Animal husbandry

A C. teleta colony was maintained in the laboratory according to published culturing methods [31]. Embryos and larvae were recovered as previously described [32].

Cloning and sequencing of C. teleta runt, Pax3/7, and eve orthologs

Fragments corresponding to conserved regions of runt, Pax3/7 and eve orthologs were isolated from C. teleta by degenerate polymerase chain reaction (PCR) using a cDNA template prepared from mixed embryonic and larval stages. To recover a fragment of runt, two rounds of amplification were performed in a semi-nested PCR reaction using the following primers from published sequences [33]: runfw-1: 5'- RCNRYNATGAARAAYCARGTNGC -3' and runbw: 5'-CKNGGYTCNCKNGGNCCRTC-3' followed by runfw-2: 5'-MRNTTYAAYGAYYTNMGNTTYGTNGG -3' and runbw. To isolate a fragment of Pax3/7 from mixed stage cDNA, a semi-nested approach was utilized with the following published primers [34] representing the conserved paired domain and paired-like homeodomain: Prbyfout: 5'- GGNGGNGTNTTYATHAAYGG -3' and Prbyr: 5'-RTTNSWRAACCANACYTG -3' followed by Prbyfin: 5'-MARATHGTNGARATGGC -3' and Prbyr in a second round of amplification. A 226 base pair (bp) fragment of eve was recovered from a genomic DNA template using the following degenerate primers: ab-evefw: 5'-MGTTAYMGTACIGCITTYAC-3' and eve-bw1: 5'-CKYTGNCKYTTRTCYTTCAT-3' [33]. Additional sequences for Ct-eve, Ct-Pax3/7 and Ct-runt were obtained from a mixed stage embryonic and larval stage template using gene-specific primers with the Smart rapid amplification of cDNA ends (RACE) amplification kit (Clontech, Mountain View, CA, USA). Resulting fragments were cloned into the pGEM-T Easy vector (Promega, Fitchburg, WI, USA) and sequenced by the University of Hawaii sequencing facility (Honolulu, HI, USA) or by Macrogen Inc. (Seoul, South Korea). The resulting products were: a 702 bp 5' RACE fragment and a 1269 bp 3' RACE fragment for Ct-Pax3/7, a 644 bp 5' RACE fragment and a 945 bp 3' RACE fragment for Ct-runt. For eve, a 460 bp 5' RACE fragment was recovered by PCR using gene-specific primers. Sequencing of multiple cDNA clones from 5' RACE reactions revealed the presence of a five amino acid insertion in some clones. Two distinct 3' RACE fragments were recovered: a 583 bp and a 734 bp fragment. Gene-specific primers designed to each of the 5' and 3' eve sequence variants were used to amplify and characterize features of individual cDNAs. Subsequent to the isolation of Ct-runt, Ct-Pax3/7, Ct-eve1 and Ct-eve2 by degenerate PCR, sequences became available for the C. teleta genome project (Department of Energy, Joint Genome Institute, Walnut Creek, CA, USA [35]). Cloned sequences were consistent with predicted gene models from the genome. The genome was searched for additional runt, Pax3/7 and eve paralogs; none were detected. Linkage analysis of Ct-eve1 and Ct-eve2 genes was performed through searches of the C. teleta genome using nucleotide sequences obtained from degenerate and RACE PCR clones. Accession numbers for Ct-runt, Ct-eve1, Ct-eve2 and Ct-Pax3/7 are listed in Additional file 1: Table S1.

Sequence alignments and phylogenetic analyses

Sequences isolated by degenerate PCR and RACE were assigned putative orthologies based on BLASTX searches of the GenBank database from the National Center for Biotechnology Information. In addition, tBLASTn searches of the C. teleta genome (Joint Genome Institute (JGI), Walnut Creek, CA, USA) were conducted to find all homologs of eve, runt and the paired homeobox (pax) family. Two putative orthologs were found for eve, one for runt and six for pax family members. Amino acid sequences from diverse animal taxa were downloaded from the protein database in GenBank. Alignments of conserved domains were generated with ClustalX, using default parameters in MacVector 11.1.1 (MacVector, Inc., Cary, NC, USA). The domains analyzed were the homeodomain and 10 amino acids flanking the 5'- and 3'-ends of the homeodomain for Eve, the runt domain for Runt and the 127 amino acid paired domain for the Pax family. Alignments were edited by hand to correct obvious alignment errors. Nexus alignments are available upon request. Sequences, abbreviations and accession numbers used are included in Additional file 1: Table S1.

ProtTest v2.4 [36] was run for each alignment to determine the appropriate model of protein evolution. The Jones model was recommended and used for the Eve and the Pax family, and the RtRev model for Runt. Both Bayesian and maximum likelihood methods were used for each gene family. Bayesian analysis was conducted using MrBayes v3.1.2 [37], with the model determined by ProtTest. For the Runt and Pax alignments, three million generations were run, sampled every 100 generations, with four independent runs and four chains. Ten million generations were run for the Eve alignment. Once convergence was reached, majority rule consensus trees were generated with burnin values of 25,000 (Eve), 9,700 (Runt) and 5,400 (Pax). Maximum likelihood analyses were performed with RAxML v7.0.0 [38] using the same models as for Bayesian analysis with 1,000 bootstrap replicates. Trees were visualized with FigTree v1.3.1 [39] and drawn using Adobe Illustrator version CS4.

Whole mount in situ hybridization

Whole mount in situ hybridization was performed according to previously published protocols [26]. Digoxigenin-labeled riboprobes (dig-11-UTP; Roche Diagnostics, Indianapolis, IN, USA) were transcribed with the MEGAscript High Yield Transcription Kit (Ambion, Austin, TX, USA) and diluted to working concentrations of 3 ng/μL for Ct-runt, 0.25 ng/μL for Ct-Pax3/7, 0.1 ng/μL for Ct-eve1 and 0.2 ng/μL for Ct-eve2. Probe lengths were as follows: 1041 bp for Ct-runt, 1270 bp for Ct-Pax3/7, 1379 bp for Ct-eve1 and 1224 bp for Ct-eve2. Following termination of the development reaction, specimens were equilibrated and stored in 80% glycerol, 20% 5× phosphate buffered saline (PBS). The detailed protocol is available upon request.

Visualization of mid-body mesoderm

Animals were fixed and labeled according to previously described conditions (fixation #4, [40]). Specimens were exposed to a 1:500 dilution of mouse anti-histone antibody (F152, C25.WJJ; Millipore, Billerica, MA, USA) and a 1:400 dilution of goat anti-mouse rhodamine (Invitrogen, Carlsbad, CA, USA). After antibody labeling, animals were incubated in 1:100 BODIPY FL phallacidin (Invitrogen) in PBS, 0.1% Triton X-100 (PBT) for two hours at room temperature followed by several PBT washes over 45 minutes. Animals were mounted in Slow Fade Gold (Invitrogen) and analyzed.

Microscopy

Riboprobe-labeled specimens were analyzed using differential interference contrast (DIC) optics on an Axioskop 2 compound microscope (Zeiss, Munich, Germany). Digital images were captured with either a stem-mounted 4.0 megapixel Nikon Coolpix 4500 (Nikon, Inc., Melville, NY, USA) or SpotFlex digital camera (Diagnostic Instruments, Inc., Sterling Heights, MI, USA). Confocal laser scanning microscopy was performed using an Axioplan 2 LSM510 microscope (Zeiss). Three-dimensional reconstructions were generated from confocal images with ImageJ (National Institutes of Health, Bethesda, MD, USA). Figures and diagrams were constructed using Photoshop CS4 and Illustrator CS4 (Adobe Systems Inc., New York, NY, USA). Some panels in figures seven and eight (see Figure legends) are composites of multiple DIC focal planes that were merged using Helicon Focus (Helicon Soft Ltd., Kharkov, Ukraine).

Cloning and phylogenetic analyses of C. teleta runt, eve and Pax3/7 genes

Using degenerate and RACE PCR, we isolated fragments comprising 1585 bp of a C. teleta runt ortholog, designated Ct-runt. The sequence contained a 157 bp 5' untranslated region (UTR), a 570 bp open-reading frame (ORF) predicting a protein of 190 amino acids and an 857 bp 3' UTR and poly-A tail. The predicted ORF included a conserved Runt DNA binding domain of 128 amino acids. When compared to Runt sequences from other organisms, the Ct-runt predicted ORF showed strong conservation of amino acid residues throughout the Runt DNA binding domain (Figure 1A). Phylogenetic analyses of the Runt domain from Ct-Runt were performed using both Bayesian and maximum likelihood methods. Ct-Runt groups most closely with the Lozenge protein from the parasitic flatworm Schistosoma mansoni, another lophotrochozoan, and these two sequences fall within a group that contains ecdysozoan, lophotrochozoan and non-vertebrate deuterostome Runt sequences (Additional file 2: Figure S1). The vertebrate and cephalochordate Runt and Runx sequences form a well-supported separate group, and cnidarian sequences fall outside of both groups.

The C. teleta runt ortholog we recovered lacks the characteristic motif VWRPY at the C-terminal end, which is necessary for Groucho-mediated transcriptional repression activity [41]. To confirm that the missing motif was not due to a cloning artifact, 3' RACE PCR was repeated with a second primer set and an independently generated cDNA pool. The resulting fragment was identical in sequence to the original Ct-runt 3' RACE product. In addition, we identified a single runt sequence from searches of the C. teleta genome [35], which has the same amino acid sequence as the sequence isolated by PCR and lacks the C-terminal Groucho-binding domain. Searches of the Lottia gigantea genome [42] reveal that its predicted runt homolog also lacks the VWRPY C-terminal motif (Protein ID 119594).

A paired (Pax3/7) ortholog, Ct-Pax3/7, was amplified by degenerate and RACE PCR from a mixed stage cDNA template. This sequence was composed of a 99 bp 5' UTR, a 1536 bp ORF predicting a protein of 416 amino acids and a 175 bp 3' UTR that includes a poly-A tail. Pax3/7 proteins are defined by the presence of three conserved domains: a paired domain and 'paired' class homeodomain, both of which are DNA binding domains, and a short octapeptide motif, which is typically located in the linker region between the paired domain and the homeodomain. This octapeptide motif has been shown to act as a site for Groucho-mediated repression [43]. The predicted Ct-Pax3/7 ORF from the C. teleta cDNA contains a putative paired domain and a 'paired' class homeodomain (Figure 1B) but lacks the octapeptide motif in the linker region. To confirm our placement of Ct-Pax3/7 within the Pax3/7 family, we identified other putative Pax homologs in the C. teleta genome and ran phylogenetic analyses using the paired domain. Several distinct Pax families are broadly distributed across bilaterians and include: Pax1/9, Pax2/5/8, Pax3/7, Pax4/6 and Pox neuro [44]. More recently, a Paxβ family has been described, a novel Pax subfamily only found in lophotrochozoans [45]. C. teleta has six Pax family members, with a single representative in each of the Pax families Pox neuro, Paxβ, Pax1/9, Pax2/5/8, Pax3/7 and Pax4/6 (Figure 2). Both Bayesian posterior probability and maximum likelihood bootstrap show strong support for the distinct Pax gene groups. Ct-Pax3/7 groups most closely with the two Pax3/7 sequences from another annelid, the leech Helobdella sp. Austin. These lophotrochozoan Pax3/7 sequences group separately from the deuterostome Pax3 and Pax7 clade and the arthropod homologs Gooseberry, Pairberry and Paired.

A 226 bp fragment of eve was initially identified by degenerate PCR from a C. teleta genomic DNA template. Subsequent RACE PCR recovered a 5' fragment and two distinct 3' fragments. From sequence searches of the C. teleta genome, two distinct eve genes were identified, which we designate Ct-eve1 and Ct-eve2. The Ct-eve1 and eve2 predicted ORFs are 325 and 330 amino acids, respectively. The Ct-Eve1 and Ct-Eve2 predicted ORFs are largely identical within the predicted ORF; the exceptions are the two amino acids immediately 5' of the stop codon (KS for Ct-eve1 and PK for Ct-eve2), and a small five amino acid insertion present only in Ct-Eve2. This insertion immediately follows amino acid residue 85 and is located 17 amino acids 5' of the homeodomain. The homeodomain in both Eve sequences contains conserved amino acids characteristic of eve gene sequences (Figure 1C). Bayesian and maximum likelihood analyses of the homeodomain confirm the identity of Ct-eve1 and Ct-eve2 as eve gene orthologs within the homeodomain superfamily (Additional file 3: Figure S2). Ct-Eve1 and Ct-Eve2 cluster together with Eve proteins from other animal taxa and separately from the homeodomain-containing Gbx and Engrailed outgroups. There is 100% Bayesian posterior probability and 90% bootstrap support for the Eve node. Within the predicted ORFs, there are only seven nucleotide differences between Ct-eve1 and Ct-eve2, aside from sequence differences due to the five amino acid insertion in Ct-Eve2 and the two amino acids immediately 5' of the stop codon. In contrast, the two sequences diverge in sequence and length in the 3' UTR (Ct-eve1, 167 bp and Ct-eve2, 349 bp).

Both Ct-eve1 and Ct-eve2 are located on scaffold 262, approximately 18.7 kbp apart, and these two genes exhibit opposite transcriptional orientations (Figure 3). For both genes, the predicted ORFs are distributed among three exons, and the positions of the intron-exon boundaries are shared between the two genes. In contrast, the size and sequence composition of the first intron are different between Ct-eve1 (371 bp) and Ct-eve2 (652 bp). The second intron is the same in length and sequence in the two genes and is located within the homeodomain, between homeodomain positions 46 and 47. The position of the second intron within the homeodomain is conserved in the two amphioxus and mouse eve gene homologs and the single Tribolium castaneum eve gene [46, 47], and is different from the position of the single intron in the D. melanogaster eve gene, which is upstream of the homeodomain [48]. Exon 3 for both genes also contains 191 amino acids of the ORFs, as well as the gene-specific 3' UTRs. Comparisons of the 3 kbp of 5' and 3' sequence flanking the two ORFs lack regions of sequence similarity between Ct-eve1 and Ct-eve2. The C. teleta eve genes are not clustered with the Hox genes in the genome [49], although it remains a possibility that they are all on the same chromosome. In P. dumerilii, eve is on the same chromosome as the posterior class Hox gene Post2, but is not part of the Hox cluster [50], and T. castaneum eve is also not clustered with the Hox genes [47]. In contrast, eve orthologs in vertebrates and amphioxus, and in the cnidarian Nematostella vectensis, are clustered with the Hox genes [51–53].

C. teleta development

C. teleta development has been described previously and a standard staging system is used to distinguish among embryonic and larval stages (Figure 4 top) [32]. Following unequal spiral cleavage (Stages 1 to 2) and gastrulation (Stage 3), two trochal bands, the prototroch and telotroch, penetrate through the egg membrane to initiate the larval phase (Stages 4 to 9). The prototroch and telotroch delineate, respectively, the anterior and posterior boundaries of the segmented mid-body. The mouth is positioned immediately posterior of the prototroch, on the ventral side of the larva. The first four to five segments initially appear during early larval stages (Stage 5), and segmental furrows are visible on the ventral face of the larva [28, 32]. The first ten segments form progressively from anterior to posterior within a 24 hour time frame, and three additional larval segments form from a posterior growth zone (PGZ) by late larval stages (Stage 9). The final three larval segments form at a rate of approximately one per day. Following metamorphosis, 40 to 50 additional posterior segments are generated from the PGZ.

The segmented ectoderm and mesoderm have distinct embryonic origins in C. teleta[54]. All segmental ectoderm is generated from the blastomere 2d, which also generates nonsegmental posterior ectoderm. The segmented mid-body ectoderm segregates from nonsegmental ectoderm within the first few divisions of 2d, such that 2d¹¹ generates segmented ectoderm and PGZ, but not the telotroch or the posterior unsegmented pygidium [54]. Most of the mid-body mesoderm arises from the two cells 3c and 3d, whose descendants form the right and left mesodermal bands, respectively (Figure 4). Initially, a pair of large posterior mesodermal cells is visible beneath the ectoderm (Figure 4A,D). Subsequently, two mesodermal bands become visible as a row of subsurface cells in the lateral part of the body (Figure 4B,C,E,F). Over time, the bands increase in cell number and expand circumferentially (Figure 4G-L). By stage 5, when the first segments are visible, longitudinal and circular muscle fibers extend from the mesodermal bands (Figure 4G). Extension of a single circular muscle fiber per segment is one of the earliest signs of segmentation in the mesoderm and follows soon after ectodermal segmentation. These circular muscle fibers are positioned within the intersegmental grooves (Figure 4I) [32].

Ct-runt expression

Ct-runt is expressed in several distinct ectodermal tissues, predominantly in the nervous system, foregut and hindgut, but is never detected in ectodermal stripes of expression in the segmented part of the body (Figure 5). Most Ct-runt expression is transient, lasting only one to two days. Initially, Ct-runt expression is localized to two small domains in the brain in Stage 5 larvae (Figure 5A), and is not detected in Stage 3 and Stage 4 larvae. The first segments appear at Stage 5, and Ct-runt is not expressed in nascent segments or in the presumptive segmental tissue at this stage. At Stage 6, expression in the brain appears as several discrete cell clusters (Figure 5B,E). There is also expression in the anterior four to five segments in two columns of cells (Figure 5C). The more lateral runt-expressing cells are centrally positioned within the segment (Figure 5D). The medial cells are clearly in the forming ventral nerve cord. Because the lateral boundary of the ventral nerve cord is not yet distinct at this stage, it is unclear if the lateral cells (short arrows) are within or lateral to the ventral nerve cord. At Stage 7, several new domains of Ct-runt expression appear, most notably in the foregut. At late Stage 7, discrete sub-domains of the foregut become morphologically obvious, and there is expression of Ct-runt in the pharynx (Figure 5J). By Stage 8, foregut expression is no longer detectable. Expression also appears in the ventrolateral ectoderm of the PGZ during Stage 7 (Figure 5G,I,J), which persists to Stage 8 (Figure 5K). At Stage 7, there is also transient expression in a small lateral mesodermal domain, posterior to the mouth (Figure 5H). In the nervous system, expression in the ventral nerve cord expands to more posterior segments and is prominent in a single row of segmentally repeated cells (Figure 5F). Brain expression ceases, although there remain Ct-runt-positive ectodermal cells in the head, which are likely sensory cells (Figure 5H-J). In late stage larvae, expression is limited to a ring in the ectoderm of the hindgut (Figure 5L). In some cases we also observed weak expression in the PGZ (not shown), which we interpret to be residual transcript from expression during earlier stages.

Ct-Pax3/7 expression

Ct-Pax3/7 transcripts are detected at all developmental stages examined (Stage 3 to 9 and in juveniles) in two ventrolateral ectodermal bands oriented along the anterior-posterior axis. Following gastrulation at Stage 3, Ct-Pax3/7 is expressed in two small lateral domains in the posterior half of the embryo (Figure 6A). Slightly later, this expression expands anteriorly to form two bands of ectodermal expression (Figure 6B,C). This expression appears prior to the formation of the larval segments, prototroch, telotroch and the ventral nerve cord. The lateral bands of Ct-Pax3/7 expression continue to expand anteriorly and by Stage 4 they extend from the level of the mouth to the telotroch (Figure 6E,F). The longitudinal bands of Ct-Pax3/7 expression are the only detectable expression domains at every stage examined, including immediately prior to and during the appearance of the first segments (Figure 6C-H). The relative intensity of expression varies somewhat along the length of the band (Figure 6G). At Stage 6, expression in anterior segments decreases (Figure 6J), and gradually decreases from anterior to posterior. The posterior boundary of expression is stable from Stages 5 to 9; it abuts, but does not include the telotroch. Between Stage 4 and 6, the distance between the two bands of expression decreases (compare Figure 6G and 6 J), reflecting the movement of the two sides of the ventral nerve cord towards the ventral midline. By late Stage 7, Ct-Pax3/7 expression is limited to a small ventrolateral domain of ectoderm in the PGZ (Figure 6K). This expression is positioned immediately lateral to the ventral ganglia (Figure 6L), and persists through Stage 8 (Figure 6M). The Ct-Pax3/7 transcript is no longer detectable in Stage 9 larvae. In post-metamorphic juveniles, Ct-Pax3/7 expression is limited to a small ventrolateral domain of ectoderm that abuts the lateral edge of the ventral ganglia in the nascent segment and also extends into the growth zone (Figure 6N,O,P). Thus, in juveniles, Ct-Pax3/7 is expressed in a restricted area of ectoderm in the PGZ.

Expression of eve genes

Although the two C. teleta eve genes have close to 100% amino acid identity within their ORFs (see previous section), each gene has a unique expression pattern. Both genes are detected at all developmental stages examined in multiple tissues and multiple germ layers (Stages 3 to 9). However, neither Ct-eve1 nor Ct-eve2 is detected in ectodermal or mesodermal stripes in the mid-body, whether preceding or following the appearance of segments (Figures 7 and 8). Ct-eve1 is broadly expressed during cleavage stages (data not shown); here, we focus on spatial characterization of the Ct-eve1 transcript from the end of gastrulation (Stage 3) through metamorphosis (Stage 9). During gastrulation, when a blastopore is present (Stage 3b), Ct-eve1 is detected in two domains: in the mesodermal precursor cells (Figure 7A) and the dorsal side of the embryo, in the presumptive brain (Figure 7B). Soon after, Ct-eve1 is transiently expressed in the endoderm (Figure 7C). In addition, expression persists in the nascent mesodermal bands and in the forming brain (Figure 7C). During Stage 3, Ct-eve1 also is expressed in a fourth domain in a posterior ring of ectodermal cells that extend to the dorsal surface of the embryo (Figure 7D). Expression associated with the presumptive foregut and hindgut appears at Stage 4 (Figure 7E). The Ct-eve1 transcript is broadly expressed throughout the mid-body mesoderm as it expands circumferentially through Stage 5 and 6 (Figure 7F,G,I). Expression also persists in the mesodermal precursor cells, which at this stage are positioned beneath the telotroch (Figure 4C and 7I, large arrows). During this time, expression in the brain, foregut and hindgut persists from previous stages (Figure 7G,H,I). In late larval stages, there is expression in the posterior mid-body mesoderm, which gradually becomes restricted to the PGZ by Stage 9 (Figure 7J,K,L), and expression in other domains is either very weak or not detectable. In juvenile stages, there is a continuous domain of expression that includes the ectoderm and mesoderm of the PGZ and the nascent segments (Figure 9A,C). Ct-eve1 is also expressed in posterior ganglia of juveniles.

Ct-eve2 has distinct expression domains from Ct-eve1, although there are clear overlapping areas of expression. When the blastopore is present (Stage 3b), Ct-eve2 is detected in the mesodermal precursor cells (Figure 8A) and on the posterior face of the blastopore at the ventral midline (Figure 8B). Later during Stage 3, there is continued expression in the mesodermal precursor cells, and novel expression in the presumptive hindgut (Figure 8C) and in a ring of ectodermal cells near the posterior end of the embryo (Figure 8D). At Stages 4 and 5, after the telotroch has formed, it is apparent that these posterior ectodermal cells are immediately anterior to the telotroch in the position of the presumptive PGZ (not shown). Ct-eve2 expression in the mesodermal precursor cells and hindgut persists (Figure 8E,G). At Stage 5, Ct-eve2 also appears in several additional ectodermal domains, including weak expression in the foregut and mid-body ectoderm (between the prototroch and telotroch). The most prominent expression is in the forming ventral nerve cord (Figure 8F), which persists as the left and right sides fuse at the ventral midline at Stage 6 (Figure 8H). In the mid-body, there is expression in both the ectoderm and mesoderm at Stage 6. The mesodermal expression is particularly pronounced in the PGZ and in the mesodermal precursor cells (Figure 8I). In the ectoderm, small clusters of cells express Ct-eve2 more strongly relative to surrounding cells (Figure 8H arrowheads), which become refined to three columns of discrete, segmentally iterated cell clusters on each side of the animal by Stage 7 (Figure 8 J,K,L). The three columns are in the following positions: between the ventral nerve cord and the neuropodial chaetae, between the neuro- and notopodial chaetae, and dorsal to the notopodial chaetae (Figure 8J). The most dorsal column is the first to express Ct-eve2. These three columns of Ct-eve2-expressing cell clusters span the ectoderm epithelia and have a very similar expression pattern to elav[40] and synaptotagmin (Meyer and Seaver, unpublished data), conserved markers of neuronal differentiation and of differentiated neurons, respectively. These columns of cell clusters are likely peripheral sensory structures. In addition, at Stages 6 and 7, there is ectodermal expression in a single line of cells across the dorsal surface of some larvae in the PGZ (Figure 8L). This expression domain is not visible at later stages. By Stage 8, there is only weak expression in the columns of ectodermal cell clusters (not shown) and they are undetectable by Stage 9. Expression persists in the hindgut and PGZ into Stage 8 (Figure 8M,O,P), and at this stage it is clear from the subsurface position of the hindgut expression that Ct-eve2 is localized to the rectum. A new expression domain appears at Stage 8 in mesodermal cells along the dorsal midline in the posterior mid-body (Figure 8M,N,O). These Ct-eve2-expressing cells can be seen extending from the dorsal midline along the anterior edge of segments (Figure 8N). It is unknown to what structures these cells contribute. At Stage 9, expression is limited to a small domain in the rectum (Figure 8Q) and weak expression in the PGZ (Figure 8Q) and dorsal midline (not shown). In juveniles, Ct-eve2 is expressed in the PGZ, hindgut and weakly in posterior ganglia (Figure 9B,D).

In summary, Ct-eve1 and Ct-eve2 have unique spatio-temporal expression patterns, although they are both expressed in mesodermal precursor cells, foregut and hindgut, and the PGZ. Prior to the generation of the mesodermal bands, Ct-eve1 and Ct-eve2 are similarly expressed in the mesodermal precursor cells (compare Figures 7A and 8A). Once the mesodermal bands appear, however, Ct-eve1 is expressed in both mesodermal precursor cells and their descendant bands, whereas Ct-eve2 is restricted to the mesodermal precursor cells and is not expressed in their progeny. The temporal dynamics of hindgut expression between the two genes is distinct. Ct-eve2 is continuously expressed in the hindgut from Stage 3 through metamorphosis into juvenile stages (Figures 8 and 9). In contrast, Ct-eve1 expression is transient in the hindgut, is initially detected a day later than for Ct-eve2 at Stage 4, and is not detectable in late stage larvae (compare Figures 7L and 8Q). Another difference between Ct-eve1 and Ct-eve2 expression is in the mid-body, where Ct-eve2 is largely expressed in ectodermal derivatives such as the ventral nerve cord and peripheral sensory structures. This differs dramatically from Ct-eve1, which has broad expression throughout the mid-body mesoderm. The unique expression patterns for Ct-eve1 and Ct-eve2 indicates a divergence of regulatory elements between these two genes.

Runt, Pax3/7 and eve genes in C. teleta

The C. teleta genome has clear orthologs of the D. melanogaster pair-rule segmentation genes runt, paired (Pax3/7) and eve. Two eve genes are present in the C. teleta genome and, based on amino acid sequence similarity and genomic linkage, they appear to be either the result of a recent duplication or concerted evolution, possibly due to gene conversion event(s) [55, 56]. Gene conversion has previously been proposed to occur between the two Dlx genes in the C. teleta genome [57]. Both Ct-Pax3/7 and Ct-runt lack a Groucho-binding domain, which is conserved in orthologs of these genes from other species. However, the Pax3/7A ortholog in the leech Helobdella sp. [58] and the sea squirt Ciona intestinalis[59], and the D. melanogaster paired gene [60] also lack this motif, as does the predicted limpet Lottia gigantea runt homolog. The C. teleta genome contains a groucho gene (scaffold_692 1858), and other C. teleta genes contain the groucho repression motif, including the hes genes [29]. These features confirm that the groucho repression motif is present and likely utilized in the context of transcriptional co-repression in C. teleta. Although we do not know the functional consequences of the lack of a groucho domain on the C. teleta runt and paired homologs, the lack of a groucho domain in D. melanogaster paired suggests that this domain is not necessary for a segmentation function.

Relationship of Ct-runt, Ct-Pax3/7, Ct-eve1 and Ct-eve2 expression to segment formation

If annelids and arthropods shared a common segmented ancestor, the expectation is that pair-rule gene orthologs would be expressed in transverse stripes with a consistent relationship to segment boundaries in C. teleta, with either a pair-rule or segmental periodicity. However, none of the genes we examined, Ct-runt, Ct-Pax3/7, Ct-eve1 or Ct-eve2, have a spatio-temporal expression pattern consistent with a possible role in segment formation. When the first larval segments are forming, Ct-runt expression is limited to a subset of cells in the brain and there is no expression in the mid-body where new segments form. Although Ct-Pax3/7 is expressed in the mid-body prior to overt segment formation in larvae, its spatial pattern does not prefigure segments, either in a pair-rule or segmental pattern. Both Ct-eve genes are expressed in the mid-body as larval segments begin to form. At these stages, the Ct-eve1 transcript is ubiquitously expressed in the mid-body mesoderm and lacks spatially restricted positional expression expected for a role segment boundary formation. Ct-eve2 expression has more restricted expression in the mid-body, limited to mesodermal precursor cells and the ventral nerve cord, and also lacks pair-rule or segmental stripes of expression in the ectoderm or mesoderm. At late larval stages and during juvenile growth when segments form from the PGZ, all four genes are expressed in the PGZ. However, none of them have patterns that prefigure segments, such as circumferential banded expression or expression limited to either the anterior or posterior region of the growth zone. Ct-Pax3/7 and Ct-runt have restricted expression around the circumference of the PGZ, in a small ventrolateral domain and ventrolateral domain of the ectoderm, respectively. Both Ct-eve genes have broad posterior expression domains that include the PGZ and nascent segments in late larval stages and juveniles, but do not resolve into more restricted patterns. In some animals that form segments sequentially, such as in vertebrates, centipedes and spiders, segment formation is preceded by waves of dynamic gene expression [61–63]. Of the genes examined here, we did not observe dynamic expression in the PGZ, since we did not detect multiple patterns in the PGZ for a gene in similarly staged larvae.

We previously examined expression of orthologs of the pair-rule genes odd-paired and hairy in C. teleta[29, 30]. During early stages of segment formation, Ct-zic, the C. teleta ortholog of odd-paired, is expressed along the lateral edge of the ventral neural ectoderm, and in segmentally iterated domains of mesoderm expression in the mid-body soon after segments form. However, there are no stripes of expression in either the ectoderm or mesoderm. One of the hairy orthologs, CapI-hes1, is expressed in the mesoderm of presumptive larval segments in a segmentally iterated pattern, although it is very distinct from arthropod patterns of hairy expression [7, 64, 65]. Of the pair-rule orthologs for which we have examined the expression pattern in C. teleta, CapI-hes1 is the most promising candidate to function in segmentation [29]. Ct-Pax3/7, Ct-runt, Ct-eve1, Ct-eve2, CapI-hes2 and Ct-zic all have very distinct expression patterns, all of which indicate that these genes are not likely to be involved in segmentation. Perhaps a subset of these genes are post-transcriptionally regulated and play a role in establishing segmental boundaries; however, functional studies will be necessary to obtain a definitive confirmation of the role of these pair-rule homologs in C. teleta.

Comparison of runt, Pax3/7 and eve expression among species

Expression of the pair-rule genes Pax3/7 and eve have been characterized in a few other annelids, but expression patterns for runt have not been reported for any other lophotrochozoan. Runt orthologs appear to have a conserved role in D. melanogaster[66] and in Danio rerio[67], chick [68] and mammalian neural development (reviewed in [69]), and expression of Ct-runt in the brain and ventral nerve cord of C. teleta extends this conservation. For Pax3/7 orthologs, expression has been examined in two other annelids. Expression of Pax3/7 in P. dumerilii is somewhat similar to the pattern in C. teleta[70]. Both Pdu-pax3/7 and Ct-Pax3/7 are expressed in bilaterally symmetric longitudinal bands in the ventral trunk ectoderm, adjacent to the ventral nerve cord. In P. dumerilii, each band of expression is much broader than in C. teleta, possibly reflecting differences in the size of the neurogenic field at these stages. Helobdella sp. Austin has two Pax3/7 orthologs. The expression pattern has been reported for Hro-Pax3/7A, and it is expressed in the mesodermal bands and later in mesodermal derivatives [58]. This is in contrast to the longitudinal bands of Pax3/7 expression in the ventral ectoderm in C. teleta and P. dumerilii. A better understanding of the extent to which these genes have conserved expression domains in annelids will emerge as data from additional species are reported. Outside of annelids, Pax3/7 orthologs function in neural specification in the ventral nerve cord of D. melanogaster and the vertebrate neural tube. However, their expression patterns are quite distinct in that there are transverse bands of expression in D. melanogaster and longitudinal bands of expression in vertebrates [71, 72].

Comparison of eve expression patterns across bilaterian animals reveals several common expression domains. In annelids, eve is expressed in the growth zone during posterior growth of C. teleta and H. robusta, and in the PGZ during regeneration of P. dumerilii[73, 74]. However, in P. dumerilii, expression in the PGZ is ectodermal, whereas in H. robusta and C. teleta it is both ectodermal and mesodermal. Furthermore, there is prominent expression in the mesodermal bands of C. teleta and H. robusta, but not in P. dumerilii. Additionally, eve is expressed in the posterior gut of both C. teleta and P. dumerilii. Outside of annelids, posterior expression of eve has been reported in Danio rerio[75], Xenopus laevis[76], B. floridae[46], nematodes [77], and in a range of arthropods [10, 15, 17, 78–80]. Eve also appears to have an important and widespread role in neural specification or neurogenesis in bilaterians, including in insects (D. melanogaster[48, 81], Schistocerca americana[10]), Caenorhabditis elegans[82], B. floridae[46], X. laevis[83] and Mus musculus[84]. In annelids, eve is expressed in the developing ventral nerve cord of C. teleta, P. dumerilii and H. robusta, and a role in neural development appears to be the most conserved function of eve. Although less well characterized, eve has also been implicated in mesodermal patterning in Danio rerio[85] and D. melanogaster[48]. The expression of eve in pericardial progenitors along the dorsal midline in D. melanogaster and in mesodermal cells along the dorsal midline of C. teleta may indicate a shared role in protostome cardiovascular development. In summary, eve genes show conserved roles across bilaterians in multiple domains, indicating conservation of several developmental roles, and the C. teleta eve gene expression patterns appear to show conservation with several of these functions. This broad conservation for eve is in contrast with its function in arthropod segmentation, which appears to be more of a clade-specific role.