Integrins are a large family of cell surface transmembrane receptors known only from metazoans, which function in intracellular signalling as well as cell-cell and cell-extracellular matrix (ECM) adhesion 1. As the main mediators of cell-ECM interactions they are key players in early development 2 functioning in gastrulation by rapid modulation of their own adhesion between low and high affinity states and by modulating the activities of adhesion molecules (e.g. cadherins) in cell layers undergoing convergence and extension 3.

Integrins function as αβ heterodimers with several subunits of each type being present in most animals. Analyses of the whole genome sequence of Caenorhabditis elegans 45 indicate that it has a single β subunit of the β1 type that is capable of associating with two α subunits, which confer specificity for either laminin- or RGD-containing ligands, and it has been suggested that this may reflect the ancestral state. Drosophila melanogaster has five α and two β subunits 5 and mammals, eighteen α and eight β integrin subunits 6. In each case, however, integrin subunits above and beyond likely orthologs of the two αs and one β of Caenorhabditis are clearly lineage-specific. "Lower" animals are of particular significance in terms of understanding the ancestral state, but have not been extensively studied. Both α and β integrin subunits have been identified in sponges 789, and cnidarians 71011, but the extent of integrin diversity and the range of functions of these molecules in "lower" animals are unknown.

Anthozoan cnidarians such as the coral Acropora millepora and the sea anemone Nematostella vectensis appear to have retained much of the genetic complexity of the metazoan common ancestor 1213; hence these animals are likely to be highly informative with respect to the ancestral integrin complement and may provide insights into the original roles of these molecules. Known cnidarian integrins include a β integrin from Acropora millepora 7 and single α and β subunits (IntA and IntB) from the hydrozoan jellyfish Podocoryne carnea 11. Here we report the characterisation of novel β and α integrins from Acropora. The known Acropora integrins are expressed during gastrulation in patterns like those seen at the corresponding stages of echinoderm and amphibian development. However, we know from morphological observation 1415 that Acropora gastrulation is not a simple epithelial to mesenchymal transition in which the expressing cells lose their adhesivity and invaginate. Instead, in Acropora it is clear that changing cell shape also plays a major role14. Two further implications of this work are that the cnidarian integrin complement is significantly more complex than was predicted, and that functional diversification of α integrins may have occurred independently in Cnidaria and Bilateria.

Whereas Reber-Muller et al. 11 hypothesized the presence of only single α and β integrin subunits in cnidarians, the integrins identified to date in Acropora are likely to be only a subset of those present. Preliminary surveys of the genome of Nematostella imply that at least four β integrins (two resembling AmItgβ1 and two more similar to AmItgβ2), and at least three distinct α types are present (data not shown). However, gene models for only one α integrin were sufficiently complete to allow its inclusion in the phylogenetic analyses shown as Fig 5A. The integrin repertoire of this morphologically simple animal is therefore considerably more complex than that of Caenorhabditis (one β and two αs), which has often been assumed to reflect the ancestral metazoan state.

Due to structural constraints on integrin proteins during evolution, phylogenetic analyses are likely to be complicated by homoplasy effects. The position of the cnidarian sequences in the α integrin phylogeny (Fig. 5A) is likely due to both the early divergence of the Cnidaria and homoplasy effects. Hence, in this case, phylogenetics is not informative as to ligand binding properties. It is likely that in cnidarians, as in higher animals, distinct α integrin types participate in binding to laminin and RGD-containing ligands, but functional analyses are required to verify this hypothesis.

Despite having a common pattern of cysteine loss, phylogenetic analyses (Fig 5B) indicate that AmItgβ2 is only distantly related to the vertebrate β4-type. Those cysteine residues (positions 8 and 9) in the consensus absent from the Acropora and Podocoryne sequences form the c8–c9 loop of the β A domain, which has been implicated in determining integrin-ligand specificity, specificity of α-β interactions, and signalling properties 2425262728. Whilst this loss has apparently occurred independently of that leading to the vertebrate β4 type, it may result in common consequences for ligand and/or α-subunit specificity.

The presence of maternal integrin mRNAs and their relatively uniform expression through development reported here for Acropora have precedents in Podocoryne 11 as well as in higher animals. Co-localization of mRNAs for AmItgβ1, AmItgβ2 and AmItgα1 suggests that either or both of the β subunits associate with the α1 subunit. There are many precedents from bilaterians for RGD-type αs associating with β1-type βs (α5β1; αVβ1; α8β1; PS2βPS). Of these, α5β1 and PS2βPS have been implicated as regulators of gastrulation in vertebrates 2930 and Drosophila 31 respectively. In Podocoryne, IntA is assumed to associate with IntB (a possible AmItgβ2 ortholog) since they are co-expressed in a wide variety of locations over a range of life cycle stages 11, suggesting that AmItgβ2 may associate with the AmItgα1 subunit.

Although integrins clearly play important roles in gastrulation in several animal groups, including vertebrates 29, Drosophila 31, and sea urchins 32, the interactions that have been demonstrated are heterogeneous with respect to both the ligands and types of integrins involved. The expression patterns of α and β integrin genes observed in the coral are reminiscent of expression during both amphibian and sea urchin gastrulation. Modulation of integrin adhesion to the RGD domain of fibronectin plays a central role during Xenopus gastrulation 33, and in the sea urchin Lytechinus changes in laminin adhesion mediated by the epithelial α integrin α SU2 are likewise important 34. In both the sea urchin and Xenopus, changes in integrin-mediated cell adhesion during gastrulation occur independently of transcription and translation. In Acropora, integrin mRNAs are present in eggs 23 and the extent to which those visualized in the presumptive endoderm reflect zygotic transcription is unknown.

The simplest interpretation of the patterns of integrin expression is that they reflect increases in cell adhesion in the presumptive endoderm. Expression on the concave side of the "fat prawn chip" stage embryo, which is essentially a flat bilayer of cells (for a description of gastrulation in Acropora, see 14), suggests that integrin-based adhesion may constrain the presumptive endoderm whilst the cells of the presumptive ectoderm move upward and inward around them. Whilst the cells that will end up inside the embryo may adhere to each other more tightly than they do to the putative ectoderm, as is consistent with many forms of gastrulation, the inferred increased adhesion in the putative endoderm is not consistent with a typical epithelial to mesenchymal transition.

The early expression patterns of the α and β integrins in Acropora are very similar to those of two transcription factors, snailA 15 and otxB 35. Snail genes have central and conserved roles in gastrulation in Drosophila 36 and vertebrates 37 and members of the broader class of related genes are regulators of other epithelial to mesenchymal transitions (EMTs). Whilst the best understood means by which snail genes regulate cell adhesion is by acting as repressors of E-cadherin expression 363839, in human epidermal keratinocytes the snail-related gene Slug (Snail2) is a repressor of α3, β1 and β4 integrin expression, leading to decreased cell-adhesion to fibronectin and laminin 5 40. Across the Bilateria, Otx genes are conserved anterior markers 41 and, whilst most Otx genes are expressed in the nervous system, evidence from a diverse range of metazoans suggests an ancient role as regulators of cell adhesion (eg. 4243). In Hydra, high levels of CnOtx expression correspond to regions where cells are undergoing rearrangements or movement 44. Both snailA and otxB are thus candidate regulators of integrin expression in Acropora. Given the similarity of AmItgα1 to diverged RGD-type mammalian integrins, it will be of particular interest to examine the expression of ECM proteins containing fibronectin type III domains during cnidarian gastrulation in parallel with adhesion studies. Candidates identified in Nematostella include predicted proteins similar to vertebrate usherin and titin, and a likely homolog of Drosophila sidekick.

Whilst one might expect morphologically simple metazoans to have a correspondingly basic integrin complement, comprising perhaps just two αs and a single β subunit (as in Caenorhabditis elegans), the repertoire of these molecules in anthozoan cnidarians is considerably more complex. In the case of cnidarian α integrins, ligand specificity cannot be predicted by phylogenetic analysis, suggesting the possibility that specificity mechanisms arose independently in Cnidaria and Bilateria. During early development in Acropora, some of these adhesion/signalling molecules are expressed in patterns which parallel those of their amphibian and echinoderm counterparts, and which are inconsistent with gastrulation being a simple epithelial to mesenchymal transition.

BAK was responsible for conducting DNA sequencing and RT-PCR analysis, and participated in both the in situ hybridisation experiments and preparing the manuscript. AI collected and prepared coral developmental stages, assisted in DNA sequencing and, with CS, carried out in situ hybridisation experiments and photographed the results. DCH was responsible for the initial identification of clones, assisted in sequence alignment and critically reviewed both the data and the manuscript. DJM conducted the phylogenetic analyses, and he and EEB drafted and reviewed the manuscript.