A cnidarian homologue of an insect gustatory receptor functions in developmental body patterning

Insect Gustatory and Odorant Receptors (GRs and ORs) form a superfamily of novel transmembrane proteins, which are expressed in chemosensory neurons that detect environmental stimuli. Here we identify homologues of GRs (Gustatory receptor-like (Grl) genes) in genomes across Protostomia, Deuterostomia and non-Bilateria. Surprisingly, two Grls in the cnidarian Nematostella vectensis, NvecGrl1 and NvecGrl2, are expressed early in development, in the blastula and gastrula, but not at later stages when a putative chemosensory organ forms. NvecGrl1 transcripts are detected around the arboral pole, considered the equivalent to the head-forming region of Bilateria. Morpholino-mediated knockdown of NvecGrl1 causes developmental patterning defects of this region, leading to animals lacking the apical sensory organ. A deuterostome Grl from the sea urchin Strongylocentrotus purpuratus displays similar patterns of developmental expression. These results reveal an early evolutionary origin of the insect chemosensory receptor family, and raise the possibility that their ancestral role was in embryonic development.

T he insect chemosensory receptor superfamily, including gustatory receptors (GRs) and odorant receptors (ORs) (which represent an expanded clade within the GR family 1 ), are polytopic membrane proteins responsible for detecting a vast number of non-volatile and volatile chemicals in the environment 2,3 . Most sequenced insect genomes are predicted to encode several dozen to a few hundred highly divergent GRs and/or ORs 4 . Almost all of these receptors are likely to be expressed in small subsets of sensory neurons in peripheral chemosensory organs. Here they recognize external chemical cues-including esters, alcohols, ketones and volatile lipid-derived pheromones (for ORs), and sugars, bitter compounds and cuticular pheromones (for GRs)-to permit perception of food, environmental dangers, kin and potential mates 5,6 . A few exceptional GRs have either internal chemosensing or other sensory functions [7][8][9] . In the presence of their cognate ligands, these receptors induce neuronal firing, although their signalling mechanism is poorly understood. ORs are thought to function as odour-gated ion channels 10,11 , formed by a complex of an odour-specific OR and a co-receptor, ORCO 12 . GRs are less well characterized molecularly: although they are also likely to function ionotropically 13 , it remains unclear whether they function alone or in complexes of co-expressed GR subunits 14 .
Understanding the molecular mechanism by which GRs and ORs act has been hampered by their lack of sequence similarity to either known ion channels or other types of membrane protein.
The receptors are predicted to contain seven transmembrane domains (TMDs), similar to G protein-coupled receptors (GPCRs). Their membrane topology is, however, inverted compared with GPCRs, defining them as a distinct class of transmembrane protein 12,15,16 . Thus, despite the widespread function for insect chemosensory receptors in mediating perception of the external chemical world, the genesis of this protein family is unknown.
In this work, we have used comparative genomics to characterize the evolutionary origins and dynamics of this family of receptors, and experimental analyses of GR homologues in species distant from insects to reveal an unexpected role for these proteins during embryonic development.

Results
Deep conservation of Grl genes. We performed TBLASTN and PSI-BLAST searches using insect GRs and ORs as queries to identify putative homologous genes within non-insect genomes available at NCBI and other public genome and Expressed Sequence Tag (EST) databases (see Methods) ( Fig. 1a, Supplementary Table 1 and Supplementary Data 1). Within Protostomia, these searches recovered the previously described GRs of the crustacean Daphnia pulex (water flea) 17 , and the gustatory related (gur) and serpentine receptor class R (srr) genes of Caenorhabditis elegans 1,18,19 and other nematodes. Beyond Ecdysozoa, we identified what we term GR-like (Grl) genes in Lophotrochozoa, including the annelids Capitella teleta (polychaete worm) and Helobdella robusta (freshwater leech) 20 , and the molluscs, Lottia gigantea (owl limpet) 20 , Crassostrea gigas (Pacific oyster) 21 and Aplysia californica (sea slug). We also found homologous genes in Deuterostomia, including the hemichordate Saccoglossus kowalevskii (acorn worm), and several echinoderms, including the sea urchin S. purpuratus 22 , but not in any chordate (Fig. 1a). Finally, we identified a small number of Grl genes in non-bilaterian species, including Cnidaria, such as the sea anemone N. vectensis 23 (independently noted in another study 24 ), the corals Acropora digitifera 25 and Acropora millepora, and the placozoan Trichoplax adhaerens 26 . However, no sequences were recovered from the genomes of another cnidarian, Hydra magnipapillata, the ctenophores Mnemiopsis leidyi 27 and Pleurobrachia bachei 28 , or the sponges Amphimedon queenslandica 29 and Oscarella carmela 30 . No fungal, unicellular eukaryotic or prokaryotic genomes contained identifiable Grls (Fig. 1a).
Several observations support the classification of the identified genes as Grls. First, reciprocal BLAST with each Grl sequence against insect genomes identified an insect GR as the most significant hit. Second, the encoded proteins each contain multiple predicted TMDs-most commonly 6-7 TMDs in putative full-length Grl sequences-similar to GRs and ORs (Fig. 1b). (Accurate computational predictions of TMD number and topology are difficult because of the variable hydrophobicity of TMDs within this family of receptors 12 ). Third, the phase and position of three inferred ancestral introns in the 3 0 region of Drosophila melanogaster GR transcripts 1 are well-conserved in Grls (Fig. 1c). Finally, Grls share a number of characteristic amino acid residues with the most highly conserved C-terminal region of the insect receptors, including a tyrosine residue important for ion channel function in insect ORs 31 (Fig. 1c).
The overall sequence similarity of Grls and insect GRs (or ORs) is very low (B15%-20% amino acid identity, similar to the minimum identity within the GR and OR families themselves 1 ), with the highest similarity to the GR64a-f/GR61a clade of sugar-sensing receptors 14 . Phylogenetic analysis revealed no unambiguous orthologous relationships of Grls with any insect GR, or between Grls of different lophotrochozoan, deuterostome or non-bilaterian species (Fig. 2). Although the high divergence of these sequences precludes accurate reconstruction of their ancestry, many members of these repertoires appear to have formed by species-specific expansions (Fig. 2).
Early developmental expression of Grl genes in N. vectensis. To investigate the role of Grls in non-Arthropoda, we focused in this study on the cnidarian N. vectensis (here abbreviated Nvec), which is amenable to gene expression and functional analyses. Moreover, the development and anatomy of this organism are relatively well-described, and a presumed chemosensory structure, the apical organ, has been identified ( Fig. 3a) 32,33 . N. vectensis development proceeds through a coeloblastula stage to gastrulation, which occurs mainly by invagination 34,35 . After gastrulation, the embryo emerges from the egg jelly as a free-swimming planula, which moves by ciliary beating with the apical sense organ pointing forward. Subsequently, the planula transforms ('metamorphoses') into a sessile polyp, which starts feeding after formation of the tentacles that surround the single opening of the animal (Fig. 3a).
Surprisingly, analysis of the temporal pattern of expression of the two Grls we found in this species, NvecGrl1 and NvecGrl2, by quantitative reverse transcriptase-PCR (qPCR) revealed the highest level of expression of these genes in early developmental stages ( Fig. 3b and Supplementary Table 2): NvecGrl1 is expressed in the egg and blastula, gastrula and early planula stages, diminishing rapidly to very low levels later in development (Supplementary Table 2). NvecGrl2 is only robustly detected in the egg and blastula. These temporal expression profiles contrast markedly with those of D. melanogaster GRs and ORs, where expression of these receptors has been reported only in differentiated sensory neurons 5 .
We next examined the spatial distribution of N. vectensis Grl transcripts by whole-mount RNA in situ hybridization on a range of developmental stages. Although NvecGrl2 RNA was undetectable in these experiments ( Supplementary Fig. 1), we first observed NvecGrl1 transcripts in a broad aboral domain at the gastrula stage (Fig. 3c). This domain develops under the control of genes homologous to those that regulate anterior (or apical) development in bilaterians 36 , and its centre is the site of formation of the apical sensory organ in the planula larva (Fig. 3a). Genes with similarly broad aboral expression at the gastrula stage display different expression dynamics after gastrulation: 'spot genes' (for example, NvecFgfa1, NvecFgfa2 and NvecSoxB1) become restricted to a small spot at the aboral pole, from which the apical organ develops, whereas 'ring genes' (for example, NvecSix3/6 and NvecFoxQ2a) remain expressed in a ring-like domain around the apical organ 36 . Comparison of the spatial distribution of NvecGrl1 with NvecFGFa1 by double RNA in situ hybridization revealed that the NvecGrl1 expression domain remains broad at late gastrula/early planula stages when NvecFgfa1 has already restricted to a small aboral spot (Fig. 3d). Subsequently, NvecGrl1 transcripts are excluded from the apical organ-forming spot (Fig. 3e), similar to other 'ring genes'. Although the 'ring gene' domain includes some scattered sensory cells 37,38 , the broad and early expression of NvecGrl1 in the aboral domain indicates that this receptor may primarily have a non-chemosensory role in this species.
Regulation of NvecGrl1 by the aboral patterning network. Several transcription factors and signalling molecules that are   expressed in the aboral region and required for apical organ formation in N. vectensis have been identified 36 . For example, the homeodomain transcription factor NvecSix3/6 is an important determinant of the aboral territory, where it initiates an autoregulatory loop of fibroblast growth factor (FGF) signalling to pattern this domain 36 . To determine whether NvecGrl1 aboral expression is regulated by these factors, we performed RNA in situ hybridization on embryos injected with morpholino oligonucleotides (morphants) designed to block translation of these genes. In NvecSix3/6 morphants, NvecGrl1 aboral expression is reduced or absent in gastrula embryos compared with control morphants, similar to other markers of this    NvecFgfa1 and NvecFgfa2 participate in positive and negative feedback regulation of apical organ formation, respectively 33,36 : NvecFgfa1 morphants fail to develop an apical organ 33 , whereas NvecFgfa2 morphants display precocious formation of an expanded apical organ 36 . However, although effects of NvecFgfa1 knockdown become visible only after gastrulation, an expansion and/or upregulation of apical organ domain genes in NvecFgfa2 knockdown animals is detectable already at gastrula stage 36 . Consistently, we found that the expression of NvecGrl1 appears only slightly affected in NvecFgfa1 morphants at the end of gastrulation (Fig. 4b). By contrast, in NvecFgfa2 morphants, NvecGrl1 expression forms an expanded ring with a wider area of low NvecGrl1 expression in its centre and this ring-like domain appears earlier than in control animals ( Fig. 4c compared with  Fig. 3e). These data provide further evidence that NvecGrl1 is expressed in cells that do not directly give rise to the apical sensory organ.
Grl1 is required for aboral patterning in N. vectensis. To investigate the function of NvecGrl1, we injected zygotes with morpholino oligonucleotides designed to block translation of this receptor, and examined development of these animals in parallel with those injected with control morpholinos (see Methods). Early developmental processes, including gastrulation, appeared to proceed normally in both NvecGrl1 and control morphant animals (Fig. 5a). However, NvecGrl1 morphants displayed pronounced morphological defects in the apical organ domain in planula larvae (Fig. 5a,b). In control animals, the first sign of apical organ formation is the translocation of the nuclei in a small group of cells at the aboral pole to a more basal position (Fig. 5a).
These cells subsequently form a small indentation of the aboral ectoderm (Fig. 5a); this indentation is the site from which the long cilia of the apical organ emerge 37 . In NvecGrl1 morphants, the initial basal movement of nuclei is absent or strongly reduced, and no clear indentation develops (Fig. 5a). Correspondingly, the apical tuft is absent or consists of only few long cilia in 4-day-old NvecGrl1 morphants (Fig. 5b). These animals subsequently fail to develop into primary polyps (0%, n ¼ 159 NvecGrl1 morphants, compared with control morphants 63%, n ¼ 122). These defects were reproduced with a second morpholino against NvecGrl1 (o9% primary polyps, n ¼ 109).
To understand the basis for these morphological defects, we analysed expression of aboral domain patterning genes in NvecGrl1 morphants. NvecSix3/6 expression is greatly diminished-but not abolished-in NvecGrl1 morphants at the planula stage compared with control morphants (Fig. 5c). Consistent with the reduction of NvecSix3/6 expression, two of its downstream targets, the transcription factor NvecFoxQ2a and NvecFgfa2 (ref. 36), are also reduced in NvecGrl1 morphants compared with controls (Fig. 5c). Defects in gene expression were also detectable at gastrula stages ( Supplementary Fig. 2a), when NvecGrl1 transcripts are first detected in the aboral domain, although these phenotypes were very subtle, though consistent, for NvecSix3/6 and NvecFoxQ2a. Together, these observations implicate NvecGrl1 in a feedback pathway that maintains NvecSix3/6 expression to pattern the aboral domain. By contrast, patterning genes expressed elsewhere in the embryo, for example, NvecWnt2, which is found in a belt-like domain between the aboral and oral poles 39 , are not or at most mildly affected in NvecGrl1 morphants (Fig. 5c).
One of the important functions of the FGF signalling loop is to downregulate NvecSix3/6 expression after gastrulation in a small 'spot', the developing apical organ domain. Within this spot,  genes required for apical organ formation, such as the transcription factor NvecHoxF, are expressed 36 . Defects in the initiation of FGF signalling leads to a failure in formation of this NvecSix3/6and NvecFoxQ2a-negative and NvecHoxF-positive aboral spot 36 .
We observed a similar phenotype in NvecGrl1, but not control, morphants ( Fig. 5c and Supplementary Fig. 2b); these defects could account for failure of apical organ development. Together, these results support a role for NvecGrl1 in the regulatory network that controls aboral domain patterning and thereby apical sensory organ formation. and Alexa Fluor 488 Phalloidin, respectively. The arrow marks the group of nuclei at the aboral pole that have translocated to a more basal position (towards the right of the image) in control but not NvecGrl1 morphants; this results in these cells forming a small indentation of the apical ectoderm (arrowhead). Scale bar, 100 mm (applies to all images). Quantification of the phenotypes is shown below. (b) Bright-field images of living 4-day planulae derived from N. vectensis embryos injected with either control (control#1) or NvecGrl1 (NvecGrl1#2) morpholino oligonucleotides revealing the failure in apical tuft develops in NvecGrl1 morphants. Quantification of the phenotypes is shown below. Scale bars, 100 mm. (c) RNA in situ hybridization using probes against marker genes in animals injected with either control (control#1) or NvecGrl1 (NvecGrl1#1) morpholino oligonucleotides. Scale bar, 100 mm (applies to all images). Quantification of the phenotypes is shown below.  Developmental expression of a deuterostome Grl gene. To determine whether the early developmental expression of Grl genes is conserved in Deuterostomia, we used qPCR to analyse transcript levels of the five Grls of the echinoderm S. purpuratus during embryonic stages (Fig. 6a). Notably, SpurGrl1, which is the most similar sea urchin Grl to NvecGrl1 (Fig. 2), was detected during early development ( Fig. 6b and Supplementary Table 3): the highest expression of SpurGrl1 was at the blastula stage, but it remained at elevated levels in the gastrula and later embryonic developmental stages. These qPCR data are in agreement with quantitative developmental transcriptome data ( Supplementary  Fig. 3a) 40 . The other four SpurGrls were expressed at only very low levels during developmental stages (Fig. 6b), but transcripts could be readily amplified for all SpurGrls, except SpurGrl3, from adult neural tissue (Supplementary Fig. 3b). These observations suggest that SpurGrl1 has a unique function during embryonic development.

Control morpholino
By whole-mount RNA in situ hybridization, we observed that SpurGrl1 exhibits a dynamic spatial expression pattern during early development (Fig. 6c). In the blastula SpurGrl1 RNA was present ubiquitously, while in later stages it was restricted to a few groups of cells. From the late gastrula, transcripts were detected in cells in the apical domain, which develops into a neurosensory structure 41 . Strong SpurGrl1 expression was also detected in two bilaterally arrayed cells in the oral ectoderm and, at lower levels, in the hindgut and adjacent ectoderm, as well as a region of the aboral ectoderm that will develop into the vertex (Fig. 6c). The apical domain expression of SpurGrl1 represents an intriguing parallel with that of NvecGrl1 in the developmentally equivalent region of the sea anemone gastrula 36,42,43 , although the additional expression patterns suggest that the sea urchin gene has several functions.

Discussion
We have defined an early evolutionary origin and broad phylogenetic distribution of the insect chemosensory receptor superfamily. Previously only characterized in Ecdysozoa 2,17,44,45 , we show Grl genes are present in genomes across Protostomia, non-chordate Deuterostomia and non-Bilateria. These observations establish the origin of this gene family before the split of Bilateria and non-Bilateria 550-850 million years ago 46 , and suggest that these genes have been lost in the ancestral chordate.
Through analysis of Grls in the non-bilaterian N. vectensis, we found-unexpectedly-several lines of evidence supporting a role in developmental body patterning rather than external chemosensation. First, the existence of a small number of NvecGrls is inconsistent with this cnidarian receptor repertoire playing a wide-ranging role in the detection of environmental chemical signals. We note that N. vectensis has several dozen vertebrate OR-like GPCRs, which represent candidate chemosensory receptors of this species 47 . Second, in contrast to Drosophila ORs and GRs 48 , both NvecGrls are expressed in early developmental stages, before formation of sensory neurons and a putative chemosensory organ (at late gastrula and mid-planula stage, respectively 37,38 ). Moreover, NvecGrl1 transcripts are detected broadly around the aboral pole in gastrula and planula stages, but are selectively excluded from an aboral 'spot' of cells that subsequently gives rise to the apical sensory organ. Third, functional analysis of NvecGrl1 supports a role in aboral body patterning essential for developmental progression. Although our data argue that NvecGrls have a different function to insect GRs, we cannot exclude that they have a later, additional role, for example, in the detection of metamorphosis-related sensory signals.
Although loss of NvecGrl1 blocks developmental progression beyond the planula stage, the relatively limited current knowledge of aboral patterning in Cnidaria makes it difficult to define the precise role of this protein. Our data are consistent with a model in which NvecGrl1 functions downstream of the aboral territory determinant NvecSix3/6 in the feedback loop that is important to refine aboral pole patterning and promote apical organ formation 36 . Secondary structure and sequence conservation of NvecGrl1 (and other Grls) with insect GRs and ORs suggest that this protein acts as a ligand-gated ion channel and might recognize a paracrine developmental signal; testing this model will require identification of a relevant ligand. Nevertheless, it is intriguing that NvecGrl1 plays a role in the formation (albeit indirectly) rather than the function of a putative chemosensory organ. We have not been able detect transcripts for NvecGrl2 gene in situ and morpholino-mediated knockdown of NvecGrl2 did not reveal any obvious patterning defects (unpublished data), suggesting that this gene has a different role to NvecGrl1.
The similarities in the early developmental expression of the deuterostome Grl gene with highest sequence identity to NvecGrl1, S. purpuratus Grl1, hints that a role in embryonic patterning may have been an ancestral function of this family. However, SpurGrl1 displays additional spatial expression patterns beyond the apical domain and four additional SpurGrls do not show such early expression, suggesting that Grls in this species have multiple different roles. Cleanly dissecting the function of SpurGrl1 (and its paralogues) will require methods that permit tissue-specific gene knockdown.
Beyond the work presented here, very little is known about the role of members of the GR/Grl family outside insects. The large size and high sequence divergence of the GR repertoire in the crustacean D. pulex is consistent with a role in chemosensation 17 , although no expression or functional analyses of these receptors have been reported. The nematode C. elegans relies on a very large family of GPCRs for chemical detection 18 , and of the 15 Grl genes (the gur and srr families 1,18,19 , the latter not previously defined as members of this superfamily) only two have been characterized. GUR-2 (also known as LITE-1) functions in photosensitive neurons 44,45 , although it has not been determined whether this protein senses light directly. GUR-1 (also known as EGL-47) appears to regulate motoneurons that control egg laying 49 . Together with our data in N. vectensis and S. purpuratus, these observations suggest that Grl/GR genes may have been recruited only in arthropods to fulfil a predominant role in peripheral chemosensation. Future analysis of other Grls in a range of species should yield insights into the diverse functions and evolutionary history of this unusual receptor family.

Methods
Gene identification. To identify homologues of insect chemoreceptors, all known Drosophila sequences of GRs and ORs were used as queries in extensive TBLASTN and PSI-BLAST searches of public genome and EST data sets (see Supplementary  Table 1 and Supplementary Data 1), and in screens of OrthoDB 50 . Identified genes were used in iterative BLAST searches of species-specific data sets until no new sequences were identified. The numbers of genes in most species are likely to represent minimum estimates due to the challenge of identifying highly divergent Grl genes from incompletely assembled genomes with computational predictions alone. For the same reasons, most predicted sequences are incomplete, particularly at the more divergent 5 0 end. Analysis of the membrane topology (excluding partial protein sequences) was performed using TMHMM v.2.0 (ref. 51). Intron positions and phases were predicted by comparing cloned (or predicted) transcript sequences and the corresponding genomic sequences.
Alignment/phylogenetic tree building. Protein alignments were built using Clustal Omega 52 with the following settings-max HMM Iterations: 5; combined iterations: 5; max guide iterations: 5. The amino termini of the sequences were cut off after aligning using Jalview 53 and only the carboxy-terminal regions from TMD5 (starting with position 352 in DmelORCO) was used for further analysis. Tree reconstruction was performed using PhyML3.0 (ref. 54) and the LG model 55 NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7243 ARTICLE was used as a substitution matrix. Branch support was calculated using the approximate likelihood-ratio test method 56 . The tree was rooted on the Rhodopsin branch, although we acknowledge that the evolutionary relationship between these receptor families is still unclear.
Cloning of full-length Grl coding sequences. To obtain the complete coding sequence of NvecGrl1, PCR was performed on mixed embryo stage complementary DNA, with specific primers based on EST and genome data (genome.jgi-psf.org/ Nemve1/Nemve1.home.html). For NvecGrl2 5 0 and 3 0 rapid amplification of cDNA end (RACE) PCRs were performed. The following primer sets were used: NvecGrl1. 5 0 -GTGTGAGCACGCTGCGAGAATGG-3 0 3 0 -TCACTTAAACTGGACAAGAATCG-5 0 NvecGrl2. 5 0 RACE outer primer: 5 0 -AGAGCAGAAGGCTCACAAGC-3 0 5 0 RACE inner primer: 5 0 -CAAGGAATTTCATTGCGACAG-3 0 3 0 RACE outer primer: 5 0 -CGATTGGAGGAATTGTCCCAGTAACACGCGG-3 0 3 0 RACE inner primer: 5 0 -TCAACACAGAAGTGATCAGCCCAACGATTGG-3 0 Genbank accession numbers are: KP294348 (NvecGrl1) and KP294349 (NvecGrl2). For SpurGrl1, a 1,212 bp fragment of the coding sequence was amplified by PCR from 18 hpf embryonic cDNA, cloned into pGEM-T Easy (Promega) and sequenced, using the following primers designed on the WHL22.291934 transcriptome sequence: Quantitative PCR. For N. vectensis, cell lysates of embryonic stages were collected and total RNA was extracted using the RNAqueous kit (Ambion). cDNA synthesis was performed using the SuperScript II RT Reagent Kit, according to the manufacturer's protocol. Expression of NvecGrl1, NvecGrl2 and reference genes were analysed by performing PCR on an Applied Biosystems 7900HT SDS, using the cDNA as template and SYBR Premix (Roche) with the following primers: For S. purpuratus, cell lysates of appropriately staged embryos were collected and total RNA was extracted using the RNeasy Micro Kit (Qiagen). cDNA synthesis was performed using the iScript cDNA Synthesis Kit (Bio-Rad), according to the manufacturer's instructions. At least four technical replicas were run on two independent cDNA batches, using cDNA equivalent to 1 embryo per reaction (2.8 ng) and Power SYBR Green QPCR master mix (Applied Biosystems), on the 7900HT Fast Real-Time PCR system and the following primers (SpurUbq was used as internal standard 57,58 ): The two morpholinos against NvecGrl1 are partially overlapping to avoid 5 0 -untranslated region sequence that is present elsewhere in the genome. The sequence corresponding to the start codon in the NvecGrl1 morpholinos is underlined.
Histology. N. vectensis. For actin and DNA staining, specimens were fixed in 4% formaldehyde in PBT (0.1% Triton X-100 in PBS) for 1 h and washed four times in PBT. 30 ml of Alexa Fluor 488 Phalloidin (Molecular Probes) were put in an open Eppendorf tube for 30 min at room temperature, to let the solvent evaporate. The pellet was redissolved in 1.5 ml PBT and 1.5 ml TO-PRO-3 Iodide (Molecular Probes) was added. The specimens were incubated for 2-3 h in this staining solution, washed five times with PBT and mounted in ProLong Gold (Life Technologies). RNA in situ hybridization on N. vectensis was performed using antisense digoxygenin-labelled probes against the open reading frame (ORF) of the desired genes essentially as described 60 . In brief, hybridization was carried out at 60°C for 60 h, and post-hybridization stringency washes were with 0.01 Â SSCT. Antidigoxygenin Fab fragments (Roche) were diluted 1:5,000 before detection by NBT/ BCIP staining (Roche). For N. vectensis double in situ hybridization, digoxygeninlabelled NvecGrl1 and fluorescein isothiocyanate-labelled NvecFgfa1 probes complementary to the corresponding ORF were used. NvecFgfa1 was detected by NBT/BCIP staining and NvecGrl1 was detected using anti-DIG-POD and tyramide-based signal amplification (TSA plus DNP (AP) kit, Perkin Elmer), followed by anti-DNP-AP-mediated BCIP staining. S. purpuratus. Specimens were fixed in 4% formaldehyde in MOPS buffer overnight at 4°C, then washed several times in MOPS buffer and once in 70% ethanol. After rehydration in TBST, RNA in situ hybridization was performed using an antisense digoxygenin-labelled probe against the 1,212 bp fragment of the SpurGrl1 ORF (see above) as described 61 . In brief, 0.05 ng ml À 1 of probe was hybridized at 60°C for 16 h. Post-hybridization stringency washes were with 0.1 Â SSC and 0.01 Â SSC. Anti-digoxygenin Fab fragments were diluted 1:2,000 before detection by NBT/BCIP staining.