The role of macrobiota in structuring microbial communities along rocky shores

Introduction

The dynamics and interactions of the macroscopic species on rocky shores of the northeast Pacific Ocean have been well-characterized and thus have contributed significantly to our understanding of coastal ecological processes (e.g., Paine, 1966; Wootton, 1994; Estes & Duggins, 1995). Although some specialized symbiotic associations have been described in rocky shore species (Secord & Augustine, 2000; Bergschneider & Muller-Parker, 2008), we know little about multi-taxa microbial associations. There is increasing evidence that many marine macrobiota have surface biofilms (Grossart et al., 2005; Kvennefors et al., 2012) or endosymbionts (Zurel et al., 2011; Wegner et al., 2013) or both (e.g., Qian et al., 2006; Taylor et al., 2007). However, our understanding of the specificity of these associations and their functional significance remain nascent with some notable exceptions (Webster & Taylor, 2012; Fan et al., 2013; Heisterkamp et al., 2013). The shelf waters of the California Current Large Marine Ecosystem (CCLME) maintain diverse and unique microbial communities across upwelling areas (Bertagnolli et al., 2011). The relatively high productivity of this system has been attributed to the seasonal upwelling of nitrate (Barber & Smith, 1981), which can lead to significant levels of carbon fixation. The diversity of potential plant and animal ‘host’ species in the CCLME, their relatively large geographic range, their longevity and the provision of key resources suggest they provide unique microniches capable of increasing the diversity and function of microbial communities.

If macrobiota directly provide habitat for intertidal microbes (e.g., mussels Pfister, Meyer & Antonopoulos, 2010; algae, Miranda et al., 2013) or indirectly provide resources such as nitrogen, in the form of animal-regenerated ammonium, then our understanding of coastal biogeochemistry is incomplete without considering the contribution of macrobiota. In the northeast Pacific, Tatoosh Island, WA shows persistent shoreward peaks of ammonium (Pfister, Wootton & Neufeld, 2007), while the tidepools at Second Beach, WA have animal-regenerated ammonium from the California mussel with locally enhanced algal productivity, and microbial nitrification (Pfister, 2007). We employed 16S rRNA V4 amplicon sequencing to test whether microbes associated with macrobiota differ from those in the water column and on inert surfaces. We further compared the identified microbial taxa with those found when using shotgun metagenomics in a subset of samples for mussel shell biofilms. Finally, we used a manipulative field experiment of mussel presence or absence in tidepools to ask whether the increased rates of nitrogen remineralization and uptake demonstrated with mussels by tracer ammonium addition (Pather et al., 2014) resulted in changed bacterial communities on inert surfaces.

Materials and Methods

We asked whether microbial diversity and abundance were distributed differentially among intertidal microhabitats by (1) sampling distinct microhabitat types including artificial and natural surfaces at Tatoosh Island (48.32°N, 124.74°W), and (2) sampling artificial and natural substrates in the context of an animal removal experiment at nearby Second Beach (48°, 23′N, 124, 40′W). Tatoosh Island is 0.7 km off the northwestern tip of Washington State, USA and has been well studied ecologically. Previous metagenomic analysis of biofilms associated with the shells of mussels at Tatoosh Island demonstrated a rich microbial assemblage, with the genetic capacity for nitrogen cycling (Pfister, Meyer & Antonopoulos, 2010). On 6 Aug 2009, we sampled both inert and biogenic substrates in situ, as well as artificial substrates. Biogenic substrates sampled included the surface of the red alga Prionitis sternbergii (n = 2), the anemone Anthopleura elegantissima (n = 2), and gill tissue (n = 3) and siphon tissue (n = 1) of the California mussel Mytilus californianus. These biogenic hosts were chosen because they are persistent members of the community and have relatively long-lived and sessile tissues that might provide a predictable substrate. In addition to the biogenic surfaces, rocks were chipped off of bench adjacent to mussel beds. Artificial substrates (glass crucible covers, 3 cm diam, www.leco.com) were attached with epoxy on rock adjacent to mussel beds (n = 4) and in tidepools (n = 5) on 10 Jun 2009 and left in situ for 2 months. We chose glass crucible covers because they provided heterogeneous surface, yet an inert substance. Immediately after a late morning collection at low tide, samples were frozen and sent to −80 °C storage at the University of Chicago prior to extraction. Environmental characteristics of the seawater, including nutrient concentrations, were recorded as part of a regular sampling program (Wootton & Pfister, 2012) and are reported in Table S1. The Upwelling Index for this latitude (48°N) was, on average, positive (www.pfeg.noaa.gov).

We tested whether the presence of mussels had entrained different microbial assemblages by sampling natural and artificial substrates in tidepools where mussels had been removed or unmanipulated since 2002 at Second Beach (Pfister, 2007), a north-facing complex of rocks 2 km east of Neah Bay, WA, USA within the Makah Tribal Reservation. For seven years prior to our microbial sample collection, mussels have been excluded from 6 tidepools that previously contained mussels by pulling them out by hand at an approximately monthly intervals during the spring and summer months. Other tidepools that naturally had mussels served as controls. The biogeochemistry of these experimental and control tidepools have been characterized multiple times, allowing us to test for mussel effects on tidepool nutrient concentrations (nitrate, nitrite, ammonium, and phosphorus), and seawater pH and oxygen and how it related to microbial community structure. These parameters were measured 9 times in each tidepool during Aug 2009 and 12 times in Jul and Aug of 2010. We also estimated ammonium remineralization and removal rates with a tracer experiment with enriched ${}^{15}{\mathrm{NH}}_4^{+}$ in 2010 (Pather et al., 2014). A ∼2 cm² piece of rock was chiseled from 5 pools with and 5 without mussels during morning hours on 24 Aug 2009. We also sampled from glass cover slips that had incubated for 3 months in these pools (n = 7 controls, n = 6 mussel removal pools) with a copper paint barrier to exclude molluscan grazers. To compare this benthic microbial assemblage to microbes in the water column, we collected 3 plankton samples on 25 mm GF/F filters by pumping 300 mL of seawater on an incoming tide (1100–1130 h). Due to the 0.7 um pore size, we likely excluded many free-living microbes, though high species richness still resulted (see below). All samples were frozen and sent to storage at −80 °C at the University of Chicago prior to extraction.

For DNA extraction, we used the Power Soil DNA Extraction Kit (MoBio). The rock, crucible cover substrates, and mussel shells were both swabbed with sterile cotton applicators and brushed with sterile spiral dental brushes that were placed in the extraction solution. Coverslips, filters, excised mussel gill or mussel siphon, anemones, and algae were placed into the beadbeater vial and pulverized. Thus, our sampling of biogenic substrata included the potential that microbes were part of the host tissue. The PCR amplification protocol followed Caporaso et al. (2010b) for multiplexing 16S rRNA samples. The PCR products were cleaned with MoBio™UltraClean htp PCR Clean-up kit. We amplified the V4 variable region of the 16S rRNA gene from community DNA using bar-coded primers according to Earth Microbiome Project standard protocols (www.earthmicrobiome.org).

Results

Some distinction exists among microbial assemblages associated with different substrates at Tatoosh Island

Between 54,490 and 250,432 sequences per sample were generated for 26 samples from a range of materials including inert surfaces (rock and glass crucible lids) as well as mussel shells and tissues, algal fronds, sea anemones, and the filtered plankton. All samples were rarified to 50,000 sequences per sample. OTU richness (total diversity) estimates were greatest for inert substrates and the water column, while the lowest richness was associated with mussel gill and siphon tissue (ANOVA, F_7,16 = 4.968, p = 0.004). Species richness was highly correlated with other metrics of diversity including chao1 and phylogenetic diversity (r = 0.98 to 0.99, p < 0.001), as was equitability (or evenness, r = 0.780, p < 0.001). Alphaproteobacteria, including Rhodobacteraceae and Hyphomonadaceae, dominated the algal Prionitis tissue and the inert substrates, while Gammaproteobacteria, especially Vibrionaceae, dominated mussel gill and siphon tissue (Fig. 1). The communities associated with mussel shells, anemones, and filtered plankton samples were similarly dominated by Gammaproteobacteria. Mussel shells and anemones similarly had many Vibrionaceae OTUs, with shells also harboring Moritellaceae. OTUs in the Psychromonadaceae were prominent in the plankton. Given our use of a 0.7 µM filter, the OTU richness may be underestimated if the smallest bacteria were not retained.

In addition to differences in alpha diversity, the microbial community composition and structure on different biotic and inert substrates showed differences in beta diversity. First, the same substrates clustered in a PCoA analysis based on weighted UniFrac distances (Fig. 2), e.g., rock substrates clustered with the glass crucibles, while mussel gill and siphon tissue clustered together. The filtered plankton samples were highly similar to each other, while the anemone and Prionitis tissues suggest greater differences among individual hosts. Substrate differences were significant with a permuted ANOVA (F_5,18 = 6.570, p < 0.001) when we grouped substrates into 6 categories (anemone, red alga, plankton, mussel shell, mussel internal tissue, and inert substrates). Analysis of the differences among those 6 categories showed that each differed significantly from one or several others, except for filtered POM, which did not statistically differ from any other group.

A second test indicating beta diversity differences among substrates was revealed in an ANOVA on OTU abundance. There were 10 OTUs that differed significantly in abundance (Bonferroni corrected ANOVA, p < 0.05, Fig. 3) and these distinctions came primarily from their abundance in association with macrobiota. For example, Moritella and Aliivibrio were found on mussel shell and gill tissue, while Cyanobacteria were primarily associated with algal fronds.

Further, the beta diversity differences we detected among substrates were a function of the abundance of OTUs among samples. By comparing OTUs shared between substrates in a network plot, it was evident that the abundant OTUs that were detected at least 5,000 times were generally common to all substrates (Fig. 4A). In contrast, when we examined shared diversity among rare OTUs (those detected only 5 to 10 times across the dataset, Fig. 4B), there was strong differentiation among substrates, indicating that the rare OTUs were responsible for the majority of the compositional differences between the microbial communities associated with different substrates.

The presence of mussels has little impact on microbial assemblages in tide pools at Second Beach

In experiments performed at Second Beach, the biogeochemical parameters of tidepools were affected by the presence or absence of mussels (Table S1). A principal components analysis that included the ammonium regeneration and removal rates (Pather et al., 2014), the maximum seawater pH and dissolved oxygen, ammonium, nitrate, nitrite and phosphorus measured in the tidepools over both daytime and nighttime low tides indicated that the first principal component explained 81.7% of the variance and differed among mussel versus no mussel tidepools (p = 0.049, Fig. S1). Rock had more than twice the microbial diversity of coverslips (Rock = 3,727 OTUs, Coverslips = 1,750 OTUs; F_1,18 = 140.59, p < 0.001), perhaps reflecting greater time in the environment. Both types of substrata maintained more diverse and equitable community profiles in tidepools where mussels were removed, than in those with mussels present (Rock = 3,282 OTUs; coverslips = 1,343 OTUs; Fig. 1B, Two-way ANOVA, F_1,18 = 12.759, p = 0.002). However, there was no interaction between substrate and mussel presence (F_1,18 = 0.013, p = 0.909), indicating that the distinction between microbial communities associated with natural rock and glass coverslip communities did not depend upon the presence of mussels. The equitability with which diversity was distributed was strongly correlated with total diversity (r = 0.920, p < 0.001), indicating that when mussels are removed, the equitability of taxa also increases. Further, the microbial communities associated with the rock substrate in tidepools at Second Beach were similar in OTU composition with rock substrate at Tatoosh Island (Fig. 1A vs. 1B).

A high degree of OTU sharing and community structure similarity were observed between microbial communities associated with rock surface or coverslip samples regardless of whether mussels were present (Figs. 5 and 6); this was supported by the absence of OTUs with significantly different relative abundances between tidepools with or without mussels based on Bonferroni-corrected ANOVAs. However, there were fewer shared OTUs among rock samples when OTUs that were rare were considered against OTUs that were common (26.7% versus 96.9% shared). There was no relationship in the degree of OTU sharing with mussel presence or absence. Indeed, in contrast to our comparison of Tatoosh Island substrates, both rock and coverslip samples showed no differentiation between common or rare OTUs as a function of mussels, and the clustering of nodes was highly similar. Further, the presence of mussels was not associated with any changes in the relative weight of deterministic and stochastic forces governing community assembly. OTU co-occurrence was significantly non-random on rock surfaces (c-score analysis; p < 0.01), and this pattern did not change with mussels.

Discussion

All sampled substrates had microbial assemblages, which showed variation in important taxonomic and functional properties. Differences in diversity and OTU composition may reflect differences in colonization preference, temporal dynamics, the duration in the environment to accrue microbes, and host-specific interactions on the longer-lived macrobiota. The red alga, the anemone and the mussels are all long-lived and could have been present for years prior to our sampling and had ample time to accumulate a microbial assemblage. The clonal nature of the anemone Anthopleura elegantissima might even provide nearly immortal tissue for microbial proliferation. Relatively high diversity on rocks and crucible lids may indicate a number of micro-niches, perhaps as a result of surface irregularities on these substrates. The biotic substrates we sampled are especially likely to have a number of microbial niches, including microtopography and strong oxygen gradients (Heisterkamp et al., 2013).

Similarities between the filtered plankton and the animal tissue may have resulted from mussel and anemone tissue harboring planktonic microbiota, due to their feeding activities. Although increased sample size will be needed to quantify the extent of within-host heterogeneity, the animals we examined were dominated by Gammaproteobacteria, including the symbiont-bearing anemone. In contrast, coral reef invertebrates with eukaryotic photosynthetic symbionts are have been shown to be dominated by Alphaproteobacteria (e.g., Bourne et al., 2013), suggesting differential colonization drivers in different symbiont-bearing invertebrates. Nevertheless, many specialized symbionts of marine bivalves fall within the Gammaproteobacteria (Stewart & Cavanaugh, 2006; Newton, Girguis & Cavanaugh, 2008).

The great similarity in microbial community structure on inert substrates in tidepools regardless of the presence or absence of mussels (Figs. 1B, 3, 5 and 6), suggests that the enhanced nitrogen regeneration and uptake that has been demonstrated with mussels (Pather et al., 2014) could be due to the microbes on the mussels themselves. The only important functional difference that we found was a greater incidence of nitrate reducing taxa on rocks in the presence of mussels (Table 3), a result likely explained by the ability of mussels to temporarily reduce oxygen levels to the point where reducing processes are favored. OTUs associated with nitrification did not differ. Although mussel presence was associated with a decreased alpha diversity on inert substrates (Fig. 1), this did not lead to significant differences in the network of OTUs that were shared across tidepools with and without mussels and thus did not affect beta diversity (Fig. 6). Hence, the macrobiota did not drive any shifts in community structure on nearby inert substrates, even though tissue-associated microbial communities were significantly differentiated. Both 16S rRNA and shotgun metagenomic analyses (Table 2) suggest that macrobiota host OTUs that are important for nitrogen cycling. This corroborates work by Welsh & Castaldelli (2004) that showed nitrogen metabolism was hosted within a related mussel species. Similarly, deep-sea mussels are known to host nitrogen-utilizing symbionts (Lee & Childress, 1995). In contrast with the shell surface microbes, those with mussel gill and siphon were less diverse (Fig. 1). Thus, the elevated nitrite concentrations in tidepools with mussels, suggesting increased nitrification (Pfister, 2007), may be the result of a direct effect of habitat provisioning for microbes by the macrobiota, rather than simply a microbial community shift on other substrates due to nutrient provisioning. Alternatively, the nitrogen increase due to mussels may not be enough to drive microbial community differences. Indeed, neither microbial community composition nor the expression of several functional genes in coastal sediments showed major changes in response to nutrient perturbations (Bowen et al., 2011). Thus, further exploration of macrobiota as the repository of microbial function, not just as providers of nutrient resources, is warranted.

We note that nitrogen-transforming taxa were found with a higher incidence in the metagenomic data than the 16S rRNA V4 amplicon sequencing of mussel shells. Although there are technical differences between the two sequencing methodologies that could lead to detection differences, such as GC bias (Ross et al., 2013), metagenomic data directly identify genes for nitrogen transformations that are relatively conserved and identifiable (independent of host phylogeny), even if the taxa hosting these genes are uncharacterized. Many of the taxa identified in the amplicon survey are not closely related to cultured isolates, and their N-cycling status is unknown. Although this limits our ability to infer function from phylogeny, it nevertheless is an analysis that may become increasingly insightful as our knowledge of sequence-based diversity increases. While we recognize that OTU analysis using 16S rRNA data predictions may not yet be the strongest lens to detect function, the increased characterization of taxa involved in nitrogen metabolisms (Ward, Arp & Klotz, 2011; Munn, 2011), and the analyses of 16S data with PICRUSt (Table 2) suggests that functional inference is possible.

Marine microbial community structure has been shown to be composed of a multitude of rare taxa that have deep phylogenetic differences; the extent to which this ‘rare biosphere’ (Sogin et al., 2006) drives community function is unknown, though deep sequencing efforts have revealed that there could be the equivalent of a ‘seed bank’ of rare taxa that are persistent with only relative abundance changing through time and space (Lennon & Jones, 2011; Caporaso et al., 2012; Gibbons et al., 2013). In some ecosystems, rare taxa have also been shown to be as metabolically active as common species (Hamasaki et al., 2007), suggesting that rarity does not preclude functional importance. Although seawater and rock had the highest OTU diversity, the differences among filtered plankton and inert substrates versus biogenic substrates (mussels, seaweed, anemone) sampled in situ demonstrated that macrobiota enhance beta diversity by hosting unique OTUs (Figs. 3 and 4B), presumably via the provisioning of unique habitats or resources. It is possible that particular microbial taxa end up in association only with intertidal macrobiota, though the selectivity of these associations requires further temporal and spatial sampling. Our results with these benthic macrobiota are, however, in direct contrast to analyses of seawater where the patterns of beta diversity did not differ among abundant and rare taxa (Amaral-Zettler et al., 2010).

The demonstration that macrobiota host a unique microbial community compared to the water around them is supported by this study and other recent work with tadpoles (McKenzie et al., 2012) and marine algae (Michelou et al., 2013). Although it has been recognized for several decades that benthic invertebrates in deep sea environments host unique taxa e.g., DeChaine & Cavanaugh (2006), it may be that benthic macrobiota common to large parts of the ocean are also repositories for unique microorganisms (Grossart et al., 2005; Lee et al., 2011; Bengtsson et al., 2012; Jackson et al., 2012) and loci for important biogeochemistry (Martínez-García et al., 2008; Heisterkamp et al., 2013). The rocky intertidal flora and fauna, though relatively well-understood in terms of the interactions among macrobiota, likely also interact and mediate productivity via a rich microbial community that we are just beginning to describe. These intertidal macrobiota may also harbor a unique set of taxa adapted to host-associated niches, thereby promoting microbial community diversity in the coastal ocean.

Supplemental Information