Sample collection and processing

Jaera spp. are common intertidal invertebrates that are neither protected nor require sampling permits. Live individuals were collected in spring and summer 2017 from two coasts in north-east Scotland, separated by c. 200 km of coast line. Gardenstown on the north coast (57.672 °N, –2.337 °E) harbours all four European species of the species complex alongside the outgroup Jaera nordmanni in varying composition along the shoreline. Two beaches on the south-east coast in close vicinity to each other (Johnshaven: 56.796 °N, –2.328 °E; Arbroath: 56.518 °N, –2.659 °E) harbour >95 % pure single-species populations of Jaera albifrons and Jaera ischiosetosa respectively.

As soon as possible after collection, individuals were sexed and assigned to species by identifying patterns of pereiopod setation in males using light microscopy [40, 41]. Identified males were then kept at room temperature in a large tub containing filtered sea water from the collection site. Since females cannot be assigned to species morphologically [40, 41], we assigned putative species based on species composition of males collected at the same beach section. An approximately even species mix of females was added to the same tub as the males, and all individuals were starved for one week to reduce gut content. Individuals were then briefly rinsed in sterile water and immediately processed for DNA extraction.

We generated eleven DNA samples that comprised one pure male pool (6-12 individuals) for each of all five species, five mixed or presumably pure female pools (6-12 individuals), and a single exceptionally large female of unknown species (Table 1). These samples not only allowed us to screen males and females of all species for sex-ratio distorting parasites, but also to capture a broad snapshot of the Jaera microbiome across coasts and seasons. Samples were homogenized in heated (60 °C) lysis buffer (0.1 M Tris-HCl, 0.05 M EDTA, 0.1 M NaCl, 1 % SDS, 400 μg Proteinase K, 100 μg RNAse A) using an autoclaved teflon plunger. The homogenate was incubated overnight at 60 °C and DNA was extracted via standard phenol-chloroform extraction. DNA quality and quantity were checked with a NanoDrop ND-1000 spectrophotometer. Samples were submitted to Eurofins Genomics (Ebersberg, Germany) for bacterial 16S rRNA V3/V4 amplicon generation (c. 420 bp) using the standard S-D-Bact-0341-b-S-17 and S-D-Bact-0785-a-A-21 primers [51] with sample-specific barcodes, and sequencing on the Illumina MiSeq V3 platform in 300 bp paired-end mode.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of sequencing effort and sequence diversity across eleven samples.

https://doi.org/10.1371/journal.pone.0202212.t001

Sequence assembly, curation and taxonomic classification

Raw sequence reads were filtered, de-noised and assembled to unique single-end (forward reads only) as well as paired-end sequence variants using dada2 v1.4.0 [52] in r v3.4.0 [53]. Reads were trimmed at nucleotide call quality below 2, and reads with undetermined bases were discarded. Exploratory quality plots indicated a rapid decline in basecall quality towards the ends of the reads, particularly for reverse reads. Therefore, for paired-end analysis, forward reads were further trimmed to 260 bp and reverse reads to 220 bp, ensuring an overlap of at least 60 bp. Error rates were estimated and sequences were de-noised separately for forward and reverse reads in pooled-sample mode. Single-end and paired-end contigs were assembled from de-noised data and chimera sequences were removed.

Taxonomic classification to genus level was assigned from the silva nr v128 database [54] using the RDP classifier algorithm [55] with k-mer size 8, 100 bootstrap replicates and a minimum bootstrap support of 50. Species-level classification was added where possible based on 100 % sequence identity with silva nr v128 [56]. Sequences that were assigned to chloroplast, mitochondria, archaea, or eukaryota taxa were removed. Each sample was further annotated with putative functional metabolic capabilities of the identified microbial community using tax4fun v0.3.1 and associated pre-computed silva nr v123 reference data [57]. The observed sequence counts were transformed into abundances of KEGG enzymes via association with KEGG reference organisms. These enzymes were further classified with the first three levels in KEGG functional hierarchies [58].

The taxonomically classified microbiomes were then screened for reproductive parasite species in the Wolbachia, Rickettsia, Spiroplasma and Cardinium genera or relevant higher taxonomic levels. Candidate sequences were aligned with all available Wolbachia, Rickettsia, Spiroplasma and Cardinium reference sequences in silva nr v128 using mafft v7.305 [59] and clustered using neighbour-joining on Kimura-2-parameter phylogenetic distances in ape v4.1 [60]. Sequences that clustered closely with the silva reference sequences were more closely examined using ncbi megablast [61] against the non-redundant nucleotide collection (nt).

Microbiome sequence diversity and composition

Sequence diversity analyses were carried out on single-end as well as paired-end datasets, using r and the package phyloseq v1.19.1 [62]. Sequencing depths per sample were examined for circumstantial associations with the categorical sample variables season (spring and summer), region (north: Gardenstown; south: Johnshaven/Arbroath) and sex using negative binomial generalized linear models (GLM) in the mass package [63]. Rarefaction curves were obtained by computing the number of unique sequence variants in subsamples of increasing sizes in steps of 1,000 sequences without replacement [64]. Diversity indices (Chao1, ACE, Shannon, Simpson, inverse Simpson and Fisher) were computed for each sample and compared between samples grouped by season, region or sex using two-tailed Welch’s t-test. Consistency of all metrics between single-end and paired-end datasets was examined using Pearson’s correlation test.

Rarefied versions of the datasets were obtained by subsampling to the lowest sequencing depth across samples. Microbiome structure between samples was explored with Jaccard and Bray-Curtis dissimilarity indices and visualised in two-dimensional space using metric (Jaccard) or non-metric (Bray-Curtis) multidimensional scaling (MDS). Samples were then clustered hierarchically using Ward’s criterion on the dissimilarity matrix, and clusters were visualised as dendrograms using ggtree v1.6.10 [65]. Sources of variation attributed to season, region and sex were explored with distance-based redundancy analysis and permutational multivariate analysis of variance (PERMANOVA) with 9,999 permutations [66]. Finally, to further explore sex-specific differences in microbiome composition, we fitted negative binomial GLMs in a differential gene expression framework that accounts for differences in library size and dispersion, as implemented in deseq2 [67]. Fold changes were calculated between sexes accounting for season and region as covariates. P-values were corrected for multiple testing using the false-discovery rate method [68], and sequences with significant fold changes (FDR ≤ 0.1) were identified.

Taxonomic composition and diversity

De-noised single-end sequence data comprised 64,878–776,641 reads that collapsed to 1,019–1,998 unique sequence variants per sample. Across all eleven samples, 3,317 unique sequence variants were observed, which were assigned to 25 phyla, 45 classes, 94 orders, 185 families and 445 genera, based on the silva nr v128 database. However, considerable fractions of these sequence variants could not be assigned beyond particular taxon levels at the 50 % bootstrap cut-off, i.e., 0.78 % for phylum, 3.33 % for class, 8.06 % for order, 16.21 % for family, 42.15 % for genus and 94.16 % for species levels. Paired-end data recovered less diversity, comprising 29,535–517,273 reads and 329–1,800 unique sequence variants, and captured less diversity with 3,283 unique sequence variants assigned to 23 phyla, 40 classes, 89 orders, 182 families and 426 genera. However, taxonomy assignment was slightly better compared to single-end data, with non-classification rates of 0.45 % for phylum, 1.58 % for class, 4.28 % for order, 10.70 % for family, 35.93 % for genus and 93.97 % for species. Rarefaction curves approached asymptotic stages for most samples, suggesting that the sequencing effort captured the majority of sequence diversity in both types of datasets (S1 Fig).

The Proteobacteria and Bacteroidetes were the dominant phyla in all samples, accounting for 69.7–94.0 % and 5.7–28.2 % of sequences per sample, and the six most abundant phyla accounted for 99.5–99.9 % (Fig 1). The six most abundant orders accounted for 62.4–92.7 % and the six most abundant genera for 23.3–58.6 % of sequences, of which Vibrio dominated most samples with up to 46.1 %. Notwithstanding, microbial sequence diversity was high in all samples, with a Simpson index of 0.835–0.987 and Fisher index of 125–290 (Table 1; S2 Fig). All metrics were highly correlated between single-end and paired-end datasets (r = 0.902 − 0.999;p ≪ 0.001; Table 1). Diversity was similar among seasons, but signatures of region and, in particular, sex were apparent in many diversity metrics (S2 Fig). Rigorous statistical analysis beyond basic Welch’s t-tests was precluded by low sample size, but these tests supported a difference between sexes in particular (Table 1; S2 Fig), consistent with shallower rarefaction curves in females (S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Relative sequence abundances of the six most abundant phyla, orders and genera across eleven samples (S1-S11), organised by season and sex.

https://doi.org/10.1371/journal.pone.0202212.g001

Prediction of broad-brush metabolic capacity of the identified microbial communities via sequence similarity of taxonomically classified sequences to KEGG model taxa recovered 278 KEGG pathways. The proportion of sequences that could not be mapped to KEGG organisms (FTU) ranged from 55.0 % to 96.8 % (median 89.2 %) per sample. The most abundant top-level category was 09100 Metabolism (median 57.3 % across samples), followed by 09130 Environmental Information Processing (23.9 %). The most abundant pathways included 09131 Membrane transport, 09102 Energy metabolism and 09101 carbohydrate metabolism (S3 Fig).

Sex-ratio distorting reproductive parasites

No sequences were directly assigned to known silva strains of the reproductive parasite genera Wolbachia, Rickettsia, Spiroplasma and Cardinium. However, some sequences were assigned to relevant higher taxonomic ranks: one sequence to the family Anaplasmatacea, 16 sequences to Rickettsiacea, 36 sequences to Flammeovirgaceae and one sequence to the order Entomoplasmatales. The counts of these sequences ranged from 0 to 5,583, representing relative abundances of at best 1.16 % (Fig 2; S1 Table). Of these 54 sequences, three clustered reasonably closely with the Rickettsia, Spiroplasma and Cardinium clades, but no sequence clustered closely with Wolbachia (Fig 2).Megablast broadly supported these classifications, matching an uncultured Rickettsiaceae bacterium (accession JQ701668.1) at 90 % identity to the Rickettsia sequence, and an uncultured bacterium from the Cytophaga-Flavobacterium-Bacteroides group (DQ812543.1) at 98 % identity to the Cardinium sequence. However, the presumed Spiroplasma sequence matched with 94 % identity an uncultured bacterium from the Mycoplasmataceae family (EU646196.1), which is situated in a different order than Spiroplasma.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Neighbour-joining dendrogram (K2P phylogenetic distance) of SILVA NR v128 16S rRNA gene reference sequences for reproductive parasites Wolbachia, Rickettsia, Spiroplasma and Cardinium genera (dashed branches) and most closely related Jaera 16S rRNA gene sequences (families Rickettsiaceae and Flammeovirgaceae, and order Entomoplasmatales; solid branches).

The relative sequence abundances of the Jaera sequences are summarised alongside. Tip labels correspond to sequence identifiers in S1 Table.

https://doi.org/10.1371/journal.pone.0202212.g002

Sources of variation in microbiome composition

We further explored microbiome composition for sex-specific, seasonal and regional signatures. Since sequencing depth was different between male and female samples (negative binomial GLM: z = 3.115;P = 0.002), the datasets were rarefied (64,878 sequences for single-end data and 29,535 sequences for paired-end data) to avoid spurious sex-specific signatures in microbiome composition.

Ordination and hierarchical clustering of Jaccard dissimilarity among samples suggested two major clusters that correspond to samples collected in spring and summer. Within both seasons, samples are further clustered by geographic region, and the northern region is further subdivided by sex (Fig 3). Distance-based redundancy analysis ascribed 39.6 % of the total variance to season, 26.9 % to region and 14.1 % to sex, and all three hierarchical variance components were statistically significant (Table 2). Bray-Curtis dissimilarity broadly supported these patterns (Table 2), but hierarchical clustering did not consistently recover the same nested structure (S4 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Hierarchical clustering (left) and metric multidimensional scaling (right) of Jaccard dissimilarity among samples.

Sample categories (season, region and sex) are indicated by line type, symbol shape and colour, respectively.

https://doi.org/10.1371/journal.pone.0202212.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Permutational multivariate analysis of variance (PERMANOVA) in Jaccard and Bray-Curtis dissimilarity indices among samples.

https://doi.org/10.1371/journal.pone.0202212.t002

Since all samples from the southern regions were females, it cannot be ruled out that variance ascribed to region is in fact sex-specific variation. However, this is quite unlikely since variance among males and females in the northern region in summer was considerably smaller than the putative variance among regions in summer (Fig 3). Similarly, although not all species are represented within each cluster, it appears that structure among sexes outweighs structure associated with species or microgeography at the same beach. Jaera nordmanni and Jaera ischiosetosa males collected from the same set of rocks in Gardenstown (north) in summer clustered most closely, followed by Jaera praehirsuta males collected further down the shore at the same time. However, the single large female (unknown species) collected at the same beach and time did not cluster as closely with any of these three male single-species samples. Instead, she clustered most closely with a mix of females collected from the same rocks as the three male samples (Fig 3).

Exploring the identified sex-specific signal in microbiome composition further with negative binomial models indicated that eleven sequences were significantly (FDR ≤ 0.1) more abundant in males and six sequences were more abundant in females, after accounting for differences in season and region (Fig 4). Of these 17 sequences ten had taxonomic annotation, representing ten genera in nine families: Aureispira, Peredibacter, Loktanella, Winogradskyella and Pibocella genera were more abundant in males, and Tenacibaculum, Marinomonas, Aliiroseovarius, Leisingera and Pelagibius genera were more abundant in females (Fig 4). A simplified analysis within the northern region only recovered similar patterns, corroborating Pibocella as male-associated and suggesting Vibrio and Owenweeksia as additional female-associated taxa (S5 Fig). Likewise, using non-rarefied data and a stricter significance threshold, a similar set of differentially abundant taxa was identified, highlighting Aliivibrio, Flavirhabdus and Polaribacter as further female-associated taxa (S6 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Sequence variants with differential abundance between sexes.

The left panel summarises fold change and statistical significance for each sequence variant. The following two panels illustrate total aggregated sequence counts (abundance) and taxonomic classification (family or genus) of statistically significant (FDR ≤ 0.1) sequence variants.

https://doi.org/10.1371/journal.pone.0202212.g004

Characterisation of Jaera microbiomes

The microbial communities of all samples were dominated by the Proteobacteria and Bacteroidetes phyla, which are widely described as the most abundant phyla in intertidal and open oceanic environments [36, 69–72]. The high abundance of Vibrio in particular is consistent with microbiomes of other marine invertebrates such as copepods [35] or sea urchins [73]. Vibrio is a very common endo- and epibiont in marine crustaceans and often produces chitinolytic enzymes that allow for exploiting chitinous exoskeleton as a niche for attachment and proliferation [71, 74, 75]. Some Vibrio species are pathogens and others have been implicated in biogeochemical processes, but the specific metabolic relationships between crustaceans and Vibrio are cryptic [74].

Beyond these dominant taxa, sequence diversity in Jaera was high, consistent with diversity in intertidal sponges [36] and other marine invertebrates [35], and exceeded diversity of the terrestrial isopod Armadillidium vulgare [31]. A large proportion of sequences could not be taxonomically characterised to species level, and functional classification was hampered by very low mapping rate of sequences to KEGG organisms. Rarefaction curves suggested that more sequencing effort would have detected even more diversity in most samples, particularly in males. This would suggest a wealth of uncharacterised taxonomic diversity, consistent with other studies investigating marine microbiomes [69, 76], and highlights the need for better reference characterisation of marine microbial communities [34].

In spite of capturing high microbial diversity, there was no evidence of known silva-curated strains of the reproductive parasites Wolbachia, Rickettsia, Spiroplasma and Cardinium. Although some sequences were classified to relevant higher taxonomic ranks and clustered relatively closely with known silva strains, only one sequence (presumed Rickettsia) formed a monophyletic group with known strains. Even if some of the identified sequences represented novel, somewhat diverged, strains of these reproductive parasites, all sequences in question had very low abundances and would not suggest high infection intensities. These results are difficult to reconcile with prevalent sex-ratio distortion in Jaera [46–48] and pervasive Wolbachia infection in many crustaceans [49, 77]. However, these results are fully compatible with previous studies that have failed to reliably detect Wolbachia infection in total Jaera DNA extracts using targeted PCR assays [48, 49]. Ribardière et al. [48] screened 817 individuals across the Jaera albifrons species complex using 11 PCR protocols, but found little evidence of infection beyond an ephemeral novel haplotype in some Jaera albifrons and Jaera praehirsuta individuals, identified using a nested PCR protocol. As such, infection of Jaera species by Wolbachia or other bacterial sex-ratio distorting parasites cannot be ruled out, but infection intensities and prevalence appear to be very low and difficult to detect.

The biological relevance of rare sequences is difficult to assess and may well be an artefact of working with whole-body DNA extracts. Reproductive parasites primarily infect the reproductive and digestive tracts, thus it may be possible that dissection of these tissues prior to DNA extraction and sequencing improves detection [31, 78, 79]. Nevertheless, whole-body extracts should not in principle preclude detection of Wolbachia infection [29, 80], and even low infection levels should be readily detectable [81]. We thus conclude that prokaryotic reproductive parasites are unlikely to explain pervasive sex-ratio biases among the Jaera albifrons species complex, but note that the role of eukaryotic sex-ratio distorters (such as Microsporidian fungi [23]) remains uninvestigated in Jaera.

Sources of variation in microbiome composition

Microbiome composition in Jaera varied considerably between spring and summer despite similar species richness, suggesting that the microbiome in Jaera undergoes extensive temporal changes in species composition, yet remains fairly consistent in complexity. Seasonal changes in marine microbiomes are well-documented and are primarily driven by abiotic environmental factors such as temperature and biogeochemical processes [69, 82, 83]. In copepods, such seasonal changes have been reported even across a few weeks in early summer and may be linked to a rise in water temperatures and concomitant changes in temperature-sensitive marine microbial communities [35]. Any vacated ecological niches would then be taken up by different microbe species such that the overall species richness would not be greatly affected. For example, Vibrio form the core microbiome of copepods in subtropical locations, but Vibrio abundance is lower in temperate regions where a similar chitinolytic niche could be taken up by Pseudoalteromonas species [35].

The Jaera microbiome also showed regional structure in species richness and species composition across the two sampling regions in north-east Scotland. Although we sampled only females from the two southern coasts and therefore cannot rule out that at least part of the regional variance is attributable to species rather than geography, spatial effects on littoral and intertidal microbiome composition are, in fact, commonly reported at large and small scales. This is illustrated by vastly different microbiomes in Hymeniacidon heliophila sponges in subtidal and intertidal habitats at the same site [36], as well as spatial structure in microbiomes in populations of sand hoppers across the Tuscan coast [70] and benthic amphipods in the Great Lakes [84]. The regional structure in the Jaera microbiome could be due to differences in abiotic factors such as aspect (north-facing versus east-facing beaches), geology (cliff coast vs. plain coast) and human activity (proximity to harbours or sewage outlets). One particular factor could be differences in salinity, either intrinsic to the substrate or extrinsic from nearby freshwater streams. Both beaches in the south were affected by freshwater drainage from streams, whereas the beach in the north had no freshwater influx. The observed lower microbial diversity in the freshwater-affected sites contradicts the pattern found in a freshwater-marine transect in Greece [76], but is consistent with an increase in species richness and considerable changes in composition of the skin microbiome of Atlantic salmon after transitioning from freshwater to seawater [85].

Finally, overall microbiome composition showed a signature of sex, nested within the larger environmental variance components. Males tended to have more diverse microbiomes than females and the sexes also differed in the presence and abundance of a range of taxa, including a range of putative pathogens such as Aliivibrio, Aliiroseovarius, Vibrio and Tenacibaculum [86–88], and common environmental species typical of marine arthropods with no immediate functional link to reproductive processes, such as Loktanella, Glaciecola, Aureispira, Winogradskyella, Pelagibius and Marinomonas [75, 89]. An interesting finding was the high abundance of Leisingera in females in summer, alongside Vibrio and Tenacibaculum. Secondary metabolites produced by Leisingera are known to have antimicrobial effects and are used by cephalopods to protect their eggs against pathogens such as Vibrio [90]. Nevertheless, the specific functional roles of the taxa assemblage in male and female Jaera remains obscure and will require more detailed functional assays and experimentation.

Sex-specific differences in diversity and composition have been reported, for example, in whole-body microbiomes of phloem-feeding whiteflies, aphids and psyllids [29], and cloacal microbiomes of the striped plateau lizard Sceloporus virgatus [91]. In arthropods, these differences are often attributed to infection with reproductive endoparasites [29, 32]. For example, uninfected males and females of the terrestrial isopod Armadillidium vulgare have similar microbiomes, but Wolbachia-infected females carry higher total bacterial loads [30, 31]. Since we found little evidence of reproductive parasites in Jaera, alternative explanations need to be considered. One hypothesis could be that sex-specific microbiomes are functionally linked to intra-specific reproductive processes such as sex recognition or sexual selection that may have knock-on effects on reproductive isolation and speciation [12, 13, 92]. For example, reproductive isolation via pheromones is documented in allopatric populations of the marine polychaete Neanthes acuminata [93] and sympatric populations of the amphipod Eogammarus confervicolus [50]. The idea that microbiomes may be linked to host developmental processes and co-diverge tightly with host speciation events—a phenomenon termed “phylosymbiosis”—is a hotly debated topic [14–16, 38]. Although we were unable to separate species-specific signals from sex-specific signals with the present set of samples, the Jaera albifrons species complex would be an excellent study system for testing these ideas with more extensive sampling across both sexes within all species.

Conversely, instead of enhancing metabolic function or causing behavioural changes in the host, sex-specific patterns in microbiont abundance could simply be attributed to differences in host body size or behaviour that circumstantially cause differential uptake and proliferation of episymbiont communities. Male Jaera are usually smaller than females despite evidence of sexual selection for body size and size-assortative mating [40, 94], suggesting that the effect of body size on microbiome composition would be worth investigating further [95]. Male Jaera could also potentially occupy different microhabitats than females as a consequence of sexual dimorphism [96]. Although all samples were collected from underneath rocks, which always harbour mixed sex populations [46, 47], sex-specific differences in substrate microhabitat occupation cannot be ruled out but are yet to be investigated. Similarly, behavioural differences could also affect both epi- and endosymbiont communities, for example through differences in feeding rates or food preferences among sexes, which is well supported in copepods [97]. Gravid females in particular would be expected to change feeding habits or even cease feeding altogether, as is the case in the amphipod Ligia [98]. All sampled Jaera females were not gravid, but sexual receptivity or other ongoing reproductive processes may well have caused sex differences among microbiomes, particularly during the reproductive peak in summer where these differences were most pronounced [46].

Outlook and conclusions

In summary, the Jaera microbiome is highly diverse and appears to be subject to multiple spatio-temporal environmental sources of variation, which is typical of marine intertidal microbiomes. A surprising result was the weak evidence of sex-ratio distorting reproductive parasites, which suggested very low infection levels at best in spite of pervasive sex-ratio distortion. However, the finding of sex-specific patterns in overall microbiome composition warrants closer scrutiny and establishes the Jaera albifrons species complex as an intriguing study system for the effects of microbiomes on host reproductive processes.

Our study has provided a snapshot assay that highlights the vast amount of variation in microbiomes from highly dynamic and complex environments. No doubt much of the ephemeral variation that has been characterised is attributed to ephemeral epibionts that may not necessarily be linked to host metabolism. Variation in this fraction could be reduced by maintaining Jaera long-term under controlled common-garden conditions [38]. Similarly, dissecting digestive or reproductive tracts could be a worthwhile avenue for targeting more specialised symbionts, since these tissues are often dominated by few taxa in tight association with host metabolism [73, 99–101]. Notwithstanding, our study highlights that, in order to properly understand the causes and consequences of phylosymbiosis or other effects of the hologenome it is essential to characterise microbiomes in situ with an appropriate sampling design that allows for appreciation of all sources of extrinsic and intrinsic variation. As such, more descriptive studies are essential for generating hypotheses of how the hologenome may operate in complex environments beyond classic model systems or controlled laboratory environments [34].