Benthic invertebrates and algae were collected during four Antarctic cruises: two in the Eastern Weddell Sea on board the RV Polarstern (ANT XV/3 and ANT XXI/2 cruises); a third one on board the RV BIO Hespérides around the South Shetland Islands (ECOQUIM-2 cruise); and the fourth at Deception Island (ACTIQUIM-1 campaign). Sampling took place at a total of 24 stations between 0 and 1524 m depth by using bottom and Agassiz trawls (AGTs) and an epibenthic sledge (ES) and, at shallow sites, by scuba diving (SD). At collection, all invertebrate and macroalgal samples exhibited a healthy, fresh appearance; none of them showed signs of decomposition. Specimens belonging to the same species and from the same collecting station were grouped as a single sample, representing the mean population of each site. Fresh organisms were then photographed on deck, and a representative portion of each sample was fixed in 10% formalin solution for further taxonomical identification at the University of Barcelona. The rest of the material was subsequently frozen at −20°C and shipped back to our laboratory, where it was preserved until processed.

Invertebrate samples with large enough volumes to allow dissection were divided into different body parts in order to determine the potential chemical defences present in these different parts. Samples were dissected as follows: sponges were divided into internal/external and apical/basal regions; pennatulacean cnidarians into poliparium/axe/peduncule; and ascidians into external/internal parts. All whole and dissected invertebrate samples and algae were then chemically extracted with acetone using a pestle at room temperature. The resulting extract was fractioned in diethyl ether and afterwards in butanol, always evaporating the solvent in vacuo using a rotary evaporator. All steps were repeated three times, except for the butanol fractionation, which was done only once. At the end, we obtained a dry ether and a dry butanol fraction and an aqueous residue. The dry ether and butanol fractions were weighted and divided into the total dry weight of the corresponding sample, as follows: DW (total dry weight)=EE (ether extract)+BE (butanol extract)+DR (dry extracted residue). This yielded extracts at the natural sample concentration: N=EE/DW (Table 1). See Núñez-Pons et al. (2012) for detailed information regarding sample collection and extraction yields.

Individuals of the Antarctic asteroid Odontaster validus, measuring between 6 and 10.5 cm diameter, were collected at different sites in Port Foster Bay, Deception Island, South Shetland Islands, Antarctica (62° 59.369' S, 60° 33.424' W), for feeding-repellence assays during the ACTIQUIM-1 (December–February 2008/09) and ACTIQUIM-2 (January 2010) cruises by SD from 3 to 15 m depth. Once testing was over, the sea stars were returned to the sea.

Odontaster validus is a predator commonly used in feeding acceptability studies (for review see Avila et al. 2008). This asteroid is readily available, and its feeding response lends itself quite well to laboratory bioassays. The collected sea stars were kept in large tanks of seawater at Spain's Gabriel de Castilla Station on Deception Island and were deprived of food for 5 days. The methodology followed in the experiments is detailed elsewhere (Avila et al. 2000; Iken et al. 2002). In brief, the assays consisted of 10 replicates, each with a 2.5 L container filled with seawater and one sea star, which was presented with a shrimp feeding cube sufficiently small (5×5×5 mm; 13.09±3.43 mg of dry mass) to be fully consumed by the asteroid. These tiny shrimp food items contained 12.4% protein, 9.1% carbohydrates and 1.5% lipids, and 17.8 kJ g⁻¹ in dry wt and 4.1 kJ g⁻¹ wet wt, according to the manufacturer's nutritional information and the Atwater factor system (Atwater & Benedict 1902). They were loaded either with lipophilic extracts from Antarctic invertebrates and algae, applied at their respective natural concentration in the treatment test, or with solvent carrier alone (diethyl ether) in the control tests. In both cases, the solvent was left to totally evaporate under a flow hood. The factor to normalize tissue concentrations of each fraction (hereafter referred to as natural concentration) was calculated on a dry-weight basis employing the total dry weight of each sample, as previously mentioned. Dry weight was chosen, rather than volume or wet weight, because it eliminates the water content, which may entail notable deviations in aquatic porous samples (Table 1). When the sea stars everted the cardiac stomach and bolted down whole food cubes, we could measure the “defence per shrimp cube.” After 24 h, the food items that had been eaten were noted, and feeding repellence was evaluated by applying Fisher's exact tests for each assay referred to the control run simultaneously (Sokal & Rohlf 1995; Fig. 1). Uneaten extract-treated shrimp pieces were preserved frozen for further extraction and analysis by thin layer chromatography to check for possible alterations in the extracts. No changes were observed. Because EEs are not hydrophilic and the water temperature was fairly cold (ca. −1°C), little, if any, loss to the water column was expected.

The assays with the sea star Odontaster validus demonstrated a widespread presence of lipid-soluble deterrents within cnidarians and ascidians. The opposite pattern is shown in sponges and algae. The low deterrence we found in sponges conflicts with other published findings. Past studies, however, usually tested demosponge species, which are prolific sources of bioactive natural products and are presumably far better defended than the hexactinellids, which represent most of our Porifera (McClintock & Baker 2001; Avila et al. 2008; McClintock et al. 2010; Taboada et al. 2012).

When comparing the current study with parallel work testing the same fractions against the omnivore lyssianasid amphipod Cheirimedon femoratus (Núñez-Pons et al. 2012), we observed that out of the 49 lipophilic extracts assessed in both bioassays, 22 (44.9%) were deterrent and five (10.2%) were suitable for both predators. Therefore, 55.1% (27 fractions) elicited the same response in the two predators. Twenty-two fractions had different results for the two species, 19 (38.8%) being discouraging for the amphipod but palatable for O. validus, and the remaining three fractions were repellent just for the sea star. Overall, there was a 32.7% higher incidence of deterrency towards amphipod feeding than against sea star ingestion, especially in the groups of sponges and algae (Figs. 1, 2).

Even if O. validus and C. femoratus have both circumpolar-eurybathic distributions, and extensive generalist diets (scavenger, detritivore, planktivore; Bregazzi 1972; McClintock 1994; De Broyer et al. 2007), the distinct ecological habits of these predators may promote variable defensive responses in potential prey. In general, sea stars are mobile macropredators that start extra-oral digestion from the surface of the victims (Sloan 1980). Amphipods instead feed with minute peripheral bites, and only prolonged feeding would allow access to internal tissues. However, in prey organisms with evident openings, these small crustaceans may reach the inner regions to forage (Peters et al. 2009). We compare the results of the present study with those from Núñez-Pons et al. (2012), estimating prey protective adaptations against two different predators.

The experiments with Odontaster validus were no-choice feeding repellence tests, whereas those performed with Cheirimedon femoratus by Núñez-Pons et al. (2012) were preference assays with-choice. When allowed to choose between extract-treated versus palatable control food pearls, predators may be better able to discriminate repellent agents. Another difference in the two studies is that the diets were distinct in energetic content: alginate pearls used with the amphipod C. femoratus were less nutritious (19 kJ g⁻¹ dry wt and 1.5 kJ g⁻¹ wet wt) than the shrimp cubes presented to O. validus (17.8 kJ g⁻¹ dry wt and 4.1 kJ g⁻¹ wet wt; by Atwater factors; Atwater & Benedict 1902). Therefore, some repellent extracts might have been less evident in the sea star tests due to a higher nutritional value of shrimp compared to alginate pearls (and potentially compared to some of the samples too). Despite these modifications, we are still able to make comparative approximations related to deterrence potential since nutritious foods can interact with defensive metabolites and constrain the types or concentrations that might be efficacious in nature for discouraging predators. An idea of the potency of deterrents can be drawn by measuring their activity in the two kinds of experiments using assay foods of distinct energy content (Duffy & Paul 1992; Cruz-Rivera & Hay 2003). Most of the fractions deterrent to amphipods but accepted by sea stars came from less nutritious samples: hexactinellid sponges; and fibrous body parts: tunics of S. adareanum (B&W) and the stalk from U. antarctica. In these cases, moderate to poor chemical defence along with low nutritional value may operate in tandem to make these body parts less appealing. In general, the optimal assay foods are those energetically equivalent to the studied prey; however, O. validus would not feed on gelified diets (authors’ pers. obs.). This study tested only ether fractions, and possible defences exclusive to other fractions were not analysed.

Lipophilic deterrence in algae was shown to be low towards sea stars and high towards amphipods. This could be due, in part, to a higher ingestion of algal material by C. femoratus ovigerous females and juveniles during summer (Bregazzi 1972). Hydrophilic deterrents (i.e., phlorotanins), often contributing to asteroid repellence in algae, could have been missed in our analysis (Amsler et al. 2005).

Antarctic benthic amphipods generally associate with living substrata with no obligate relationships (De Broyer et al. 2001), obtaining tri-dimensional habitat and food as well as chemical refuges from prospective enemies, like fish, when hosts are chemically defended (McClintock et al. 2009; Zamzow et al. 2010). Cheirimedon femoratus may be found as deep as 1500 m (Krapp et al. 2008). At Deception Island, it is both abundant and commonly found on macrophytes and poriferans (authors’ pers. obs.), similar to other northern Cheirimedon amphipods (Vader 1984). Such a niche is filled in some Antarctic areas by several other species (Huang et al. 2007; Amsler et al. 2009). Organism associations adapt in diverse interactions depending on the chemical potential of the host and the feeding adaptations of the grazers. As part of its opportunistic habits, C. femoratus may graze on host tissues directly (authors’ pers. obs.), while also profiting from adhered detritus and diatoms (Bregazzi 1972; Graeve et al. 2001). This leads to a more constant localized pressure on host-and-prey organisms than that exerted by wandering macropredators (Hay et al. 1987; Toth et al. 2007), like O. validus, which forages on ubiquitous prey with less recurrent encounters (McClintock 1987, 1994).

Our algae are comprised of Antarctic brown and red seaweed, which are rich sources of food (11–13 kJ g⁻¹ dry wt; Montgomery & Gerking 1980; Gomez & Westermeier 1995) and bioactive molecules. Except for one species, all sponges were hexactinellids with no reported secondary metabolites (Blunt et al. 2013 and previous reports). Their high spicule content and low energetic value (5–6 kJ g⁻¹ dry wt; McClintock 1987; Barthel 1995) are believed to discourage predators. Average ether-lipophilic yields showed low values in hexactinellids (ca. 1.3%) and in Rodophyta (0.6%), and higher ones (ca. 4.4%) in brown algae (Table 1). The small ether fraction yield in red algae may reflect their small lipid content (Montgomery & Gerking 1980). Antarctic poriferans and macroalgae hold high diversities of amphipods (Kunzmann 1996; Huang et al. 2007; Amsler et al. 2009). Moreover, some algae act as a chemical refuges, including Desmarestia menziesii with lipophilic deterrents and D. anceps with hydrophilic defences; this might have been missed in our study, which only tested ether partitions. Our inactive extract of D. menziesii from deeper waters (ca. 110 m) may have different chemistry due to the variation in depths or the extraction protocol, which yielded unexpected inactivity. Other algae, like Palmaria decipiens, are more preferred as food (though not for our amphipod) but are not preferred as hosts (Amsler et al. 2005; Aumack et al. 2010; Bucolo et al. 2011). Analogous chemical refuges have not been described in Antarctic sponges, yet some defended species host dense amphipod populations (Amsler et al. 2009). Most of the seaweed and sponge fractions reflected no activity against O. validus, but were rejected by C. femoratus, which may use them as substrata and occasionally as prey. The few fractions that repelled both asteroids and amphipods were those from the macroalgae Ascoseira mirabilis and Georgiella confluens, the demosponge Isodictya toxophila and the hexactinellid Rossella nuda. Rossella nuda is abundant and primarily foraged by spongivorous generalists, including O. validus. The richer fractions and repellent chemistry of I. toxophila supports the greater bioactivity of demosponges compared to glass sponges (McClintock 1987; Barthel 1995). Extracts that showed palatability in this study and also towards C. femoratus in our previous study were those from the algae Adenocystis utricularis and Desmarestia antarctica (second-year) epiphyted by Geminocarpus austrogeorgiae (Table 1, Fig. 1). Amsler et al. (2005) proposed that in second-year D. antarctica mild acidity could assist weak hydrophilic repellents against sea stars, while in first-year algae the lipophilic extract was already effective, suggesting a loss of lipophilic defences with senescence. Glass sponges are readily preyed upon by Odontaster and Acanthaster sea stars (McClintock 1987), and as it turns out, no lipophilic repellents or external defence allocation were shown to assist these sponges in asteroid avoidance (Furrow et al. 2003). Instead, efficient defences towards amphipods were found distributed throughout the volcano-shaped rossellids. Opportunistic amphipods, through lax associations, may influence the chemical ecology of macroalgae and sponges and change the expectations of the ODT (Rhoades & Gates 1976), replacing, in some cases, asteroids as the main inducers of defence distribution.

Antarctic Cnidaria and Ascidiacea are rich in bioactive metabolites (Blunt et al. 2013 and previous reports) and feeding repellents (Koplovitz et al. 2009; Núñez-Pons et al. 2010). Our cnidarian samples comprised hydrozoans and anthozoans whereas the ascidians included colonial species. Soft corals and colonial ascidians are highly energetic (16 and 15 kJ g⁻¹ dry wt, respectively; Slattery & McClintock 1995; McClintock et al. 2004), while hydroids and gorgonians are also quite nutritious (Coma et al. 1998). Ether fraction yields may illustrate this outcome: ascidians ca. 4.9%; soft corals ca. 3.3%; gorgonians ca. 1.2% (Table 1). Antarctic cnidarians and ascidians represent transitory biosubstrata for amphipods in shelf communities not dominated by sponges or algae (Dayton et al. 1974; De Broyer et al. 2001). In the present study, most samples were chemically defended, but other tactics might co-occur in these groups.

Hydrozoans use lipophilic as well as nematocyst-based defences even if considered redundant by the OT (Herms & Mattson 1992; Stachowicz & Lindquist 2000; Lindquist 2002). Staurotheca antarctica yielded a rich fraction deterrent to amphipods but inactive against sea stars. Nematocysts may protect against asteroids, which seem particularly vulnerable when stinging happens on sensitive ambulacral feet or cardiac stomach mucosa, and Syntheciidae hydroids are generally armoured with penetrating cnidos, often injecting polar proteinaceous venoms (Shostak 1995; Ostman 2000). Thus, S. antarctica, in using repellent lipids against amphipods and nematocyst against asteroids, might contravene the OT. Nonetheless, polar fractions may contain other defences not tested here. The hydroid Symplectoscyphus glacialis (Family Sertulariidae) rarely presents penetrating nematocysts (Shostak 1995), and its extract was significantly deterrent to both predators. Regarding Anthozoa, Umbellula antarctica is a pennatulacean with a fibrous axis and a distal crown of giant (3–4 cm long) polyps (Pasternak 1962; Dolan 2008). Both regions revealed distinct bioactivities and thin layer chromatography profiles, suggesting that lipid repellents were present only in the axis. The chemically unprotected and exposed polyparium is presumed to either rely on hydrophilic repellents and/or on nematocysts for defence, which actively participate in capturing prey (Dolan 2008). Autozooid nematocysts in the pennatulid sea pansy, Renilla kollekeri, keep sea star predators away in a similar way (Kastendiek 1976). The fibrous stalk of U. antarctica, denuded of nematocysts, did possess deterrents for amphipods. Since alternative defences seem to occur in different body regions, OT and ODT could both be sustained for U. antarctica. Soft corals and gorgonians generally lack stinging nematocysts and likely rely on chemical defence (Sammarco & Coll 1992), as reflected in our results. Only one fraction from Primnoisis antarctica sample 1 was accepted by both consumers, and one extract from Thouarella laxa was only suitable for the star. Both fraction yields were less abundant than active conspecific extracts (Table 1; Fig. 1).

Many colonial ascidians protect their common external tunic with alternative tactics other than organic deterrents. For example, the Antarctic species Distaplia cylindrica and D. colligans combine lipidic defences and inorganic chemistry (acid, heavy metals; McClintock et al. 2004; Koplovitz et al. 2009). None of the species studied here showed a similar kind of defence (Koplovitz et al. 2009; Lebar et al. 2011). Moreover, tunics may be nutritiously unattractive (McClintock et al. 1991; Pisut & Pawlik 2002). Among our Synoicum adareanum samples, as is common in this species (Varela 2007), there were three colour morphs; black and white (B&W), brown (Br), and orange (O) morphotypes. The dissected O sample exhibited no defence allocation. In contrast, the internal extracts of both B&W samples caused strong rejection by O. validus and C. femoratus, while the external colonial tunic fractions were accepted by sea stars. B&W morph tunics were thick and tough and yielded poorer lipophilic fractions than their inner regions (Table 1, Fig. 1). To ward off predators, the B&W samples rely on lipophilic distasteful metabolites in the inner tissues. Smaller quantities of these deterrents present in the external tunic may work together with low nutritional value to discourage external asteroid attack. The participation of polar products, again, cannot be ruled out. Finally, the rich apical fraction from a B&W sample containing siphon mouths and common cloaca, was rejected only by the sea star. The allocation of deterrent activities prevalent in internal regions of B&W S. adareanum samples might provide support for the ODT. Colonial ascidians invest much energy for reproduction and tend to produce complex, chemically-defended larvae (Lambert 2005). External damages due to eventual predation attacks rarely kill colonies but instead cause reversible injuries on the outer protective common tunic and death of some clonal zooids. Hence, storing repellents in inner tissues (gonads), which, in this case, are more vital for species survival than the outer colonial tunic, is consistent with the ODT (Rhoades & Gates 1976).

In less represented taxa, the abundant extract from the hemichordate pterobranch displayed no repellency in any assay. These animals may not need chemical protection since they live hidden from major enemies inside the coenoecia, colonial encasements hardened with agglutinated foreign material (Ridewood 1911; Brusca & Brusca 2003). Again, the existence of polar deterrents cannot be discarded. In contrast, the bryozoan and holothurian fractions reflected the presence of feeding deterrents against O. validus. The calcified bryozoan Isoschizoporella secunda afforded poor extract yields. It harbours sessile “trap-door” avicularia that are proposed to function as active mechanical deterrents to zooid-level predators (amphipods, nematodes, polychaetes) and as secretion sites for bioactive compounds (Carter et al. 2010). Hence, this branched bryozoan could avoid amphipod attacks through entrapments of appendages by avicularia and rely on chemistry against asteroids (Bryan et al. 1998). Finally, the elpidiid holothurian Peniagone vignioni, with an ability to swim (Wigham et al. 2008), may escape from many bottom-dwelling consumers. Thus, repellents against sea stars would only be useful while the animal is feeding on the substrate surface (Table 1; Fig. 1).