One of the most important roles of sponges in the marine ecosystem is that they function as ecosystem engineers or "living hotels”, providing micro-habitat to diverse organisms (e.g. Koukouras et al. 1985, Ribeiro et al. 2003, Padua et al. 2013). This ecological role of sponges has gained increasing research interest over the last few decades for both scientific and conservation purposes. Sponge-associated fauna has been studied in a variety of habitats, ranging from shallow rocky beds in temperate seas (Koukouras et al. 1985, 1992, 1996) to tropical coral reefs (Villamizar and Lauchlin 1991), seagrass meadows (Çinar et al. 2002), polar regions (Amsler et al. 2009) and deep-sea ecosystems (Klitgaard 1991).

Although sponges constitute the dominant sessile organisms in marine cave environments (Gerovasileiou and Voultsiadou 2012, 2016), their functional role as ecosystem engineers in this habitat type has received little attention. Recently, Navarro-Barranco et al. (2016) studied the crustaceans associated with invertebrate hosts, among which the sponge Ircinia variabilis, in a marine cave in the Alboran Sea and found that all the examined species were equally used as refuges by crustaceans.

Sessile assemblages of marine caves are characterized by a decrease in diversity, biomass and three-dimensional complexity towards inner dark cave sectors, driven by the extreme oligotrophy due to the reduction of light and the increasing water confinement (Harmelin et al. 1985, Bianchi and Morri 1994). On the other hand, sessile invertebrates have been shown to increase the structural complexity of hard substrates (Voultsiadou et al. 2010, Navarro-Barranco et al. 2014, 2016). This complexity in the space-limited oligotrophic environment of the dark caves could be particularly significant for the motile cave dwellers. However, the motile invertebrate diversity of Mediterranean marine caves has received limited interest compared with the sessile biota; the few relevant studies have mainly taken place in the northwestern Mediterranean Sea (Ledoyer 1965, Navarro-Barranco et al. 2014, 2016), while such data are lacking from the eastern basin (but see Gerovasileiou et al. 2015b).

From the trophic point of view, marine caves are considered simple systems in which organic supply depends on the circulation of water entering the cave, given the absence of primary production due to lack of light (Bianchi et al. 2003). The trophic structure of Mediterranean caves is insufficiently studied, concerning only the feeding habits of polychaetes (Bianchi 1985) and amphipods (Navarro-Barranco et al. 2013). Equally understudied is the trophic structure of benthic invertebrate-associated assemblages, except for that of the anthozoans Cladocora caespitosa (Chintiroglou 1996) and Astroides calycularis (Terrón-Siglera et al. 2016). Taking into account the above, it would be interesting to investigate the role of sponges for the cave environment from a trophic point of view.

The aim of this study was (a) to contribute to the knowledge of motile macrobenthos in caves of the eastern Mediterranean Sea and (b) to examine the hypothesis that different sponge species, previously found to host rich macrofaunal assemblages in shallow rocky beds of the same biogeographic area, maintain their role as ecosystem engineers or "living hotels” in the cave environment. To this end, we examined the taxonomic and trophic structure of the sponge-associated assemblages in two eastern Mediterranean marine caves.

A total of 3037 macrofaunal specimens were collected from the canals and cavities of the 15 examined sponges. The identification of sponge symbionts yielded 86 taxa belonging to 8 major groups (Table 1). The sponge A. aerophoba (Agios Vasilios cave) showed higher abundance of macrofaunal symbionts (2587 specimens) than A. oroides (Fara cave) (450 specimens). However, almost equal numbers of taxa were found in both sponge species (57 taxa in A. aerophoba and 56 taxa in A. oroides), with 27 taxa being present in both sponges (Table 2). Crustaceans, mainly Amphipoda and Isopoda, dominated in terms of abundance in both sponges (91% of symbionts in A. aerophoba and 65% in A. oroides), while polychaetes prevailed regarding the number of taxa (51% and 65% of symbiotic taxa respectively) (Fig. 2).

A positive correlation was found between the sponge volume and the number of symbiotic taxa for both sponges (Table 3), but it was significant only for A. oroides. A significant positive correlation was found between the sponge volume on the one hand, and the abundance and number of symbiotic taxa on the other, when all 15 specimens of the two sponge species were examined together. The cumulative number of symbiotic taxa did not approach an asymptotic value when plotted against cumulative sponge volume, for either of the two sponges (Fig. 3).

Resemblance analysis showed clear differentiation in the assemblage structure of the associated fauna between the two sponge species, with a dissimilarity of 75% (Fig. 4). Specimens of A. aerophoba from different zones of Agios Vasilios cave showed similarity higher than 56% among each other, while those of A. oroides from Fara cave showed a similarity of only 35.5%. Results of one-way PERMANOVA revealed significant difference in the composition of the associated assemblages of the two sponge species (Table 4). When each sponge was examined separately, no significant differences were found in their associated assemblage structure in the different cave zones. SIMPER analysis between the associated fauna of the two sponge species showed that nine macrobenthic species (i.e. Janira maculosa, Colomastix pusilla, Gnathia vorax, Leptochelia savignyi, Boccardia polybranchia, Eualus occultus, Liljeborgia dellavallei, Leucothoe spinicarpa and Pseudopotamilla reniformis) contributed 55% to the Bray-Curtis dissimilarity index.

The specimens of A. aerophoba examined across Agios Vasilios cave showed higher density and number of symbiotic taxa than those of A. oroides across Fara cave; the opposite trend was found for species diversity (H’) and evenness (J’) (Fig. 5). The total number of symbiotic taxa decreased from the entrance to the interior of the two caves for both sponge species (Table 2); samples of A. oroides from the dark interior of Fara cave showed half the number of associated taxa (18) compared with those from the entrance (37). The mean abundance of polychaetes associated with the sponge A. oroides increased towards the dark zone of Fara cave, with a parallel decrease of crustaceans (Fig. 6). The mean number of associated taxa decreased (Fig. 5A) towards the cave interior for both species, while the mean density of symbionts remained stable (Fig. 5B). Values of species diversity (Fig. 5C) and evenness (Fig. 5D) increased towards the cave interior for the species A. oroides, while the opposite trend was found for A. aerophoba. However, results of perANOVA showed that none of the above changes were statistically significant.

The trophic characterization of the sponge-associated polychaetes and crustaceans revealed five feeding groups: carnivores (19 taxa), deposit-feeders or detritivores (18), omnivores (17), filter-feeders (11) and one non-feeder (see Table 1). Deposit-feeders (mainly polychaetes) were the species-richest feeding group (33%) in the sponge A. oroides, while carnivores (basically crustaceans) were the most abundant (48%) (Fig. 7). On the other hand, in A. aerophoba carnivores dominated in terms of both number of taxa (35%) and abundance (55%). The isopod G. vorax, which in the adult, sponge-associated, stage is non-feeding, was found only in A. oroides.

The trophic composition, in terms of number of taxa, of the sponge-associated polychaetes and crustaceans did not change significantly across the distinct cave zones for both sponge species (Fig. 8). The mean density of carnivores decreased towards the dark cave zone in the sponge A. oroides, while that of filter-feeders and non-feeders increased (Fig. 9). However, none of these changes was statistically significant, with the exception of the non-feeding isopod G. vorax, which showed the highest density in the semidark zone (Pseudo-F=2.5543, p<0.05, df=8). No changes were found for the symbiont density of any feeding group in the sponge A. aerophoba.

Although sponges dominate in marine caves, their role as ecosystem engineers is being studied here for the first time. Our results revealed that the sponge species A. aerophoba and A. oroides in the two studied caves supported a considerable associated macrofauna. Twelve macrofaunal species found in this study are reported for the first time as sponge symbionts, 22 as symbionts of A. oroides, and 17 as symbionts of A. aerophoba (Table 1). Most of these species (around 70%) have been previously recorded in Mediterranean marine caves (Gerovasileiou et al. 2015b). Characteristic examples include the polychaete Vermiliopsis monodiscus, which is considered a typical species of cave communities in the northwestern Mediterranean (Zibrowius 1968), and the Liechtenstein’s goby, Corcyrogobius liechtensteini, which finds shelter in shadowy micro-habitats within caves (Gerovasileiou et al. 2015a). Both species are reported here for the first time as sponge symbionts.

The dominance of crustaceans (mostly amphipods and isopods), followed by polychaetes, in terms of abundance in both sponge species, is in accordance with previous studies (Pansini 1970, Koukouras et al. 1992, 1996, Ribeiro et al. 2003). The numeric dominance of amphipods in sponges could be related to their small size providing ease in movement along the sponge canals (Koukouras et al. 1985, Voultsiadou-Koukoura et al. 1987). Koukouras et al. (1985) found that crustaceans (mainly amphipods) dominated in A. aerophoba, but that the infaunal assemblage of A. oroides was equally represented by crustaceans and polychaetes.

Mean density of symbionts of A. aerophoba was 5-6 times higher than that of A. oroides, as previously found for other habitat types in the same biogeographic area (Koukouras et al. 1992, 1996). Koukouras et al. (1996) estimated that A. aerophoba has higher complexity (in terms of canal volume and mean canal diameter) than A. oroides, to which they attributed the higher abundance and species number of its symbiotic assemblage. In our case, despite the fact that density and number of taxa showed higher values in A. aerophoba, Shannon-Wiener and species evenness indices had higher values in A. oroides, as found by Koukouras et al. (1992). This is probably related to the dominance of two peracarids, Colomastix pusilla and Janira maculosa, in A. aerophoba. The convergence of our results regarding quantitative taxonomic composition of the sponge-associated fauna with those of studies conducted four decades ago at different sites, depths and habitat types of the same geographic area, the North Aegean Sea (Koukouras et al. 1985, Voultsiadou-Koukoura et al. 1987, Koukouras et al. 1992, 1996), verifies that sponges support well-structured macrofaunal assemblages, consisting of species with particular ecological habits.

A notable portion of the macrofauna found in this study (organisms identified to the species level) has been recorded before in the sponges A. oroides and A. aerophoba (45% and 67% of the species, respectively). Typical sponge-endobionts include the isopod J. maculosa, the amphipods C. pusilla and Leucothoe spinicarpa and the bivalve Hiatella arctica; the ophiurid Ophiothrix fragilis has also been commonly reported as a sponge-epibiont (Koukouras et al. 1985). The high abundance of some of these species has been attributed either to parasitism (e.g. the amphipod L. spinicarpa according to Connes 1967) or to specific stages of their life cycle, such as recruitment (Turon et al. 2000). Sponges have been reported to function as "reproductive centres” or "resting places” for larvae and adults of several macrobenthic species (Klitgaard 1991, Turon et al. 2000). This was also evidenced in our study but only for sponge specimens collected near the cave entrance; juveniles of the amphipod C. pusilla and egg-carrying individuals of the tanaid Paradoxapseudes intermedius were found in the cavities of A. oroides, while egg-carrying individuals of the decapods Athanas nitescens and Eualus occultus were found in the sponge A. aerophoba.

The analysis of the trophic structure of the sponge-associated polychaete and crustacean assemblages showed an almost even distribution of taxa among feeding types. The slight prevalence of deposit-feeders in A. oroides and carnivores in A. aerophoba highlights the differences between the two sponge species, as discussed above. Concerning abundance, the dominance of carnivores in both sponges, which is mainly attributed to the amphipod C. pusilla, is in accordance with the findings of Navarro-Barranco et al. (2013) on amphipod assemblages in cave sediments. The absence of herbivores is attributed to the lack of primary producers in the cave environment (Harmelin et al. 1985). Non-feeders were represented only by the isopod G. vorax, which has a biphasic life cycle, with the adults inhabiting cryptic environments like sponges (Tanaka 2007). Although the trophic structure of the entire sponge-associated assemblage did not change between cave zones, non-feeders showed an increase in density inwards; this might be due to the fact that they are not affected by the decrease in trophic input towards the cave interior but only by the availability of suitable habitat (i.e. the sponge-dominated semidark cave zone). The feeding preferences of a given species may vary with environment, as shown by Terrón-Siglera et al. (2016), who found that typical carnivorous amphipods mainly fed on detritus inside the colonies of the coral Astroides calycularis. On the other hand, Navarro-Barranco et al. (2013) found no significant differences between caves and nearby open habitats in the gut content of sediment amphipods. A thorough study of the feeding habits of typical sponge associates could enhance understanding of the trophic web in sponge-dominated habitats such as marine caves.

The results of multivariate analyses in this study highlighted differences in the structure of macrofaunal assemblages of the examined sponges. This high dissimilarity was mainly attributed to several species, which could be divided into the following categories: (a) species with higher abundance in A. aerophoba (e.g. the crustaceans J. maculosa, C. pusilla, Leptochelia savignyi and E. occultus); (b) species with higher abundance in the sponge A. oroides (e.g. the polychaetes Boccardia polybranchia and Pseudopotamilla reniformis); (c) species found only in A. oroides (e.g. the isopod G. vorax); and (d) species found only in A. aerophoba (e.g. the amphipods Liljeborgia dellavallei and Leucothoe spinicarpa). Previous studies on the associated fauna of these two sponges in sublittoral rocky reefs showed that the Bray-Curtis similarity index between the two sponges was over 50%, as half of the symbionts were commonly found in both sponges (Koukouras et al. 1992, 1996). In this study, the number of common symbionts was slightly lower (47-48%).

The considerable number of newly recorded macrofaunal species in the examined sponges (30% of A. aerophoba and 39% of A. oroides symbionts) indicates the potential influence of the marine cave environment on the associated assemblage structure. Interestingly for both sponges, the majority (59%) of the newly recorded symbionts were found exclusively in one cave. Thus, dissimilarity between the associated assemblages of the two sponges could be partially attributed to the fact that they were collected in different caves with different geomorphological and ecological characteristics (Gerovasileiou et al. 2013, Gerovasileiou and Voultsiadou 2016). Marine caves are commonly acknowledged for their "idiosyncratic behaviour” regarding the prevailing environmental features and their sessile and motile fauna (Martí et al. 2004, Bussotti et al. 2006, Navarro-Barranco et al. 2013). The influence of the surrounding environment on the composition of sponge-associated assemblages has been observed for temperate sublittoral rocky beds and Posidonia meadows (Koukouras et al. 1996, Gherardi et al. 2001). In support of this, none of the macrofaunal species found associated with the sponges A. aerophoba and A. oroides in the two caves can be considered an obligate sponge symbiont: they are either sciaphilic species finding shelter in cryptic micro-habitats, including caves, or euryoecious species dwelling in a wide range of marine habitats (Ledoyer 1965, Koukouras et al. 1985, 1996).

The fact that the cumulative number of symbiotic species did not approach an asymptotic value when plotted against cumulative sponge volume for either of the two sponges possibly shows that further sponge sampling in the studied caves would increase the number of associated species. The positive relationship of sponge volume with number of taxa and abundance of associated fauna has been previously shown for A. aerophoba and A. oroides (Koukouras et al. 1992), as well as for other sponge species (Gherardi et al. 2001, Çinar et al. 2002).

The decrease in the total number of sponge-associated taxa towards the inner dark zone of Fara cave seems to follow the general diversity patterns described for several Mediterranean caves (Balduzzi et al. 1989, Martí et al. 2004, Gerovasileiou and Voultsiadou 2016). This is most likely due to the fact that environmental features, such as algal coverage and hydrodynamic regime, seem to affect diversity of sponge-associated fauna, as has been shown for open environments (Voultsiadou-Koukoura et al. 1987, Çinar et al. 2002). In accordance with this finding, Navarro-Barranco et al. (2014) suggested that the absence of macroalgae along with the increasing oligotrophy were responsible for a significant reduction in the diversity of motile epifauna from the open rocky substrates close to the cave mouth towards the inner semidark cave sectors. However, in our case, the mean density and diversity, as expressed by the examined indices, did not vary significantly along the horizontal axis of the surveyed caves, implying that sponges provide quite stable micro-habitat for the associated organisms. This is further supported by the observation that the qualitative and quantitative trophic structure of the sponge-associated fauna did not significantly change between cave zones. These findings become particularly important in the inner impoverished parts of the caves, where micro-habitats are generally lacking due to the low biotic coverage and the local dominance of few sessile invertebrates with encrusting growth forms (Gerovasileiou and Voultsiadou 2016 and references therein). The presence of habitat-forming organisms, such as the massive/tubular sponges, becomes critical for motile diversity by adding three-dimensional complexity to the assemblages and providing a habitat for a variety of species, as Navarro-Barranco et al. (2016) have shown for crustaceans in semidark caves. The scarcity or even absence of benthic invertebrates with a massive or dendritic form, other than Porifera, in eastern Mediterranean marine caves (Gerovasileiou et al. 2015b) could be indicative of their significance for these habitats. Furthermore, this might have interesting implications for the energy-transfer by motile fauna, as shown for mysids (Coma et al. 1997), in dark cave zones, which have been characterized as "extreme” oligotrophic environments (Harmelin et al. 1985).

In conclusion, the results of this study support the initial hypothesis that sponges maintain their functional role as ecosystem engineers across the studied marine caves, as shown by the study of both the taxonomic and trophic structure of their associated assemblages. However, in order to better understand the engineering function of sponges in marine caves, further study is required in different caves and areas, including comparisons of their associated macrofauna with the cave-dwelling motile macrofaunal assemblages. Such research, besides shedding light on the role of sponges, might also reveal interesting findings about the functioning of the peculiar cave environment itself.