Microbial epibiotic community of the deep-sea galatheid squat lobster Munidopsis alvisca

Life at hydrothermal vent sites is based on chemosynthetic primary producers that supply heterotrophic microorganisms with substrates and generate biomass for higher trophic levels. Often, chemoautotrophs associate with the hydrothermal vent megafauna. To investigate attached bacterial and archaeal communities on deep-sea squat lobsters, we collected ten specimens from a hydrothermal vent in the Guaymas Basin (Gulf of California). All animals were identified as Munidopsis alvisca via morphological and molecular classification, and intraspecific divergence was determined. Amplicon sequencing of microbial DNA and cDNA revealed significant differences between microbial communities on the carapaces of M. alvisca and those in ambient sea water. Major epibiotic bacterial taxa were chemoautotrophic Gammaproteobacteria, such as Thiotrichaceae and Methylococcaceae, while archaea were almost exclusively represented by sequences affiliated with Ca. Nitrosopumilus. In sea water samples, Marine Group II and III archaea and organoheterotrophic Alphaproteobacteria, Flavobacteriia and Planctomycetacia were more dominant. Based on the identified taxa, we assume that main metabolic processes, carried out by M. alvisca epibiota, include ammonia, methane and sulphide oxidation. Considering that M. alvisca could benefit from sulphide detoxification by its epibiota, and that attached microbes are supplied with a stable habitat in proximity to substrate-rich hydrothermal fluids, a mutualistic host-microbe relationship appears likely.


Integrative taxonomic identification of Munidopsis alvisca
. We identified all ten collected squat lobsters as representatives of the species Munidopsis alvisca Williams, 1988 based on the following morphological characteristics: absence of pairs of epigastric spines on the carapace, a smooth carapace with transverse rugae, rostrum without pairs of lateral spines, absence of epipods on P1-P4, cornea not depressed, long eyespine strongly produced, posteromedian margin of the sixth pleomere transverse, dactyli terminal claw of P2-P4 strongly curving 20 . The M. alvisca specimens collected at Rebecca's Roost agree well with the original description and occur locally in high densities (Fig. 1A). None of the animals showed any lesions or other morphological alterations of the carapace that would indicate shell disease (Fig. 1B, C). Out of the ten specimens selected for analysis of epibiotic microbial communities, five were identified as males and five as females. The intraspecific sequence divergence (p-distance), calculated by comparison of mitochondrial cytochrome c oxidase subunit I (COI) gene sequences of collected M. alvisca specimens, ranged between 0.15-1.07% (Fig. 2, Supplement T1). The mean interspecific sequence divergence, when compared with other Munidopsis vent species, ranged between 2.4% (with Munidopsis lauensis) and 25.0% (Munidopsis lentigo) (Fig. 2, Supplement T1).
Sequencing data of the attached microbial community on M. alvisca specimens. Using amplicon sequencing of bacterial and archaeal 16S rRNA genes and their transcripts, five male and five female M. alvisca specimens were investigated for their epibiotic microbial community and compared to the free-living community in the ambient sea water. After bioinformatic quality control, a total of 237,723 and 211,909 reads were obtained for bacterial DNA and cDNA, respectively, and 225,323 and 77,820 reads for archaeal DNA and cDNA, respectively. Transcribed RNA (cDNA) was sequenced and analysed as a measure for microbial activity. Amplicon sequence variants (ASVs) with ≥ 5 reads were considered for further analysis, yielding 2037 and 1965 bASVs for bacterial DNA and cDNA, and 364 and 111 aASVs for archaeal DNA and cDNA, respectively. Diversity of free-living and epibiotic microbial communities of Munidopsis alvisca and differences between sample types. Our dataset features several bacterial and archaeal ASVs that are highly abundant in the cDNA dataset, but are not detected in the DNA dataset. To avoid underrepresentation of rare taxa with high 16S rRNA transcript numbers, in the comparison of different sample types (female/male/environment) total ASVs (combined DNA and cDNA datasets) are shown and referred to. Both, bacterial and archaeal ASVs showed very low overlap between epibiota and sea water (Fig. 3). Most ASVs were exclusive to either sample type (male, female, sea water) except epibiotic bASVs, of which 38% were found on both sexes. The most pronounced difference between the bacterial and archaeal dataset was that the majority of bASVs (90.3%) was exclusively found on squat lobsters (Fig. 3A), while the majority of aASVs (58.3%) was only present in the sea water samples (Fig. 3B).
Further, each sample was characterised by a variety of distinct oligotypes of the same taxonomic unit, thus showing a unique "fingerprint" of detected sequences, some of which were highly active (Fig. 4). This microdiversity was extremely pronounced in several bacterial families like Rhodobacteraceae or Methylomonaceae and in the archaeal dataset, in which all samples were dominated by the genus Nitrosopumilus (Fig. 4). alvisca carapaces were significantly different from those in the free-living fraction. This was found in both, the RNA (p Archaea = 0.001 and p Bacteria = 0.007) and the DNA dataset (p Archaea = 0.006 and p Bacteria = 0.006). Additionally, in all samples the archaeal diversity was distinctly lower than the bacterial diversity. While the bacterial epibiotic diversity was higher than that of sea water samples, the opposite was found for the archaeal communities (Supplement T2).
Comparison of microbial communities: assemblages on male and female M. alvisca specimens. Between sexes, significant differences were only observed for the composition of the active archaeal communities (p cDNA = 0.015), but not for the other sample sets (present bacterial/archaeal communities; active bacterial communities). Further, differences between present (DNA-derived) and active (cDNA-derived) epibiotic communities were significant for bacteria, but not archaea (p Bacteria = 0.003, p Archaea = 0.203). Hence, the difference between active and present community was more pronounced than the difference between sexes. Non-metric multidimensional scaling (NMDS) highlights the relative similarity of bacterial epibiotic samples within the DNA and cDNA datasets and the dissimilarity between the two datasets ( Fig. 5A). In contrast, epibiotic archaeal DNA and cDNA samples form one dense cluster, which indicates a rather high degree of similarity (Fig. 5B). Four archaeal cDNA samples are displayed very distant from the main cluster, most likely due to highly abundant sequences occurring exclusively in individual samples (Fig. 5B). NMDS further emphasises distinct differences between epibiotic and sea water samples for both bacterial and archaeal datasets (Fig. 5A, B).
Major microbial taxa on M. alvisca and in ambient sea water. Bacterial community composition. Carapaces of the M. alvisca specimens as well as the sea water samples were clearly dominated by the three major bacterial phyla Proteobacteria, Planctomycetes and Bacteroidia, together accounting for 70-80% of the entire bacterial community (Fig. 6, Supplement T3-4). Their most prominent representatives are highlighted in the following, percentages refer to the share the respective taxon constitutes in a given sample set. Taxonomy was assigned against a curated SILVA 132 database 26 . In all samples, Proteobacteria was the most abundant (35-54%) phylum with the highest number of rRNA transcripts (60-95%). Gammaproteobacteria (DNA > 40%, cDNA > 75%) were dominant in M. alvisca samples, and to a lesser degree in the sea water samples (DNA > 15%, cDNA > 25%). On lower taxonomic ranks, Methylomonaceae and Methylococcaceae were the most abundant gammaproteobacterial families, accounting for almost one third of the epibiotic community in individual M. alvisca samples (Fig. 6). In contrast, these two families played little to no role in the bacterial community in sea water samples. Beggiatoacea (Gammaproteobacteria) were only present on M. alvisca (abundance < 3%) and not in sea water samples.
In sea water samples, Proteobacteria were represented mostly by Alphaproteobacteria (DNA and cDNA 24-29%). In M. alvisca samples, Alphaproteobacteria represented the second most abundant class (< 10%) and they additionally made up between 2.5 and 9.2% of total 16S rRNA gene transcripts in this sample type. The www.nature.com/scientificreports/ Alphaproteobacteria were mostly represented by the family Rhodobacteraceae that accounted for 2-3% of the DNA sequences in the epibiotic communities and for 5% in sea water. While their share of the cDNA sequences in sea water was > 5%, they constituted > 1% in only one out of ten M. alvisca samples. However, one specimen exhibited exceptionally high shares of Rhodobacteraceae (64% of rRNA transcripts), mainly attributable to the genus Paracoccus (Fig. 6). In sea water samples, unclassified Sphingomonadaceae (Alphaproteobacteria) were found in both DNA (< 10%) and cDNA (> 10%) datasets but were not detected in M. alvisca samples. Interestingly, while no Enterobacteriaceae (Gammaproteobacteria) were detected in any of the DNA samples, the family was strongly represented in epibiotic cDNA. In individual samples, bASVs affiliated with Escherichia-Shigella contributed > 70% of all 16S rRNA gene transcripts (Fig. 6). In contrast, no 16S rRNA transcripts of Escherichia-Shigella were detected in sea water samples. We consider these highly abundant bASVs as true members of the epibiotic communities since we tested our primer sequences against the complete target sequences of the Escherichia-Shigella organisms obtained from NCBI to exclude discrimination and mismatching errors during the transcription of RNA into cDNA, but sequences matched completely. Moreover, negative controls (chemicals only) implemented in extraction, transcription and amplification did not show any contamination.
In all DNA and cDNA samples, Bacteroidetes and Planctomycetes constituted the second and third most dominant bacterial phyla, respectively (Fig. 6). They were mostly represented by Bacteroidia and Planctomycetacia, with Bacteroidia being detected almost two times more abundant on M. alvisca samples than in sea water. The majority of bASVs affiliated with Bacteroidia were Flavobacteriaceae and split into eight main genera. None of them was found in all three sample types (male/female/sea water), and the genera Formosa and Lutimonas were exclusively found in samples from female M. alvisca specimens. Flavobacterial genera contributed < 1-6% of DNA sequences and either zero or < 1% of cDNA sequences, except for Mesonia in sea water samples (RNA transcripts: ≤ 8%). In epibiotic DNA samples, Planctomycetacia were mainly represented by the genus Planctomicrobium (Rubinisphaeraceae) and several genera belonging to the Pirellulaceae, whereas this family occurred in minor proportions (mostly < 0.5%) in cDNA samples. In sea water samples, ASVs matching with Pirellulaceae were well-represented in the DNA sequences but showed very low cDNA frequencies and were not further characterised on genus level.
The fourth most abundant phylum in the DNA dataset, Actinobacteria, accounted for 2-8% and 6% of the bacterial sequences from M. alvisca samples and sea water, respectively and in all cDNA samples Actinobacteria accounted for up to 7%. Actinobacteria in M. alvisca samples were mainly attributed to Acidimicrobia, and to the class Actinobacteria in sea water samples.
Sea water samples were also dominated by Nitrososphaeria, most strongly represented by Ca. Nitrosopumilus. In contrast to > 90% Nitrososphaeria in the epibiotic dataset, this class accounted for only 64-68% of the archaeal DNA sequences in sea water. Second and third most abundant in sea water were also MGII and Woesearchaeia, constituting 28-29% and 3-8%, respectively (Fig. 7). In addition, aASVs representing Marine Group III (MGIII, Thermoplasmata, Euryarchaeota) made up 5% of the free-living archaeal sequences, but they were absent from the epibiotic dataset. The cDNA dataset showed a similar composition, with Nitrososphaeria accounting for two thirds of the sequences. Woesearchaeia were less abundant in cDNA compared to DNA, while Thermoplasmata accounted for almost one third of archaeal cDNA sequences in sea water samples. otic aASVs (criteria: found in ≥ 70% of all M. alvisca samples with ≥ 0.95% abundance) and were subjected to phylogenetic analysis (Supplement F1, T6). Phylogenetic trees calculated with the most abundant ASVs and their best nBLAST hits are presented in Supplement F1 and F2. The dataset comprises of eight Nitrosopumilusrelated sequences of which aASV_8 appeared in the sea water samples as well, although at lower abundance (0.37%). Only aASVs from the DNA dataset matched the defined criteria, due to the high microdiversity within the archaeal cDNA dataset. The bacterial dataset consisted of 25 bASVs that matched the criteria stated above. They were affiliated with the groups Acidimicrobiia (1 bASV), Bacteroidia (6 bASVs), Planctomycetes (4 bASVs), Alphaproteobacteria (1 bASV) and Gammaproteobacteria (13 bASVs) (Supplement F2, T7). The latter class was dominant in all compared sample sets (DNA/cDNA, male/female). None of these bASVs was found in sea water samples. BLAST-analysis revealed high similarities between our epibiotic archaeal and bacterial ASVs and sequences from various geographic locations. The distribution of the sequences related to those from M. alvisca included samples from the Mid Atlantic Ridge, European coastal waters, the Mediterranean Sea, the East and West Pacific, as well as Arctic and Antarctic waters. The respective habitats were often influenced by hydrothermal activity, cold or methane seeps or hydrocarbon and oil spills (Supplement F1, F2). In addition, we found many similarities with bacterial and archaeal sequences originating from sponges, corals, tube worms, fish and other invertebrates from different habitats and geographic locations. Further, bASV_14 and bASV_113 shared a ≥ 98% sequence similarity with sequences from the setae of the deep-sea crab Shinkaia crosnieri.

Discussion
Despite the large distribution of Munidopsis spp. in global deep-sea environments, detailed information on their ecological role and interactions is scarce 27 . Here, we present the first comprehensive study on epibiotic microbial communities of the genus Munidopsis (Decapoda) and the first record of intraspecific genetic divergence within the species Munidopsis alvisca. Since the 1980s, M. alvisca specimens have been sighted and collected at different locations in the Guaymas Basin, including mainly hydrothermal vent and cold seep sites, but also abyssal plains 17,28-31 . However, many reports of Munidopsis spp. were often based solely on morphological identification from photographs taken in situ 32 . Considering the minute morphological differences by which some Munidopsis spp. are distinguished, the occurrence of misclassifications cannot be ruled out 29 . Consequently, sightings of M. alvisca have been questioned and revaluated over the past years 33 . Nowadays, traditional morphological  www.nature.com/scientificreports/ identification is corroborated by molecular biological methods, allowing not only an integrative classification, but also investigation of phylogenetic relationships 18 . For instance, local morphological variants can be identified, by comparing specimens from different geographic locations that genetically belong to the same species. This might even allow for the discovery of recent or ongoing speciation 32 , or the identification of cryptic species 18 . As indicated by previous, comprehensive studies on phylogenetic relationships within the genus Munidopsis 18,32,34 , adequate sample size is crucial to this endeavour. To date only a single M. alvisca COI sequence was deposited in GenBank (MZ197621), which did not enable investigation of intraspecific divergence across habitats via comparison with our samples. Hence, our records provide a basis for future phylogenetic studies on M. alvisca in relation to other Munidopsis species. The significant differences found between epibiotic and free-living microbial communities hint at an origin of epibionts other than the ambient sea water. Distinct differences between free-living and decapod-associated microbial communities have been highlighted previously 3 . An alternative source of epibionts often suggested for deep-sea hydrothermal vent crustaceans like R. exoculata, K. hirsuta or S. crosnieri is the horizontal acquisition from conspecifics 9,10,35,36 . This coincides with our findings that microbial communities found on all M. alvisca specimens were highly similar (Fig. 6). Several authors further assumed that specific epibionts are selected by the crustacean host via recognition of microbial cell surface modifications by the host immune system [37][38][39][40] . Other transmission routes include the vertical transfer from parent to offspring 16,40 , or acquisition from other locations (migration), which is facilitated in the Guaymas Basin, as different habitats are not strictly separated by biogeographic barriers 4 . We did not find significant differences between epibiotic communities on male and female specimens, which coincides with a previous report 41 . The authors compared the microbial community on 48 male and female (sex ratio 1:1) Chinese mitten crabs (Eriocheir sinensis, Decapoda). Where feasible, a combination of molecular biological studies on epibiotic community composition and observations of host behaviour might provide additional insights. Important factors, like migratory behaviour, which may differ between male and female decapods 41 , could be revealed and evaluated for impact on epibiotic community composition. Further, other environmental samples i.e. sediment or components of the chimney precipitates and plumes could be included in the comparison of free-living and host-associated microbial communities to identify the potential source of M. alvisca epibionts.
On M. alvisca carapaces we found an archaeal community that almost exclusively consists of aASVs affiliated with Ca. Nitrosopumilus (Thaumarchaeota). Members of this genus are known mainly as ammonia oxidisers 42 . Hence, we assume, that ammonia oxidation is one of the major metabolic processes carried out within the archaeal epibiotic community on M. alvisca. Frequently, Nitrosopumilus spp. have been reported in hydrothermal vent and other deep-sea environments, including sediments and hydrothermal plumes of the Guaymas Basin 15,43 . While associations of Nitrosopumilus spp. with hydrothermal vent sponges and corals are documented in the literature 42 , our study is the first to report Ca. Nitrosopumilus from a hydrothermal vent decapod. Only few previous attempts were made to study archaeal epibiotic communities on hydrothermal vent crustaceans and none of them yielded any sequences 44,45 . In the ambient sea water, we also found Ca. Nitrosopumilus in high relative abundances, however much less dominant in comparison to epibiotic communities. Hence, association with a host seems to be preferable for these microorganisms. One reason may be that the proximity of M. alvisca to the ammonia-enriched sediments and hydrothermal plume ensures direct access to substrate for chemosynthesis. Further, exposure to water mixing, dilution and resulting chemical alteration of fluid components, i.e. substrates, is prevented 14,42 . Vice versa, we assume that the increased abundance and activity of MGII and MGIII archaea in the water column may be related to the availability of sinking particulate organic matter (POM) 46 . However, little is known about the ecology of these two taxa in the deep-sea and information on metabolic capacities mainly originate from metagenomic studies 46 . As reported previously, MGII archaea show an affinity to particles and are prolific degraders of proteins, lipids and other macromolecules 47 , which corroborates our hypothesis.
Within the bacterial epibiotic community on M. alvisca carapaces, we found several taxa known for their abilities to oxidise methane (Methylococcaceae 14 , Methylomonaceae 42 ) sulphide (Marithrix (formerly Thiotrix, Thiotrichaceae) 48 ) or ammonia (Pirellulaceae 49 ). All of the latter components occur in elevated concentrations in hydrothermal plumes at Rebecca's Roost 1,15 and their removal by microbial oxidation could contribute to the survival of M. alvisca. Sulphide oxidisers have been found frequently on other hydrothermal vent crustaceans, like Alvinocaris longirostris, Rimicaris exoculata and Shinkaia crosnieri 7,9,45 . Associations with sulphide "detoxifiers" are favourable for macro-and megafauna in sulphide-rich environments 45 , as many organisms suffer at elevated sulphide concentrations due to the inhibition of respiratory processes 10 . In return, as explained for Nitrosopumilus, the epibiotic partners are provided with a stable habitat in immediate proximity to the chimney. Also Methylococcaceae are frequently reported as ecto-and endosymbionts of deep-sea decapods 3,7,9 . In addition, the growth-promoting impact of methane-oxidising gill symbionts on their host, the hydrothermal vent mussel Bathymodiolus childressi, has been determined experimentally 50 . In the framework of this study, it remains speculative whether methane oxidisers provide direct advantages for M. alvisca. However, as previous studies have identified bacterial biofilms as a relevant food source of Munidopsis spp. and M. alvisca in particular 31,51 , we assume that epibiotic microbes to contribute to the M. alvisca diet.
In our study, heterotrophic bacterial degraders of labile organic matter, peptides, lipids and other high molecular weight compounds, like Flavobacteriaceae, Rhodobacteraceae and Sphingomonadaceae, were enriched in the sea water relative to the epibiotic community. As stated above, we assume that this is due to the availability of high amounts of POM in the water column. Members of the Flavobacteriaceae are known as major components of picoplankton and many representatives are specialised on the degradation of polysaccharides 52 . Moreover, all three families harbour several members that have been described to degrade (aromatic) hydrocarbons, crude oil or other xenobiotics [53][54][55] . This is reflected in our phylogenetic analysis of the most abundant bASVs, showing high sequence similarity with clones and isolates obtained from a variety of hydrocarbon-or oil-enriched habitats, highlighting the relevance of hydrocarbon degradation at Rebecca's Roost. Extreme relative abundances of www.nature.com/scientificreports/ Paracoccus (Rhodobacteraceae) in our cDNA dataset may be due to locally increased concentrations of petroleum and hydrocarbons. Some members of the family, like Paracoccus are well-known as prolific degraders of those compounds 56 . Further, a rapidly rising dominance of Roseobacter clade members (marine Rhodobacteraceae 57 ) has been found in response to the sedimentation of oil-rich POM after the Deepwater Horizon blowout in 2010 55 . Our findings on free-living hydrothermal vent microbial communities are in accordance with previous studies 14,58 . In contrast, microbial communities in deep-sea, hydrothermal sediments share only few key members with epibiota identified in our study. While Nitrosopumilaceae and aerobic Gammaproteobacteria are still abundant in surface sediments that are influenced by sea water chemistry, their abundance rapidly decreases with depth 23 . After the depletion of oxygen in the first few millimetres to centimetres of sediment, thermophilic archaea, sulphate-reducing Deltaproteobacteria and other anaerobic taxa become dominant 2,23,59 . The benthic microbial community is hence strongly influenced by the availability of electron acceptors, as well as the increasing impact of heat and hydrothermal fluid components with depth 23,60 .
In contrast to previous studies, we found highly active bASVs affiliated with Pasteurella and Escherichia-Shigella on all M. alvisca specimens. As stated in the results section, all methodological and technical controls were free from contamination and allow the assumption that these taxa are true members of the bacterial epibiotic community. We hypothesise that they may originate from sloppy feeding, a behaviour which has been documented for other decapods 61 and which results in the release of gut bacteria from prey animals. Even though information on marine members of Pasteurella and Escherichia-Shigella is scarce, other Enterobacterales have been found before in marine sediments and waters 62,63 . Enterobacteriaceae have been reported as epibionts of shallow-water sponges 64 , as intracellular pathogens of the hydrothermal vent mussel Bathymodiolus azoricus and as obligate biofilm-associates 65 . Despite their roles in anoxic environments where they reduce sulphate or thrive on fermentation, they may play a role in recycling of organic matter in oxic environments 66,67 , such as the carapace of M. alvisca. The observed discrepancies between activity and abundance of individual ASVs may be caused by one or several factors, including cell size, growth rate or nutrient availability. Depending on those aspects, RNA content can reach > 10 000 copies per cell 68 . We assume that the highly abundant cDNA sequences represent rare taxa with high metabolic activities that escape detection in the DNA samples due to their low abundances, which is not unusual for microbial assemblages in nutrient-restricted habitats 69,70 . Other authors have found 295 bASVs associated with sub-Arctic lichen that occur only in their cDNA dataset 71 . Most of these sequences were affiliated with Proteobacteria as well. Another study reports that a single bacterial species accounting only for 0.3% of the present community in a freshwater lake, constitutes 70% of the total carbon uptake 72 . Moreover, Alpha-and Gammaproteobacteria are known for their high growth rates which also leads to accumulation of RNA inside the cells and an enrichment relative to the DNA content 73 . Pasteurella or Escherichia-Shigella spp. might have gone undetected in previous investigations of decapod epibiota due to methodological or conceptual aspects. Most existing studies on epibiota of deep-sea crustaceans are based on clone libraries, FISH and morphological observations using electron microscopy and have limited their taxonomic resolution to the class level 74 . In addition, previous screenings have often focused on specific community members such as ammonia oxidisers 43 . However, to understand distribution and ecology of deep-sea microorganisms and observe distinct differences between habitats and hosts, a high taxonomic resolution and a holistic investigation of the microbial community are crucial 75 .

Conclusion
By identification of all squat lobster specimens as Munidopsis alvisca, we demonstrated that Rebecca's Roost, as many other hydrothermal vent environments, harbours a low-diversity, high-abundance decapod megafauna. We found a low intraspecific divergence, indicating no recent or ongoing diversification of the species at this location. However, future investigations should include collection and sequencing of M. alvisca specimens from other sites and in adequate numbers, to enable identification of local genetic variants. Microbial epibiotic communities on M. alvisca carapaces were dominated by the same taxa found abundant on other deep-sea hydrothermal vent decapods. We assume that M. alvisca benefits from sulphide detoxification by their epibiota, while attached microbes are supplied with a stable habitat in proximity to substrate-rich hydrothermal fluids. Hence, a mutualistic host-microbe relation is likely. The high microbial microdiversity, especially within cDNA samples from M. alvisca carapaces, indicates a high potential for rapid adaptation to minute changes in the immediate environment of epibiotic communities. Comparisons with literature further revealed that M. alvisca epibiotic communities, as well as sea water communities at Rebecca's Roost are dominated by different taxa than benthic communities in the Guaymas Basin. Due to significant differences between attached and free-living microbial communities, we consider it unlikely that M. alvisca epibionts are acquired from ambient sea water. Possibly, epibionts are transferred between conspecifics or obtained from other environmental sources like sediment or precipitate surfaces which should be included for comparison in future studies. To trace back the origin of M. alvisca epibionts an individual study would be neccessary including also different molting and age stages as well as eggs.

Material and methods
Sample site and collection of samples. Galatheid  www.nature.com/scientificreports/ stored (4 °C overnight, − 80 °C afterwards) in RNAlater. To collect free-living microbial cells, one litre sea water was filtered through a 0.2 µm filter, which was frozen immediately. All samples, including the squat lobster specimens, were stored at − 80 °C until further use. Processing of COI sequences. Forward and reverse sequences from each specimen were assembled using Geneious v.9.1.7 (www. genei ous. com). The assembled COI sequences were aligned using MAFFT v7.309 79 and compared with sequences from other vent-associated Munidopsis species available in GenBank (Supplement T1). Sequence divergences (intra-and interspecific) were estimated as uncorrected p-distances using MEGA7. A neighbour-joining tree based on the COI sequences was constructed in MEGA7 using a p-distance substitution model, treating gaps and missing data with "pairwise deletion" and by running 1000 bootstrap replicates. Assembled COI sequences from this study were deposited in GenBank (MW591339-MW591351).

Microbial nucleic acid extraction and amplification. Nucleic acid extraction and purification.
All chemicals used during the nucleic acid extraction were RNAse-free, devices were cleaned thoroughly with RNAse AWAY before use. An extraction blank (chemicals only) was included to ensure purity of used chemicals. DNA and RNA were extracted simultaneously following a phenol-chloroform protocol (modified after 80 , cf. Supplementary Information). The extracted RNA was further purified using the RNeasy Mini Kit (Qiagen, Hilden, Germany). Residual genomic DNA was removed with the Turbo DNA-free kit (Invitrogen, Carlsbad, USA) before the purified RNA was transcribed into cDNA using the SuperScript Double-Stranded cDNA synthesis kit (Invitrogen, Carlsband, USA). All commercial kits were applied according to the manufacturer's instructions.
Amplification and purification of archaeal and bacterial DNA and cDNA for Illumina sequencing. The 16S rRNA gene of bacterial and archaeal DNA and cDNA was amplified using the primer pair of 81 for bacteria, and the primers 340f 82  Processing of microbial amplicon sequences. Based on the primers, samples were split into a bacterial and an archaeal dataset and analysed separately. Both datasets were processed in parallel with the same workflow. Initially, primer sequences were trimmed with cutadapt, discarding any sequence pairs lacking primer sequences. Up to 20% mismatches within the primer sequence were allowed to retain amplicons with single mismatches. Sequences were processed within QIIME 2 84 . Reads were trimmed to remove low quality ends and filtered according to overall read quality. Filtered reads were processed within QIIME using the dada2 plugin, www.nature.com/scientificreports/ which calculates run-specific error models and denoises sequences according to the latter. Denoised sequences were merged and checked for chimeric reads. Resulting ASVs were classified using the Naïve Bayesian classifier, implemented in the classify-sklearn plugin of QIIME 2, against a curated SILVA 132 database 26 , which was fitted to the primer region of the respective dataset. Finally, reads were exported into count tables for further statistical analysis. ASVs with less than five counts were excluded from the data. Sequencing data is deposited in ENA (European Nucleotide Archive) under the accession PRJEB46145.
Statistical analysis. Prokaryotic communities were visualised with barplots. For diversity measures, samples were rarefied to the lowest common count number before calculating richness, evenness and Shannon index using the R package vegan 85 . We used NMDS to display sample variability within each dataset. Therefore, count tables were transformed into Bray-Curtis distances and NMDS was calculated using the metaMDS function from vegan. Read data was standardised using Hellinger-transformation before inferring Bray-Curtis-Distance. Bray-Curtis distances were then tested against the sex of the host for various group combinations (DNA, RNA, Environment) by using a permutational MANOVA (function adonis, vegan package). Significant effects occurring in cases with high group dispersion were judged by inspecting their mean differences in relation to internal group variation. Beta-diversity was calculated as distance to group centroids using betadisper function from the vegan package. Significant differences between beta-diversity of groups were inferred from ANOVA. Venn-diagrams were used to display ASVs shared between the different sexes or the habitat. All analyses were conducted within the R computing environment 86 . Heatmaps displaying microdiversity within taxonomic groups were calculated using features of the pheatmap package. For display, ASVs were selected that were present in an abundance > 0.001% and (i) were present in at least one sample at a relative abundance > 1% or (ii) were present in at least 2% of samples at a relative abundance > 0.1% or (iii) were present in at least 5% of samples at any abundance level and (vi) minimum of 1000 counts per sample.
Phylogenetic analysis of amplicon sequences. The most abundant ASVs in the archaeal and bacterial datasets were subjected to phylogenetic analysis via nBLAST comparison with closely related sequences in the GenBank database. ASVs were selected for comparison according to the following criteria: The most abundant ASVs of one M. alvisca dataset (bacteria OR archaea AND DNA OR cDNA AND male OR female) that were found in at least 70% of the respective sample type (environment OR animals) and make up ≥ 0.95% of the respective community. Neighbor-joining trees were calculated using the ARB software package 87 . Trees were calculated with 1000 replicates using only sequences ≥ 1200 bp. Further sequences were added subsequently to the existing tree.

Data availability
The datasets generated and analysed in the current study are available in the following repositories: Sequences, trace files, collection data and photographs of each