Marine organisms are a prolific source of natural products, many of which have become the target of pharmacological investigations. As well as being bioactive, these compounds can play an important role in the survival of the organisms and potentially affect community structure. Sessile invertebrates such as sponges, cnidarians, bryozoans and tunicates, in particular, are an especially rich source of ecologically relevant compounds. These organisms are typically soft-bodied and unable to escape predators, and as such, they rely on chemical defence for their persistence. These invertebrates are also frequently hosts for microbial symbionts.
Marine microbes are a prolific source of bioactive natural products, many of which can be allelopathic and prevent the growth of pathogens. For sessile marine invertebrates, which often have both bioactive natural products and microbial symbionts, it is logical to hypothesize that microbial symbionts may produce the secondary metabolites.
While symbionts are often thought to be responsible for producing bioactive natural products that play a role in host defence, relatively few studies have experimentally demonstrated that symbiont-produced compounds defend the hosts. One reason for this is the difficulty of manipulating the host–symbiont relationship, which is often obligate for one or both partners. Given the importance of natural products to marine invertebrates for defence, the prevalence of symbiosis in the marine environment and the diverse metabolic capabilities of micro-organisms, symbiont-mediated host chemical defence may be more prevalent than currently understood.
In this review, I document evidence supporting chemical defensive symbioses in marine organisms, discuss commonalities and differences among the diverse relationships, and provide future research directions. Epibiont-produced defensive compounds would seem to result in the most efficient manner of protection, as the compounds are most accessible to the predators. By contrast, endosymbiont-produced compounds may need to be transported to exposed tissues. Elucidating mechanisms to transport symbiont-produced defensive compounds within the host will provide greater insight into the breadth and complexity of host–symbiont interactions. Another important unexplored issue is how the hosts are able to tolerate symbiont-produced metabolites that are often eucaryotic cell effectors. A greater understanding of defensive symbiosis will result from detailed studies of the co-evolution of predator, host and symbiont, and if or how predators can influence the production of defensive metabolites.
The marine environment has provided some of the most ecologically important and well-studied examples of mutualism between eucaryotic hosts and microbial symbionts. For instance, the interaction between the bioluminescent bacterium, Vibrio fischeri, and the bobtail squid, Euprymna scolopes, has become a model system for investigating the establishment of relationships between a host and an environmentally acquired symbiont [reviewed in (Visick & McFall-Ngai 2000; Nyholm & McFall-Ngai 2004)], as well as the role of microbes in invertebrate development (Montgomery & McFall-Ngai 1995). The relationship between anthozoans and symbiotic dinoflagellates, also known as zooxanthellae, serves as a model for host–symbiont nutrient exchange [reviewed in (Yellowlees, Rees & Leggat 2008)]. While many symbiotic interactions are studied in the context of nutrition, there are a few examples from terrestrial environments of symbionts protecting their host from predators via deterrent natural products. For example, fungal endophytic symbionts in grasses produce toxic alkaloids that reduce feeding on the hosts by herbivores such as insects and mammals (Cheplick & Clay 1988; Clay 1988, 2001). Larvae of the rove beetle, Paederus spp., are protected from predatory spiders by pederin, a toxic polyketide molecule produced by a γ-proteobacterial symbiont in the host (Kellner & Dettner 1996; Kellner 2001, 2002). These systems illustrate that the basis for mutualistic symbiosis can be protection rather than nutrition.
In the 1970s, as researchers were able to access more remote marine habitats, invertebrates, plants and algae became the focus of natural product chemists searching for novel compounds with potent bioactivity. The sessile nature of invertebrates such as sponges, cnidarians and ascidians may have favoured the evolution of structurally diverse natural products. Many bioactive compounds shown to have potential pharmaceutical activity also have ecological activity such as predator deterrence or pathogen inhibition [reviewed in (Paul et al. 2007; Hay 2009; Paul, Ritson-Williams & Sharp 2011)]. In addition, constant exposure of filter-feeding invertebrates to micro-organisms in sea water likely facilitates the establishment of symbiotic relationships between animal hosts and beneficial microbes. In the 1990s, it was recognized that some natural products isolated from marine invertebrates had structural similarities to compounds produced by cultured, free-living terrestrial microbes. While many symbiotic microbes are recalcitrant to cultivation, innovations in techniques such as confocal microscopy and molecular biology facilitated visualization and identification of potential symbionts. Advances in understanding microbial natural product biosynthetic gene clusters and enzymes meant that natural product biosynthetic genes could be predicted for some compounds and used to confirm biosynthesis by a particular symbiont (Piel 2002; Hildebrand et al. 2004a; Schmidt et al. 2005; Sudek et al. 2007). This approach is bound to increase as genome sequencing becomes more accessible and routine.
Despite these technological advances, there are still relatively few examples of symbiont-mediated chemical defence in marine organisms. To confirm that a relationship is a chemical defensive symbiosis, a natural product must be shown to be produced by a microbial symbiont and to defend the host from natural enemies through comparisons of symbiont-cured and control hosts (Fig. 1 illustrates current and future research approaches to characterize chemical defensive symbiosis using the relationship between the bryozoan Bugula neritina and its uncultured symbiont as an example, see below). This requires collaboration between chemical ecologists, natural products chemists and organismal biologists or microbiologists. Beyond interdisciplinary collaboration, this research often requires fieldwork in remote locations to collect samples for natural product characterization and bioassay-guided fractionation. Sophisticated equipment for compound structural elucidation, symbiont visualization and localization, and microbial genome or natural product biosynthetic gene cluster sequencing may also be required. Furthermore, there are often technical difficulties associated with separating the symbiont from the host to assess the source of the compounds, to identify possible symbiont biosynthetic genes, or to sequence symbiont genomes. Curing hosts of the symbiont, but not other associated microbes, or a specific natural-product-producing symbiont if there are more than one, may also be difficult or impossible if the relationship is obligate. In this case, establishing the deterrence of symbiont-produced natural products is the next best evidence for defensive symbiosis. Finally, analogous to Koch's final postulate, reintroduction of the symbiont to the cured host and subsequent restoration of the defence would confirm the role of the symbiont. Most marine symbionts are uncultured and therefore difficult to reintroduce into a cured host. Further, the symbiont may have undergone genomic deterioration (Moran, McCutcheon & Nakabachi 2008) and lost its capability to infect the host. All of these hurdles make identifying and documenting defensive symbiosis challenging.
The first studies to unequivocally demonstrate a symbiont-produced natural product protected a marine invertebrate host involved the protection of crustacean embryos from a pathogenic fungus (Gil-Turnes, Hay & Fenical 1989; Gil-Turnes & Fenical 1992). Experiments demonstrated that a bacterium coating embryos of the estuarine shrimp, Palaemon macrodactylus and the lobster, Homarus americanus, produces small compounds that inhibit the growth of the fungal pathogen, Lagenidium callinectes. Removal of the epibiotic symbiont by antibiotic treatment resulted in embryos with reduced survival rates when exposed to L. callinectes, compared to control embryos. Furthermore, the survival rate of embryos recolonized with the symbiont after the antibiotic treatment increased. Finally, the survival rate of aposymbiotic embryos incubated with both the pathogen and the purified natural products increased, confirming these compounds as protective agents. While instances of allelopathy between microbes are important and can lead to defence against pathogens, cases of a symbiont producing a natural product that defends the host against predators represents a multitrophic interaction suggestive of a long co-evolutionary history among all participants. Hundreds of novel marine natural products are discovered every year (Blunt et al. 2010, 2011, 2012), and many of these compounds are integral to the chemical defence of the source organism (Hay 2009; Paul, Ritson-Williams & Sharp 2011), so discovery of novel chemical defensive symbioses is likely to increase in the future. Several detailed studies of these associations reviewed below help provide insights into host–symbiont interactions and how microbes can influence relationships between eucaryotic organisms.
The temperate marine bryozoan Bugula neritina (Fig. 2a,b), has been of interest to researchers since its crude extract was shown to have potent anticancer activity (Pettit et al. 1970). In 1982, the structure of bryostatin 1 was elucidated (Fig. 2c; Pettit et al. 1982), and to date, 20 different bryostatins have been characterized from different populations of the bryozoan (Pettit 1996; Lopanik, Gustafson & Lindquist 2004). In the mid-90s, researchers found that B. neritina larvae were distasteful to a wide array of predators (Lindquist 1996; Lindquist & Hay 1996; Tamburri & Zimmer-Faust 1996). As the larvae lacked obvious morphological defences, and the crude extract of larvae was unpalatable, it was hypothesized that the larvae were chemically defended. Furthermore, experiments demonstrated that larvae tasted and subsequently rejected by fish and anemones had a very high rate of settlement and metamorphosis, suggesting that this chemical defence was very potent (Lindquist 1996; Lindquist & Hay 1996). Using bioassay-guided fractionation, several of the bryostatins (bryostatin 10 and 20, Fig. 2d,e), which coat the outer surface of larvae (Lopanik, Lindquist & Targett 2004; Sharp, Davidson & Haygood 2007), were shown to defend larvae from predators (Lopanik, Gustafson & Lindquist 2004; Lopanik, Lindquist & Targett 2004). Further experiments demonstrated that solitary larvae have much greater concentrations of bryostatins than adult colonies, which appear to be protected by increased levels of structural material (Lopanik, Targett & Lindquist 2006b) and, since they are colonial and can regenerate, refuge in size [i.e. removal of a portion of the colony by a predator is not necessarily catastrophic (Jackson 1985)].
Microscopy studies initially revealed rod-shaped bacteria in both adult (Woollacott & Zimmer 1975) and larval (Woollacott 1981) B. neritina that were later characterized as a γ-Proteobacterium based on its 16S rRNA sequence and named ‘Candidatus Endobugula sertula’ (Haygood & Davidson 1997). Furthermore, this bacterium was the only one consistently associated with B. neritina larvae, indicating a persistent relationship between the two organisms. Haygood et al. also demonstrated that ‘Cand. E. sertula’ was the likely source of the bryostatins, as antibiotic-cured colonies had a 50% reduction in bryostatin levels and a 95% reduction in symbiont quantity (Davidson et al. 2001). Next-generation larvae from antibiotic-cured colonies had a 98·6% reduction in bryostatins levels, almost undetectable levels of ‘Cand. E. sertula’, and did not significantly deter feeding by a predator when compared to control larvae with symbionts (Lopanik, Lindquist & Targett 2004), clearly demonstrating that microbial symbionts can defend their hosts from predators via the production of deterrent natural products, illustrating a tritrophic interaction.
While increased levels of bryostatins defend larvae, and adult colonies rely on physical defences and refuge in size, the relative importance of bryostatins in the protection of juvenile B. neritina was unknown. Early, postsettlement life-history stages of marine invertebrates can be particularly susceptible to abiotic and biotic factors, such as UV radiation, temperature, desiccation, water motion, predation, competition, energy depletion and disease (Gosselin & Qian 1997). Mortality can be high once larvae have settled and metamorphosed (Gosselin & Qian 1997; Hunt & Scheibling 1997) and can therefore affect the population structure of the organism (Keough & Downes 1982; Osman, Whitlatch & Malatesta 1992; Osman & Whitlatch 1995). Feeding assays revealed that postsettlement juvenile extracts (from 4 h postsettlement to 7-day-old juveniles) were deterrent to Lagodon rhomboides, although palatability increased, and bryostatin levels decreased with increasing age (Lopanik, Targett & Lindquist 2006b). Further, Haygood et al. colocalized bryostatins and ‘Cand. E. sertula’ in the very early developmental stages of B. neritina (larvae to first zooid with next zooid bud, 96 h; Sharp, Davidson & Haygood 2007). Juveniles lose their bryostatin coating while metamorphosing from the larval stage into the first zooid, before the chitinous outer layer develops. Bryostatins were detected when the next zooid bud was developing (96-h postsettlement), presumably to protect the growing tissue. These data support the finding that structural material in the form of chitin and mineral material increase from the early stages of B. neritina development to the adult colony (Lopanik, Targett & Lindquist 2006b), suggesting a shift from chemical defence in the early stages of B. neritina, to, at least in part, physical defence in the older stages. Differing levels of deterrent bryostatins in different life-history stages of B. neritina implies that there may be coordination between the host and symbiont resulting in accumulation of bryostatins on the vulnerable larval stage. The pattern of brooding portions of the colony being protected by defensive compounds is evocative of the optimal defence theory of plant/herbivore interactions [i.e. a defence is allocated within an organism in proportion to the predation risk; reviewed in (Stamp 2003)], but it is unclear as to how the natural products become concentrated on different portions of the colony. For bryozoans, transport of bryostatins and/or symbionts through the funicular cords, which feed the developing embryo with nutrients, is likely. ‘Cand. E. sertula’ in funicular cords that connect the developing embryo to the maternal zooid appear to produce bryostatins that protect the released larva (Sharp, Davidson & Haygood 2007). The mechanism, however, resulting in varied bryostatin concentrations on the different life stages is unclear. Do symbionts found in portions of the colony not brooding larvae contribute to bryostatin production? Does the symbiont detect signals from the host at the onset of embryogenesis and commence bryostatin biosynthesis? Does symbiont titre increase in the ovicells, resulting in quorum-regulated bryostatin biosynthesis? Does the host regulate symbiont populations within the colony? Identifying molecular interactions between the two partners may answer some of these questions, but will be challenging as very little genomic information is available on these organisms (and most other organisms in this review).
Bugula neritina inhabits temperate marine waters around the world. Investigations into the population structure have defined three distinct sibling species of B. neritina-’Cand. E. sertula’. Two sibling species co-occur in southern California (CA, USA). Colonies from deep waters (>9 m) have bryostatins with different substituents (longer unsaturated acyl chains) on C20 such as bryostatin 1 (Fig. 2c), although the bryostatin core remains similar (Davidson & Haygood 1999). Genetically, the B. neritina mitochondrial cytochrome c oxidase subunit I (COI) diverges by ~8% in the two populations, and the ‘Cand. E. sertula’ 16S rRNA gene diverges by ~0·4%. McGovern & Hellberg (2003) showed that colonies collected from the Gulf of Mexico were genetically similar to those in shallow waters in NC and CA, and that colonies collected from Delaware comprised another sibling species (11·5% difference in COI). Interestingly, the northern colonies did not possess a candidate symbiotic bacterium. Although extracts of northern larvae were significantly unpalatable to the pinfish L. rhomboides (Lopanik, Lindquist & Targett 2004), no bryostatins were detected by HPLC (Lopanik, unpubl. data), suggesting the presence of endogenous defensive metabolites. COI sequence analysis of B. neritina colonies revealed that the shallow sibling species is widely distributed throughout the world (Mackie, Keough & Christidis 2006). As B. neritina larvae settle quickly out of the water column, widespread distribution by larval dispersal is not likely. B. neritina is often found among fouling communities on ships and boats, and anthropogenic dispersal is likely responsible for the widespread distribution of the shallow sibling species. Although B. neritina has a cosmopolitan distribution and encounters different predators, it is unknown if predator/prey co-evolution results in differing bryostatin compositions among the populations with similar genotypes. Minor variations in bryostatin structure elicited quantitatively different responses from pinfish [i.e. bryostatin 10 was more deterrent than bryostatin 20 (Lopanik, Gustafson & Lindquist 2004)]. It is possible that variations in bryostatin composition are due to differing communities of predators selecting for symbiont-produced bryostatins that are most unpalatable to those local predators. This intriguing aspect of defensive symbiosis has not been explored.
The putative bryostatin biosynthetic gene cluster (bry) was sequenced from two B. neritina-’Cand. E. sertula’ sibling species, the deep CA and the shallow NC species (Hildebrand et al. 2004b; Sudek et al. 2007). There are several lines of evidence suggesting that this large polyketide synthase (PKS) gene cluster produces the bryostatins, although it has not been demonstrated unequivocally. First, portions of this gene cluster are expressed in ‘Cand. E. sertula’ in the pallial sinus of B. neritina larvae (Davidson et al. 2001) and are absent in hosts cured of the symbiont (Davidson et al. 2001; Lopanik, Targett & Lindquist 2006a). Furthermore, bioinformatic analysis of the PKS enzymatic domains supports the putative biosynthetic scheme (Nguyen et al. 2008; Trindade-Silva et al. 2010). Finally, in vitro biochemical assays with bry genes expressed in a heterologous host demonstrated that the encoded proteins are functional (Lopanik et al. 2008; Buchholz et al. 2010). Despite the differences in bryostatin composition in deep and shallow sibling species, there are very few differences in the sequences of the two gene clusters (>98% similarity). The primary difference is that the gene cluster in deep colonies is located on two loci of the chromosome flanked by a transposon, while that of the shallow colonies are contained on one locus (Sudek et al. 2007). According to models of polyketide biosynthesis (Hill 2006), these differences do not account for the differences in bryostatin composition in the two sibling species. The mechanism resulting in the different bryostatin compositions among sibling species is still unknown.
Interestingly, several other Bugula species have symbionts, including B. pacifica, B. simplex and B. turbinata, while others (B. stolonifera and B. turrita) do not (Woollacott 1981; Lim-Fong, Regali & Haygood 2008). Lim & Haygood (2004) fully characterized the symbiont associated with B. simplex and while there were some obvious morphological and distribution differences between ‘Candidatus Endobugula glebosa’, the symbiont in B. simplex, and ‘Cand. E. sertula’, they are 95% identical in 16S rRNA sequence and are closely related. LC/MS and binding assays revealed that B. simplex has compounds that may be bryostatins or similar to bryostatins (Lim & Haygood 2004). Another congener, B. pacifica, has a symbiont closely related to ‘Cand. E. sertula’, but extracts were negative for bryostatin activity (Lim-Fong, Regali & Haygood 2008). In another study, extracts of B. pacifica had antimicrobial activity against two marine microbial isolates, as well as Bacillus subtilis, Staphylococcus aureus and Escherichia coli (Shellenberger & Ross 1998), suggesting that it may have bioactive natural products besides the bryostatins. It is still unclear whether these other Bugula sp.-symbiont relationships constitute defensive mutualisms. Do symbionts that do not produce bryostatins in Bugula hosts provide another benefit for their partners or produce defensive compounds other than bryostatins? Do Bugula hosts with no symbionts have an intrinsic defence, as was suggested by the deterrence of Northern B. neritina? Are variations in bryostatin composition a result of the local predator populations? Identifying patterns in closely related hosts with closely related symbionts that do and do not produce bryostatins may provide insight into ecological and evolutionary influences on these relationships.
Isopod and epibiotic cyanobacteria
An unique mutualistic interaction exists between isopod crustaceans (Santia spp.) and cyanobacteria in reefs off of Papua New Guinea (Lindquist, Barber & Weisz 2005) where predators ignored apparent, red isopods in fully exposed regions of the reefs during the day. Microscopic examination revealed the presence of a lawn of photosynthetic epibionts on the dorsal side of the isopods. Pigment and molecular analyses of the epibionts demonstrated that the community is composed of cyanobacteria related to the genera Synechococcus, Prochlorothrix and Synechocystis, of which, Synechocystis is known to produce natural products such as terpenoids, fatty acids, and lipopeptides (Burja et al. 2001). Microscopic examination of the cyanobacterial cells revealed the presence of heterotrophic bacterial cells surrounding the cyanobacterial cells, which could potentially be involved in the production of deterrent natural products. In order to differentiate between the roles played by the cyanobacteria and the heterotrophic bacteria, the isopods were shaded for 2 days to reduce levels of cyanobacteria and then offered to fish predators in the field. While control isopods with their full complement of cyanobacteria were significantly rejected (~98% rejection), a nonsignificant 20% of cyanobacteria-depleted isopods were rejected. Extracts of the control isopods incorporated into artificial food also were significantly rejected compared to control pellets, although specific compounds responsible for the chemical defence have not been identified. To identify the defensive compounds, traditional approaches such as bioassay-guided fractionation followed by metabolite structural interrogation can be used, but these traditionally require large amounts of tissue. Symbiont genomic sequence analysis, which would require less host tissue, can provide hints as to the types of metabolites present so that chemical methodology could be refined and targeted. Bioassays, however, are still necessary to prove that the isolated metabolites are, in fact, deterrent as there may be more than one natural product. Interestingly, the researchers also observed the isopods utilizing the cyanobacteria for food. This is further supported by the observation of the isopods spending time in exposed reef regions, presumably optimizing growth conditions for their epibionts. Host behaviour maximizing symbiont growth has recently been hypothesized to be employed by yeti crabs to increase the density of their food-source symbionts (Thurber, Jones & Schnabel 2011). Utilization of the epibiotic cyanobacteria that presumably produce unpalatable compounds as both a food source and defence against generalist predators suggests a close co-evolutionary process between the partners. This leads to additional questions of how the host tolerates the metabolites, and if there is a balance between consuming the symbiont and defence.
Antipredator defence in soft corals
The relationship between dinoflagellates and anthozoans has been well documented, especially that of zooxanthellae and reef-building corals [reviewed in (Smith & Douglas 1987)]. Many species of soft corals closely associate with dinoflagellates, which contribute photosynthetically derived nutrients to the host (Muscatine & Porter 1977). Soft corals are a rich source of terpenoid natural products [reviewed in (Harper et al. 2001)]. Several diterpenes isolated from soft corals have anti-inflammatory properties, and one in particular, pseudopterosin (Fig. 3a), is currently an ingredient in cosmetics (Kijjoa & Sawangwong 2004). Surveys of soft coral chemistry have demonstrated that many compounds are deterrent to predators. O'Neal & Pawlik (2002) found that all crude extracts of 32 species of soft corals tested in the Bahamas were deterrent to the generalist predator, Thalassoma bifasciatum, although there were no further investigations into the identity of the metabolites or the role that symbiotic dinoflagellates play in the production of the compounds. Other researchers showed that Symbiodinium spp. dinoflagellates produce the pharmacologically active diterpenes isolated from the soft coral Pseudopterogorgia spp. (Mydlarz et al. 2003). Pseudopterosins comprised 5% of the lipid extract of Pseudopterogiorgia elisabethae branches, but 11% of the purified dinoflagellate lipid extract. Furthermore, the presence of elisabethatriene, the first committed precursor of pseudopterosins, in isolated dinoflagellate cells suggested that they were the source of the pseudopterosins. Finally, to confirm the source of these compounds, radiolabelled elisabethatriene and pseudopterosins were purified from Symbiodinium sp. cells incubated with radiolabelled NaH14CO3 (Mydlarz et al. 2003). Similar experiments with Pseudopterogorgia bipinnata and its Symbiodinium sp. symbionts demonstrated that the dinoflagellates were the source of the kallolide family of diterpenes (Boehnlein, Santiago-Vazquez & Kerr 2005). Extracts of both P. elisabethae and P. bipinnata were significantly unpalatable to the carnivorous T. bifasciatum in assays (O'Neal & Pawlik 2002), but the deterrent compounds were not identified. While the ecological significance of the gorgonian diterpenes has not been definitively established, these studies demonstrate that the dinoflagellate symbionts associated with soft corals are capable of biosynthesizing these pharmaceutically active natural products.
The benefit of having a symbiont with two roles (nutrition and defence) in the host could be context-dependent. When the host is under high predation pressure, the symbiont-provided defence would be positive. If predation pressure decreases, symbiont resources allocated to defence production may result in less nutrition for the host. This may also reduce the ability of the host to compete with others that have nutritional, but not defensive symbionts. For hosts with only defensive symbionts, like Bugula, it seems that the contextual benefit of the relationship is similar; hosting a symbiont is likely to cost some amount of energy, and in the absence of the ecological pressure, a symbiont-free host will out-compete the symbiotic individuals. The host with only a defensive symbiont, however, is more likely to lose its symbiont than one with a symbiont that also contributes nutrition.
Host biotransformation of metabolites produced by a symbiont represents another strategy to maximize the host defence. For example, chrysomelid beetles transform salicin, a compound produced by their plant food source, into salicylaldehyde, which is a more potent deterrent metabolite against beetle predators (Pasteels et al. 1983). The soft coral Pseudopterogorgia americana possesses a deterrent chemical extract (O'Neal & Pawlik 2002), and several related novel secosteroids, including 9,11-secogorgosterol, 7-hydroxy-9,11-secogorgosterol and 9,11-secodinosterol (Fig. 3b,c; He et al. 1995). Extracts of two chemotypes of P. americana were assayed for deterrency with T. bifasciatum (Epifanio et al. 2007). One chemotype (B) was more unpalatable than the other (A); TLC of the extracts revealed that chemotype B had a polar compound that was not present in chemotype A extracts. Bioassay-guided fractionation resulted in identification of 9,11-secogorgosterol and 9,11-dinosterol as the major and minor components, respectively, of the active fraction. The two isolated compounds were not significantly deterrent in field assays, but when they were combined, the mixture was highly unpalatable, suggesting a synergistic defensive role for the secosterols. It is hypothesized that gorgosterol is the precursor for 9,11-secogorgosterol, and dinosterol for 9,11-secodinosterol. These two sterols were identified in some cultures of zooxanthellae that were separated from their invertebrate hosts (Withers et al. 1982), suggesting that the dinoflagellates are capable of de novo synthesis of these sterols. No secosterols were identified from dinoflagellates isolated in this study, but subsequent research by Kerr, Rodriguez & Kellman (1996) demonstrated that cell-free extracts of whole P. americana were able to convert gorgosterol into 9,11-secogorgosterol, indicating that symbiont-produced precursor sterols are biotransformed by the host into the deterrent secosterols.
Bioactive tambjamines in bryozoans and ascidians
One of the first compounds discovered from a marine invertebrate and described as potentially having a microbial source is a blue tetrapyrrole pigment (Fig. 4a) isolated from a colonial ascidian in Western Australia (Kazlauskas et al. 1982). The pigment had been previously identified as a product of a mutant strain of the γ-Proteobacterium, Serratia marcescens, which normally produces prodigiosin, a related tripyrrole pigment (Wasserman, Friedland & Morrison 1968). Subsequent studies demonstrated that the bryozoan Bugula dentata also contained the tetrapyrrole (Matsunaga, Fusetani & Hashimoto 1986). The presence of tetrapyrrole pigments in phylogenetically diverse marine invertebrates (an ascidian and a bryozoan) and in a mutant bacterium led to the speculation of a microbial source of the compounds in the invertebrates (analogous to the discovery that endophytic fungi produce ergot alkaloids in both grasses and morning glories as discussed in Pannicione et al., this volume). Closely related bipyrrolic compounds were isolated from three nembrothid nudibranchs (Tambja abdere, Tambja eliora and Roboastra tigris; Carteé & Faulkner 1983). These compounds, the tambjamines (Fig. 4b), were shown to be unpalatable to the generalist predator, the spotted kelpfish Gibbonsia elegans (Carteé & Faulkner 1986). Higher concentrations of tambjamines in mucus secreted by T. abdere in response to attack by the predatory nudibranch R. tigris also suggested a defensive role for these compounds. The bryozoan Sessibugula translucens, a food source of the Tambja spp. nudibranchs, also possesses the tambjamines. It was hypothesized that the tambjamines found in S. tranluscens could result from conversion of the microbial tetrapyrroles, and the high concentrations in the Tambja spp. nudibranchs (3·4 and 2·2% dry weight for T. abdere and T. eliora, respectively) resulted from sequestration from a diet of S. translucens (0·45% dry weight). In choice assays, T. eliora was attracted to sea water exposed to S. translucens and tambjamines A and B, demonstrating that the nudibranchs utilize the molecules as feeding stimulants (Carteé & Faulkner 1986). Similarly, investigation of the chemical defences of Atapozoa spp. (later identified as Sigillina signifera) ascidians from the Indo-Pacific demonstrated that several tambjamines found in the ascidian from different locations were deterrent to generalist fish predators in field assays (Paul, Lindquist & Fenical 1990). These studies also demonstrated that the Nembrotha spp. nudibranchs, which are specialist predators on S. signifera, had sequestered sufficient quantities of deterrent tambjamines for defence. Furthermore, the tetrapyrrole previously isolated from the microbe S. marcescens, and the bryozoan B. dentata was found in mucus exuded from the nudibranchs at levels that were sufficient for deterrency. Nudibranchs have often been shown to utilize natural products from their diet for defence (Cimino, Fontana & Gavagnin 1999), similar to defensive utilization of cardiac glycosides from milkweed by monarch butterflies and other insects (Harborne 1997). The utilization of tambjamines and the tetrapyrrole pigment as defensive compounds by two groups of nudibranchs from different diet sources (the bryozoan S. translucens and the ascidian Atapozoa sp.), suggests specialization by the nudibranchs and co-evolution between predator and prey containing the deterrent metabolites. Furthermore, it indicates that diverse hosts may utilize similar symbiont-produced chemistry in their defence.
As the tetrapyrrole pigment, a possible product of tambjamine condensation, was identified as a metabolite from a mutant strain of the microbe, S. marcescens (Wasserman, Friedland & Morrison 1968), a microbial origin of the tambjamines and the tetrapyrrole was proposed. Franks et al. (2005) isolated a novel tambjamine (YP1, Fig. 4b) from the marine bacterium Pseudoalteromonas tunicata. This compound had antimicrobial, antifungal (Franks et al. 2006) and antifouling (Holmstrom et al. 2002) activity, suggesting that the tambjamines can be effective in several ecological roles. Recently, a biosynthetic gene cluster from P. tunicata that produces the tambjamine YP1 was sequenced (Burke et al. 2007). Colonies from a large-insert genomic DNA library generated in E. coli were screened for antifungal activity against Candida albicans. The sequences of three active clones were aligned to the P. tunicata draft genome sequence to identify a region of 21 open reading frames (ORFs) that may be involved in YP1 biosynthesis. Transposons were used to disrupt the ORFs and showed that at least 19 of the ORFs are critical for tambjamine production. As multiple microbial sources of tambjamines have been identified, and since many bryozoans and tunicates harbour microbial symbionts, it is reasonable to hypothesize that a microbial symbiont is likely responsible for their production in S. translucens and S. signifera. It is interesting to note that the tambjamines likely defend both the sessile, microbial-symbiont hosting invertebrates (B. dentata, S. translucens, S. signifera), as well as the mobile nudibranch, although the nudibranch likely acquires the defensive tambjamines from its diet of ascidians. Research into the microbiomes of tambjamine-containing invertebrates could reveal if the taxonomically diverse organisms have closely related symbionts, or if the symbioses evolved independently followed by the acquisition of tambjamine biosynthetic genes via horizontal gene transfer. Comparison of the bryostatin-producing symbionts of B. neritina and B. simplex and the tambjamine-producing symbiont of B. dentata may help identify specific symbiotic traits that appear to be necessary for the establishment of the relationship within this group of invertebrates.
Natural products and Didemnidae ascidians
The didemnid ascidians are characterized by both their abundance of natural products (Davidson 1993; Schmidt et al. 2012), and their possession of photosynthetic cyanobacterial symbionts of the genera Procholoron or Synechocystis (Lambert, Lambert & Waaland 1996; Sings & Rinehart 1996). For example, the Caribbean ascidian Trididemnum solidum is the host of the cyanobacterial symbiont Synechocystis trididemni and the source of the potent antiviral and antitumor cyclic peptides, the didemnins (Rinehart et al. 1981a, b). Didemnin B (Fig. 5a) was one of the first marine natural products isolated to be tested in clinical trials (Phase I and II), although it was discontinued because of toxicity [reviewed in (Lee et al. 2012)]. In studies of the ecological role of didemnins, as well as other tunicate secondary metabolites, didemnin B and the closely related nordidemnin B were significantly deterrent to an array of coral reef fishes when incorporated into palatable feeding pellets at concentrations 40% below their natural levels in T. solidum adult colonies (Lindquist, Hay & Fenical 1992). Furthermore, the palatability of T. solidum larvae and larval natural products was assessed in an elegant experiment in which the larval natural products were transferred to a krill eye, a larval mimic in both size and coloration, using capillary action of solvent. Krill eyes with impregnated larval compounds reduced feeding by the generalist predator T. bifasciatum 89% compared to control krill eyes, demonstrating that natural products can defend brooded, nutrient-rich, visually apparent larvae (Lindquist 2002). Furthermore, later studies by Lindquist & Hay (1995) demonstrated that while these metabolites were not distasteful to allopatric fish predators, they induced regurgitation, and the predators were able to learn an aversion to these metabolites. Previous studies of T. solidum larval release and settlement had demonstrated that larvae are released midday by adult colonies and settle onto habitats with intermediate light intensity that allow photosynthesis in the cyanobacterial partner (van Duyl, Bak & Sybesma 1981). Having the distasteful didemnins can reduce the threat of predation as these large larvae search for an appropriate settlement site.
The origin of the didemnins (host or symbiont) was suspected to be the cyanobacterial symbiont (Sings & Rinehart 1996) given that didemnin B is also found in a geographically and phylogenetically distant ascidian, Aplidium albicans, and that didemnin B has structural similarities with natural products isolated from the free-living cyanobacterium, Lyngbya majuscula (Marner et al. 1977). The recent finding that an α-Proteobacterium, Tistrella mobilis YIT 12409, isolated from marine sediments in Japan, produces didemnin B and nordidemnin B during cultivation (Tsukimoto et al. 2011) provides further evidence that a microbial symbiont may be responsible for the production of the didemnins. Simultaneously, another group isolated a didemnin-producing T. mobilis strain (KA081020-065) from the Red Sea (Xu et al. 2012). The sequenced genome of this strain is composed of a 4 Mb chromosome and four plasmids ranging in size from 83 kb to 1·1 Mb. Bioinformatic analysis of the genome revealed that the putative didemnin biosynthetic gene cluster resides on the largest plasmid, pTM3, and interestingly, the products of the biosynthetic gene cluster are two precursor metabolites didemnin X and Y, which are later converted into didemnin B. It is unknown if T. solidum has a microbial symbiont other than the cyanobacterial S. trididemni described in 1979 (LaFargue & Duclaux 1979), but as the putative didemnin biosynthetic gene cluster is located on a plasmid in the free-living T. solidum strain, horizontal gene transfer to S. trididemni is possible. T. solidum is a newly described genus (Shi et al. 2002), and there are not any documented examples of symbiotic interactions between this microbe and eucaryotic hosts. Time-course MALDI imaging mass spectrometry of T. mobilis KA081020-065 growing on a solid substrate showed that didemnin X and Y were exported out of the cells and then converted to didemnin B by a secreted protein (Xu et al. 2012). Exportation and conversion of a metabolite to an active form are a resistance strategy exhibited by some antibiotic-producing microbes (Walsh 2003), and suggests that the didemnin B can potentially affect the growth of the producer. While the mode of action of didemnin B with eucaryotic cells is not completely elucidated, it binds to human elongation factor 1 (EF-1α) [reviewed in (Lee et al. 2012)]. The general toxicity of didemnin B to the host (and potentially the producer) leads to questions of how the host is able to survive in the presence of the symbiont-produced metabolites, and if the molecule is transported to parts of the colony that are the most susceptible to predation.
Another didemnid ascidian, Lissoclinum patella, is the source of the bioactive peptides, the patellamides (Fig. 5b; Sesin, Gaskell & Ireland 1986; Degnan et al. 1989; Fu et al. 1998) and the thiazole-containing polyketides, the patellazoles (Fig. 5c; Zabriskie, Mayne & Ireland 1988; Richardson, Aalbersberg & Ireland 2005), although the patellazoles are much more sporadically distributed in L. patella individuals. The number and bioactivity of the natural products isolated from L. patella have led to recent investigation into their biogenesis. The L. patella cyanobacterial symbiont, Prochloron didemni, was proposed to be the source of the patellamides and patellazoles, as micro-organisms often biosynthesize these classes of metabolites via nonribosomal peptide synthetases (NRPS) and PKS, respectively (Schmidt 2008). These types of biosynthetic genes are modular, and often have highly conserved regions, which allow them to be targeted in DNA libraries. In the search for the genes responsible for prescribing biosynthesis of patellamide, a Prochloron spp.-dominated metagenomic library was probed for the presence of NRPS genes (Schmidt, Sudek & Haygood 2004). Although researchers identified an NRPS gene that could potentially produce a portion of patellamide, presence of the gene did not correspond with L. patella colonies that contained the metabolites, suggesting that this gene was not involved in patellamide biosynthesis. As the source of the patellamides remained enigmatic, the metagenome of the symbiont community (again, dominated by Prochloron spp.) was sequenced (Schmidt et al. 2005). Bioinformatic analysis of the Prochloron genome sequence did not reveal any other NRPS genes that could possibly dictate biosynthesis of the patellamides. Researchers then searched for DNA sequences encoding for different amino acid combinations that could result in the patellamide A peptide macrocycle and identified a gene encoding the peptide comprising patellamide A as well as patellamide C. Six flanking genes were hypothesized to be involved in the biosynthesis of the patellamides based on homology to other biosynthetic genes. The small size of the putative biosynthetic pathway (11 kb, compared to other microbial natural product biosynthetic gene clusters) allowed cloning and expression of the genes in E. coli. During fermentation, the clone produced patellamide A, thus confirming this gene cluster as the patellamide producer. Furthermore, the presence of this gene cluster corresponded with colonies of L. patella that contained patellamides. Another research group demonstrated patellamide D production by E. coli with a plasmid containing the pat gene cluster from Prochloron spp. symbionts from Australian L. patella individuals (Long et al. 2005), confirming the findings by Schmidt et al.
To identify the patellazole producer, the microbiome of multiple L. patella individuals was investigated through metagenomic sequencing (Donia et al. 2011). A gene cluster hypothesized to prescribe biosynthesis of the patellazoles was identified in the symbiotic metagenomic sequences, but is not part of the P. didemni genome, suggesting that another member of the host microbiome is responsible for patellazole production. Further investigation into the L. patella microbiome resulted in the full genome sequence of a α-Proteobacterium designated ‘Candidatus Endolissoclinum patella’ that possesses a PKS gene cluster for patellazole biosynthesis (ptz; Kwan et al. 2012). Remarkably, in the genome, ptz remains intact, while the rest of the genome appears to be undergoing degradation; in fact, ptz comprises 10% of the symbiont genome, suggesting that the sole purpose of this microbe is to produce patellazole. Interestingly, L. patella is readily preyed upon by fish predators (Olson & McPherson 1987) and patellamide C is palatable to an array of reef fishes (Lindquist, Hay & Fenical 1992), suggesting that the patellamides may be involved in processes other than defence, or are effective against a group of predators not assayed in these studies. It is unclear whether the patellazoles are deterrent to predators, but the maintenance of the ptz gene cluster while the rest of the genome undergoes degeneration suggests that the molecules are important to the relationship. Sporadic distribution of patellazole in L. patella individuals in different locations in the Indo-Pacific highlights the sometimes–ephemeral relationship between host and symbiont, which may be due to the defensive nature of the secondary metabolites. If predation on the host is reduced, the defensive metabolites may not be necessary, and the physiological cost of hosting a symbiont may result in breakdown of the relationship. The differences in symbiont and bryostatin occurrence in the B. neritina sibling species also support this idea. Variation in the presence of the symbionts, as well as the types of natural products produced by a symbiont (e.g. different patellamides are found in L. patella individuals), indicates a complexity in these relationships that should not be overlooked.
Conclusions and future directions
Marine invertebrates are a prolific source of bioactive natural products, as evidenced by the number of novel compounds identified every year (>1000 each year from 2009 to 2011; Blunt et al. 2010, 2011, 2012). The potent activity and pharmaceutical potential of these compounds have resulted in investigations into the source of the natural products, which have demonstrated that many are produced by associated microbes (Piel 2009). It is, however, rarely shown experimentally that the symbiont-produced molecule defends the host from predators. For soft-bodied sessile marine invertebrates that cannot escape predators, having a potent chemical defence can ensure their survival. In symbiotic associations between microbes that produce deterrent compounds and the sessile hosts, the microbes can benefit from host-provided nutrients and stable habitats. When considering the benefit to both partners and the ability of microbes to produce diverse natural products, it seems likely that symbiont-produced chemical defences are more common than currently recognized in the marine environment and may play a significant role in structuring communities. For the host, having a symbiont that produces defensive natural products, vs. an endogenous defence, may seem immensely beneficial, as no host genetic resources will have to be dedicated to the defence. It is apparent, however, that hosting a beneficial microbe can come at a physiological cost, and that the host may have to modify its immune response, among other things, to accommodate the symbiont. Under certain circumstances, the physiological cost may be too high or the benefit too low to facilitate the association. The physiological cost of hosting a symbiont may also lead to the heterogeneous distribution of symbionts and defensive compounds within the host.
Chemical defensive symbiosis reflects a multitrophic co-evolutionary history and complexity, as illustrated in the examples considered here. Much remains to be learned about the host–symbiont interactions that facilitate production of these compounds, and their utilization as a defence. For instance, differences in bryostatin concentrations in adult and larval B. neritina suggest that there is some mechanism for concentrating the bryostatins on the vulnerable larvae: does the host provide a signal to the symbiont at the onset of reproduction to initiate bryostatin biosynthesis? or does the host control symbiont density and bryostatin biosynthesis is regulated by quorum sensing? Alternatively, is there a mechanism to transport the bryostatins to the larvae? In these relationships, host tissues are in continuous contact with symbiont-produced compounds, which have potent bioactivity towards eucaryotic cells. Does the presence of the bioactive natural products affect the host or has the host adapted to being in continuous contact with these compounds? For instance, fungal hosts of rhizoxin-producing microbial symbionts have an amino acid substitution in their β-tubulin so that rhizoxin does not inhibit host cell replication (Partida-Martinez & Hertweck 2005; Schmitt et al. 2008). Are there such modifications in the host cellular targets of bryostatin, patellamide or patellazole? Other questions to be answered include: does the host play a role in biosynthesis of or modify the compounds, as is proposed for diterpenoids in gorgonian corals? Does the host pay a cost for housing the symbiont? Can geographical differences in predator distribution affect symbiont metabolite biosynthesis? If predation pressure is removed, is the relationship still viable? Answering these questions will require a significant effort in molecular characterization of both partners and further field and laboratory-based experimental manipulations of these nonmodel organisms. In turn, these efforts have the potential to reveal the complexity and co-evolution of host–symbiont relationships, and how microbes can affect interactions between organisms at different trophic levels. Advances in sequencing technology may provide insight into the degree of horizontal gene transfer of these natural product biosynthetic gene clusters from free-living microbe to symbiont, from symbiont to symbiont or even from symbiont to host. Furthermore, future may lead to a better understanding of the biosynthesis of natural products by recalcitrant micro-organisms, and allow for future production and exploitation of these bioactive compounds.
I would like to thank Keith Clay for the invitation to participate in this Defensive Symbiosis special feature and two anonymous reviewers for their helpful comments that improved the manuscript. Thanks to Niels Lindquist for interesting discussions on this topic and for facilitating our ongoing research on defensive symbiosis in Bugula neritina.