Temporal molecular and isotopic analysis of active bacterial communities in two New Zealand sponges


Correspondence: Peter Deines, Institute of Natural & Mathematical Sciences, Massey University, Private Bag 102904, Auckland, New Zealand. Tel.: (+64) 9 414 0800 ext. 41493; fax: (+64) 9 443 9779; e-mail: p.deines@massey.ac.nz


The characterization of changes in microbial communities is an essential step towards a better understanding of host–microbe associations. It is well established that sponges (phylum Porifera) harbour a diverse and abundant microbial community, but it is not known whether these microbial communities change over time. Here, we followed two sponge species (Ancorina alata and Tethya stolonifera) over a 2-year sampling period using RNA (16S rRNA)–based amplicon pyrosequencing and bulk stable isotope analysis (δ13C and δ15N). A total of 4468 unique operational taxonomic units (OTUs) was identified, which were affiliated with 26 bacterial phyla. Bacterial communities of both sponge species were remarkably stable throughout the monitoring period, driven by a small number of OTUs that dominated their respective communities. Variability of sponge-associated bacterial communities was driven by OTUs that were low in abundance or transient over time. Stable isotope analysis provided evidence of both bacteria- and host-derived nutrients and their variability throughout the season. While δ15N values were similar, significant differences were found in δ13C of sponge tissue, indicative of a varying reliance on particulate organic matter as a carbon source. Further temporal studies, such as those undertaken here, will be highly valuable to identify which members of a sponge bacterial community are truly symbiotic in nature.


Marine sponges (phylum Porifera) are an ecologically important reef taxon as they are diverse, abundant, and they link benthic and pelagic habitats due to their high filtration rates (Diaz & Ruetzler, 2001; Bell, 2008). Sponges are a rich source of bioactive metabolites and have attracted much research interest from microbiologists due to their diverse and abundant microbial communities (Hentschel et al., 2002; Taylor et al., 2007; Webster & Taylor, 2012). Microorganisms can occupy up to 35% of sponge volume with all three domains of life (Bacteria, Archaea and Eukarya) known to live symbiotically within sponges (Hentschel et al., 2012), with affiliations to more than 32 bacterial phyla and candidate phyla (Schmitt et al., 2012c; Webster & Taylor, 2012). These associations are often specific to sponges, with many phylotypes found exclusively in sponges and not in the surrounding environment, leading to the so-called ‘sponge-specific’ or ‘sponge-and-coral-specific’ clusters (SCs, SCCs) (Hentschel et al., 2002, 2006; Taylor et al., 2007; Simister et al., 2012a; but see Taylor et al., 2013). Some species, particularly demosponges, harbour extraordinarily dense and diverse microbial communities. These types of sponges have been termed ‘high-microbial-abundance’ (HMA) sponges or ‘bacteriosponges’ and host between 108 and 1010 microbial cells per gram of sponge wet weight (Hentschel et al., 2006). In other species, where the mesohyl is largely free of microorganisms, the concentration of microbiota reflects that of the surrounding seawater (105–106 microbial cells per gram of sponge wet weight) (Hentschel et al., 2006). These species are referred to as ‘low-microbial-abundance’ (LMA) sponges. HMA and LMA sponges coexist in the same habitats, and the reasons for the differences in microbial abundances are unknown. To date, it remains unclear as to whether the majority of sponge-associated microorganisms are commensal (one species benefits, the other is not harmed), mutualistic (both species benefit) or parasitic (one species benefits and the other is harmed). In this article, the term ‘symbiosis’ is therefore used in a broad definition, to refer to the long-term association of two or more different organisms (Taylor et al., 2007).

Understanding the variability of microbial communities is a fundamental goal when examining any microbe–host association. A recent surge in the number of sponge microbiology studies has greatly increased our understanding of sponge microbial diversity. However, many studies have been based on a single time point, with the temporal stability of sponge-associated microorganisms remaining relatively unknown. From the few studies that have investigated temporal stability, the general consensus is that communities are for the most part stable over time. In one of the first such studies, neither microbial abundance nor community composition changed significantly over 11 days in the captively held Mediterranean sponge, Aplysina aerophoba (Friedrich et al., 2001). Similarly, a specific alphaproteobacterium dominated the culturable fraction of the microbial community of the Great Barrier Reef sponge Rhopaloeides odorabile, over four consecutive seasons (Webster & Hill, 2001). Furthermore, little variability was recorded in bacterial community composition of three south-eastern Australian sponges, both on short time scales and over five consecutive seasons (Taylor et al., 2004). Most recently, high temporal stability was observed in the bacterial symbionts of three Ircinia spp. over a 1.5-year monitoring period in the Mediterranean Sea (Erwin et al., 2012). One of the main caveats of these previous studies is that they employed techniques such as denaturing gradient gel electrophoresis (DGGE), terminal-restriction fragment length polymorphism (T-RFLP) and/or cloning of the 16S rRNA gene to define microbial populations. These approaches are limited by the fact that they only detect the most abundant taxa in these complex and diverse communities. However, a recent 454 pyrosequencing study of the Caribbean reef sponge, Axinella corrugata, did identify shifts in several bacterial taxa (White et al., 2012). In addition, most studies use DNA to define microbial populations that does not differentiate active cells from inactive (but viable) cells, dead cells, extracellular DNA or degrading/lysed cells (Gaidos et al., 2011). In contrast, cellular concentrations of rRNA are correlated with growth rate and activity (Novitsky, 1986; DeLong et al., 1989; Poulsen et al., 1993), so sequencing microbial populations using reverse transcripts of RNA (complementary DNA, cDNA) can yield information about which community members are active (Frias-Lopez et al., 2008; Urich et al., 2008; Kamke et al., 2010), although this method is still subject to nucleic acid extraction and PCR biases.

Temporal variation of host tissue and its symbiotic community can also be tracked by stable isotope analysis. Natural abundances of isotopes such as 13C and 15N have been used to study organic matter sources utilized by microorganisms in many marine symbioses (Kline & Lewin, 1999; Boschker & Middelburg, 2002; Maier et al., 2010). In combination, 13C and 15N isotopes can allow one to infer the trophic position of the consumer and, additionally, nitrogen and carbon sources used by the consumer can be indicated (Peterson & Fry, 1987; Vander Zanden et al., 1999; Zehr & Ward, 2002; Deines et al., 2009). This technique has also been applied to investigate the importance of symbiont-derived nutrition to the host sponge (Behringer & Butler, 2006; Weisz et al., 2007; Mohamed et al., 2008; Bergmann et al., 2009; Freeman & Thacker, 2011).

The composition of bacterial communities and the dynamics of their change over time are not well understood in marine sponges. Moreover, analysis of microbial diversity from a single time point does not distinguish between members of the community that are transient and those that are permanently associated. Considering the specificity and complexity of these associations, it seems likely that the microbial communities within sponges may be sensitive to environmental perturbations (Hentschel et al., 2006; Webster, 2007; Webster et al., 2008; Simister et al., 2012b). Long-term monitoring data would provide a starting point to document the variability of sponge microbial communities, which will be crucial to understanding how symbiosis will be affected by environmental change. In addition, these types of data will be useful to sponge aquaculture efforts, in which studies have highlighted the need to investigate natural host microbial variability prior to aquaculture (Webster & Taylor, 2012). This study presents the results from a 2-year sampling regime of bacterial community dynamics of two New Zealand sponges: Ancorina alata, an HMA sponge and Tethya stolonifera, an LMA sponge. The specific objectives were: (i) to employ 454 pyrosequencing of the 16S rRNA (derived from cDNA), in order to gain deeper insights into the stability/variability of the active sponge-associated bacterial community and (ii) to analyse the stability/variability of stable isotope signatures of bulk sponge tissue (δ13C and δ15N), in order to infer the importance of symbiont-derived nutrition to the host.

Materials and methods

Sample collection and processing

The marine sponges A. alata (phylum Porifera; class Demospongiae; order Astrophorida) and T. stolonifera (phylum Porifera; class Demospongiae; order Hadromerida) were collected by SCUBA at a depth of 3 m at Jones Bay, northeastern New Zealand (36°23′S, 174°49′E). Samples (n = 3 individuals per time point) were taken every month from November 2009 to November 2010, then every 3 months from November 2010 to November 2011 (Supporting Information, Table S1). Collected sponge samples were transferred underwater into plastic ziplock bags containing seawater and then brought to the surface. Each sample was split into two portions, half for molecular analysis and half for stable isotope analysis. Samples for molecular analysis were cut into pieces of no more than 0.125 cm3 (0.5 × 0.5 × 0.5 cm) using sterile scalpel blades, then transferred into sterile polypropylene tubes containing RNAlater (Ambion) in the field and transported back to the laboratory on ice, incubated overnight at 4 °C, then stored indefinitely at −20 °C (according to the manufacturer's instructions). Samples for stable isotope analysis were kept on ice, then transferred to −20 °C (within 2–3 h of collection), then subsequently freeze-dried. Seawater samples (n = 3 × 1 L, per time point) were filtered onto 0.2-μm polycarbonate filters (diameter 47 mm, Millipore Filter Corp.) to collect bacteria and frozen at −80 °C until further analysis.

Extraction of nucleic acids from sponge samples

RNA extraction methods were chosen based on Simister et al., 2011). Samples frozen in RNAlater (Ambion) were thawed on ice and cut further into smaller pieces (to ensure subsequent complete homogenization via bead-beating). All tissue samples were homogenized using lysing matrix E tubes (MPBio) in combination with the FastPrep FP120 bead-beating system (Bio-101). For RNA extraction, a Qiagen AllPrep DNA/RNA Mini kit (Cat. #80204) was used according to the manufacturer's instructions. Pure RNA was obtained by treatment with 3U RQ1 RNase-free DNase (Promega) according to the manufacturer's instructions. Purity and quantity of RNA was assessed using a NanoDrop 1000 spectrophotometer (Thermo Scientific), gel electrophoresis of a 5 μL aliquot on a 1% agarose gel containing 0.5 μg mL−1 ethidium bromide and Agilent 2100 Bioanalyzer RNA chips (Agilent Technologies). RNA was reverse-transcribed into cDNA using the SuperScript III First Strand Synthesis System (Invitrogen). Reverse transcription was carried out with random hexamer primers in 25 μL reactions, incubated at 50 °C for 50 min, and the reactions terminated at 85 °C for 5 min. cDNA was stored at −20 °C until further use. DNA was extracted from seawater filters by bead-beating in an ammonium acetate buffer containing CTAB, as described previously (Taylor et al., 2004).

Amplicon pyrosequencing

Amplification primers were designed with FLX Titanium adapters. The forward primer contained the A adapter (CCA TCT CAT CCC TGC GTG TCT CCG AC), and the reverse primer contained the B adapter (CCT ATC CCC TGT GTG CCT TGG CAG TC). A multiplex identifier (MID) was synthesized directly onto the forward 16S primer sequence (Roche Applied Sciences). The 16S rRNA-specific sequences (targeting the V4-V5 region) were 454MID_533F (GTGCCAGCAGCYGCGGTMA) and 454_907R (CCGTCAATTMMYTTGAGTTT). Each 100 μL PCR reaction mixture contained: 10 μL 10× Invitrogen Taq Reaction buffer (200 mM Tris, pH 8.4, 500 mM KCl), 3 μL MgCl2 (50 mM), 8 μL dNTP (2.5 mM) (Invitrogen), 1 μL of each primer (10 μM), 0.4 μL BSA (10 mg mL−1), 0.2 μL of Taq DNA polymerase (Invitrogen) and 1 μL of cDNA template which was prediluted 1:100. Reactions were made up to 100 μL total volume with Milli-Q water. Touchdown PCR conditions were as follows: 3 min at 94 °C followed by 20 cycles of 30 s at 94 °C, 30 s at 60 °C (−0.5 °C per cycle), 45 s at 72 °C. This was followed by a further 10 cycles of 30 s at 94 °C, 30 s at 50 °C, 45 s at 72 °C, with a final extension of 10 min at 72 °C. For each sample, 50 μL of PCR products were pooled from multiple reactions (100 μL total) and purified with AMPure XP beads (Agencourt, Beckman Coulter). Amplicon quality was checked on an Agilent 2100 Bioanalyzer DNA 1000 chip (Agilent Technologies). The number of molecules for each sample was calculated using size (bp) and concentration (ng mL−1) data from a Qubit Quant-iT DNA high-sensitivity assay kit and a Qubit® fluorometer (Invitrogen) according to the manufacturer's instructions. Pyrosequencing was performed using a Roche GS FLX titanium system on 2 × 1/8th plate runs (Macrogen, Inc., Seoul, South Korea).

Processing of raw sequence data

Sequences were processed using a combination of Mothur (Schloss et al., 2009) and custom PERL scripts. Pyrosequencing flowgrams were filtered and denoised using the Mothur implementation of AmpliconNoise (Quince et al., 2011). Sequences were removed from the analysis if they were < 200 bp, contained ambiguous characters, had homopolymers longer than 8 bp, more than one MID mismatch, or more than two mismatches to the reverse primer sequence. Unique sequences were identified with Mothur, aligned against a SILVA alignment (available at http://www.mothur.org/wiki/Silva_reference_alignment). Sequences were chimera-checked using UCHIME (Edgar et al., 2011), then grouped into 97% operational taxonomic units (OTUs) based on uncorrected pairwise distance matrices. A representative sequence (defined in Mothur as the sequence which is the minimum distance to the other sequences in the OTU) of each OTU was used for the taxonomic assignment using custom PERL scripts, similar to a previously used approach (Webster et al., 2010; Schmitt et al., 2012c). For each tag sequence, a blast search (Altschul et al., 1990) was performed against a manually modified SILVA database (Simister et al., 2012a). Modification of the original database involved addition of SC and SCC sequences (as defined in Simister et al., 2012a) and the manual annotation of various candidate phyla that were previously unnamed in the database, such as ‘Poribacteria’ and the sponge-associated unidentified lineage (SAUL) (Schmitt et al., 2012c). A Smith–Waterman algorithm was used to create pairwise global alignments between the 10 best hits against a tag sequence. For assignment, the most similar sequence to the tag sequence (or multiple sequences if within a range of 0.1% sequence divergence) was used. Sequence similarity thresholds of 75%, 80%, 85%, 90% and 95% were applied for assignment at phylum, class, order, family and genus level, respectively (Webster et al., 2010; Schmitt et al., 2012c). In cases where the taxonomy of the most similar sequences was inconsistent, a majority rule was applied, and the tag was only assigned if at least 60% of all reference sequences shared the same taxonomic annotation at the respective taxonomic level. All previously published, sponge-derived sequences in the SILVA reference database were labelled as such (Simister et al., 2012a), and it was noted when a tag sequence was assigned to a sponge-specific (SC) and/or sponge-coral-specific sequence cluster (SCC). A sequence read was only assigned to a SC and/or SCC cluster if it was more similar to the members of that cluster than to other sequences outside the cluster AND its similarity to the most similar sequence within that cluster was above 75%.

Processing of quality data

Mothur was used for diversity and richness estimation using Chao1 estimates, inverse Simpson diversity index and rarefaction curves on 97% OTUs. The magnitude of change in bacterial abundance was calculated for each phylum or OTU by first normalizing the number of reads per phylum or OTU, per sample. Values for all samples were then ranked, and the top 10 OTUs with the highest abundance were chosen for further analysis. Data at both phylum and OTU level were visualized using Microsoft Excel.

Stable isotope analysis

To test for inorganic carbon and inter-species variability between samples at a fixed time point, sponges (n = 3 per sponge species) were divided into two subsamples. The first portion was analysed without any pretreatment to serve as a control, while the second portion was treated with 15 mL 1M HCl for 24 h (Jacob et al., 2005). Acidified samples were neutralized with deionized water (Schubert & Nielsen, 2000; Kang et al., 2003), centrifuged three times (5400 g for 5 min), and the supernatant discarded before the sample was freeze-dried and homogenized again. We chose this method above other methods (e.g. dropwise addition of HCl and evaporation or fuming) (Jacob et al., 2005; Ng et al., 2007) to ensure the preservation and longevity of the chemicals inside the combustion tube in the elemental analyzer during the analytical phase. Carbon and nitrogen content and isotopic composition of T. stolonifera (n = 3) and A. alata (n = 3) from each time point were analysed at the Stable Isotope Laboratory, GNS Science, New Zealand, using a Europa Geo 20/20 (PDZ Europa Ltd, UK) isotope ratio mass spectrometer, interfaced to an ANCA SL elemental analyzer (PDZ Europa Ltd) in continuous flow mode (EA-IRMS). Powdered samples (0.95–1.15 mg) were weighed in duplicate into 4 × 6 mm tin capsules. The carbon dioxide gas was resolved from nitrogen gas using gas chromatographic separation on a column at 65 °C and analysed simultaneously for isotopic abundance, as well as total organic carbon and nitrogen. International and working reference standards (NIST-N1, IAEA-CH6, leucine, bovine liver, EDTA and Caffeine) and blanks were included during each run for calibration. Isotopic ratios (13C/12C and 15N/14N) are expressed as isotopic deviations (δ) defined as:

display math

where Rs is the isotopic ratio measured for the sample and Rref that of the international standards. The δ13C value is relative to the international Vienna Pee Dee Belemnite (VPDB) standard, and the δ15N value is relative to atmospheric air. Results are expressed in δ (‰) vs. the specific reference. Analytical precision is within ±0.2% for carbon and within ±0.3% for nitrogen (1 σn).

Statistical analysis of stable isotope values

Statistical analyses of stable isotope values (δ13C and δ15N) were performed using JMP® (Version 9, SAS Institute Inc). Differences between each month and individual isotope values, per sponge, were determined statistically using one-way analysis of variance (anova) [incorporating Levene's statistic to test for the equality of group variances, and Tukey's honestly significant difference (HSD) test (at P < 0.05)].


Samples collected for amplicon pyrosequencing and stable isotope analysis are presented in Table S1. One of the three replicate T. stolonifera RNA samples (extracted in November 2009 and November 2010 for amplicon pyrosequencing) was excluded from analysis as RNA was not of sufficient quality (see 'Materials and methods').

Bacterial community structure

For the pyrosequencing analysis, after noise reduction and quality filtering, a total of 141 254 sequences, with a mean of 3210 (±969, 1 SD) sequences per sample, was recovered (Table S2). On average, < 0.33% (±0.39, 1 SD) of reads were taxonomically unassigned per sample (see Materials and Methods). Rarefaction curves (Fig. S1) indicate that diversity coverage was high, with most curves approaching asymptotes (although further sequencing would still have yielded a greater number of OTUs). Chao1 diversity estimates (Table S2) for each sample were calculated at the 97% sequence similarity level and indicated that estimated richness ranged from 440 to 1172, 128–603 and 506–818 OTUs in seawater, T. stolonifera and A. alata-derived samples, respectively. In total, 4468 unique OTUs (97% sequence similarity) were identified, affiliated with 26 bacterial phyla in total – 24 from seawater, 20 from T. stolonifera and 15 from A. alata samples. The number of reads per phylum was normalized and expressed as a percentage of the total for each sample (Fig. 1). The proportion of reads assigned at phylum level was highly consistent across all time points, for both sponge species and seawater samples. The most dominant bacterial phyla in seawater samples were Proteobacteria (61%), Bacteroidetes (30%) and Actinobacteria (4%) (Fig. 1a). Within the Proteobacteria, most sequences were Alpha or Gammaproteobacteria (39 and 19%, respectively, of the total bacteria sequences) (Figs 1a and S2a). In T. stolonifera samples, the majority of 16S rRNA transcripts were assigned to Proteobacteria (approximately 93% of sequences per sample), most of which were Betaproteobacteria (78% of the total bacteria sequences) (Figs 1b and S2b). All other phyla identified made up < 0.5% of total reads, with the exception of Bacteroidetes (4%) and Spirochaetes (1%) (Fig. 1b). The most dominant phyla from all A. alata samples were Chloroflexi, Proteobacteria, ‘Poribacteria’, SAUL and Actinobacteria, representing approximately 32, 32, 9, 6 and 6% of sequences, respectively (Figs 1c and S2c). Within the Proteobacteria, most sequences were Delta, Gamma or Alphaproteobacteria (10, 10 and 9%, respectively, of total bacterial sequences). Phyla that were less abundant but found in similar proportions across all sampling points included: Gemmatimonadetes (4%), Acidobacteria (3%), Spirochaetes (3%), Bacteroidetes (2%) and Nitrospira (2%) (Fig. 1c). Although seawater and T. stolonifera samples showed higher phylum level diversity than A. alata samples, the abundance of many of these phyla was < 1% of the total reads assigned.

Figure 1.

Distribution of 454 amplicon reads per phylum across (a) Seawater samples, (b) Tethya stolonifera samples and (c) Ancorina alata samples. The number of reads per phylum is calculated as a percentage of the total reads in each sample. *SAUL (sponge-associated unidentified lineage) (Schmitt et al., 2012c). **NA (not assigned). ***Other (reads were only present in one sample from seawater, A. alata or T. stolonifera at < 0.2% of total reads; Chlamydiae, Deferribacteres, Deinococcus-Thermus, Fibrobacteres, OD1, OP3, Tenericutes, TM6, TM7, WS3).

Assignment of sequences into ‘sponge-specific’ clusters

The proportion of reads assigned into one of the previously described SCs/SCCs (Simister et al., 2012a) was highly similar across all sampling time points in both sponge species (Fig. 2a), with average values of 78% of A. alata reads and 78% of T. stolonifera reads assigned into SCs/SCCs. As expected, very few sequences from seawater were assigned to SCs/SCCs (on average 3% of total reads per sample), however, these reads were affiliated with 43 of 90 SCs/SCCs identified in this study. Clearly, clusters containing sequences derived from seawater cannot be strictly referred to as ‘sponge-specific’ anymore, but for consistency, we have retained the SC/SCC nomenclature from Simister et al. (2012a). That is, when we refer to seawater sequences being assigned to clusters, we are referring to those SCs/SCCs that were identified in the 2012 study. SCs/SCCs were identified to phylum level and the number of ‘sponge-specific’ reads in each phylum was expressed as a proportion of the total reads that were assigned into SCs/SCCs (Fig. 2b and c). Of the small proportion of SCs/SCCs found in seawater samples, the majority were Proteobacteria (Alpha, Beta and Deltaproteobacteria), Actinobacteria and Chloroflexi, the proportions of which were highly variable over sampling time points. Most SC clusters in A. alata belonged to Chloroflexi, Deltaproteobacteria and ‘Poribacteria’ (Fig. 2b) and the majority of SCCs to Chloroflexi, Gemmatimonadetes, Bacteroidetes and Alphaproteobacteria (Fig. 2c). The proportion of SCs/SCCs was highly consistent across the majority of sampling time points; however, there were some exceptions. Samples from November 2009, 2010 and 2011 showed substantial increases in Betaproteobacteria, although this was not consistent across all replicates. All T. stolonifera samples were dominated by a single Betaproteobacteria cluster (SC112), comprising 99% of all reads that were assigned into clusters (Fig. 2b). SCs were also assigned to Spirochaetes and Alphaproteobacteria, albeit at low abundances (on average < 0.5% per sample of the total reads assigned into clusters). In addition, very few SCCs were assigned to T. stolonifera (on average < 0.05% per sample of the total reads assigned into clusters) (Fig. 2c).

Figure 2.

(a) The proportion of reads that were assigned (blue) or not assigned (red) to an SC and/or SCC per sample (b) The proportion of reads that were assigned to an SC per bacterial phylum and (c) The proportion of reads that were assigned to an SCC per bacterial phylum. The number of reads per phylum (b and c) is calculated as a percentage of the total reads that were assigned to a SC/SCC in each sample. Samples are grouped according to type (Seawater followed by Tethya stolonifera and then Ancorina alata) in order of sampling month (November 2009–November 2011).

Identification of OTUs with the largest change in abundance

The top 10 most abundant OTUs represented a combined average total of 41% of all sequences per seawater sample (Fig. 3a), the majority of which were assigned to Proteobacteria (in particular Alpha). The most abundant OTUs varied over the monitoring period; however, there were no correlations in highest/lowest abundance with any particular season. For example, OTU4387 (Alphaproteobacteria, class Rhodobacteria) was recorded with the greatest seasonal variation, with abundance being the highest in summer (February 2010) and lowest in spring (November 2010). In contrast, OTU4292 (Alphaproteobacteria, class Rhodobacteria) peaked in abundance in spring (November 2010) and decreased to its lowest abundance in summer (February 2010). The top 10 most abundant OTUs in T. stolonifera represented on average 85% of total sequences per sample (Fig. 3b). The most abundant OTU was OTU3891 (Betaproteobacteria, SC112) and in contrast to A. alata, this OTU was highly stable at all time points, with little effect of season. The next nine most abundant OTUs were rare or transient members of the bacterial community (approximately < 3% of total sequences per sample) and were variable with season. Similar to seawater, most T. stolonifera OTUs were affiliated with Proteobacteria (in particular Gamma), and few were assigned into SCs/SCCs. In A. alata, the top 10 most abundant OTUs represented on average 35% of total sequences per sample. The majority of these OTUs were highly stable over the monitoring period (Fig. 3c), and eight of 10 were assigned into SCs/SCCs. There were, however, exceptions; for example, OTU3891 (Betaproteobacteria, SC112) showed significant variation with season. The lowest abundance of this OTU was recorded in summer (February 2010), increasing throughout autumn and winter to peak abundance in spring (November 2010).

Figure 3.

Log scale graphs of the 10 most abundant OTUs for (a) Seawater samples, (b) Tethya stolonifera samples and (c) Ancorina alata samples. The most abundant OTUs were chosen by first normalising the number of reads per OTU, per sample/timepoint. Values for all samples were then ranked and the top 10 OTUs with the largest abundance were chosen for further analysis. Averages of three replicates are plotted, with SE bars ±1 SD of the mean.

Effect of season on stable isotope composition of A. alata and T. stolonifera

Statistical analysis revealed that A. alata and T. stolonifera had highly similar δ15N values to each other (A. alata 10.0 ‰ (±0.3, 1 SD), T. stolonifera 9.9 ‰ (±0.2, 1 SD), P = 0.6094, n = 102, d.f. 100), whereas δ13C values were significantly different (A. alata -17.2 ‰ (±0.11, 1 SD), T. stolonifera -19.2 ‰ (±0.2, 1 SD), P = <0.001, n = 102, d.f. 100) (Fig. 4a). Ancorina alata δ13C and δ15N values were maintained at similar levels from the start of measurements throughout 2009, 2010 and 2011 (δ13C R2 = 0.17 and δ15N R2 = 0.011) (Fig. 4b and c). δ15N values for A. alata over the period of monitoring ranged from 9.2 to 10.4 ‰ and δ13C values ranged from −16.5 to −17.9 ‰. There were only 2 months in which δ13C values deviated significantly from the mean [−17.2 ‰ (±0.11, 1 SD)]; autumn samples were less enriched (March 2010, 0.63% less than the mean), while spring samples were more enriched (November 2011, 0.78% more than the mean). Tethya stolonifera δ13C and δ15N values were also maintained at similar levels from the start of measurements to final values recorded in 2011 (δ13C R2 = 0.0084 and δ15N R2 = 0.0546) (Fig. 4b and c). Tethya stolonifera samples did exhibit more statistically significant temporal variations in both δ13C and δ15N values (Fig. 4b and c), with δ15N values ranging over the monitoring between 9.4 and 10.5 ‰ and δ13C values from −18.4 to −19.8 ‰. However, these differences were not cyclical over the 2-year monitoring period, that is, sampling times that led to significant differences recorded in 2009 were not always significant in the following year. For example, δ13C values were significantly less enriched in autumn 2010 than autumn 2011 values, which were significantly enriched.

Figure 4.

Isotope values for Tethya stolonifera (white squares) and Ancorina alata (black squares). (a) The mean δ13C and δ15N of sponge tissues from all time points (n = 3 per month = 102 total), SE bars ±1 SD of the mean. δ13C values were significantly different between the two species (P < 0.001, n = 102, d.f. = 100). δ15N values were not significantly different between the two species (P < 0.6094, n = 102, d.f. = 100). (b) Changes in the δ13C of the sponge tissues over the time period November 2009 to November 2011 (significant differences P < 0.05 are marked by *, per sponge n = 51 and d.f. = 50). Both T. stolonifera, (R2 = 0.0084) and A. alata (R2 = 0.171) maintained similar levels of δ13C from November 2009 to final values recorded in November 2011. (c) Changes in the δ15N of the sponge tissues over the time period November 2009–November 2011 (significant differences P < 0.05 are marked by *, per sponge n = 51 and d.f. = 50). Both T. stolonifera, (R2 = 0.0546) and A. alata (R2 = 0.0109) maintained similar levels of δ15N from November 2009 to final values recorded in November 2011.


Temporal stability of sponge-associated bacterial communities

At bacterial phylum level, the structure of communities in A. alata, T. stolonifera and seawater were highly conserved throughout the 2-year monitoring period. Consistent with other HMA sponge amplicon pyrosequencing studies (Webster et al., 2010; Lee et al., 2011; Schmitt et al., 2012ac), the bacterial community of A. alata was dominated by Chloroflexi, Proteobacteria, ‘Poribacteria’, Actinobacteria and SAUL. Bacterial communities of seawater samples were highly diverse but dominated by few phyla, mainly Proteobacteria (Gamma and Alpha) and Bacteroidetes, indicating clear compositional differences between seawater samples and A. alata samples. Differences between bacterial community profiles of sponges and those of the surrounding seawater have been reported in numerous other studies (Lee et al., 2011; Fan et al., 2012; Jackson et al., 2012) and are consistent with the idea that sponge-associated symbionts play significant roles in the host sponge (Hentschel et al., 2002; Taylor et al., 2007). Bacterial communities of T. stolonifera samples were also highly diverse, but were dominated throughout the sampling period by very few phyla. The dominant phyla were mostly Betaproteobacteria (over 90% of sequences) and to a lesser extent Bacteroidetes. The presence of one dominant Proteobacteria phylotype has been reported previously from many other studies with LMA sponges, although the reasons for this are as yet unknown (Sipkema et al., 2009; Kamke et al., 2010; Luter et al., 2010; Erwin et al., 2011; Giles et al., 2013). Previous studies on HMA and LMA sponges have also found differences in their bacterial species richness. A 2007 study showed that LMA and HMA sponges from the Caribbean differ in their 16S rRNA DGGE banding patterns (Weisz et al., 2007). HMA sponges contained more 16S rRNA DGGE bands, and many of these were closely related to previously identified sponge-derived sequences. In contrast, LMA sponges had a DGGE profile more similar to that of seawater. Clone library analysis of A. alata and the co-occurring LMA sponge Polymastia sp. (Kamke et al., 2010) detected only three bacterial phyla in the latter, compared with eight in A. alata. Although the results of this study show that T. stolonifera was more diverse by ~5 phyla per sample than A. alata, these phyla were often represented by very few sequences. Our results are consistent with a more recent study of LMA sponges (Giles et al., 2013), which showed that the bacterial community profiles of five LMA sponges were also diverse, but many of the bacterial phyla identified were only represented by very few sequences.

Despite the high diversity of bacterial phyla found in seawater, A. alata and T. stolonifera samples, the observed phylum level stability across all sampling time points was driven by the persistence of a small number of dominant OTUs. In seawater and A. alata samples, the 10 most abundant OTUs represented > 25% of the observed bacterial community. Small numbers of OTUs, accounting for a large proportion of a sponge bacterial community, have also been reported in another 454 tag pyrosequencing study (Webster et al., 2010) and T-RFLP studies (Erwin et al., 2011, 2012). In T. stolonifera, one Betaproteobacteria OTU (OTU3981) accounted for more than three-quarters of the observed bacterial community. In LMA sponges, a single large OTU that represents the majority of sequences has recently been reported, with four of five LMA sponges containing one large Proteobacteria OTU (Giles et al., 2013). In contrast to other temporal stability studies in sponges (Friedrich et al., 2001; Webster & Hill, 2001; Taylor et al., 2004; Erwin et al., 2012), a recent 454 pyrosequencing study of the Caribbean reef sponge, A. corrugata, did identify shifts in several bacterial taxa (White et al., 2012). However, while this study used ‘universal’ 16S rRNA primers, they were not modified to include core groups of the sponge-associated bacterial community, for example, only a single sequence was loosely affiliated with ‘Poribacteria’. Furthermore, in samples from both seasons, over 50% of sequences could not be identified to phylum level, in contrast with the current study in which only 0.33% (±0.39, 1 SD) of sequences were unassigned. In addition, we used a manually modified SILVA reference database where all previously published sponge sequences, candidate phyla and SCs/SCCs (at the time) were labelled (Simister et al., 2012a).

Various environmental (temperature, light and nutrient fluctuations) and biological (planktonic community interactions) parameters are more likely to influence free-living planktonic microbial communities than those in sponges (Gilbert et al., 2012). This is reflected by the higher seasonal variation in abundant seawater OTUs compared with the most dominant sponge OTUs, suggesting that the sponge host may buffer against some environmental fluctuations. The most variable or transient sponge OTUs over time (and between replicates) were those that were relatively ‘rare’ (abundances of 0–1%). Marine sponges are capable of massive filtration rates, with sponge species pumping over half a litre of water s−1 kg dry mass−1 (Weisz et al., 2008). In turn, large amounts of particulate matter (from seawater), which includes microbial cells, pass through the host. These low abundance and sometimes transient taxa may therefore represent food that has been filtered by the sponge (Taylor et al., 2007), or environmental bacteria present in the sponge when sampled, rather than being true symbionts. Two recent studies also found that temporal variation in sponge-associated microbial communities was restricted to low-abundance taxa (Anderson et al., 2010; Erwin et al., 2012). Studies of microbial communities in the vertebrate gut have found an inverse correlation between population size and temporal stability, with transient members of a microbial community more likely to be detected than a persistent, but very low abundance, population (Walter & Ley, 2011). If the same holds true for marine microbial communities, then studies which repeatedly sample over time are highly valuable to identify which members of the ‘rare’ sponge microbial community are truly symbiotic, that is, which members are allochthonous vs autochthonous. We acknowledge that some variation in OTU abundance over time will be due to differences between communities associated with individual sponges of the same species, as sponges were not sampled repeatedly. However, collecting consecutive samples from the same sponge is highly destructive, particularly to T. stolonifera, which at approximately 6 cm in diameter, would remove most, if not all, of the sponge biomass.

Temporal stability of ‘sponge-specific’ bacteria

Sponge-specific clusters (SC) and sponge-coral-specific clusters (SCC) (Hentschel et al., 2002; Taylor et al., 2007; Simister et al., 2012a) have been defined as monophyletic clusters of 16S rRNA sequences found only in sponges (or sponges and corals) and not in the surrounding environment such as seawater or sediments. Here, the assignment of seawater-derived sequences into previously defined SCs/SCCs (Simister et al., 2012a) shows that supposedly ‘sponge-specific’ bacteria can actually occur outside the host. This is consistent with other recent pyrotag sequencing studies (Webster et al., 2010; Taylor et al., 2013), raising important questions about the origin of sponge-symbiont associations. Seawater samples sequenced here and in Webster et al. (2010) are DNA derived so whether ‘sponge-specific’ bacteria identified outside the host are active cannot be determined. Although some seawater-derived sequences were assigned to previously described (Simister et al., 2012a) ‘sponge-specific’ clusters, these were at low levels (< 7%) and half of the total clusters identified were still exclusive to sponges based on these data (47/90). Clearly, in the light of these data, the 43 clusters that were identified among seawater samples can no longer be deemed ‘sponge-specific’ or ‘sponge-and-coral-specific’. For consistency's sake, we will continue to refer here to the SCs/SCCs as defined by Simister et al. (2012a). The proportion of reads from both sponge species that were assigned to SCs or SCCs was high – approximately three-quarters of all reads. Despite the high assignment of sequences into SCs/SCCs in T. stolonifera, nearly all sequences were assigned to one Betaproteobacteria (SC112) cluster. This is consistent with previous studies that have reported significantly more SCs/SCCs within HMA sponges compared with LMA sponges (Kamke et al., 2010; Erwin et al., 2011; Schmitt et al., 2012b; Giles et al., 2013). We saw no consistent change in the proportion of reads assigned to SCs or SCCs over time for all sample types. In addition, phylogenetic assignment of SCs/SCCs in both sponge species was also highly consistent over the sampling period, with little indication of shifts in cluster assignment in response to changing environmental conditions. Phylogenetic assignment of SCs/SCCs in seawater samples was highly variable through the monitoring period, suggesting that environmental fluctuations affect free-living vs. host-associated bacterial communities differently (Schauer et al., 2003; Erwin et al., 2012; Gilbert et al., 2012).

Temporal stability in isotopic signatures

Previous research on stable isotope values in marine sponges has shown differences between HMA and LMA sponges (Weisz, 2006; Weisz et al., 2007; Southwell et al., 2008; Freeman & Thacker, 2011), purportedly due to differences in the diversity and abundance of their microbial communities. The enrichment of δ13C in A. alata samples compared with T. stolonifera samples suggests that the two species differ in their nutritional sources of carbon (Peterson & Fry, 1987). Heterotrophy, through the consumption of microorganisms from seawater or by microbial uptake of dissolved organic carbon (DOC), is a common form of carbon metabolism in sponges (Yahel et al., 2003, 2007; De Goeij et al., 2008ab). Our field site is characterized by macroalgal communities, so that macroalgae-derived organic matter could play an important role as a carbon source. Trophic transfer of organic matter derived from benthic macroalgae to sponges has been reported previously (Behringer & Butler, 2006; Granek et al., 2009; van Duyl et al., 2011). Possible effects of Cyanobacteria (Wilkinson, 1983), which not only perform photosynthesis but also fix nitrogen and take up dissolved inorganic nitrogen (Wilkinson et al., 1999; Davy et al., 2002; Pile et al., 2003), on stable isotope signature can be excluded here due to the lack of large populations of photosynthetic symbionts (A. alata: 0.01%, T. stolonifera: 0.44% Cyanobacteria sequences per sample). Alternatively, Chloroflexi, which were found in different proportions in the two sponges, may have contributed to the differences in δ13C values. Some members of the phylum Chloroflexi use mixotrophic metabolism (Eiler, 2006) but nothing is known as yet on the quantity of heterotrophic vs autotrophic growth and carbon isotopic fractionation. The different abundance of Chloroflexi, coupled with different forms of heterotrophy, is a likely cause for the observed differences in δ13C between the two sponge species surveyed here. The similar pattern of δ13C values of the two sponge species over the monitoring period indicates that variation in δ13C is largely due to seasonal variation. For more detailed analysis of carbon transfers, sponge and microbial cell fractions must be analysed separately, as performed previously for four Caribbean sponge species (Freeman & Thacker, 2011).

Previous research on LMA and HMA sponges has shown that they can also differ in their δ15N values (Weisz et al., 2007). In a Florida sponge survey, δ15N values separated sponges into three distinct groups: HMA sponges with low δ15N values, HMA sponges with high δ15N values and LMA sponges with high δ15N values (Southwell, 2007; Weisz et al., 2007). It is thought that HMA sponges with low δ15N values are indicative of microbial transformations of nitrogen, as symbiont-derived nutrition enriches for 14N (i.e. by nitrogen fixation and ammonium uptake) (Zehr & Ward, 2002). By contrast, LMA sponges with higher δ15N values are indicative of nitrogen acquired from allochthonous (e.g. anthropogenic) dietary sources (Zehr & Ward, 2002; Weisz et al., 2007). In this study, both sponges had comparable δ15N values, which may indicate that nutritional uptake of nitrogen occurs through similar pathways. The δ15N values were high throughout the monitoring period for both A. alata [10.0 ‰ (±0.25, 1 SD)] and T. stolonifera [9.9 ‰ (±0.18, 1 SD)], and suggest that both sponges obtain nitrogen solely from external sources, most likely of anthropogenic origin (Costanzo et al., 2005; Swart et al., 2005; Watanabe et al., 2009). Alternatively, the 15N-enrichment in sponges could also be linked to the use of macroalgae-derived organic matter (Riera et al., 2009). Similar evidence is provided by Kaehler et al. (2000), who showed that kelp-associated filter feeders (bivalves and sponges) cover about 50% of their feeding requirements with kelp-derived organic matter. Another explanation for these positive δ15N values could be the uptake of isotopically enriched inline image, resulting from fractionation during the process of assimilation or denitrification (Sigman & Casciotti, 2001; Swart et al., 2005).


This study examined the temporal stability of bacterial communities in two different sponge species and ambient seawater over a 2-year sampling period. By pyrosequencing the 16S rRNA gene from RNA-derived sponge samples, we were able to assess the stability of the active bacterial community within the sponge host over time. The overall bacterial community is stable in both sponge species and seawater samples, driven by OTUs that are persistent and dominate over all time points recorded. There is, however, a component of the bacterial biosphere in both sponges and seawater that is more variable, with some OTUs transiently present over time. These fluctuations in OTU abundance could be caused by environmental factors such as changes in currents, sea surface temperatures and nutrient levels. Long-term monitoring data, as undertaken here, will be valuable to assess whether environmental perturbations cause shifts in the rare sponge-associated microbial community. Isotopic investigations revealed variations in δ13C, but consistently high values of δ15N, suggesting that only some carbon and nitrogen nutritional pathways are shared between the two sponge species. In this study, we provide extensive molecular and isotopic data on the sponge microbiota through time, and highlight the importance of long-term temporal studies to enhance our understanding of these dynamic marine symbioses.


Research was supported by a University of Auckland Doctoral Scholarship to RLS and University of Auckland grant 3622989 to MWT. PD was supported by an Alexander von Humboldt Fellowship.