Abstract
Animals acclimate to changes in their environment through diverse responses, including phenotypic plasticity and shifts in their microbiome. These microbial communities are also taxonomically distinct across the geographical distribution of the host. It is less known, however, whether taxonomic differences in host-associated bacterial communities between geographically distinct populations mask shifts due to environmental changes within a population. We tested for potential ecological masking using larvae of the echinoid Strongylocentrotus droebachiensis from three coastal locations in the Pacific and Atlantic Oceans that were exposed to four feeding regimes. When considering OTU membership and the relative proportion of those taxa, the composition of the larval-associated bacterial communities were best explained by location, not feeding regime. Similarly, predicted metagenomic gene profiles from these bacterial communities were congruent with population specificity and may suggest a role in metabolism. We hypothesize that, while much of the differences in the bacterial communities is related to the large geographic distances between these locations, the predicted overlapping functions of the microbiome may relate to responding to ecological variation experienced by these larvae. Taken together, these results suggest that differences in community composition between populations masks local variation, and that scaling should be considered in when studying microbiome dynamics.
Introduction
Acclimating to environmental variability through morphological, developmental, and/or physiological plasticity is a common trait of animals (Boidron-Metairon 1988, Bradshaw 1965, DeWitt et al 1998, Miner et al 2005, Schlichting and Smith 2002, Sterns 1989, West-Eberhard 2003). Over the past decade, the appreciation for the role that animal-associated microbial communities play in ameliorating environment-induced stress has grown profoundly (Apprill 2017, Carrier and Reitzel 2017, Carrier and Reitzel 2018, Kohl and Carey 2016, Macke et al 2016, Shapira 2016, Theis et al 2016). When experiencing a heterogeneous environment, the animal host may recruit, expel, and/or shuffle the relative proportion of associated microbiota (Bordenstein and Theis 2015, Zilber-Rosenberg and Rosenberg 2008), to assemble a community with particular molecular pathways for the environmental conditions (Burke et al 2011, Louca et al 2016, Roth-Schulze et al 2018).
Microbial communities associated with animals often vary in response to diverse abiotic and biotic factors, including temperature, salinity, diet quality and quantity, season, and habitat-type (see reviews by Carrier and Reitzel 2017, Kohl and Carey 2016). Of these, dietary responses are best studied and have a major impact on the composition of and potential mutualistic functions for this community (David et al 2014, Kohl and Dearing 2012, Rosenberg and Zilber-Rosenberg 2016, Sonnenburg et al 2016). When faced with prolonged food deprivation, for example, the community composition and diversity of microbiota associated with both invertebrate and vertebrate hosts shift considerably (Carrier and Reitzel 2018, Kohl et al 2014), a response hypothesized to buffer against reduced exogenous nutrients.
Microbial communities associated with animals are also species-specific (Carrier and Reitzel 2018, Fraune and Bosch 2007, Schmitt et al 2012) and taxonomically variable across the geographical distribution of the host species (Dishaw et al 2014, Huang et al 2018, Marino et al 2017, Marzinelli et al 2015, Mortzfeld et al 2015). Habitat-specific microbial communities are primarily controlled by environmental conditions (Pantos et al 2015) and diverge with respect to dispersal limitations (Moeller et al 2017). Despite this taxonomic variation, microbial communities can remain functionally similar due to shared genes across bacterial species (Roth-Schulze et al 2018). The bacterial communities of the green alga Ulva spp., for example, are too variable to define a ‘core’ community; however, nearly 70% of the microbial genes are biogeographically consistent (Roth-Schulze et al 2018). How components of host ecology attribute to the taxonomic variation in these bacterial communities is less understood and are needed to identify the relative strength and importance of each abiotic or biotic factor. Planktotrophic (feeding) larvae are one biological system to compare the components of host ecology and their dynamics on animal microbial communities. At a local scale, many planktotrophic larvae (e.g., the pluteus of sea urchins) inhabit heterogeneous feeding environments and are morphologically and physiologically plastic to food availability (Adams et al 2011, Boidron-Metairon 1988, Byrne et al 2008, Carrier et al 2015, Hart and Strathmann 1994, McAlister and Miner 2018, Miner 2004, Miner 2011, Soars et al 2009). Feeding-induced plasticity in the echinoid Strongylocentrotus droebachiensis, specifically, is correlated with phenotype-,diet-,and development-specific bacterial communities (Carrier and Reitzel 2018). At a regional scale, adult S. droebachiensis have an Arctic-boreal distribution (Scheibling and Hatcher 2013) and can be divided into genetically distinct populations across multiple oceans (Addison and Hart 2004, Addison and Hart 2005). Attributes of the reproductive biology of S. droebachiensis (e.g., sperm morphology) have significant phenotypic variation between populations, suggesting potential directional selection (Manier and Palumbi 2008, Marks et al 2008). How these differences in reproductive characteristics relate to the variation in other phenotypic traits, such as larval morphological plasticity or symbioses with bacteria have not been studied, although population-specific variation and differential selection would not be surprising given the differences experienced in their natural environments.
The ability of S. droebachiensis larval holobionts to acclimate to local feeding variation across its broad geographic distribution was used as a biological system to evaluate local versus regional effects. Specially, using S. droebachiensis larvae, we tested the hypothesis that host geographical origin better correlates with community composition than does local variations on food availability, and that the predicted functional gene profiles converge between host habitats. To test these hypotheses, S. droebachiensis larvae from three sites (Figure 1) were differentially fed, and the associated bacterial communities were assayed and used to coarsely predict functions of the metagenome.
Experimental Procedures
Adult urchin collection and larval rearing
Adult S. droebachiensis were collected from populations in the North Sea in March 2015, the Salish Sea in April 2016, and the Gulf of Maine in February 2017 (Figure 1). Individuals from the North Sea were collected by divers in Droebak, Norway (59°39’ N, 10°37’ E) and transported in cold and aerated seawater to the Sven Lovén Centre for Marine Infrastructure (Kristineberg, Sweden). Urchins were maintained in natural seawater and fed ad libitum on a live mix of Ulva lactuca and Laminaria spp. collected from the Kristineberg shoreline. Urchins from the Salish Sea were hand-collected at low tide at Cattle Point, San Juan Island, USA (48°27′ N, 122°57′ W), transferred to the Friday Harbor Laboratories within one hour, suspended in subtidal cages off the dock at FHL, and fed Nereocystis spp. ad libitum until spawning two weeks later. Lastly, individuals from the Gulf of Maine were collected from Frenchman Bay, Maine (44°25′ N 68°12′ W), shipped overnight to the Darling Marine Center, and were maintained in flow-through aquaria and fed Saccharina latissima ad libitum until spawning within one week.
Adult urchins were spawned with a one-to two-mL intracoelomic injection of 0.50 M KCl. For each population, gametes from three males and three females were separately pooled. Fertilization of eggs and larval rearing followed Strathmann (1987), except, to include the environmental microbiota, embryos and larvae were reared using 5.0-μm filtered seawater (FSW). Briefly, embryos were incubated in one-liter of FSW at ambient temperature and salinity, and two hours post-fertilization were transferred to three or five-L of FSW, divided into triplicates, and larval density was adjusted to two larvae•mL-1 and subsequently diluted as larvae reached the 6-and 8-armed stages. Larval cultures were given 90 to 95% water changes every other day by reverse filtration.
Monocultures of Rhodomonas lens were grown at room temperature with f/2 media and a combination of ambient and artificial lighting (Guillard 1975).
Experimental feeding and larval morphometrics
At 48 hours post-fertilization, prism-stage larvae were divided into three replicate jars for each of the four experimental feeding treatments varying in R. lens quantity: 10,000, 1,000, 100, or 0 cells•mL-1. For each experiment, larvae fed 10,000 cells•mL-1 were reared through metamorphosis while diet-restricted and started larvae were cultured until developmental stasis (Supplemental Table 1). Larvae (n=100) from each replicate for each treatment were sampled weekly. Immediately after sampling, larval samples were concentrated into a pellet using a microcentrifuge and all seawater was removed. Pelleted larvae were then preserved in RNAlater and stored at-20 °C before DNA extractions.
Complementary to sampling S. droebachiensis larvae, the environmental microbiota from the seawater was also sampled. When larval cultures were sampled weekly, triplicates of ∼1-L of seawater was filtered onto a 0.22-µm Millipore filter to retain the environmental microbiota. Full filter disks were then preserved in RNAlater and stored at-20 °C before DNA extractions.
In addition to sampling larvae to assay the associated bacterial communities, twenty larvae (n=20) from a single replicate from each dietary treatment were sampled for morphometric analysis. Larvae were imaged using a compound microscope (Salish Sea: Nikon Eclipse E600; camera: QImaging MicroPublisher 5.0 RTV; Gulf of Maine: Zeiss Primo Star HD digital microscope; North Sea: Leica stereomicroscope) and morphometrics (length of larval body, postoral arms, and stomach area; Supplementary Figures 1-2) were measured using ImageJ, v. 1.9.2 (Schneider et al 2012). We tested whether larval morphology and stomach volume were influenced by differential feeding over time using a two-way ANOVA (JMP Pro v. 13), and a whether this pattern was site-specific using a one-way ANOVA. Where statistical differences were observed (p<0.05), we used a posthoc test to determine the affect at each time point and for each diet.
Assaying microbial communities
We extracted total DNA from larval samples using the GeneJet Genomic DNA Purification Kit (Thermo Scientific). For filtered seawater samples, we extracted eDNA using the FastDNA Spin Kit for Soil (MP Biomedical). DNA was then quantified using the NanoDrop 2000 UV-Vis Spectrophotometer (Thermo Scientific) and diluted to 5 ng•μL-1 using RNase/DNase-free water.
Bacterial sequences were amplified using ‘universal’ primers for the V3/V4 regions of the 16S rRNA gene (Forward: 5′ CTACGGGNGGCWGCAG, Reverse: 5′ GACTACHVGGGTATCTAATCC) developed by (Klindworth et al 2013). Products were purified using the Axygen AxyPrep Mag PCR Clean-up Kit (Axygen Scientific), indexed via PCR using the Nextera XT Index Kit V2 (Illumina Inc.), and then purified again. At each of these steps, fluorometric quantitation was performed using a Qubit (Life Technologies) and libraries were validated using a Bioanalyzer High Sensitivity DNA chip (Agilent Technologies). Illumina MiSeq sequencing (v3, 600 cycles) was performed at the University of North Carolina at Charlotte.
Forward and reverse sequences were paired and trimmed using PEAR (Zhang et al 2014) and Trimmomatic (Bolger et al 2014), respectively, converted from fastq to fasta using custom script (Supplemental Note 1), and, prior to analysis of bacterial 16S rRNA sequences, chimeric sequences were detected using USEARCH (Edgar et al 2011) and removed using filter_fasta.py. Using QIIME 1.9.1 (Caporaso et al 2010), bacterial 16S rRNA sequences were analyzed and grouped into operational taxonomic units (OTUs) based on a minimum 97% similarity. The biom table generated by the pick_open_reference_otus.py script was filtered of OTUs with less than ten reads as well as sequences matching chloroplast for cryptophytes (i.e., R. lens; after (Carrier and Reitzel 2018).
Using the filtered biom table and “biom summarize-table” function to count total sequences per sample, the rarefaction depth of 3,193 was determined and applied to all subsequent analyses (Supplemental Figure 1). Beta diversity was calculated using the weighted UniFrac (Lozupone and Knight 2005), and principal coordinate analyses (PCoA) were visualized in EMPeror (Vazquez-Baeza et al 2013) and recreated using PhyloToAST (Dabdoub et al 2016) or stylized for presentation in Adobe Illustrator CS6. Community composition was generated using summarize_taxa_through_plots.py script and visualized using Prism 7 (GraphPad Software). Community similarity across phenotypes, dietary states, and developmental stages were compared statistically using an ANOSIM as part of the compare_categories.py script.
A step-by-step listing of QIIME scripts used to convert raw reads to OTUs for visualization of the data is located in Supplementary Note 1.
Functional predictions using PICTUSt
For the QIIME-generated OTU (i.e., biom) table to be compatible with PICTUSt (Langille et al 2013), all de novo OTUs were filtered according to the Greengenes (v. 13.5) database. Closed OTU tables (that retained 57.8% and 88.2% of OTUs from full and ‘shared’ communities, respectively) were normalized by copy number, upon which metagenomic gene profiles were predicted and categorized by biological function. The PICTUSt output was made compatible for STAMP using biom_to_stamp.py from Microbiome Helper (Comeau et al 2017). These metadata were subsequently analyzed using STAMP (Parks et al 2014) to test for a population-specific predicted functional profile. The principle coordinate from STAMP were compared statistically using a one-way ANOVA (JMP Pro, ver. 13), as part of the STAMP package. We then generated taxa plots for Gene Ontology groups of interest using metagenome_contributions.py and custom scripts.
A step-by-step listing of PICTUSt scripts used to convert from QIIME and the subsequent data analysis is located in Supplementary Note 1.
Results
Larval holobionts and the feeding environment
Diet-induced morphological plasticity was recorded for S. droebachiensis larvae from each population, with the pattern of expression being location-specific (ANOVA, p<0.0001; Supplemental Figures 2-3). Diet-restricted larvae from the Salish Sea and Gulf of Maine exhibited a higher post-oral arm to mid-body line ratio than ad libitum counterparts (ANOVA, p<0.0001; Supplemental Figure 2; Supplemental Table 2), even though analyses of Gulf of Maine larvae were confounded by developmental stage (Supplemental Table 1). Larvae from the North Sea, however, exhibited the opposite response: ad libitum feeding induced a higher post-oral arm to mid-body line ratio than diet-restriction (ANOVA, p<0.0001; Supplemental Figure 2; Supplemental Table 2).
S. droebachiensis larvae from each population associated with a diet-specific bacterial community (Supplemental Figures 4-6; ANOSIM, Supplemental Table 4). Larvae from the Salish Sea and Gulf of Maine exhibited similar diet-specific community-level patterns (Supplemental Figures 4-5; ANOSIM, Supplemental Table 4), where the bacterial consortium of food-restricted individuals generally were more similar to each other than to wellfed counterparts. Larvae from the North Sea, on the other hand, exhibited the opposite response (Supplemental Figure 6), where dietspecific bacterial communities were still observed (ANOSIM, Supplemental Table 4) except that all food rations were more similar to each other than to starved individuals (Supplemental Figure 6). In addition to diet-specificity, larvae from each population associated with bacterial communities that were specific to phenotype (Supplemental Figure 7; ANOSIM, Supplemental Table 4) as well as varied with developmental stage and/or ecological/stochastic drift (Supplemental Figure 8-10; ANOSIM, Supplemental Table 4) and were distinct from the environmental bacterial community (Supplemental Figure 11; ANOSIM, Supplemental Table 4).
Location-specific bacterial communities
Variation in OTU diversity and the relative proportions of those taxa associated with S. droebachiensis larvae were best correlated with geography (ANOSIM, p<0.001; Figure 2), where larvae from the Western and Eastern Atlantic Ocean were more similar to each other than to those from the Pacific Ocean (Figure 2; Supplemental Figure 12A). Site-specific bacterial communities of S. droebachiensis larvae were independent of plasticity state, developmental stage, and feeding regime (Figure 2; Supplemental Figures 4-10, 13-14; Supplemental Table 2), even though larvae at each site associated with phenotype-(Supplemental Figure 7), diet-(Supplemental Figure 8-10), and development-(Supplemental Figure 4-6) specific bacterial communities (Supplemental Table 2). Moreover, the structure of the bacterial community associated with Gulf of Maine larvae are taxonomically richer and the most diverse while larvae from the Salish Sea were the least rich and diverse, leaving those from the North Sea as intermediate (Table 1).
Of the thousands of OTUs associated with S. droebachiensis larvae across sites, phenotypes, developmental stages, and diets, ∼32.7% were found in at least one sample at each of the three sites (Supplemental Figure 15). Moreover, ∼8.1% to ∼13.0% of all OTUs were shared between two sites, and ∼10.8% to ∼12.7% were unique to a single site (Supplemental Figure 15). When clustered by bacterial classes, S. droebachiensis larvae from the North Sea primarily associate with γ-proteobacteria (34.2%; Phylum: Proteobacteria), α-proteobacteria (26.8%; Phylum: Proteobacteria), Flavobacteriia (19.5%; Phylum: Bacteroidetes), and Saprospirae (12.3%; Phylum: Bacteroidetes), while larvae from the Gulf of Maine primarily associated with γ-proteobacteria (49.1%;), α-proteobacteria (17.8%), Flavobacteriia (17.7%), and, lastly, larvae from the Salish Sea primarily associate with Flavobacteriia (44.3%), α-proteobacteria (23.2%), and γ-proteobacteria (20.0%) (Supplemental Figure 15).
Dynamics of shared taxa
For balanced inter-population comparisons, only OTUs in at least one sample from each population were retained. This restriction yielded 4,502 shared OTUs (Supplemental Figure 15), which were divided by population and subsequently filtered to only include ‘core’ taxa (i.e., those found in all samples for a given urchin population). Inclusion of the ‘core’ OTUs for each population totaled 178 OTUs (Supplemental Figure 16). An unweighted and weighted comparison of these OTUs suggest that these ‘core’ communities associated with S. droebachiensis larvae is, again, best correlated with geography (ANOSIM, p<0.001; Figure 3), with larvae from the Atlantic Ocean being more similar to each other than larvae from the Pacific Ocean (Figure 3; Supplemental Figure 12B).
Of the combined ‘core’ bacterial communities associated with S. droebachiensis larvae, three OTUs (∼1.8%) were found in all samples within and between populations: an unclassified species in the class γ-proteobacteria (OTU number: 1106577), an unclassified species in the family Flavobacteriaceae (OTU number: 1105269), and an unclassified species in the genus Polaribacter (OTU number: 586650). Of the 178 OTUs, 27 OTUs were shared between North Sea and Gulf of Maine samples, four between all Gulf of Maine and Salish Sea samples, and three between all Salish Sea and North Sea samples (Supplemental Figure 16). Furthermore, 6, 52, and 78 OTUs were specific to S. droebachiensis larvae from the Salish Sea, Gulf of Maine, and North Sea, respectively (Supplemental Figure 16). When clustered by bacterial classes, these ‘core’ communities associated with S. droebachiensis larvae, individuals from the North Sea primarily included α-proteobacteria (41.8%; Phylum: Proteobacteria), γ-proteobacteria (34.8%; Phylum: Proteobacteria), and Flavobacteriia (20.9%; Phylum: Bacteroidetes), while larvae from the Gulf of Maine primarily associated with γ-proteobacteria (57.0%), α-proteobacteria (17.4%), and Flavobacteriia (17.1%), and larvae from the Salish Sea primarily associate with Flavobacteriia (67.3%) and α-proteobacteria (30.0%) (Supplemental Figure 16).
Predicted community function and representative taxa
Similar to the 16S rRNA assays, PICRUSt-generated metagenomic gene profiles of the full (ANOVA, p<0.0001; Figure 4A) and shared (ANOVA, p<0.0001; Figure 4B) bacterial community associated with S. droebachiensis larvae were specific to urchin biogeography. Predicted gene content of PICRUSt-generated metagenomes were, on average, ∼55.1% metabolism, ∼17.9% genetic information processing, ∼12.5% environmental information processing, and ∼7.2% cellular processes and signaling (Supplemental Figures 17-18). Of these, total gene content was significantly different between locations for metabolism (ANOVA, p<0.0001), cellular processes (ANOVA, p<0.0001), and cellular processes and signaling (ANOVA, p<0.0001) (Supplemental Figure 17-18). Moreover, of only the shared or ‘core’ community ∼55.8% metabolism, ∼17.3% genetic information processing, ∼12.8% environmental information processing, and ∼6.8% cellular processes and signaling (Supplemental Figure 19-20), with total gene content being significantly different between locations for metabolism (ANOVA, p<0.0001), cellular processes (ANOVA, p<0.0001), and cellular processes and signaling (ANOVA, p<0.0001).
More than half of the gene content of the predicted S. droebachiensis larval metagenome was related to metabolism (Supplemental Figures 17-18). Several of the bacterial classes from the Salish Sea, Gulf of Maine, and North Sea are predicted to contribute to metabolism. For S. droebachiensis larvae from the Salish Sea this group of bacteria primarily included the α-proteobacteria (41.5%), γ-proteobacteria (29.7%), and Flavobacteriia (21.8%); larvae from the Gulf of Maine primarily included α-proteobacteria (41.9%), γ-proteobacteria (39.1%), Flavobacteriia (16.5%); and larvae from the North Sea primarily included γ-proteobacteria (40.5%), α-proteobacteria (29.8%), and Flavobacteriia (25.8%) (Figure 6).
Discussion
Comparisons of the bacterial communities associated with S. droebachiensis larvae across dietary treatments and host biogeography suggests three primary findings. First, the composition of the bacterial community associated with S. droebachiensis larvae is best correlated with location. Second, for each of the geographical regions, urchin larvae associated with a microbial community specific to phenotype, developmental stage, and dietary state. Lastly, the predicted metagenomic profiles are site-specific and primarily related to metabolism.
Marine invertebrate larvae experience a feeding environment that varies in space, time, and composition (Bidigare and Ondrusek 1996, Chevez et al 1996, Cloern and Jassby 2010, Milici et al 2016, Needham and Fuhrman 2016). In response to this variation, planktonic larvae can arrest their development and/or increase the frequency of encounter rates by enlarging their feeding structure (Boidron-Metairon 1988, Byrne et al 2008, Carrier et al 2015, Carrier and Reitzel 2018, Hart and Strathmann 1994, McAlister and Miner 2018, Miner 2004, Miner 2011, Soars et al 2009). Plasticity in development and morphology have historically been viewed as a means for the host to acclimate (Bradshaw 1965, Hart and Strathmann 1994, McAlister and Miner 2018, Miner 2011, Soars et al 2009, West-Eberhard 2003). Recent work suggests this response is also linked to the associated bacterial communities (Carrier and Reitzel 2018).
Inter-population comparisons of the dynamics of the bacterial community associated with S. droebachiensis larvae suggest three additional, not mutually exclusive inferences. First, composition of the bacterial communities is seemingly a product of the host feeding environment. Second, while both community composition and predicted functional profiles are dynamic across and specific to host feeding environment, the functional profiles are more informative for understanding hologenomic acclimation. Third, urchin larval holobionts may be locally adapted. Of these, the data presented here largely support the first inference, while the latter two are supported but require specific validation (e.g., using molecular, genomic, and physiological assays and manipulations (Williams and Carrier 2018).
Population-specific bacterial communities is an emerging theme of animal-microbiome ecology (Dishaw et al 2014, Huang et al 2018, Marino et al 2017, Marzinelli et al 2015, Mortzfeld et al 2015). For the three populations of S. droebachiensis used here, the bacterial communities were region specific, with Gulf of Maine and North Sea individuals being more similar to each other than to Salish Sea larvae. The environmental conditions of these regions are different, and the selective pressures on the microbial partners, larva, and holobiont likely vary (Bordenstein and Theis 2015). Differential selection on multiple components of the S. droebachiensis larval holobiont may result in local adaption (Pespeni et al 2013, Sanford and Kelly 2011). The feeding environment-specific differences in the microbial community and predicted metagenomic gene profiles are suggestive of potential adaptation, where the functional microbial community may aid in acclimating to unique oceanographic conditions of these three S. droebachiensis larval populations face.
Previous studies on the populations of S. droebachiensis larvae documented that phenotypic traits varied across host geography. Manier and Palumbi (2008), for example, reported significant differences in sperm morphology between urchins across this spatial scale and found evidence of strong directional selection for sperm traits between locations, particularly between S. droebachiensis in the Pacific and Western Atlantic. Population genetic studies of S. droebachiensis, on the other hand, show significantly higher FST values between the Eastern and Western Atlantic than the Pacific and Western Atlantic due to more frequent genetic exchange through the Bering Strait (Addison and Hart 2004, Biermann et al 2003, Manier and Palumbi 2008, Palumbi and Wilson 1990). Consistent with sperm morphology, the population-specific differences in bacterial communities and predicted metagenome is suggestive of environmental influence shaping this variation over, perhaps, the last few hundred thousand years. Future population genomic studies of S. droebachiensis should identify outlier loci that correlate with specific differences in OTUs and characterize the larval metagenome to provide a window into how animal genetic variation and the environmental conditions may shape the associated microbial community.
A growing body of literature suggests that planktotrophic larvae utilize diverse ‘alternative’ nutritional resources (Feehan et al 2018, Manahan et al 1993, Rivkin et al 1986). Based on our predicted metagenomic gene profiles, we propose that planktotropic larvae are aided by metabolites derived from their bacterial symbionts. This is of particular importance because larvae often inhabit food-limited environments (Fenaux et al 1994, Olson and Olson 1989, Pauley et al 1985). To decrease mortality due to starvation (Morgan 1995, Rumrill 1990, Young and Chia 1987), a metabolic input from the symbiont community may serve as a physiological buffer and complement metabolic depression induced by diet-restriction (Carrier et al 2015).
Our comparisons of the bacterial communities associated with S. droebachiensis larvae suggest that geographic location better correlated with community composition than local biological (e.g., phenotype) and ecological variation (e.g., diet quantity). This type of specific comparison suggests that in studying the dynamics of animal—and perhaps plant—interactions with their associated bacterial community, location may drive the taxonomic profiles, and that in transitions towards functional or predicted functional comparisons, the potential for local adaptation should be considered (Kelly et al 2014, Pespeni et al 2013, Sanford and Kelly 2011).
Conflict of Interest
The authors declare that they have no conflict of interest.
Acknowledgements
We thank the staffs of the Friday Harbor Laboratories, Darling Marine Center, and Sven Lovén Centre for Marine Infrastructure for facility access and logistical assistance; Jason Hodin and Billie Swalla for providing laboratory space; Richard Strathmann for collection assistance; Daniel Janies for sequencing resources; Karen Lopez for technical assistance with sequencing; and Jason Macrander for bioinformatics assistance.
T.J.C. was supported by an NSF Graduate Research Fellowship, Charles Lambert Memorial Endowment fellowship from the Friday Harbor Laboratories, and a Sigma Xi Grants-in-Aid of Research grant; S.D. was funded by the Centre for Marine Evolutionary Biology and supported by a Linnaeus grant from the Swedish Research Councils VR and Formas; and A.M.R. was supported by NSF DEB1545539 and Human Frontier Science Program Award RGY0079/2016.