OTU (corresponding to the cyanobacterium S. spongiarum) accounted for over one-fourth of all recovered clones and the top three bacterial OTUs accounted for nearly half (48.1%) of all recovered clones. Approximately half of the OTUs recovered from I. variabilis (n =19, 55.9%), I. oros (n =15, 45.5%), and I. fasciculata (n =15, 51.7%) were singletons.
Minimal overlap in OTU composition was observed between seawater and sponge-derived sequences, consisting of only two shared OTUs between I. variabilis and seawater that accounted for a small portion of the total clones from each library (2.5% and 5.3%, respectively). Comparisons among the sponge-associated bacterial communities revealed four dominant bacterial OTUs (ca. 25% of each clone library) present in all three host sponge species, with an additional seven OTUs shared between I. fasciculata and I. variabilis and six OTUs shared between I. variabilis and I. oros (Fig. 3). No additional bacterial OTUs were shared between I. fasciculata and I. oros. The majority of OTUs recovered for each host species’ community consisted of symbionts recovered from a single host species (Fig. 3) and represented rare OTUs, commonly appearing once (65.5% of specific OTUs) or twice (25.5%) in clone libraries.
Consistent with patterns of symbiont OTU overlap among host sponges, overall bacterial community similarity values were lowest between I. fasciculata and I. oros (16.7%) and the symbionts in these two hosts differed significantly in community structure (LIBSHUFF, P <0.001; Table 3). The microbiota in I. variabilis exhibited higher similarity to I. fasciculata symbiont communities (36.6%), where no significant difference in symbiont structure was detected (P >0.639; Table 3), than to bacterial symbionts in I. oros (31.1%), where significant differences in community structure were detected (P <0.05; Table 3). Further, distinct phylogenetic lineages of symbionts (P-test, P <0.001) were observed among all pairwise comparisons between source communities (Table 3), and amova revealed significant genetic differentiation among all four sources (P <0.001; 62.1% of genetic variation) and among replicates within each source (P <0.001; 37.9% of genetic variation).
Phylogenetic analysis of bacterial symbionts
Phylogenetic analysis revealed that sequences recovered from I. fasciculata, I. variabilis, and I. oros formed 56 monophyletic sequence clusters (Fig. 4). Nearly half (48.2%) of these clusters were comprised exclusively of sponge-associated bacterial sequences, with an additional 11 clusters (19.6%) consisting of sponge and coral-associated clones. The remaining 18 clades (32.1%) contained nonsymbiotic representatives, most commonly derived from sediment (n =8) and seawater bacterioplankton (n =5). Previously described bacterial sequences from Caribbean Ircinia spp. were present in 21 of the 56 clades (37.5%), with I. felix from Florida (n =16) and I. strobilina from the Bahamas (n =12) more commonly presenting related symbionts than I. strobilina from Florida (n =3). Only 1 of the 56 clades (1.8%) was comprised of sequences exclusively from Ircinia spp. (Fig. 4c). Notably, symbiont sequences from the unrelated host species Aplysina aerophoba (Order Verongida) from the Mediterranean, Ancorina alata (Order Astrophorida) from New Zealand, Rhopaloeides odorabile (Order Dictyoceratida, Family Spongiidae) from Australia, and Xestospongia muta (Order Haplosclerida) from the Caribbean were also prevalent in these 56 clusters (32.1%, 19.6%, 14.3%, and 10.7% of clusters, respectively; Fig. 4).
Phylogenetic analysis was also used to compare the host specificity of bacteria observed among the three Ircinia spp. with host specificity on a broader scale. The four common bacterial OTUs (i.e. generalist symbionts) present in all three Mediterranean Ircinia spp. were related to sequences derived from unrelated host sponge species, nonsponge (coral) hosts, and environmental (sediment) clones (Fig. 4). These generalist symbionts corresponded to one Deltaproteobacterium (IRC001) and three Gammaproteobacteria (IRC006, IRC012, and IRC019), with only IRC006 forming a sequence cluster comprised exclusively of sponge-derived clones. The remaining generalist symbiont OTUs formed sequence clusters with not only sponge-derived clones but also coral-derived (IRC001) and sediment-derived (IRC012 and IRC016) bacteria. Further, even the bacterial OTUs identified as specific to a single species of Ircinia in the clone libraries constructed herein (Fig. 3) were closely related to sequences derived from unrelated sponge hosts and environmental samples (Fig. 4).
Molecular identification of host sponges
Consensus COI sequences from I. fasciculata (n =3) and I. variabilis (n =3) and I. oros (n =2) individuals yielded a 1213-bp fragment encompassing the standard barcoding region (‘Folmer’ partition; 676 bp) and the I3-M11 region (537 bp). No intraspecific variation was observed among species, and I. fasciculata and I. variabilis exhibited identical sequences across the entire fragment length (Table 4). Ircinia oros was clearly differentiated from I. fasciculata and I. variabilis, with higher variability observed in the Folmer partition of COI (Table 4). Interestingly, I. fasciculata and I. variabilis sequences were more closely related to the Caribbean species I. strobilina (0.6% divergence) than the sympatric Mediterranean species I. oros (1.8% divergence).
Table 4. Pairwise genetic distance (p-distance) among Ircinia spp. and Hippospongia lachne (order Dictyoceratida, family Spongiidae) for the ‘Folmer’ partition (upper right; 676-bp fragment) and the I3-M11 partition (lower left; 537-bp fragment) of the mitochondrial gene cytochrome oxidase subunit I
Values shown as percentages.
Consensus rRNA gene sequences from I. fasciculata (n =6), I. variabilis (n =7) and I. oros (n =8) clones yielded a 650- to 654-bp fragment encompassing the 3′-end of the 5.8S subunit (50 bp), the entire ITS-2 region (243–248 bp), and the 5′-end of the 28S subunit (356–357 bp). Low levels of intragenomic polymorphisms were observed and variable sites occurred in all rRNA gene subunits and regions as inconsistent point mutations (i.e. occurring at different positions). Partial 28S rRNA gene sequences differentiated I. oros from I. fasciculata (1.16% divergence) and I. variabilis (1.10% divergence) but did not resolve the latter two species. ITS-2 sequences exhibited the highest variability and clearly differentiated I. fasciculata from I. variabilis (2.97% ±0.38 divergence), with variability among species (range = 2.43–3.63%) consistently greater than variability within I. fasciculata (range = 0.40–2.02%) and I. variabilis (range = 0.00–0.00%). Phylogenetic analysis of combined ITS-2 and 28S rRNA gene sequences resolved each species into monophyletic clades with high bootstrap support and showed that I. fasciculata and I. variabilis were more closely related to each other than to I. oros (Fig. S3), similar to the bacterial symbiont communities inhabiting these host species (Fig. 5).
The Mediterranean sponges I. fasciculata, I. variabilis, and I. oros were shown to host-dense communities of phylogenetically diverse bacterial symbionts, as occurs in congeneric hosts in the Caribbean and more distantly related HMA host sponges. Bacterial communities associated with Ircinia spp. were clearly differentiated from ambient bacterioplankton communities, in terms of richness, diversity, and composition, and were comprised primarily of sponge-specific symbionts related to bacterial sequences derived from related and unrelated host sponge species. Despite the generalist nature of these symbionts among sponge hosts, phylogenetic metrics revealed that each species of Ircinia harbored a unique bacterial community. These differences in symbiont communities among congeneric and sympatric species suggest that factors specific to each host play a role in structuring bacterial symbiont communities in marine sponges.
Host-specific symbiont structuring is particularly notable as the vast majority of the bacterial symbiont communities found in Mediterranean Ircinia spp. were closely related (> 97% identity) to sequences derived from other sponges, often hosts from different oceans. The observed generalist nature of bacterial symbionts, combined with the host-specific structure of entire symbiont communities, suggests that each Ircinia species harbors a specific mix of sponge generalist symbionts. A similar trend was observed in a meta-analysis of bacterial symbionts in 10 sponge species that revealed significant pairwise differences in symbiont community structure among all hosts, despite the predominance of generalist symbionts comprising each microbiota (Taylor et al., 2007). It is important to note that the widespread distribution of generalist symbionts does not imply that they are ubiquitous components of the sponge microbiota. On the contrary, unrelated and geographically distant hosts often share symbionts that are absent from the microbiota of neighboring sponges with close taxonomic affiliations. In the current study, for example, an Actinobacteria-affiliated symbiont (IRC046) of I. oros matched closely (99%) to a symbiont sequence from the Indo-Pacific sponge Theonella swinhoei yet was absent from the microbiota of I. fasciculata and I. variabilis. As a result, the patchy distribution of generalist sponge symbionts can contribute to the differentiation of bacterial communities among hosts.
The persisting question is which factors dictate the distribution of generalist symbionts and account for the observed dissimilarities in microbial community structure among different sponges. To classify the numerous physical, chemical, and biological conditions that may structure symbiont communities, a framework is presented which divides putative factors into four categories based on their source (environment vs. host related) and prevalence among host sponge species (shared vs. specific; Fig. 6). The influence and number of factors in each category depends on the geographic and taxonomic scope of each study. Congeneric and sympatric host species were examined herein; thus, numerous environmental and host-related factors were shared among these species and cannot account for the reported differences in symbiont communities. However, factors specific to different host species may represent a source of differentiation for Ircinia-associated bacterial communities.
Environmental factors specific to distinct habitats (i.e. habitat-specific factors; Fig. 6) may play a role in structuring symbiont communities, as the host species investigated exhibit different zonation patterns within littoral benthic landscapes. Ircinia variabilis and I. oros occur preferentially in semi-sciophilous communities, commonly inhabiting vertical relief structures where much of the sponge diversity in these habitats exists. In contrast, I. fasciculata is more prevalent in photophilic communities, characterized by high-irradiance conditions and dense algal growth. Irradiance levels are primary factor in structuring sponge assemblages in the Mediterranean, likely as an indirect consequence of stimulating algal growth and competitive pressures (Uriz et al., 1992). The distribution of host species within the local landscape may thus have ecological consequences for sponge–symbiont interactions and dictate symbiont composition by imposing functional performance pressures. For example, the dense populations of cyanobacteria in I. fasciculata may contribute to host nutrition and growth via the transfer of photosynthetic byproducts (Arillo et al., 1993; Erwin & Thacker, 2008; Freeman & Thacker, 2011), thereby enhancing competitive ability and allowing this species to thrive in algal-dominated habitats.
Host factors specific to each sponge species (i.e. species-specific factors; Fig. 6) may also influence the composition of symbiotic bacterial communities in sponges, as suggested by the observed correlation between Ircinia spp. phylogenies and symbiont community similarity. Microbial symbionts maintained by vertical transmission, or direct parent-to-offspring passage, are linked to the evolutionary trajectory of host sponges and thus more similar among hosts sharing a more recent ancestor. Vertical transmission has been observed in several marine sponges (Usher et al., 2001; Ereskovsky et al., 2005; Oren et al., 2005; Enticknap et al., 2006; de Caralt et al., 2007; Sharp et al., 2007; Steger et al., 2008; Lee et al., 2009a; Webster et al., 2010), including Ircinia species where successive generations are seeded as larvae with diverse bacterial symbionts (Schmitt et al., 2007, 2008). Therefore, while periodic horizontal transmission of microbial symbionts (i.e. environmental acquisition) may contribute to the homogenization of symbiont communities in unrelated sponge hosts (Schmitt et al., 2008), recurrent vertical transmission may dictate symbiont composition over shorter evolutionary time scales owing to legacy effects of the microbial inheritance. Notably, environmental and host-related factors are not mutually exclusive and may act in concert to structure bacterial symbiont communities in marine sponges.
Discerning the contribution of host ancestry to symbiont structure requires the accurate resolution of evolutionary relationships among sponges based on molecular data and phylogenetic analyses. The current study showed that COI mtDNA sequences (including extended coverage of the I3-M11 partition) were unable to differentiate I. fasciculata from I. variabilis. The utility of the I3-M11 region in differentiating related sponge species (Reveillaud et al., 2011) and as a population genetics marker for intraspecific differentiation in sponges (López-Legentil & Pawlik, 2009; Xavier et al., 2010) suggests that this gene region is either highly conserved within the genus Ircinia or that I. fasciculata and I. variabilis are closely related species. Indeed, some degree of controversy surrounds the taxonomic status of I. fasciculata and I. variabilis (Pronzato et al., 2004), and these species can be difficult to differentiate in the field (Maldonado et al., 2010). Analysis of the second internal transcribed spacer region (ITS-2) of the nuclear ribosomal operon revealed consistent genetic differentiation between I. fasciculata and I. variabilis, indicating that some molecular markers are able to delineate these species. The variable resolution of different genetic markers among sponge taxa indicates that multiple target genes are required to accurately resolve the relationships among species and determine how these relationships may affect the community structure of bacterial symbionts in sponges.
Based on the sponge-specific nature of individual symbionts and host-specific structure of symbiont communities, our results reveal a complex picture of host–symbiont specificity in HMA sponges and suggest that a ‘specific mix of generalists’ framework characterizes bacterial communities in Ircinia spp. from the Mediterranean. Further, by comparing symbiont communities in congeneric sponges from the same environment, our results suggest that factors specific to the each sponge species may structure the distinct symbiont mix observed in Ircinia hosts, including habitat-specific conditions (e.g. irradiance) and host evolutionary history. Additional research on the spatiotemporal dynamics of bacterial symbionts in sponges and controlled experimental manipulations of sponge holobionts are required to further unravel the multiple and potentially interactive factors that structure the complex sponge microbiota.
We thank Dr. M. Uriz (CEAB) for help with sponge taxonomy and sample identification and F. Crespo (CEAB) for field assistance. This research was supported by the Spanish Government projects CTM2010-17755 and CTM2010-22218 and by the US National Science Foundation under grant 0853089.