The innate immune system is the first line of defense against pathogens, and in non-chordates is assumed to be the sole means by which any non-self cells are detected and either killed or contained 1. Innate immunity in vertebrates is essentially a two-tier system consisting on one hand of phagocyte activation by the interaction of specialized surface receptors with pathogens or pathogen-derived components, and on the other of the direct opsonization and lysis of pathogens via the complement cascade. Whilst the vertebrate innate immune system has been the subject of intense investigation and is relatively well understood, studies of invertebrate immunity, which have focused primarily on the arthropods Drosophila and various horseshoe crab species 234, have revealed some striking similarities. For example, in both Drosophila and vertebrates, the Toll/Toll-like receptor (TLR) mediates the activation of appropriate response genes to microbial challenge 56.

Toll and the TLRs are transmembrane proteins with a characteristic domain structure consisting of an extracellular amino-terminal domain containing leucine-rich repeats (LRRs) responsible for pattern recognition and an intracellular Toll interleukin receptor (TIR) domain that mediates signal transmission. Although the Toll and TLR families of arthropods and mammals are thought to have independently diversified 78, all Tolls and TLRs signal via a common pathway that is conserved between Drosophila and mammals. The ultimate step in this pathway is translocation of nuclear factor (NF)-κB or its fly counterpart (the Dif/Rel heterodimer) into the nucleus, where it stimulates transcription of appropriate response genes. The immune repertoire of the horseshoe crab Carcinoscorpius includes a complex complement pathway that has both opsonic and lytic effector functions 9. Horseshoe crab complement C3 is functionally homologous with mammalian C3, mediating phagocytosis of bacteria (by hemocytes) in a strikingly similar manner.

Whilst these specific studies imply that at least some innate immune mechanisms have been conserved, broader comparative studies highlight the extent of gene loss and divergence in various metazoan lineages. For example, although Carcinoscorpius clearly uses a vertebrate-like complement system, none of the central components of the cascade (C2, C3, C4, C5) are encoded by the genomes of the ecdysozoans Drosophila, Caenorhabditis or Anopheles. Moreover, the sole Toll/TLR in Caenorhabditis elegans and C. brigssae is not known to function in the context of immunity, nor does that reported in the horseshoe crab Tachypleus tridentatus 10. There are also important differences between the Toll/TLR systems of Drosophila and mammals. For example, some mammalian TLRs themselves act as pattern recognition receptors (PRRs) upon microbial challenge, whereas in fly this is not the case 11. Moreover, whereas most of the ten or so vertebrate TLRs function primarily in immunity, only one of the nine fly (and ten mosquito) Tolls functions in this context. The others play a role in development 10, most famously in controlling differentiation in the dorsal/ventral axis.

The significance of gene loss in animal evolution has recently been brought into focus by preliminary expressed sequence tag (EST) and genomic analyses of some 'basal' animals (Figure 1), particularly the anthozoan cnidarians Acropora millepora and Nematostella vectensis 1213 and the planarian Dugesia japonica 14. Paradoxically, the genomes of these morphologically simple animals contain many genes previously thought to have evolved much later in the context of vertebrate complexity, and most of the complexity of signaling pathways and transcription factors associated with higher animals is represented in the anthozoan datasets 13151617. In contrast to Drosophila and Caenorhabditis, which have undergone substantial gene loss, for at least some groups of genes Acropora and Nematostella appear to have preserved much of the genetic complexity of the common metazoan ancestor. For example, whereas fly and worm have each lost approximately half of the ancestral Wnt complement, all but one of the 12 known Wnt subfamilies is represented in Nematostella 15. The emerging cnidarian EST and genomic datasets are, therefore, potentially highly informative with respect to the ancestral immunological repertoire. In addition to this basic evolutionary significance, understanding the bases of cnidarian immunity is of major applied significance in light of the dramatic decline in coral health that has occurred on a global scale over the past 20 years. Increasing human activity in coastal zones throughout the world has led to declines in water quality with increased sediment, nutrient and heavy metal concentrations. These have all had detrimental effects on corals with an associated increase in the prevalence of disease, and perhaps led to some new diseases, although this is less certain, as coral diseases are very poorly understood. Most are named for their symptoms, for example, 'white band disease', 'black band disease', and 'rapid wasting syndrome' and causative agents are frequently unknown. In the face of this uncertainty it is important that coral defense mechanisms should be better understood.

Although cnidarians have no specialized immune cells, at least some display highly specific allorecognition characteristics. Allorecognition, xenorecognition, and killing mechanisms have been demonstrated in several hydrozoans 181920 and anthozoans 212223. Allorecognition is thought to protect colonial cnidarians from fusion with genetically different individuals and to prevent germ line parasitism. The effector mechanisms range from contact avoidance involving chemical sensing, to barrier formation, or usage of nematocysts. For example, the sea anemone Anthopleura xanthogrammmica will 'tolerate' adjacent clonal individuals, but will attempt to 'reject' heterogenic clones with which it comes into contact 2425. In the anthozoans, Stylophora pistillata and Montipora verrucosa branches within one colony will readily fuse while branches of genetically different individuals never undergo fusion 222627. Fusion of two conspecific individuals is occasionally referred to as 'natural transplantation'. Observations in sea anemones (Anthopeura elegantissima, Phymactis clematis) and gorgonians (Eunicella stricta) indicate that individual colonies possess unique sets of histocompatibility elements, which are recognized as nonself by all other conspecific colonies 2328. In Hydrozoa, the same phenomenon was reported for Millepora dichotoma 29 and studied in great detail in the colonial marine hydroid Hydractinia echinata 30. Hydractinia in fact was among the first invertebrates shown to display a genetically based system of intolerance against allogenic tissue: for more than 50 years it has been known that allorecognition and the inability to fuse stolons of different colonies is under the control of one polymorphic gene 3132. Recent efforts using defined genetic lines of the hydroid Hydractinia symbiolongicarpus have confirmed this and shown that the single chromosomal region contains at least two loci 33.

The availability of whole genome sequences of two basal cnidarians at respectably high levels of redundancy - the hydrozoan Hydra magnipapillata (>6-fold coverage) and the anthozoan Nematostella vectensis (>7.6-fold coverage 17) - together with large-scale EST datasets for these and for the coral Acropora millepora potentially offer new perspectives on the origins of mammalian immune functions. Here we report the results of a screen of the available genomic and EST resources for the cnidarian counterparts of key components of the vertebrate innate immune repertoire.

These preliminary analyses of the newly available genomic and EST datasets indicate that a surprising number of key components of the innate immune system, including the Toll/TLR pathway and some complement cascade components, were in place at the base of the Eumetazoa. Also represented in the datasets are proteins strongly matching both the tumor necrosis factor (TNF) and Nod-like receptors and with the same domain structures (data not shown).

The analyses presented here are consistent with the idea of a genetically complex common metazoan ancestor 1213; clearly the Toll/TLR, MyD88 and IL-1R protein families were distinct prior to the divergence of the Cnidaria from the Bilateria, and a Toll/TLR pathway may predate even the Porifera/Eumetazoa split. The discovery of proteins with the same domain structure as the IL-1R in Nematostella indicates that this receptor type predates chordate origins and that its original ligands may not have been interleukins. However, as the TIR domains in the cnidarian IL-1R-like and mammalian IL-1R proteins are divergent, separate evolutionary origins cannot yet be ruled out despite the similarities at the level of domain structure.

It is also likely that a prototypic complement effector pathway involving C3 and multiple MAC/PF proteins was in place in Ureumetazoa. Comparisons of the TIR and MAC/PF complements of Hydra and Nematostella highlight the likely extent of gene loss and sequence divergence in the former - not only does Hydra appear to have lost the Toll-receptor, but NF-κB has also either been lost or has diverged beyond recognition. Moreover, only a few TIR proteins could be identified in Hydra compared to the rich representation of these in the anthozoans. In addition, Hydra appears to have lost a number of MAC/PF proteins and lacks an equivalent of the ancestral complement C3 protein, implying that although anthozoans most likely have a prototype complement effector pathway, this has undergone degeneration in Hydra. Hydra is an efficient producer of a number of potent antimicrobial peptides (Bosch, unpublished data), indicating that at least some of the functions of the complement effector system may have been subsumed by pathways involved in synthesis of these peptides.

The observation that genes of two of the three classes of cnidarian immune-repertoire genes examined here are expressed in the endoderm (Figure 2) was unexpected, but is consistent with the primary sites of expression of such genes in many animals. In Drosophila, the fat body (an endodermal derivative, and the functional homolog of the mammalian liver) is the predominant source of the antimicrobial peptides whose synthesis is under the control of the Toll and Imd pathways 44 and in Caenorhabditis the gut is the primary source of antimicrobials 4546. Although mammalian skin contains immune sensory cells (Langerhans cells, dendritic cells and so on), these are also endodermal derivatives, hence there is a deep evolutionary link between the immune system and the endoderm.

Other specific cases of gene loss from Hydra have been documented 47 (Seneca, unpublished data), hence there are precedents for the apparent loss of immune repertoire components reported here. More surprising is the implication that gene losses may also have occurred in Nematostella, as evidenced by the apparent absence of the apextrin gene that is present in both Hydra and Acropora. This loss may be associated with differences between Nematostella and Acropora during the planula to polyp transition. Metamorphosis from the motile planula to the sessile polyp involves dramatic tissue remodeling 48 and gene expression changes in Acropora (Grasso, unpublished data), and the massive up-regulation of apextrin leading up to and during metamorphosis (Figure 5k-o) is consistent with a role in this process. It is worthy of note that apextrin is expressed in the tissue that is least remodeled at the time of metamorphosis. By contrast, it appears that there are no dramatic events comparable to Acropora metamorphosis occurring during this same period of Nematostella development; rather, Nematostella undergoes a simple and continuous transition from planula to polyp and appears to show some neotenous features, continuing to glide over the bottom aboral end first for some time after metamorphosis 49, and never forming a pedal disc 50. Thus, loss of this gene may be related to the different modes of development of these two animals. This comparison between two anthozoans with different modes of development is clearly reminiscent of the original identification of apextrin as a gene specifically expressed in the 'direct' developing sea-urchin Heliocidaris erythrogramma during metamorphosis, but which is not expressed during the development of H. tuberculata, a congeneric 'indirect' developer 4142. Because indirect development is considered to be ancestral in sea urchins, these authors hypothesized that apextrin had been coopted to function in metamorphosis in H. erythrogramma. The specific ectodermal expression of homologous proteins during metamorphosis in a cnidarian and an echinoderm is either a remarkable example of convergence or reflects conservation of function at some level. This latter possibility carries with it the implication that the 'simple' continuous planula to polyp transition seen in Nematostella might represent a derived state, and that the more dramatic metamorphosis seen in Acropora might more closely reflect the ancestral condition.

An important general implication that flows from these data is that gene loss may occur stochastically. If the genes in a pathway function only in that pathway, then one might expect that entire pathways would disappear following loss of one key component. However, the Hydra data appear to contradict this; most of the intracellular intermediates of Toll/TLR signaling are present despite loss of all of the major receptor types (both Toll/TLR and the IL-1R types known from more basal cnidarians are apparently absent from Hydra). Moreover, simple comparisons between related animals (for example, Nematostella versus Acropora) are unlikely to be informative in terms of understanding the origins of genes 50 - loss of a gene from Nematostella and loss of a specific function in the indirect developing urchin seems more likely than the independent co-option of the same protein to related roles in Acropora and a direct developing urchin. Stochastic loss underlying the distribution patterns of genes across the Metazoa may account for some cases of assumed lateral gene transfer - chance retention of an ancestral gene in one or a few animal lineages might easily be confused with lateral transfer. Reconstructing the genome of the common animal ancestor will not, therefore, be a simple undertaking; it will require whole genome data for a wide range of 'lower' as well as 'higher' animals.

The following additional data are available with the online version of this paper. Additional data file lists accession numbers for additional sequences identified within the database searches that were not further characterized in the present study.