Antarctic sponges from the Terra Nova Bay (Ross Sea) host a diversified bacterial community

Sponges represent important habitats for a community of associated (micro)organisms. Even if sponges dominate vast areas of the Antarctic shelves, few investigations have been performed on Antarctic sponge-associated bacteria. Using a culture-dependent approach, the composition of the bacterial communities associated with 14 Antarctic sponge species from different sites within the Terra Nova Bay (Ross Sea) area was analyzed. Overall, isolates were mainly affiliated to Gammaproteobacteria, followed by Actinobacteria and CF group of Bacteroidetes, being the genera Pseudoalteromonas, Arthrobacter and Gillisia predominant, respectively. Alphaproteobacteria and Firmicutes were less represented. Cluster analyses highlighted similarities/differences among the sponge-associated bacterial communities, also in relation to the sampling site. The gammaproteobacterial Pseudoalteromonas sp. SER45, Psychrobacter sp. SER48, and Shewanella sp. SER50, and the actinobacterial Arthrobacter sp. SER44 phylotypes occurred in association with almost all the analyzed sponge species. However, except for SER50, these phylotypes were retrieved also in seawater, indicating that they may be transient within the sponge body. The differences encountered within the bacterial communities may depend on the different sites of origin, highlighting the importance of the habitat in structuring the composition of the associated bacterial assemblages. Our data support the hypothesis of specific ecological interactions between bacteria and Porifera.

Sponges (phylum Porifera) are one of the most ancient extant multicellular animals and can provide valuable insights into the origin and early evolution of Metazoa 1 . As sessile filter feeders, they are capable of removing microorganisms (including bacteria, yeasts, microalgae) from the surrounding water by pumping many thousands of litres of water through their aquiferous system within the mesohyl matrix 2 . From an ecological perspective, marine sponges provide a protected and nutrient-rich niche where extensive interactions among the diverse microbial populations are fostered and probably inevitable [3][4][5][6] , allowing the establishment of microbial consortia (the overall associated community is structurally defined as a sponge holobiome) within the holobiont body [5][6][7] . Sponges can benefit from nutrition supply, transport of waste products and active metabolites, chemical defence against predators and biofouling, and contribution to mechanical structure [8][9][10] . Consequently, the microbial colonization of sponges often plays an important role in the development and evolution of the holobiont.
In this context, the extreme and remote Antarctic environment offers a unique opportunity to study the peculiar and often strict interactions that are established between Porifera, as well as other benthic hosts, and their symbionts 11 . To date, studies on the association between microbial communities and Antarctic sponges have only rarely performed [see for review 11,12 and mainly addressed to bacterial symbionts 6,[13][14][15][16][17][18] . Altogether, results highlighted that Antarctic sponge-associated bacterial communities might be sponge-specific 13,17 . Interestingly, the occurrence of the different bacterial populations inhabiting the sponge body may be inter-regulated by bacterium-bacterium interactions 6 , or intra-regulated by the production of N-Acyl homoserine lactones in the quorum sensing phenomenon 15 . Furthermore, the production of bioactive metabolites by bacterial symbionts has been demonstrated 18,19 and it could be responsible for the selection of symbiotic bacteria. formed the fourth cluster, characterized by a relative abundance of Actinobacteria >50%. When comparing the bacterial communities more in details using "phylotypes" as a factor, bacterial communities were grouped into two main clusters, with 20% similarity (Fig. 3b), and a strong separation of Tedania spinata, mainly due to the absence of Pseudoalteromonas sp. SER45 -which is present in all the other sponge species -and for the exclusive presence of the Frigoribacterium sp. SER10 and Nocardioides sp. SER16. The other sponge species formed four sub-clusters with similarity of 40%, among which the sponge Haliclona rudis grouped alone for the exclusive presence of Erythrobacter sp. SER34 and Mycetocola sp. SER15, and the higher relative abundance of Bizionia sp. SER20 and Citricoccus sp. SER9.
Overall comparison of the associated bacterial communities at site level. With respect to the sampling site, most isolates derived from Caletta and Thetys Bay (481 and 228 isolates, respectively), followed by Adelie Cove, Gondwana and Road Bay (93, 6 and 1 isolates, respectively). Gammaproteobacteria predominated at all sites (ranging from 45.7 to 100% of total isolates per site) (Fig. 2b), followed by Actinobacteria (ranging from 24 to 14% of total isolates per site). Alphaproteobacteria were represented by 13% of total isolates at Adelie Cove, while Bacteroidetes were represented mainly at Caletta, with 26.6% of total isolates. The few isolates obtained from Road Bay and Faraglioni (six isolates and one isolate, respectively) were all affiliated to Gammaproteobacteria. The bacterial community associated with sponges appeared different also among the sampling sites. At it is shown in Fig. 3c, using "phyla" as a factor, Thetys Bay and Adelie Cove (similar at 80%), together with Caletta and Gondwana (similar at 80%, too) grouped in a bigger cluster with 60% of similarity, while Road Bay and Faraglioni grouped in a separate cluster. A similar clustering appears when comparing the bacterial communities more in details using "phylotypes" as a factor, as well as Thetys Bay, Adelie Cove, Caletta and Gondwana formed a big cluster again with 40% similarity, while the site Faraglioni grouped alone (Fig. 3d).
The one-way ANOVA analysis showed that no difference occurred in a significantly pattern between sites, but the distribution of OTUs evidenced a significantly statistical predominance of Pseudoalteromonas sp. SER45 respect to all the other detected OTUs (p < 0.05).
The nMDS analysis was performed also by considering both sponge species and site ( Figure S2). The ANOSIM Pairwise Test, computed by setting site as factors, showed that a significant difference between the all sites didn't occur, while the SIMPER analysis underlined that the average dissimilarity occurring between them presented a cumulative value of 84.14% for the sites Caletta and Adelie Cove and a cumulative value of 83.38% for the sites  Gondwana and Adelie Cove. The cluster analysis and nMDS were computed also on the bacterial communities of the same sponge species sampled from different sites, and on bacterial isolates from different sponge species sampled from the same site, as described below.
OTU-sharing among sponge specimens belonging to the same sponge species, but collected from different sites. The comparison (when applicable) between the same sponge species sampled from different sampling sites at Terra Nova Bay is shown in Fig. 4. Hemigellius pilosus (Fig. 4a) was obtained from the sites Caletta, Gondwana and Thetys Bay, which all shared Pseudoalteromonas sp. SER45and Psychrobacter sp. SER48, both occurring at the highest relative abundances at Gondwana (52% and 35% of total isolates, respectively). Shewanella sp. SER50 and Arthrobacter sp. SER44 were further shared between specimens collected at Caletta and Thetys Bay. The sites didn't show significant differences in terms of OTU composition, even if Caletta presented two non-shared OTUs (i.e., Roseovarius sp. SER39 and Leifsonia sp. SER12), while Thetys Bay showed three non-shared OTUs (i.e., Bizionia sp. SER20, Citricoccus sp. SER9, and Microbacterium sp. SER14). It could be highlighted that the OTU Pseudoalteromonas sp. SER45 presented the statistical significantly higher relative abundance (p < 0.05), followed by Arthrobacter sp. SER44 and Psychrobacter sp. SER48, and then by Shewanella sp. SER51 and Colwellia sp. SER1 among sites. All the remaining OTUs occurred at relative abundances significantly lower.
Haliclonissa verrucosa (Fig. 4b) specimens were sampled from Caletta, Adelie Cove and Faraglioni. No significant differences among OTUs were detected in the three sites, and Colwellia sp. SER1 was the unique OTU that was shared between specimens collected from Caletta and Faraglioni. Conversely, Psychrobacter sp. SER48 and Shewanella sp. SER51 were shared between Caletta and Adelie Cove. Four non-shared OTUs were detected in this case at Caletta (i.e., Arthrobacter sp. SER44, Rhodococcus sp. SER17, Staphylococcus sp. SER33, and Marinobacter sp. SER4) and two at Adelie Cove (i.e., Sulfitobacter sp. SER42 and Tateyamaria sp. SER43). Pseudoalteromonas www.nature.com/scientificreports www.nature.com/scientificreports/ sp. SER45 confirmed its predominance in all the three sampling sites (i.e. Caletta, Adelie Cove and Road Bay) also in the case of Myxodoryx hanitschi (Fig. 4c), with a relative abundance significantly higher than the other OTUs (82%, 100%, 100%, respectively; p < 0.05). Psychrobacter sp. SER48, Polaribacter sp. SER27 and Arthrobacter sp. SER44 were the three non-shared OTUs among sponges collected from Caletta site.  OTU-sharing among different sponge species collected from the same sites. Obtained data were also analyzed (when applicable) by comparing the diversities of the bacterial communities associated with different sponge species collected from the same sites (Fig. 5).
The different sponges sampled from Caletta formed three main clusters (Bray-Curtis similarities of 20%) using phylotypes as a factor: a first cluster grouped Calyx arcuarius, Hemigellius pilosus and Haliclonissa verrucosa; the higher abundance of Pseudoalteromonas sp. SER45 probably determined the formation of a second cluster grouping Tedania charcoti, Myxodoryx hanitschi and Haliclona virens, while Haliclona dancoi grouped alone, probably due to the higher relative abundance of Gillisia sp. SER23 (Fig. 5a).
The Bray-Curtis similarities, calculated on the relative abundance of phylotypes associated with sponges collected from Thetys Bay, clustered together Lyssodendoryx nobilis, Haliclona sp., Hemigellius pilosus, with values of 60%. This was probably due to the co-occurrence of Pseudoalteromonas sp. SER45 and Psychrobacter sp. SER48. Conversely, the sponge Phorbas glaberrimus grouped alone due to the high relative abundance of Arthrobacter sp. SER44 and the exclusive presence of Planococcus sp. SER32 (Fig. 5b).
A similarity of 60% was shown by the sponges Haliclona dancoi and Hemigellius pilosus from Gondwana, which formed a single sub-cluster due to the comparable abundance values of Pseudoalteromonas sp. SER45 and Psychrobacter sp. SER48 (Fig. 5c).
Finally, sponges collected from Adelie Cove grouped by forming a subcluster including Myxodoryx hanitschi and Lyssodendoryx nobilis with the 60% of similarity, probably due to a higher relative abundance of the OTU Pseudoalteromonas sp. SER45, occurring also in Phorbas glaberrimus and Haliclonissa verrucosa, even if at lower abundances (Fig. 5d).
Sponge bacterial community vs seawater bacterial community. Sponge-associated bacterial communities were compared to that reported by Lo Giudice et al. 21 for seawater. A total of 21 phylotypes were detected in seawater and resulted absent in the sponge samples. In particular, seawater samples recorded the absolute and exclusive predominance of Psychromonas sp. COL1 (44% of total isolates), followed by Pseudoalteromonas sp. SER45(relative abundance 21%), which was shared with sponges.
Overall, when using "phyla" as a factor, seawater and sponge communities showed a similarity of 80% in a cluster including Anoxycalyx joubini, Tedania charcoti, Tedania sp., Calyx arcuarius and Haliclonissa verrucosa (Fig. 6a). A separate cluster was formed by Haliclona dancoi and Haliclona rudis, while Tedania spinata grouped alone. When comparing the bacterial communities more in details using "phylotypes" as a factor, seawater community strongly differed from those observed for sponges (Fig. 6b). This separation was due to the exclusive presence of some OTUs, and for the concomitant higher abundance of common OTUs (i.e. Sulfitobacter sp. The percentage of OTUs shared between the seawater community and at least one sponge species was about the 34% of total OTUs detected, and were more abundant than the sponge specific OTUs, as well as they together showed a relative abundance about 68%. Relative abundances of the detected OTUs and their correspondent taxonomic affiliation are shown in the heat map in Fig. 7. The picture highlighted clearly two aspects, i.e. that the most abundant OTUs among sponges and water samples were really different, and that the OTU composition among sponges was also diversified, with no groups shared among all samples. The OTUs Pseudoalteromonas sp. SER45 and Psychrobacter sp. SER48 were the only shared between all sponge samples and water ( www.nature.com/scientificreports www.nature.com/scientificreports/ of total isolates, respectively). Generally, the most abundant OTUs among sponges corresponded to CFB and Actinobacteria.

Discussion
Sponges have proven to be unique and highly selective environments to host microbial communities, often of ecological and biotechnological value 3,10,18,[22][23][24][25][26] . In polar environments the association between microorganisms and macroinvertebrates, the physiological roles and the diversity of bacteria isolated from marine sponges are an under-investigated aspect 11 . At the time of writing, only few studies deeply characterized, through advanced sequencing techniques, the whole bacterial communities associated to different sponge species inhabiting the Antarctic coastal marine environment 13,17,27 . Similarly, few data are available concerning their cultivable fraction 6,[14][15][16]18 . Although methodological biases could occur when applying a cultivable-based approach, bacterial cultivation could furnish complementary information to molecular-based methods in terms of community composition and it could be particularly useful in the biotechnology field in the search for novel active compounds.  www.nature.com/scientificreports www.nature.com/scientificreports/ Using a culture-dependent approach, we analyzed the composition of the bacterial community associated with 14 Antarctic sponge species. Overall, the bacterial community composition at phylum level (which included Proteobacteria, Bacteroidetes, Actinobacteria and Firmicutes) was similar to those previously reported by other authors 13,17,28 . Similarly, the predominance of Proteobacteria has been commonly observed within sponge-associated (analyzed by culture-dependent or culture-independent methods) bacterial communities from temperate and tropical environments [29][30][31] .
It is noteworthy that future experimentations are needed on Antarctic sponge-associated bacterial isolates to verify the functional diversity (which was not verified in this study) that emerged from obtained results, if considering the affiliation of retrieved bacteria. For example, Proteobacteria, generally involved in biogeochemical cycles, could have varied effects on sponge hosts, such as nitrogen fixation, sulfate-reduction function, production of low molecular-weight biological active compounds with antimicrobial and surface-active properties 3,32 . Proteobacteria were also found to produce enzymes at high levels for degrading protein and polysaccharides 33 . Among Proteobacteria, Gammaproteobacteria represented a large fraction of Antarctic sponge-associated bacterial isolates. This is not surprising, being r-strategists with the ability to rapidly grow on nutrient-rich media and successfully compete under heterotrophic conditions 21 . They were mainly affiliated to the genera Psychrobacter, Pseudoalteromonas and Shewanella. In particular, Pseudoalteromonas spp. were ubiquitarious in the analyzed samples. Pseudoalteromonas and Shewanella, isolated from both cold and temperate environments, are listed among the major producers of bioactive compounds and extracellular polymeric substances 11 , thus suggesting that they could play a pivotal role in the sponge ecology. Alphaproteobacteria isolated in this study were phylogenetically diversified, even if they were not numerically abundant (resulting 4.75% of total isolates). This is in contrast with previous investigation reporting the dominance of this taxon in association with marine invertebrates, but such result could be related to the utilization of a culture medium not suitable for their growth 34 . Among Alphaproteobacteria, members of the Roseobacter clade would seem to play an important role in the sulfur cycle, and the Roseovarius genus would be a producer of biofilms and secondary metabolites. Through the production of biofilm, they allow the adhesion to the surfaces, facilitating the colonization and acting on the ecological competition by preventing the establishment of other microorganisms 15,35 . Other alphaproteobacterial genera detected in this study, i.e. Sulfitobacter and Octadecabacter, were previously isolated from sponges 36 and from ice and seawater, respectively 37,38 .
Following Proteobacteria, in this study Actinobacteria represented the second more abundant phylogenetic group. Actinobacterial members are well known producers of biologically active secondary metabolites, such as antibiotics or other therapeutic compounds, as well as vitamins or enzymes, that could be involved in the structuring of the bacterial community associated with Antarctic sponges 6,[39][40][41][42] . It is probable that the sponges can derive different benefits from this symbiotic relationship as nutriment, degradation of substances that otherwise would not be able to dispose of and that would accumulate inside their bodies causing toxic effects on the host, as well as defense and protection thanks to the microbial production of secondary metabolites. Undoubtedly, a greater understanding of the diversity and distribution of Actinobacteria associated with sponges could contribute to the comprehension of their ecological role, in order to improve their biotechnological potential 43 . Actinobacterial www.nature.com/scientificreports www.nature.com/scientificreports/ isolates were affiliated to genera that were previously detected in marine or cold environments, e.g. Arthrobacter (which predominated in this study), Microbacterium, Rhodococcus, Leifsonia and Citricoccus 21,44-46 . Contrary to Zhang et al. 46 , who reported the genus Streptomyces as the most abundant actinobacterium isolated from sponges from temperate environments, such genus was not found among Antarctic sponge associated isolates, suggesting that environmental conditions of the sampling sites did not support its growth.
Bacteroidetes were also well represented within the Antarctic sponge-associated bacterial community. They play a key role in the carbon cycle, as they represent a group specialized in the degradation of high molecular weight compounds of the dissolved organic matter pool in the sea 47 . Among Bacteroidetes, the genus Gillisia, which is involved in the process of remineralization of organic matter in the ocean and is an efficient producer of secondary metabolites 48 , was particularly abundant.
The first evidence of the occurrence of Firmicutes in association with Antarctic sponges was reported by Mangano et al. 6 and Papaleo et al. 18 (data have been included in the present study). However, they remain scarcely represented and affiliated to the genera Oceanobacillus, Planococcus and Staphylococcus, all belong to non-marine bacterial groups. Firmicutes represent a fraction of the isolated microbial community also in sponges of other environments such as the Great Barrier Reef and South China Sea, which most abundant genus was represented by Bacillus, which has been shown to possess an efficient antibacterial activity 49 . Their involvement in biogeochemical cycles and in several degradative processes is well known 50 , therefore sponges are likely to use their metabolic products as an energy source 51 .
Even if we analyzed the sole cultivable fraction of the sponge-associated bacterial communities, a cross comparison between the different Antarctic sponge species (from the same or different sites in the Terra Nova Bay) put on evidence a possible host-specificity of associated bacteria and/or the possible influence of the sampling site. The cluster analysis showed some differences/similarities in the bacterial community composition, as previously observed by other authors 29,30 . Differences/similarities were detected mostly in terms of abundance and, in some cases, in terms of presence/absence of phylotypes. The gammaproteobacterial Pseudoalteromonas sp. SER45, Psychrobacter sp. SER48, and Shewanella sp. SER50, and the actinobacterial Arthrobacter sp. SER44 occurred in association of almost all the analyzed sponge species. However, except for Shewanella sp. SER50, the OTUs mentioned above were retrieved also in seawater 21 , indicating that they may be transient within the sponge body.
Accordingly to Cleary et al. 30 , the differences encountered within the bacterial communities may depend on the different sites of origin, highlighting the importance of the habitat in structuring the composition of the associated bacterial assemblages. The bacterial community in sponges collected from the site Caletta resulted the most abundant and diversified, and included members of great part of phyla detected, especially in the case of the sponges Haliclona dancoi and Calyx arcuarius. No bacteria isolated from Adelie Cove sponges were affiliated to Bacteroidetes, whereas Road Bay and Faraglioni specimens hosted only Gammaproteobacterial isolates at a very low percentage. However, the evident separation of the sites Road Bay and Faraglioni probably is a reflection of their scarce representativeness at level of both phyla and phylotypes. A similar bacterial community composition was observed for Tedania charcoti, Myxodoryx hanitschi and Haliclona virens from Caletta, Lyssodendoryx nobilis, Haliclona sp. and Hemigellius pilosus from Thetys Bay, Haliclona dancoi and Hemigellius pilosus from Gondwana, and Myxodoryx hanitschi and Lissodendoryx nobilis from Adelie Cove, suggesting that the site of origin could be a determinant factor in shaping the sponge associated bacterial community. At this regard, such similarity was mainly driven by the presence of Pseudoalteromonas SER45 and/or Psychrobacter SER48, which were also abundant in seawater. It could be noted, however, that a number of phylotypes (as reported in Table 2) were found only in association with a few sponge species (and often they were not present in seawater), suggesting that a close relationship between the host and the microorganisms could exist. Among them, the OTU Shewanella sp. SER50 presented a total abundance of 90% in the sponges, and was retrieved in eight sponges out of fourteen. While some of these phylotypes are considered quite common in Antarctic sponges, as such as Psychromonas and Pseudoalteromonas 11 , some others have been only recently reported as sponge associated genera, as it is the case of Brevibacterium 52 . The existence of specific microbial communities in sponges from different environments with deep differences from the surrounding waters have been previously evidenced by several authors 9,13,22,53 . Here the presence of many phylotypes detected only in sponges confirms the great potential of Antarctic sponges as diversity reservoirs, as just observed by Rodríguez-Marconi et al. 17 .
Seawater 21 and sponge bacterial communities at Terra Nova Bay shared 21 phylotypes (out of 62), but the cluster analysis highlighted a clear separation between the two bacterial communities, which both included exclusive phylotypes or phylotypes that were particularly abundant. Our results are in line with those reported by Webster et al. 13 and Rodríguez-Marconi et al. 17 for McMurdo Sound and Fildes Bay, respectively, applying molecular approaches on the whole bacterial community. Taylor et al. 3 suggested two main possible theories underlying the maintenance of the symbiotic relationship between bacteria and sponges, according to which microorganisms coming from the surrounding water during filtration could establish symbiosis with the host, or by vertical transmission with the reproductive processes. While in the first case a partial overlap between sponge and water bacterial community is expected, in the second one the co-evolutionary processes that would result would go to support the sponge specificity, so that a separation between sponges and water community composition occurs 29 . Our data would seem to suggest the selection by the host organism of certain taxonomic groups, supporting the hypothesis of specific ecological interactions between microorganisms and Porifera 6,15 . In order to perform a critical analysis of the bacterial community structure in terms of host-specificity, we attempted to classified each OTU as part of the core microbiome (i.e. a set of OTUs, absent in seawater, that were shared by at least 6 sponges), variable community member (i.e. a set of OTUs, absent in seawater, but shared by 2 to 5 sponges), or as part of the host-specific community (i.e. occurring only in a single sponge species, but absent in seawater). Based on the classification above, none of the OTUs detected in this study were part of the core microbiome as the most abundant (i.e. Pseudoalteromonas sp. SER45 and Arthrobacter sp. SER44) were retrieved also in seawater samples. Shewanella sp. SER50 presented the highest degree of specificity as it was shared by eight sponge species and Scientific RepoRtS | (2019) 9:16135 | https://doi.org/10.1038/s41598-019-52491-0 www.nature.com/scientificreports www.nature.com/scientificreports/ was absent in seawater. The strong relation of this phylotype with Antarctic sponges could be correlated with its proven ecological role in the relationships with Porifera hosts. Several isolates affiliated to Shewanella spp. have been reported as involved in the production of extracellular polymeric substances, important regulators of bacterial adhesion processes, heavy metal resistance, emulsifying activities and cryoprotection 16,54 .

Concluding Remarks
Although the Antarctic harsh environmental conditions make sponges a particularly attractive model for the study of symbiosis, the association between Antarctic invertebrates (not only Porifera) and microorganisms remains a very underexplored topic. Our results on the cultivable fraction of sponge-associated Antarctic bacteria suggest that, despite the ubiquitous presence of a number of phylotypes (both in sponges and seawater), some sponges could select associated bacteria. To confirm this assumption, a better characterization of the composition of the whole prokaryotic community associated with selected sponge species from the Terra Nova Bay is in progress. Even if cultivation-dependent methods have several well-known limiting factors, cultivable bacteria may represent a valuable tool in the elucidation of the main processes at the basis of sponge-bacteria interactions (e.g., bacterial adhesion, biosynthesis of molecules involved in sponge-bacteria communication and sponge bacterial selection). Further, their biotechnological value should not be underestimated, as they may represent an untapped source of novel active and useful biomolecules.

Materials and Methods
Sample collection. Specimens (3 to 5) of 14 Antarctic sponge species (Table 1) (Fig. 1). Sampling depths ranged between 25 and 200 m. Sponge specimen collection was authorized by the PRNA project. Sponge specimens were treated as previously described by Mangano et al. 6 . Briefly, organisms were immediately washed at least three times with filter-sterilized natural seawater to remove transient and loosely attached bacteria and/or debris. Specimens were then placed into individual sterile plastic bags containing filter-sterilized natural seawater and transported directly to the laboratory at 4 °C for microbiological processing (within 2 h after sampling). A fragment of each specimen was also preserved in 70% ethanol for taxonomic identification.
Bacterial isolation and phylogenetic identification. Bacterial isolation from sponge was carried out as previous described by Mangano et al. 6 . Briefly, a central core of the sponge body was cut using an EtOH sterilized corkborer or a sterile scalpel. The sponge fragment was then aseptically weighted and manually homogenized in 0.22 µm filtered seawater in a sterile mortar. Sponge extracts were serially diluted using filter-sterilized seawater. Aliquots (100 µl) of each dilution were spread in triplicate on Marine Agar 2216 (MA, Difco). Plates were incubated in the dark at 4 °C for 1 month under aerobic conditions. Bacterial colonies grown on MA were randomly isolated and streaked at least three times before being considered pure. Cultures were routinely incubated at 4 °C. All the bacterial strains isolated from the sponges were included in the Italian Collection of Antarctic Bacteria (CIBAN) of the National Antarctic Museum (MNA) "Felice Ippolito", and kept at the University of Messina (Italy).
16S rRNA gene PCR amplification of bacterial isolates. PCR-amplification of 16S rDNA from bacterial isolates was carried out as described previously by Michaud et al. 55 . Briefly, a single colony of each strain was picked-up with a sterile toothpick from an MA plate, re-suspended in 20 µL of sterile distilled water and lysed by heating at 95 °C for 10 min. Cell lysates were rapidly cooled in ice and then subjected to a brief centrifugation before amplification.
The PCR program was set as follows: (1)  Amplified rDNA restriction analysis. Sponge-associated bacterial isolates were grouped by the Amplified rDNA Restriction Analysis (ARDRA) 56 . Each amplicon (5 µl), containing approximately 1.5 µg of amplified 16S rDNA, were digested with 3 U of the restriction enzyme AluI (Fermentas, Italy) in a total volume of 20 µl at 37 °C for 3 h. The enzyme was inactivated by heating at 65 °C for 15 min and the reaction products were analyzed by agarose (2.5%, w/v) gel electrophoresis (at 90 mV for 90 min) in TAE buffer containing 1 µg/ml of ethidium bromide 21,55 . (2019) 9:16135 | https://doi.org/10.1038/s41598-019-52491-0 www.nature.com/scientificreports www.nature.com/scientificreports/ A GeneRuler ™ 100 bp DNA Ladder (Fermentas, Italy) was applied to each gel as a band reference. On the basis of the restriction patterns obtained (and visually compared one to each other), Antarctic sponge associated isolates were grouped into Operational Taxonomic Units (OTUs), assuming that one OTU was made up of strains belonging to the same species. Isolates showing identical ARDRA patterns were also checked for colony morphology on agar plates 21 .
16S rRNA gene sequencing. For each OTU, one to three representative strains (where possible) were selected for sequencing by using the primer 27F. The amplicons were purified using QIAquick PCR purification KIT (Qiagen, Italy) and the subsequent sequencing was carried out at the Macrogen Laboratory (The Netherlands). Next relatives of the bacterial isolates were determined by comparison to 16S rRNA gene sequences in the NCBI GenBank and the EMBL databases using BLAST 20 . Sequences were further aligned using the program Clustal W 57 to the most similar orthologous sequences retrieved from the database. Each alignment was checked manually and corrected. A phylogenetic tree was constructed using the MEGA X (Molecular Evolutionary Genetics Analysis) software 58 . The tree was performed using the Maximum composite Likelihood with Tajima-Nei Model (with Rate Variation and Pattern Heterogeneity) model and the Neighbour-Joining algorithm. Robustness of the inferred trees was evaluated by 1000 bootstrap re-samplings.
Nucleotide sequence accession numbers. Nucleotide sequences have been deposited in the GenBank database under the Accession Nos MK660285-MK660324.
Data analyses. All statistical analyses were performed using PRIMER v6 for Windows (PRIMER-E Ltd, Plymouth, UK). Data were analyzed for eventual differences/similarities between the bacterial communities, using as a factor (1) the sponge species (without taking into consideration the eventual different sampling sites) or (2) the sampling sites (in the case of sponge specimens belonging to the same species, but collected from different sites).
Relative abundances of bacterial phyla and OTUs were opportunely transformed, and then used to calculate pairwise similarities among samples using the Bray-Curtis similarity coefficient. Bray-Curtis similarity matrices were used to perform cluster analysis and nMDS analysis of the bacterial communities from the different sponges. Analysis of Similarity (ANOSIM) was calculated to test the significance of differences among different sites using PRIMER v6 for Windows (PRIMER-E Ltd, Plymouth, UK).
Heatmaps were constructed to show the OTU distribution and clustering by using Heatplus version 2.24.0 59 and Gplots packages version 3.0.1 60 in R environment version 3.4.4 61,62 , using square root transformed data of relative abundance of the detected OTUs. Hierarchical clustering was generated with group average method.
The sponge-associated bacterial community composition was further compared with that previously reported by Lo Giudice et al. 21 for seawater collected from the Terra Nova Bay area during the same Antarctic Expedition, and treated exactly as sponge samples for bacterial isolation. Briefly, the seawater bacterial community included 606 bacterial strains, mainly belonging to Gammaproteobacteria (68.5%; with Psychromonas and Pseudoalteromonas as predominant genera), followed by Actinobacteria (16.6%; main genera: Microbacterium and Arthrobacter), Alphaproteobacteria (9.4%; main genera: Octadecabacter and Sphingomonas), Bacteroidetes (5.8%; main genera: Gillisia and Bizionia) and Firmicutes (0.8%; main genus: Oceanobacillus). Each sequence from representative bacterial isolates from seawater was pairly aligned by BLAST 20 to those obtained in this study to check for similarity. Isolates showing a similarity ≥97% where grouped in the same OTU/phylotype.
In order to evaluate the sharing or specificity level of each OTU, a threshold of presence was established and used to classify them as variable, host-specific and core sponge microbiome.

Data availability
Raw data are available upon request.