Expression and function of an even-skipped homolog in the leech Helobdella robusta

The evolutionary origin(s) of segmentation in bilaterally symmetric animals is a topic of interest and controversy (Davis and Patel, 1999). Until relatively recently, a mainstream view of this topic was that segmentation arose independently at two points in evolution: once in the deuterostome ancestor of the chordates, and once in the protostome ancestor of annelids, arthropods and onychophorans. Paradoxically, recent discoveries in comparative development and in molecular phylogeny have been interpreted as supporting two different and mutually contradictory scenarios. On one hand, the discovery that vertebrate homologs of Drosophila segmentation genes are expressed in segmentally iterated patterns has led some to propose that the last common ancestor of protostomes and deuterostomes was already segmented (De Robertis, 1997; Holland et al., 1997; Kimmel, 1996). On the other hand, molecular phylogenies now organize most or all bilaterians into three clades, Deuterostomia, Ecdysozoa and Lophotrochozoa (Aguinaldo et al., 1997; Ruiz-Trillo et al., 1999), that have been evolving separately since before the Cambrian radiation (Adoutte et al., 2000). As each of these clades contains only a minority of segmented taxa (Brusca and Brusca, 1990), it is more parsimonious to conjecture that segmentation evolved independently in each of the three clades.

Analyzing the mechanisms of segmentation in glossiphoniid leeches such as Helobdella robusta may contribute to resolving this paradox by either of two routes. For example, finding that segmentation is homologous between leeches (a highly derived but experimentally tractable taxon of segmented lophotrochozoans) and insects (a highly derived but experimentally tractable taxon of segmented ecdysozoans) would suggest that the urprotostome was already segmented, and would be consistent with the hypothesis that the urbilaterian was already segmented as well. Conversely, concluding that segmentation is not homologous between representatives of the two clades of protostomes would support the hypothesis that segmentation arose multiple times.

The primary pair-rule gene even-skipped (eve) is among the first genes to exhibit regular, spatially iterated patterns in Drosophila (Frasch et al., 1987; Macdonald et al., 1986). Homologs of eve have also been described in arthropods, such as the beetle Tribolium castaneum and the grasshopper Schistocerca americana, and the spider Cupiennius salei, that form segments sequentially. Tribolium-eve and Cupiennius-eve are expressed in striped patterns that correlate with segmentation (Brown et al., 1997; Damen et al., 2000; Patel, 1994), while Schistocerca-eve is expressed in the posterior growth zone without forming stripes (Patel et al., 1992). Neuronal expression of eve-class genes is conserved throughout the arthropods (Damen et al., 2000; Duman-Scheel and Patel, 1999; Frasch et al., 1987; Patel et al., 1992).

We report the identification of eve-class genes from two species of glossiphoniid leeches and the initial characterization of Hro-eve, the eve homolog in Helobdella robusta. Semi-quantitative RT-PCR revealed that the highest levels of Hro-eve transcription occur during organogenesis. In situ hybridization revealed earlier expression, during cleavage and the early phases of segmentation, but we found no evidence for a pair-rule pattern. In early development, Hro-eve transcripts seem to be associated with mitotic chromatin in the stem cells of the posterior growth zone (teloblasts) and in their progeny (primary blast cells) that are the founder cells for the five distinct lineages of segmental mesoderm and ectoderm. In more advanced embryos, Hro-eve is expressed in segmentally iterated subsets of the neurons that arise from the N teloblasts. To assess the function of Hro-eve, we examined embryos in which selected blastomeres had been injected with antisense Hro-eve morpholino oligonucleotides (AS-Hro-eve MO). Focussing on the N lineage, we found that teloblasts injected with AS-Hro-eve MO continued to divide at about the normal rate, but other aspects of their division were disturbed, and primary blast cell divisions were also disrupted. Presumably because of these early effects on teloblasts and blast cells, segmentation was also perturbed in the injected lineage, as evidenced by misalignment of left and right hemiganglia, fusion of serially adjacent hemiganglia and/or missing groups of cells. Ganglionic neurons expressing Hro-eve arose, though in an abnormal pattern; serotonergic neurons failed to form. In the O lineage, injection of AS-Hro-eve MO disrupted both neural and epidermal cell fates. Our results suggest that Hro-eve is important in regulating cell divisions in stem cells and segmental founder cells in early development and that it also regulates the differentiation of a subset of ganglionic neurons. However, we find no evidence that Hro-eve plays a pair rule function similar to that of eve-class genes in arthropod segmentation.

Embryos of Helobdella robusta were obtained from a laboratory colony and those of Theromyzon trizonare from specimens collected in the ponds of Golden Gate Park, San Francisco, and cultured at 23°C in HL (Helobdella) saline (Blair and Weisblat, 1984). The embryonic staging system and cell nomenclature are as reviewed elsewhere (Fig. 1) (Weisblat and Huang, 2001).

To mark specific cell lines, cells were pressure injected with rhodamine-conjugated dextran amine (RDA, Molecular Probes, catalog number D-1827) or fluorescein-conjugated dextran amine (FDA, Molecular Probes, catalog number D-1820) at a final concentration of 75 mg/ml in 0.2 N KCl with 1% Fast Green as described previously (Smith and Weisblat, 1994).

To perturb Hro-eve expression, cells of interest were injected with an antisense morpholino oligonucleotide (MO; Genetools) complementary to the downstream end of the 5′ UTR and first five bases of coding sequence of Hro-eve, designated as AS-Hro-eve MO (5′-ATCATTTTACTTTTCGATTCAGCGG-3′; anti-start codon underlined). For control injections, we used a generic MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) or MM-Hro-eve MO (5′-ATCtTTTaACTTTTCGATaCAGgGG-3′) that had the same overall nucleotide composition as AS-Hro-eve MO, but with four mismatched nucleotides (Nasevicius and Ekker, 2000).

MO concentrations ranged from 0.06-1.3 mM in the micropipette and the volume of material injected corresponds to roughly 1% of the teloblast (Bissen and Weisblat, 1987). Thus, we estimate that the MO concentrations ranged from 0.6-13 μM in the cytoplasm of the injected cells. MO was co-injected with lineage tracer (∼50 mg/ml in the pipette) and Fast Green (∼0.5% in the pipette).

Degenerate oligonucleotides (upstream, 5′-MGIYAYMGIACIGCITT-3′; downstream, 5′-YGIYAYYTTRTCYTTCAT-3′) corresponding to nucleotides 7-24 and 108-125 of the eve-class homeobox were designed by comparing the sequences for eve-class genes from zebrafish (Joly et al., 1993), fly (Macdonald et al., 1986), mouse (Bastian and Gruss, 1990), human (Faiella et al., 1991), coral (Miles and Miller, 1992), grasshopper (Patel et al., 1992), nematode (Ahringer, 1996) and frog (Ruiz i Altaba and Melton, 1989). Candidate gene fragments were amplified from a cDNA library (Stratagene) of stage 7-10 embryos by degenerate PCR.

To obtain additional sequence of Hro-eve, we performed PCR on the cDNA library and on first strand cDNAs. For the first strand cDNA, ∼500 embryos from stages 6-10 were homogenized in RNAwiz (Ambion) and total RNA was extracted. Polyadenylated mRNAs were isolated using Oligotex mRNA mini kit (Qiagen). For rapid amplification of the cDNA ends, the Marathon cDNA amplification kit (Clontech) was used, with exact gene specific oligos designed from the gene fragments described above. Sequence analyses were performed by the MacDNASIS pro v3.5 program and the BLAST program (http://www.ncbi.nlm.nih.gov).

At each stage, total RNAs were extracted from 50 embryos and treated with DNase (DNA-free, Ambion), and then reverse-transcribed using random decamers (Ambion). As an internal standard to adjust for differences in efficiency of RNA extraction between samples, a 488 bp fragment of 18S rRNA was amplified in parallel to each sample. To quantitate PCR products, each sample was electrophoresed in 2% agarose gel and stained with ethidium bromide. Band intensity was measured with an Alphaimager (Alpha Innotech) using Alphaease (v3.3b) program.

To confirm the relative levels of expression, we amplified two separate regions of Hro-eve cDNA: nucleotides 784-1243 (which spans an intron site, to control for the possibility of genomic DNA contamination of the template) and nucleotides 1724-1939. Both produced equivalent results, and the identity of the amplified fragments was confirmed by sequencing one sample of each fragment.

Digoxigenin (Dig-11-UTP, Roche)-labeled riboprobes were made in vitro using MEGAscript kit (Ambion) and hydrolyzed. T7 RNA polymerase (Ambion, cat No. 1334) was used for all probes. Sequence for Hro-eve riboprobes includes 3′ coding region and most of the 3′-UTR (nucleotides 1211-3140).

To localize Hro-eve mRNA in situ, two different protocols were used, depending on the age of the embryo. Embryos at stages 9-11 were processed as described previously (Harland, 1991; Nardelli-Haefliger and Shankland, 1992). For early stages (through stage 8), embryos were fixed in 4% formaldehyde, 1×phosphate-buffered saline (PBS, diluted from 10× PBS stock, pH 7.4) for 1 hour at room temperature. After fixation, all incubations were performed at room temperature with constant rocking. Fixed embryos were rinsed with 0.1% PBTw (PBS, 0.1% Tween 20), devitellinized and rinsed in serial changes of 0.1% PBTw, 0.1% PBTw:HYB (1:1) and hybridization solution [HYB; 5×SSC, tRNA 0.5 mg/ml, heparin 50 μg/ml, 0.1% Tween 20, 50% formamide, pH 6.0 (adjusted with 1 M citric acid)], then pre-hybridized in HYB for 2 hours at 68°C. HYB was replaced with fresh HYB containing riboprobes and hybridized at 68°C for 28-40 hours. Embryos were washed with pre-warmed HYB:2×SSC/0.3% chaps (1:1), and then 2×SSC/0.3% Chaps, followed by successive 20 minute washes with 2×, 0.2× and 0.1× SSC/0.3% Chaps at 68°C. Embryos were incubated in blocking solution [0.1% PBTw, 2% sheep serum (Sigma, St Louis, MO), 2 mg/ml bovine serum albumin] for 2 hours at room temperature and further incubated in blocking solution containing anti-digoxigenin alkaline phosphatase-conjugated polyclonal Fab fragments (1:5000, Roche) at 4°C overnight. Embryos were rinsed with frequent changes of 0.1% PBTw for 5 hours at room temperature, followed by a brief rinse with alkaline phosphatase buffer (AP buffer; 100 mM Tris-HCl, pH 9.5, 100 mM NaCl, 50 mM MgCl₂, 0.1% Tween 20). Embryos were then incubated in AP buffer containing 4-nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indoyl-phosphate (Roche) for color reaction. When the embryos reached the desired color, they were rinsed with double distilled H₂O and dehydrated in a graded ethanol series, then cleared in benzyl benzoate:benzyl alcohol (3:2) for observation.

Embryos at stage 11 were fixed in 4% formaldehyde in 0.75×PBS overnight at 4°C and washed in 1×Hepes-buffered saline (HBS, diluted from 2×stock), then incubated for 24 hours in HBS containing 0.5 units/ml chitinase (Sigma C-1525, St Louis, MO) at RT with gentle shaking. Embryos were rinsed in HBS and incubated in a solution of PBS, 1% Triton X-100 and 10% NGS (PTN) overnight at 4°C, then further incubated in PTN containing rabbit anti-serotonin (1:300, Sigma S-5545, St. Louis, MO) overnight at 4°C. Embryos were rinsed in PTN and incubated in PTN containing peroxidase-conjugated goat anti-rabbit (1:300, Jackson ImmunoResearch Labs 111-036-003, West Grove, PA) overnight at 4°C. Embryos were rinsed in PTN, then in PBS and incubated in 0.75×PBS containing 0.5 mg/ml 3, 3′-diaminobenzidine (Sigma) and 0.008% NiCl₂ for 15 minutes at room temperature. The color reaction was initiated by adding 0.06% H₂O₂ and monitored visually. For observation, germinal plates were dissected and mounted in ∼75% glycerol.

For more detailed analyses, selected embryos were sectioned prior to microscopic examination. For this purpose, embryos labeled with lineage tracer only were dehydrated and embedded in glycol methacrylate resin (JB-4, Polysciences) according to the manufacturer’s instructions. This resin dissolves the colored in situ hybridization reaction product. Thus, in situ-stained embryos were embedded in an epoxide resin (Polybed 812, Polysciences). Embedded embryos were sectioned at ∼10 μm thickness (MT2-B Ultramicrotome, Sorvall) and mounted on glass slides using a non-fluorescent media (Fluoromount, BDH Laboratory Supplies, England). Intact, dissected or sectioned embryos were examined and photographed with DIC and/or epifluorescence optics (Zeiss Axiophot and 35 mm film camera or Nikon E800 and Princeton Instruments cooled CCD camera controlled by UIC Metamorph software) or by confocal microscopy (BioRad MRC 1024). Images were processed, montaged and composed digitally (Metamorph, UIC; Photoshop 5.0, Adobe).

Using degenerate PCR, we amplified fragments of the homeobox of eve-class genes from Helobdella robusta and Theromyzon trizonare, then used 5′- and 3′-RACE on the cDNA library and first strand cDNAs to obtain additional sequence for these genes, designated Hro-eve (Accession Number, AF409098) and Ttr-eve (Accession Number, AY050275), respectively. The homeodomains of Hro-eve and Ttr-eve show 73-91% amino acid identity with those of other eve-class genes (Fig. 2); the closest non-eve-class homeodomain is from Drosophila proboscipedia, with 63% identity. Thus, we are confident that these genes are bona fide eve-class genes.

The ∼1.4 kb of Ttr-eve cDNA sequence encodes a polypeptide of at least 421 amino acids. There is no in frame stop codon in the 96 bp of 5′ sequence prior to the first methionine in the ORF and no polyadenylation signal in the 59 bp of 3′-UTR. The sequence was confirmed by comparison with genomic DNA, which revealed a 269 bp intron within the homeodomain, following amino acid residue 46 of the homeodomain. This intron site is conserved in mouse, human, frog, nematode and Tribolium (Fig. 2) (Ahringer, 1996; Bastian and Gruss, 1990; Brown et al., 1997; Faiella et al., 1991; Ruiz i Altaba and Melton, 1989).

For Hro-eve, ∼3.1 kb of cDNA was obtained, encoding an ORF of at least 761 amino acids, taking the 5′-most in frame AUG as the start codon. There was no in-frame stop codon in the 171 bp upstream of the candidate start codon; three possible polyadenylation sites occurred in the 676 bp of 3′-UTR. In addition to the homeodomain, the Hro-eve polypeptide contains a polyalanine stretch and another region rich in serine and proline residues, both of which are common features in other eve-class genes (Ahringer, 1996; Bastian and Gruss, 1990; Brown et al., 1997; Faiella et al., 1991; Macdonald et al., 1986; Patel et al., 1992; Ruiz i Altaba and Melton, 1989). Comparison with genomic DNA revealed one 257 bp intron at the same site within the homeodomain as in other eve-class genes and another, 100 bp intron following amino acid residue 51 of the putative polypeptide.

To compare Ttr-eve and Hro-eve to other eve-class genes, cladograms were constructed [PAUP 4.0 b4a (PPC)] using only homeodomains. Both Ttr-eve and Hro-eve clearly fell within a strongly supported clade of eve-class genes (98% bootstrap value), but there was no well-supported resolution within this clade (data not shown but available upon request). Ttr-eve and Hro-eve showed less amino acid identity than expected for orthologous genes from closely related species (Fig. 2) and were highly diverged outside the homeodomain. These results could reflect either an ancient duplication of the protostome eve-class gene or the rapid divergence of the Ttr-eve; we cannot distinguish between these possibilities. In any event, for the following analyses of gene expression and function, we focussed exclusively on Hro-eve.

In Helobdella, each segment arises from the interdigitating clones of seven distinct classes of bilaterally paired segmental founder cells (m, nf, ns, o, p, qf and qs blast cells) (reviewed by Weisblat and Huang, 2001) (Fig. 1). Blast cells arise in columns (bandlets) by unequal divisions from five bilateral pairs of large identified stem cells: the M, N, O, P and Q teloblasts. Teloblasts divide with a cell cycle time of about 1 hour. Each lineage contributes a stereotyped set of neurons and other cell types to each segment. In the N (and Q) lineages, two classes of blast cells, nf and ns (qf and qs) arise in exact alternation; one of each is required to generate one segment’s worth of progeny (Fig. 1). In the M, O and P lineages, each blast cell makes one segment’s worth of progeny. In each teloblast lineage, the first-born blast cells contribute to anterior segments and blast cells born later contribute to progressively more posterior segments.

Thus, in leech, in contrast to vertebrates and insects, there is a strict correlation between the ‘cell cycle clock’, by which blast cells arise from the teloblasts, and the ‘segmentation clock’, by which segmental tissues arise in anteroposterior progression. From this description, what might one expect to see if Hro-eve was expressed in a pair-rule pattern? One possibility was that Hro-eve would be expressed in alternate blast cells, or blast cell clones, in the M, O and P lineages, and in alternate pairs of blasts cells in the N and Q lineages. Another possibility was that it would be expressed in alternate cells within the N and Q lineages, for example, in all nf and qf cells.

As a first step in characterizing the expression of Hro-eve, we used semi-quantitative RT-PCR to estimate the relative levels of expression of Hro-eve during development (Fig. 3). By this technique, Hro-eve transcripts were present in embryos at stage 7, which roughly corresponds to the onset of blast cell production. But in contrast to predictions based on pair-rule expression patterns, in situ hybridization of embryos at stages 7-8 revealed Hro-eve mRNA in all primary blast cells and in the micromere cap (Fig. 4A). The level of staining was uniform among different lineages, and among individual blast cells within each bandlet (Fig. 4A,B). Thus, we found no evidence for either lineage-specific or pair rule expression of Hro-eve.

In these experiments, some teloblasts in each embryo exhibited punctate staining that resembled chromatin morphology during mitosis (Fig. 4). These features suggested that Hro-eve transcripts had colocalized with the chromatin of mitotic teloblasts. We were unable to obtain nuclear counterstaining under the conditions for in situ hybridization process, but punctate staining also occurred within bandlets and germinal bands at the specific positions where primary blast cells undergo mitosis (Bissen and Weisblat, 1987; Zackson, 1984). This pattern of chromatin staining was most easily seen in the mesodermal (m) bandlet because m blast cells undergo their first two divisions prior to entering the germinal bands (Fig. 5A,B).

The specificity of this chromatin staining was confirmed by control experiments. Experiments using both sense probe for Hro-eve and no probe gave no staining (data not shown). By contrast, in situ staining for other Helobdella robusta genes gave distinct patterns: Hro-nos (a nanos homolog) (Pilon and Weisblat, 1997; Kang et al., 2002) gave very faint, diffuse staining; Hro-cycA (a cyclinA homolog) (Chen and Bissen, 1997) gave strong cytoplasmic staining in teloblasts and blast cells (data not shown). Thus, we conclude that the Hro-eve in situ signal represented an association of Hro-eve transcripts with the chromatin of teloblasts and blast cells during mitosis. We did not examine all the stages during cleavage, but similar staining was seen as early as the two-cell stage (Fig. 5C), indicating that Hro-eve is expressed throughout cleavage at levels beneath the sensitivity of our RT-PCR protocol.

RT-PCR experiments indicated that Hro-eve expression increased sharply beginning in stage 9, by which time the germinal plate is complete and the differentiation of ventral ganglia and other segmental tissues is under way. In situ hybridization using embryos at stages 9-11 revealed Hro-eve expression in segmentally iterated sets of neurons in the ventral nerve cord (Fig. 6). The Hro-eve transcripts in these cells were cytoplasmic rather than nuclear (Fig. 6).

Two lines of evidence showed that neuronal Hro-eve expression was dynamic over time, and that it undergoes the same progression in each of the midbody segments. First, within individual embryos, there were differences in staining pattern along the AP axis that correlate with the sequential segmentation process in leech (Fig. 6A). Second, the same changes in Hro-eve expression were seen by examining any given region of the germinal plate at progressively later stages of development (Fig. 6B-D). Thus, the following description applies to any midbody ganglion and is presented in terms of the clonal age of the n blast cells that contribute to the ganglion, as neurons expressing Hro-eve arise from the nf and ns blast cells (see below).

Neuronal expression of Hro-eve was first observed when n blast cell clones were ∼70 hours old. This corresponds to early stage 9 in the anterior germinal plate. Expression was seen in two spots, one anterior and another more posterior, in each hemiganglion (Fig. 6B). The anterior spot consisted of two or three adjacent cells in the ventrolateral portion of the ganglion; the more posterior spot lies about halfway back in the ganglion and consists of two or three cells in the dorsal portion of the ganglion, roughly midway between the ventral midline and the lateral edge of the ganglion. By clonal age ∼90-100 hours, the anterior spot of Hro-eve expression started to disappear, so that some ganglia had only one anterior spot (Fig. 6C). By clonal age ∼130 hours, the anterior spots had disappeared, but the posterior spot of Hro-eve expression remained in each ganglion through clonal age ∼130-150 hours. Therefore, the posterior spot was evident throughout the germinal plate in individual embryos at late stage 10 (Fig. 6D).

In glossiphoniid leeches, ganglionic neurons arise from each of the five teloblast lineages, with half or more arising from the N lineage (Kramer and Weisblat, 1985; Weisblat et al., 1984). To identify the lineage(s) of origin of the neurons expressing Hro-eve, we carried out in situ hybridization on embryos in which one or more cells (N, OP or OPQ) had been injected with lineage tracer. The confocal images of sectioned embryos revealed that the in situ label colocalized with lineage tracer when the N lineage was labeled and not when any other lineage was labeled (Fig. 7). Thus, we conclude that the cells expressing Hro-eve arose from the N teloblasts. Moreover, by comparison with more detailed lineage analyses (Shain et al., 1998), we concluded that anterior ventrolateral cells exhibiting transient expression of Hro-eve arose from the ns.a clone, and that the posterior neurons arose from the nf.a clone in each segment.

The expression of Hro-eve in the teloblasts and primary blast cells suggested that this gene might be involved in some early aspects of segmentation. To investigate the developmental role of Hro-eve, we injected teloblasts with antisense or control MO (AS-Hro-eve MO and MM-Hro-eve MO, respectively), together with Fast Green and lineage tracer to monitor the injections and to follow the development of the injected cells, respectively. In most embryos, one teloblast was injected with AS-Hro-eve MO and the contralateral teloblast with lineage tracer alone, to monitor normal development, or in combination with a control MO. Here, we focussed primarily on the N teloblast lineage (Fig. 1).

We estimate that the MO concentrations ranged from 0.6-13 μM in the cytoplasm of the injected teloblasts. When AS-Hro-eve MO was injected into N teloblasts at lower concentrations, the overall pattern of development was normal until stage 10 (∼140 hours after the injection). But fluorescence microscopy revealed marked abnormalities in the distribution of progeny from the experimental N teloblasts. The usual segmental pattern of RDA-labeled ganglionic neurons was disrupted and was often out of register with the control side (Fig. 8A-C). At higher concentrations, development was more severely perturbed; these embryos suffered higher than normal mortality during gastrulation (37/120 embryos died in six experiments), which was due to a failure of germinal plate formation and consequent rupture and leakage of the yolk cell. Many of the surviving embryos exhibited dorsolaterally directed hairpin loops in the posterior portion of the germinal plate (Fig. 9) that was not seen with control injections. Further analyses of this phenomenon are described later.

Two assays were used to assess the effects of AS-Hro-eve MO on specific neuronal phenotypes. For one, we used in situ hybridization for Hro-eve to mark the subsets of nf- and ns-derived neurons described above. Cells expressing Hro-eve arose from N teloblasts injected with AS-Hro-eve MO, but the pattern was abnormal; such cells were often out of register with their counterparts in the control half of the germinal plate (Fig. 8D,E). For another assay, we used immunostaining to screen for the appearance of three pairs of serotonergic neurons, the anteromedial giant Retzius neurons, which normally arise from the ns.a blast cell clones, and smaller ventrolateral and dorsolateral neurons (cells 21 and 61, respectively), which normally arise from nf.a clones (Stuart et al., 1987). These three neurons arose in their normal positions in control bandlets, but no serotonergic neurons were detected in the experimental bandlets (Fig. 8F).

Further examination of lineage tracer in cells derived from the experimental N teloblasts revealed that they had failed to generate appreciable numbers of neurites. In normal development, three prominent segmental nerves exiting each side of the ganglion (Ort et al., 1974; Stent et al., 1992), contain neurites derived from contralateral N-derived neurons (Shain et al., 1998); these nerves were not detected contralateral to AS-Hro-eve MO-injected N teloblasts, in contrast to the case with control injections (data not shown). Observations of ganglionic Hro-eve expression were made at stage 10, whereas scoring for serotonergic neurons and segmental nerve formation was carried out at stage 11. Together, these results suggest that AS-Hro-eve MO injections resulted in a widespread failure of neural development at some point prior to terminal differentiation. Consistent with the fact that Hro-eve is expressed in all five teloblast lineages in early development, the defects induced by AS-Hro-eve MO injections were not restricted to neuronal tissues. For example, the O lineage normally makes a mixture of neurons and epidermal cells (Fig. 8G). Here, AS-Hro-eve MO injections resulted in an almost total loss of epidermal cells; neural precursors could be recognized by their ganglionic locations, but as in the N lineage, failed to complete differentiation (Fig. 8H).

To investigate the origins of the hairpin loops that occurred in the experimental side of germinal plates, we further examined embryos in which N teloblasts were injected with AS-Hro-eve MO or MM-Hro-eve MO as above, but fixed earlier, at mid stage 8 (∼42-48 hours after the injection). At this stage, control germinal bands reached smoothly around the equator of the embryo, as in uninjected embryos (Fig. 10). By contrast, germinal bands on the experimental side often exhibited a prominent dorsally directed kink in the posterior part of the germinal band (Fig. 10).

Did the kinked germinal bands result from elongated n bandlets, reflecting increased rates of cell division in the teloblasts and/or primary blast cells? To address this possibility, embryos in which the N teloblasts had been injected with AS-Hro-eve MO or MM-Hro-eve MO were fixed at late stage 7 (24 hours after the injection) to examine the divisions of teloblast and primary blast cells in segmentation. At this stage, most of the primary blast cells still lie beneath the surface of the embryo; therefore, cell division patterns in labeled bandlets were reconstructed from sectioned embryos, making use of the facts that the teloblast cell cycle is ∼50 minutes long at 23°C in H. robusta and that the primary blast cells remain in coherent bandlets. Thus, the most recently produced blast cells lie next to the teloblast and older clones lie progressively more distal within the bandlet. Moreover, in normal development, the nf and ns blast cells exhibit prolonged cell cycles (∼26 hours and ∼28 hours, respectively), then both undergo distinct, but stereotyped patterns of cell divisions; the first division is slightly unequal in both nf and ns (Fig. 1) (Bissen and Weisblat, 1989; Zackson, 1984) and takes place within ∼6 hours of the cells entering the germinal band. It is possible to identify those clones that have undergone their first division by the size differences of the cells and nuclei within the bandlet.

This normal pattern was observed in bandlets derived from N teloblasts injected with MM-Hro-eve MO or with lineage tracer alone. Each control bandlet contained between 26 and 28 clones and, with two exceptions, the blast cells were labeled uniformly throughout each bandlet (Fig. 11). Of the two exceptional bandlets, one contained a single blast cell with noticeably fainter lineage tracer than the others, while in the second bandlet 18 out of 28 cells were noticeably fainter. In all but one of the control bandlets, the oldest clone had undergone its unequal first mitosis (Fig. 12). In two of these bandlets, another blast cell had divided, giving an average of 1.1 mitoses per bandlet.

Bandlets derived from experimental N teloblasts contained about the same number of primary blast cell clones as did controls, indicating that the cell cycle duration of the teloblast was not affected (Fig. 12). However, the experimental bandlets differed from controls in two respects. First, the division pattern of the blast cells was disrupted, in that there were more clones in each bandlet that had already undergone mitosis (an average of 2.1 per bandlet) than in control bandlets. Moreover, the two-cell clones were not confined to the distal ends of the bandlet, indicating that some cells had divided much earlier than normal. Second, the bandlets invariably comprised an irregular sequence of brightly and faintly labeled blast cells (Figs 11, 12). One explanation for the variation is that the experimental teloblasts generate blast cells with variable proportions of yolk-rich and yolk-deficient cytoplasm (lineage tracer is largely excluded from yolk platelets), but we observed no difference in the cytoplasm of faintly labeled versus brightly labeled blast cells (data not shown). Two other possibilities are: (1) that the experimental teloblasts produce blast cells of varying volumes, and that the smaller cells then grow, diluting the inherited lineage tracer; or (2) that some blast cells are metabolically altered so that they secrete or sequester the lineage tracer. We have no evidence bearing on these possibilities, but in any case, injection of AS-Hro-eve MO had somehow disrupted the normal stem cell divisions in teloblasts. Thus, we conclude that injecting teloblasts with AS-Hro-eve MO perturbed cell divisions and fate decisions directly or indirectly throughout the subsequent development of the N lineage, beginning with the stem cell divisions of the teloblasts themselves.

In the work presented here, we identified eve-class genes, Hro-eve and Ttr-eve, from two species of glossiphoniid leeches. We have begun to characterize the expression and function of Hro-eve by semi-quantitative RT-PCR, in situ hybridization and injection of antisense MO.

In situ hybridization reveals that Hro-eve is expressed during cleavage and in all teloblast lineages up through the first division of the primary blast cells. No pair-rule or striped pattern is observed for Hro-eve during segmentation.

In teloblasts and primary blast cells, the in situ signal for Hro-eve is strongest over the chromatin of cells in mitosis. One interpretation of these results is that pre-existing Hro-eve transcripts localize to chromatin upon nuclear envelope breakdown during mitosis. An alternative explanation is that Hro-eve is transcribed during mitosis; in this case, the transcripts might remain associated with chromatin because the biochemical machinery required for splicing and polyadenylation is not operative during mitosis (Alberts et al., 1994). The generalization is sometimes made that transcription is shut down during mitosis, but this is not an absolute condition; in yeast, ∼300 genes have been identified as being transcribed during mitosis (Krebs et al., 2000; Spellman et al., 1998).

Semi-quantitative RT-PCR reveals that Hro-eve transcripts are most abundant during the period in which ganglia are differentiating, and in situ hybridization reveals two sets of bilaterally paired, segmentally iterated neurons that express Hro-eve during neurogenesis. An anterior set of neurons, which are derived from the ns (probably from the ns.a) clone in each hemiganglion, expresses Hro-eve transiently, while a more posterior group of neurons, derived from the nf (probably from the nf.a) clone, continue to express Hro-eve in the oldest embryos studied (early stage 11).

Many genes that were first identified as regulating segmentation in Drosophila are also expressed later during neural development, including even-skipped (Doe et al., 1988; Frasch et al., 1987; Patel et al., 1989). The expression patterns of eve-class genes suggest similar dual functions in other arthropods (Damen et al., 2000; Patel et al., 1992). In nematodes (Ahringer, 1996), a taxon of unsegmented ecdysozoans, and in segmented deuterostomes, including amphibians (Ruiz i Altaba, 1990), fish (Thaeron et al., 2000) and mammals (Dush et al., 1992; Moran-Rivard et al., 2001), eve homologs are involved in neurogenesis. These results, together with our present findings, are consistent with the idea that eve-class genes functioned in neurogenesis in the last common ancestor of the deuterostomes, ecdysozoans and lophotrochozoans. However, we cannot yet rule out an alternative explanation for these observations, based on convergent evolution. In this scenario, the evolution of the many neuronal phenotypes present in higher metazoans entailed the independent recruitment to neuronal functions of transcription factors that were used for other purposes in ancestral animals. Further information on this issue may be gained by studying the function of eve-class genes from cnidarians (Finnerty and Martindale, 1997; Miles and Miller, 1992).

The expression of Hro-eve in the early embryos, including the teloblasts and primary blast cells (segmental founder cells) of all five lineages (M, N, O, P and Q) suggests the possibility that this gene is somehow linked to the general regulation of cell proliferation, but not the specification of teloblast lineages or segment identity.

Evidence for the function of Hro-eve in the early Helobdella embryo was obtained by knockdown experiments, in which antisense oligonucleotides were injected into the N teloblasts, along with fluorescent lineage tracers to monitor the development of the injected cells. Lacking an antibody for Hro-eve, we are unable to measure the extent of the resulting knockdown, but do observe disruptions of normal development, beginning with teloblast and primary blast cell divisions. AS-Hro-eve MO injections do not affect the average rate or general asymmetry of the teloblast divisions, but the blast cells produced by the experimental teloblasts show variations in lineage tracer content that are not seen in control MO injections. Moreover, primary blast cells produced by the N teloblasts injected with AS-Hro-eve MO sometimes divided several hours earlier than normal, and germinal bands containing bandlets derived from the experimental N teloblasts frequently exhibited a dorsally directed kink. We speculate that the kinks may result from mechanical deformations resulting from excess or ectopic N-derived cells that might be caused by the accelerated blast cell divisions, or from changes in biochemical properties that render these cells unable to undergo normal gastrulation movements.

The definitive fates of cells arising within the blast cell clones are also perturbed by AS-Hro-eve MO injections: there appear to be fewer than the normal number of cells; the usual groupings of cells within the hemiganglia are no longer evident; the three N-derived peripheral neurons in each hemisegment cannot be found. Neurons expressing Hro-eve are produced, but in ectopic positions compared with their contralateral homologs; by contrast, the serotonergic neurons that arise from the N teloblasts fail to differentiate.

Thus, we conclude that interfering with the expression of Hro-eve perturbs cell divisions of teloblasts and the subsequent divisions within the blast cell clones, consistent with the expression of Hro-eve in these cells during early development. The extent to which the effects on teloblasts and blast cells represent separate roles for Hro-eve remains to be determined.

Until recently, the notion that segmentation is homologous between annelids and arthropods was generally accepted, as were phylogenetic trees in which variously defined clades of annelids and arthropods were placed as sister taxa. In addition to the segmentation per se, similarities between the posterior teloblastic growth zones in annelids and crustacean arthropods (Dohle and Scholtz, 1988) were taken as evidence supporting these assumptions. Logically, however, using phylogenetic trees based on developmentally derived traits as the reference framework for interpreting developmental comparisons introduces an inherent circularity to the process.

Molecular phylogenies, which offer a partial escape from this circularity, suggest that protostomes comprise distinct ecdysozoan and lophotrochozoan clades (Aguinaldo et al., 1997; Ruiz-Trillo et al., 1999) and that these have been separated since long before the Cambrian radiation (Adoutte et al., 2000). If so, given the paucity of segmented taxa among Ecdysozoa and Lophotrochozoa, it is more parsimonious to assume that segmentation has arisen independently in annelids and arthropods.

Consistent with this interpretation, work to date has revealed little similarity between the embryological functions of Drosophila segmentation genes and those of their homologs in Helobdella. For example, although it was initially concluded that Helobdella engrailed-class genes play an essential role in early segmentation (Ramirez et al., 1995; Wedeen and Weisblat, 1991), subsequent experimental analyses indicate that this is not the case (Seaver and Shankland, 1999; Seaver and Shankland, 2000; Shain et al., 1998; Shain et al., 2000). Also in contrast to Drosophila, the translation of maternally derived mRNA for the Helobdella nanos-class gene occurs after the AP axis is already established and correlates with the mesoderm-ectoderm fate decision (Pilon and Weisblat, 1997). In the work reported here, we find that Hro-eve is not expressed in a pair-rule pattern as in Drosophila. Moreover, we find no evidence of a link between the early expression of eve- and engrailed-class genes in leech. The early expression of Hro-eve shows no spatial or temporal correlation with that of Hro-engrailed.

eve-class genes are expressed in the posterior segmentation zone of annelids (Lophotrochozoa), arthropods (Ecdysozoa) and vertebrates (Deuterostomia) (Brown et al., 1997; Damen et al., 2000; Joly et al., 1993; Patel, 1994; Patel et al., 1992). One explanation for this similarity is that it reflects evolutionary convergence, i.e. independent recruitment of this transcription factor into the segmentation mechanisms in the three taxa. Alternatively, similarities in gene expression patterns between arthropods and vertebrates have led some to propose that the last common ancestor of all bilaterally symmetric animals was already segmented (De Robertis, 1997; Holland et al., 1997; Palmeirim et al., 1997). This latter explanation seems unlikely, on the basis of parsimony and, as reported here, because of the differences in the expression of eve-class genes between arthropods and leech.

We thank Sharon Amacher for suggestions regarding the in situ hybridization protocol for early gene expression, Alexa Bely for assistance with phylogenetic analyses, and members of the Weisblat laboratory for helpful discussions. This work was supported by NIH grant RO1 GM 60240 to D. A. W.