A hedgehog homolog regulates gut formation in leech(Helobdella)

The gene hedgehog (hh) was originally identified inDrosophila as a segment polarity gene(Heemskerk and DiNardo, 1994;Ingham et al., 1991). Subsequently, hh-class genes have been found in many organisms including chordates [Amphioxus, zebrafish, frog, chick, mouse and human (reviewed by Shimeld,1999)] and mollusc [limpet(Nederbragt et al., 2002)]. In addition to their role in patterning larval segments in fly(Nusslein-Volhard and Wieschaus,1980), HH-class proteins have been shown to play key roles in intercellular signaling during limb development in fly and vertebrates(Vervoort, 2000), neural tube patterning in chordates (McMahon,2000; Patten and Placzek,2000; Shimeld,1999), gut formation in fly and vertebrate, cancer in vertebrates(Toftgard, 2000), cell survival and cell proliferation (Britto et al., 2000) (see alsoHammerschmidt et al., 1997;Ingham and McMahon, 2001;McMahon, 2000).

The consensus HH protein contains a secreted domain(NH₂-terminal fragment; HH-N) that functions in extracellular signaling pathways and a C-terminal domain (HH-C) that is involved in intracellular post-translational autoprocessing and covalent attachment of cholesterol (Chuang and Kornberg,2000; Lee et al.,1994). Sequence alignments among known hh homologs reveal that the HH-N domains are usually more conserved than HH-C domains.

Current molecular phylogenies support the organization of all or most extant bilaterian animals into 3 ancient clades, Deuterostomia, Ecdysozoa and Lophotrochozoa (Adoutte et al.,2000; Aguinaldo et al.,1997). Until recently, all of the hh-class genes reported in the literature were from members of the former two clades. The recent discovery that a hh-class gene from the limpet Patella vulgata (Lophotrochoazoa) is expressed along the ventral midline has been interpreted as supporting the dorsoventral axis inversion theory and as supporting a role for hh-class genes in neural patterning(Nederbragt et al., 2002), but functional tests of these ideas are yet to be reported. Thus, the question of which aspect(s) of hh-class gene function are conserved between Lophotrochozoans and the other two groups remain of interest, both for deducing the nature of early bilaterian ancestors and for understanding the evolution of body plans by modification of developmental mechanisms.

For example, in Drosophila embryos hh plays a pivotal role in segmentation. The hh gene is co-expressed withengrailed (en) in the posterior compartment of each nascent segment, and the secreted HH protein functions to specify the fates of cells in the adjoining anterior compartment in a concentration-dependent manner(Heemskerk and DiNardo, 1994;Lee et al., 1992). Later inDrosophila development, the hh and en genes are also expressed in identical compartment-specific patterns in the imaginal discs (Tabata and Kornberg,1994). An en-class gene has been described in glossiphoniid leeches, a group of segmented lophotrochozoans (phylum Annelida,class Hirudinea). But while leech-en is expressed during segmentation and neurogenesis (Lans et al.,1993; Wedeen and Weisblat,1991), more recent work suggests that it does not play a direct role in segmentation (Shain et al.,1998) and that the cells that express leech en are not required to specify the fates of cells in the remainder of the segment(Seaver and Shankland, 2001). These observations raise the question as to whether hh-class genes are expressed coincident with en-class genes in the leech embryo,and, if so, how that expression might relate to segmentation.

Here, we report the identification and characterization of a hhhomolog (Hro-hh) from the glossiphoniid leech, Helobdella robusta. Hro-hh is expressed zygotically in gut and some other tissues,but Hro-hh RNA was not detected in the cells that expressHro-en during segmentation. In addition, we found that bath-application of cyclopamine, a known blocker of hh signaling in vertebrates (Helms et al.,1997; Incardona et al.,1998; Incardona et al.,2000) induced malformation of foregut, anterior midgut and coelomic mesenchyme in Helobdella, but had no apparent effect on the segmental patterning of mesoderm and ectoderm. Along with data on the function of hh-class genes in vertebrates and insects, our results support a scenario in which the ancestral role of hedgehog family genes in bilaterian animals was associated with gut formation and/or neural differentiation, rather than segmentation.

Helobdella robusta embryos were obtained from a laboratory breeding colony (Shankland et al.,1992) or from specimens collected in a minor tributary of the American River in Sacramento, California. For some experiments, embryos were obtained from a separate breeding colony, founded with individuals collected from the mouth of Shoal Creek in Austin Texas. This second population closely resembles H. robusta from Sacramento, both morphologically and as judged by comparisons of nuclear gene sequences (see below), but exhibits significant divergence as judged by differences in mitochondrial gene sequences (A. E. Bely, personal communication). Embryos were cultured at 23°C in HL saline (Blair and Weisblat,1984). The embryonic stages are as defined previously(Weisblat and Huang, 2001), or for greater precision, in terms of the time elapsed after zygote deposition(AZD).

The relevant aspects of glossiphoniid leech development are illustrated inFig. 1. During cleavage (stages 1-7), 5 bilateral pairs of stem cells (M, N, O/P, O/P and Q teloblasts), 3 macromeres (A′′′, B′′′ and C′′′) and 25 micromeres are formed. Teloblasts undergo repeated, highly unequal divisions to generate columns (bandlets) of founder cells (blast cells) for segmental mesoderm and ectoderm in an anterior-to-posterior progression (stages 5-8). The five bandlets on each side of the embryo join together in parallel to form the germinal bands, which then coalesce from anterior to posterior along the ventral midline into the germinal plate (stage 8); the segmented nervous system, nephridia, epidermis,and musculature differentiate from the germinal plate during stages 9-11. Micromeres contribute definitive progeny to non-segmented tissues such as the supraesophageal ganglion and the foregut of the adult leech, and also a squamous epithelium that covers the germinal bands and the intervening prospective dorsal territory during gastrulation(Fig. 1)(Bissen and Weisblat, 1989;Weisblat et al., 1984). The midgut epithelium forms by cellularization of a syncytial yolk cell (SYC),that arises by the stepwise fusion of the A′′′-C′′′ macromeres, teloblasts (after they have completed blast cell production), and supernumerary blast cells(Fig. 1)(Desjeux and Price, 1999;Isaksen et al., 1999;Liu et al., 1998;Nardelli-Haefliger and Shankland,1993).

For cell lineage tracing, selected teloblasts or proteloblasts (stages 4-7)were pressure-injected with a solution of tetramethylrhodamine dextran amine(RDA, Molecular Probes Inc.) at a final concentration of 50 mg/ml in 0.2 N KCl, with 1% Fast Green to permit visual monitoring of the injections(Weisblat et al., 1980). Injected embryos were cultured to stages 9-10, then fixed and processed for in situ hybridization as described below. Prior to fixation, all embryos at mid stage 9 or beyond were relaxed in a solution of 8% ethanol in 4.8 mM NaCl, 1.2 mM KCl, 10 mM MgCl₂ (relaxant HL saline). Some embryos were stained with Hoechst 33258 (1 μg/ml) to visualize cell nuclei.

An initial fragment of Hro-hh was amplified from an H. robusta cDNA library (embryonic stages 7-10; Stratagene) by the polymerase chain reaction (PCR), using the following degenerate primers:

• Hro-hh1: 5′-GCGTIACIGARRGNWGRGAYGARGA-3′ [VTE(SG)(WR)DED]

• Hro-hh2: 5′-ACCCARTCRAAICCIGCYTCNACNGC-3′ [VEAGFDWV]

The PCR reaction mixture contained 50 mM KCl, 10 mM Tris, pH 9.0, 2.5 mM MgCl₂, 0.1% Triton X-100, 0.2 mM each dNTP, 0.8 mM each primer, 1μl H. robusta cDNA library (∼10¹⁰ pfu/ml;Stratagene), and 1 unit Taq polymerase (Perkin-Elmer) in a 50 μl volume. For the amplification, we employed a 3-step `touchdown' protocol: 5 cycles of 94°C for 1 minute, 53°C for 1 minute, 72°C for 1 minute; 10 cycles of 94°C for 1 minute, 50°C for 1 minute, 72°C for 1 minute; then 20 cycles of 94°C for 1 minute, 48°C for 1 minute, 72°C for 1 minute. The amplified DNA was purified from an agarose gel using Gel Extraction Kit (Qiagen) and cloned into pGEM T easy vector (Promega).

The 150 bp Hro-hh sequence was amplified and labeled with[³²P]dCTP from cloned pGEM T easy plasmid by PCR. We used this probe to screen an H. robusta cDNA library (stages 7-10; Stratagene). Phage plating and hybridization were carried out according to manufacturer's instructions with minor modifications. Hybridization was performed at 45°C in 6× SSC, 5× Denhardt's solution, 0.5% SDS, 500 μg/ml herring sperm DNA and 50% deionized formamide. The filters were washed in 2×SSC, 0.5% SDS at 65°C for 1 hour and in 0.1× SSC, 0.1% SDS for 3 hours. pBlueScript SK(–) phagemids were excised in vivo according to the manufacturer's instructions. A plasmid (pHrohh) containing 3.49 kbHro-hh fragment was sequenced. This sequence (GenBank accession number AF517943) was confirmed by RT-PCR from cDNA prepared from total RNA as described below.

Independently, another hh-class gene fragment was isolated by degenerate PCR from the Helobdella population collected in Austin,Texas, using primers corresponding to nucleotides 934-952 of Hro-hhin the forward direction and nucleotides 1060-1077 in the reverse direction,as follows:

• hh5: 5′-ACNGARGGNTGGGAYGAAG-3′

• hh3: 5′-CCARTCRAANCCNGCTTC-3′

From the 107 nucleotide sequence obtained, there was only a single synonymous change (position 1053). In keeping with this high degree of sequence identity, the same probe was used for in situ hybridization experiments on both types of embryos and the results obtained were indistinguishable.

Total RNA samples were prepared from H. robusta embryos at selected stages with RNA Wiz (Ambion) according to the manufacturer's instructions, using 40 embryos for each sample. The RNA samples were added to solutions containing 1× reverse transcription buffer (Gibco), 3.33 mM DTT, 0.33 mM dNTP, 3.33 μM random decamer (Ambion), and 200 Units reverse transcriptase (Gibco). The mixture was incubated at 42°C for 50 minutes and the resultant first strand cDNA (30 μl) was stored at –20°C. PCR conditions were as described by the manufacturer of ampliTaq Polymerase(Perkin-Elmer Cetus, Norwalk, CT) except that 3 μl of cDNA template were used in 50 μl of reaction mixture.

To amplify an Hro-hh-specific fragment (96 bp in length), a pair of PCR primers was designed as follows:

• hh-F: 5′-GAAACTCATCCAGAAGACTCC-3′

• hh-R: 5′-TCTAGCCAACATCCCATACTTG-3′

PCR conditions for amplifying the Hro-hh fragment were: 1 minute at 94°C, 1 minute at 60°C, and 1 minute at 72°C for 5 cycles,followed by 1 minute at 94°C, 1 minute at 58°C, and 1 minute at 72°C for 17 cycles. A 15 μl aliquot of each sample was removed after 22 cycles and the remaining material underwent 5 more cycles of amplification.

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 Hro-hh sample. For this purpose, and to attenuate the signal obtained from the abundant rRNAs, bona fide and non-extending 18S primers (competimers; Ambion) were used in a 3:7 ratio, respectively. PCR conditions for amplifying the 18S rRNA fragment were: 1 minute at 94°C, 1 minute at 58°C and 1 minute at 72°C for 1 minute for 5 cycles,followed by 1 minute at 94°C, 1 minute at 56°C, and 1 minute at 72°C for 10 cycles. A 15 μl aliquot of each sample was removed after 15 cycles and the remaining material underwent 5 more cycles of amplification.

To quantitate PCR products, each sample was run out in a 2% agarose gel and stained with ethidium bromide. Band intensity was measured with an Alphaimager(Alpha Innotech Corp.) using Alphaease (v3.3b) program.

In situ hybridization was performed using a digoxigenin-labeled RNA probe with some modifications of methods described previously(Harland, 1991;Nardelli-Haefliger and Shankland,1992). Sense and antisense probes (each of 3.6 kb length) were obtained by T7 and T3 in vitro transcription (MEGAscript kit, Ambion) using linearized pHrohh (cut with BamHI to generate the sense probe, and with SalI to generate the antisense probe) and subsequently hydrolyzed into shorter fragments (∼300 bp) in an alkaline solution(Cox et al., 1984).

Early embryos (stage 1-8) were fixed in 4% formaldehyde in PBS (130 mM NaCl, 10 mM phosphate buffer, pH 7.4) for 1 hour, then permeabilized by a 5-minute incubation with 50 μg/ml proteinase K (Gibco); late embryos(stages 9-11) were fixed as described above for 30 minutes, then treated for 25 minutes in 0.5 mg/ml Pronase E (Sigma).

Hybridization was carried out overnight at 59°C in a 1:1 mixture of deionized formamide and 5× SSC, 0.2 mg/ml tRNA, 0.1 mg/ml heparin,1× Denhardt's solution, 0.1% Tween 20 and 0.1% Chaps. Alkaline phosphatase (AP)-conjugated anti-digoxigenin Fab fragments (Roche) were added to a dilution of 1:5000 in 1× PBS, 0.1% Tween 20 for late embryos. To remove unhydrolyzed probe, embryos were treated with RNase A (Sigma; 50μg/ml for 1 hour for early embryos and 5 μg/ml for 30 minutes for late embryos) at 37°C. Alkaline phosphatase (AP)-conjugated anti-digoxigenin Fab fragments (Boehringer-Mannheim) were added (1:2000 in 1× PBS, 10%normal goat serum, 0.1% Tween 20, 0.2% Triton X-100 for early embryos; 1:5000 in 1× PBS, 0.1% Tween 20 for late embryos) and the color reaction was carried out using nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indoyl-phosphate (NBT/BCIP; Roche) by standard procedures. Intact embryos were dehydrated in ethanol and propylene oxide,followed by infiltration with plastic embedding medium (Poly/Bed 812;Polysciences).

Cyclopamine was obtained from Veratrum californicum as described previously (Gaffield et al.,1986) and was diluted to a final concentration of 10 μM, 5μM and 1 μM in HL saline (from a stock solution of 10 mM in ethanol). Experimental embryos (early or mid stage 8; 60-68 AZD) were cultured in cyclopamine for 3-6 days (sibling controls were cultured in HL saline with 0.1% ethanol). Treated embryos were then examined morphologically.

For closer examination than was possible in whole mount, some embryos were dehydrated through a series of graded alcohols to 95% and then embedded in glycolmethyacrylate resin (JB-4; Polysciences, Inc) according to the manufacturer's instructions. Embedded specimens were sectioned with glass knives on a Sorvall MT2-B microtome. Sections were mounted on glass slides and stained with Hoechst 33258 (1 μg/ml).

In our hands, the AP reaction product obtained by in situ hybridization is not stable in JB-4 embedding resin. So to examine such embryos in section, the selected specimens were dehydrated and embedded in epoxide embedding resin(Poly/Bed 812; Polysciences) following the manufacturer's instructions, and sectioned at ∼10 μm thickness with glass knives using a Sorvall MT2-B microtome or at ∼100 μm thickness using hand-held razor blades. Sections were mounted on glass slides under coverslips in a non-fluorescing medium (Fluoromount; BDH Laboratory Supplies, Ltd.), and photographed with Nomarski optics (Zeiss Axiophot) or observed with confocal microscope (MRC 600; Bio-Rad) in 0.1 μm optical sections.

Whole embryos or sections were viewed with DIC optics (Zeiss Axiophot or Nikon E800 microscope) and photographed using Ektachrome 160 film (Kodak) or with a `Spot' CCD camera (Diagnostics, Inc.). Live embryos were paralyzed in relaxant HL saline prior to photography.

Degenerate PCR primers were designed by comparing the conserved regions ofhh homologs from Drosophila and various vertebrate species(Echelard et al., 1993). These primers amplified a 150 bp fragment of a putative hh-class gene from a genomic library and from a cDNA library representing stages 7-10 H. robusta embryos, (but not from a stage 1-6 cDNA library). This fragment was cloned, sequenced and then used to screen a cDNA library. From∼120,000 plaques, one positive phage was isolated, and its insert was sequenced. This sequence was confirmed by independent PCR amplification from first strand cDNAs representing stage 9-10 embryos. The Hro-hh cDNA encodes a predicted 3219 bp (1073 amino acids) open reading frame (ORF) which is flanked by 186 bp of 3′ UTR and 75 bp of 5′ UTR. The amino and carboxy termini of the encoded HRO-HH polypeptide are well conserved with respect to those of known HH homologs (Fig. 2). This led us to conclude that we had cloned the entire coding sequence of this Hro-hh transcript.

HRO-HH contains two apparent cleavage sites, CGPG in the amino region and GCF in the carboxy region that are conserved among HH-class polypeptides and are predicted to generate a secreted peptide (HRO-HH-N;Fig. 2A) consisting of 175 amino acids, compared with 173 amino acids in Drosophila and 174 amino acids in human and mouse SHH (Fig. 2B). In other organisms, HH-N is involved in extracellular signaling pathways (Chuang and Kornberg,2000) and it seems likely that this function is conserved in HRO-HH-N.

In other organisms, HH-C is important for autoproteolysis of the HH propeptide and for covalently attaching cholesterol to the carboxyl terminus of HH-N. HRO-HH-C is less well conserved than HRO-HH-N, based on amino acid sequence comparison with other HH-class proteins(Fig. 2) and is longer than any other known HH-C peptides, but a region resembling the proposed active site motif (PASM) of HH-C peptides [including the invariant T and H residues required for autoproteolysis (Lee et al.,1994)] was also found in HRO-HH-C at residues 561-567(Fig. 2A,C).

We also note the presence in HRO-HH-C of a putative cell division sequence motif (CDSM) not found in other HH proteins. The consensus motif([ND]-C-[TES]X[DE][EDTS][DE], where X is a spacer of 1-8 amino acids) exists in the v-myc and c-myc oncoproteins, CDC25, and other proteins thought to carry out one of the required steps in the cell division competency cascade of deactivating a cell division repressor(Figge and Smith, 1988). The presence of this motif in Hro-hh(DKTCDSIDSE, residues 954-963;Fig. 2A), suggests that HRO-HH-C may also be involved in intracellular signaling related to the control of cell division.

To investigate the relationships between Hro-hh and previously described hh-class genes, we used ClustalX 1.81 to construct an unrooted phylogram, comparing that portion of the conserved HH-N domain for which sequences were available for all species of interest(Fig. 3). The best supported tree grouped Hro-hh with hh from Drosophila andPatella, separate from the deuterostome hh-class genes, in accord with the accepted metazoan phylogenetic tree based on rRNA sequence(Adoutte et al., 2000;Aguinaldo et al., 1997).

Because the relatively small number of embryos available precludes the routine application of northern blot analysis for most genes in H. robusta, we used semi-quantitative RT-PCR to obtain an initial developmental expression profile for Hro-hh. For this purpose, 18S ribosomal RNA was used as an internal standard, to control for variations in the efficiency of RNA extraction and reverse transcription reaction (see Materials and Methods). By this assay (Fig. 4), Hro-hh is an exclusively zygotic transcript. Transcripts were not detected in dissected oocytes or in zygotic stages prior to stage 8, which is many hours after all the teloblasts have been generated and are actively making blast cells. Transcript levels increase through late stage 10, which correlates with the major period of organogenesis, and then decline slowly in stage 11, as the embryo completes its transformation into a juvenile leech.

In situ hybridization was used to characterize Hro-hh expression in specific tissues and cells (Figs5,6,7). Expression was first observed in stage 8 embryos, consistent with the results of RT-PCR. But, in contrast to expectations based on the role of hh inDrosophila segmentation, Hro-hh mRNA was detected in the anteriormost, unsegmented (prostomial) region(Fig. 5A,B) and not in the segmenting germinal plate at this stage(Fig. 5A,C). Even extensively overstained embryos showed no signal other than diffuse staining presumed to be background in the segmental tissues (data not shown) until later stages when segmentally iterated organs and tissues were differentiating (described below).

By the end of stage 8, two distinct sets of cells expressingHro-hh could be distinguished(Fig. 5C,D): one at the anterior end of the germinal plate marked what we surmise to be the prospective stomodeum; the other lay beneath it, next to the SYC, at the junction between the future foregut (proboscis and esophagus) and anterior midgut. This staining demarcates the region in which the proboscis will form,and in progressively older embryos, the Hro-hh staining pattern expanded to encompass the full length of the developing proboscis(Fig. 5E,G;Fig. 6A,B). Expression in the proboscis finally disappeared during stage 11(Fig. 6H).

The proboscis of glossiphoniid leeches is a tri-radiate, muscular tube (seeFig. 7A) that is everted during feeding to pierce host or prey tissues and suck out blood or soft body parts. This process is aided by secretions from ductules arising in the salivary glands and running the length of the proboscis. As illustrated inFig. 1, the proboscis occupies the everted position during the early stages of its morphogenesis (stages 9 and early 10), and then assumes the inverted position (late stage 10 and beyond).

Hro-hh expression was largely confined to the inner layer of the developing proboscis, proximal to the lumen(Fig. 5E,G). Beginning in stage 9, staining was also seen in a ring of external tissue surrounding the base of the proboscis (Fig. 5E-G). This ring is apparently transformed into the epithelial lining of the oral opening once the proboscis has inverted (Fig. 6A,B). Hro-hh expression continued within the proboscis through early stage 11 (Fig. 6A,B), and transverse sections of such embryos revealed that the lumen proximal staining was in presumptive radial muscle cells throughout the length of the proboscis (Fig. 6C,D). Within the posterior portion of the proboscis, staining was also seen in presumptive longitudinal muscle cells, more distal to the lumen(Fig. 5F,Fig. 6D).

In addition to the expression in the foregut described above,Hro-hh expression was evident in cells associated with the midgut. Beginning in late stage 8 and continuing into stage 9, Hro-hh was seen in faint, segmentally iterated, transverse stripes at the interface of the germinal plate and the SYC (prospective midgut;Fig. 5H). These stripes appeared only after segmental morphology was established, and were associated with the developing gut wall at the dorsal edge of the intersegmental septa(Fig. 5I). During stage 9, most of these stripes fade, but five persist and become stronger during stage 10,so that by stage 11 Hro-hh staining was seen in a series of five rings around the posterior midgut derivatives: one at the boundary between crop and intestine; one near the boundary of the intestine and rectum; and three around the rectum (Fig. 6B,G). Also by stage 11, a speckled pattern of staining was evident at the surface of the crop (Fig. 6B,F,G). By mid-stage 11, much of the expression appears to have disappeared, except for the rings of presumptive muscle associated with the intestine and rectum (Fig. 6H,I).

Staining was also seen in cells at the lateral edges of the germinal plate during stage 9 (Fig. 5H), and during stage 10, in the male and female reproductive tissues(Fig. 6A,E) and throughout most of the definitive epidermis of the body wall. The epidermal staining was more intense in the posteriormost midbody segments, and absent at the ends of the embryo, corresponding to the presumptive anterior and posterior suckers(Fig. 5I,Fig. 6A,G). Finally, by stage 11, Hro-hh expression was also seen in a small number of segmentally iterated cell bodies in the ganglia of the ventral nerve cord(Fig. 6A,C).

A prominent structural feature of the proboscis is an extensive array of radial muscles (Fig. 7A,B),whose contractions open the lumen of the proboscis to draw in food; their actions are opposed by a thick, interwoven band of circumferential muscles situated roughly midway along the radius of the proboscis(Fig. 7A,B). Nuclei of the radial muscles are situated within the ring of circumferential muscle. Outside this ring, the space between the radial muscles is occupied by relatively sparse sets of longitudinal muscles and a large number of ductules that carry salivary gland secretions to the tip of the proboscis(Fig. 7B).

Thus, given the observed staining patterns, the prime candidates for cells expressing Hro-hh near the lumen of the developing proboscis are the radial muscles and/or circumferential muscles, while the longitudinally oriented fibers near the outer surface in the posterior portion of the proboscis correspond to longitudinal muscles. We have previously established that presumptive circumferential muscles of the proboscis arise from micromeres c′′′ and dm′(Huang et al., 2002). Careful examination of sectioned, in situ-stained embryos in which cell dm′ had been injected with lineage tracer, revealed no doubly labeled cells(Fig. 7G). All the cells colored with the in situ reaction product lay within or outside the ring of tracer-labeled circumferential muscles. From this, we conclude that it is the radial muscles and longitudinal muscles, and not the circumferential muscles,that express Hro-hh in the proboscis at this stage.

The embryonic origins of the radial and longitudinal muscles are less clear, but on the basis of our previous results, it seemed that the primary quartet micromeres (a′-d′) and secondary trio micromeres a′′ and c′′ were good candidates for contributing at least some of these muscles. Consistent with this interpretation,Hro-hh-stained embryos in which micromere a′′ had been injected with lineage tracer did contain some doubly labeled cells near the core of the proboscis that appear to be radial muscle precursors(Fig. 7H). Embryos in which primary quartet micromeres (a′, b′ and d′) had been injected showed double labeling of some radial muscle precursors, and also of some longitudinal fibers in the outer layers of the proboscis that we assume are presumptive longitudinal muscles (Fig. 7D-F).

The wall of the midgut comprises two closely apposed layers of cells. One,an outer layer of visceral mesoderm arises from the same m blast cells that contribute segmentally repeating mesodermal structures such as body wall and nephridia, is laid down in an anteroposterior progression and expands dorsally with the rest of the germinal plate during stages 9-10. The other layer is endoderm that arises by cellularization of the SYC(Nardelli-Haefliger and Shankland,1993), under mesodermal influence(Wedeen and Shankland, 1997). The SYC arises by a stepwise series of cell-cell fusions, and thus contains nuclei derived from the macromeres, teloblasts, and supernumerary blast cells(Desjeux and Price, 1999;Liu et al., 1998). Because the two layers of cells are thin and tightly apposed, it is difficult to tell them apart, even using lineage tracer injections. The difficulty is compounded by the fact that the M teloblasts have already fused with the SYC by the time visceral mesoderm forms, so it is not possible to label visceral mesoderm without also labeling the endoderm. The rings of staining in the posterior midgut clearly lay outside the endoderm, and presumably corresponded to M-derived muscles ringing the intestine and rectum. But examination of fortuitous sections through the anterior midgut (prospective crop) suggest that some Hro-hh expression was also occurring within the endodermal layer (not shown).

To obtain information regarding the functional significance ofHro-hh expression in the developing leech, we examined intact and sectioned stage 10-11 embryos that had been grown from early stage 8 in the presence of cyclopamine, a steroidal alkaloid that inhibits HH signaling in vertebrates (Cooper et al.,1998; Incardona et al.,1998; Taipale et al.,2000). In these experiments, epiboly, germinal plate formation and the differentiation of segmental tissues proceeded in parallel between experimental and control embryos (not shown), but the differentiation of the proboscis and crop were clearly disrupted by cyclopamine(Fig. 8).

In particular, the proboscis was shorter and often failed to invert(Fig. 8A,F,J). The esophagus was much longer and thinner than usual and the crop lacked the narrowing anterior projection normally seen during stages 9-10(Fig. 8J). The crop caeca that normally become prominent during stages 10-11 were reduced in embryos exposed to 5 μM cyclopamine and absent in those exposed to 10 μM cyclopamine(Fig. 8G,I,K). No morphological abnormalities were detected in intestine or rectum and the intestine produced its own small caeca (Fig. 8I,K).

To further test the specificity of the cyclopamine effects, we examined control and cyclopamine-treated embryos in which selected cells had been injected with cell lineage tracer. Ectodermal lineages in cyclopamine-treated embryos were indistinguishable from those in controls(Fig. 9A, D, G). In the M lineage, contributions to the provisional integument and ventral nerve cord appeared normal; mesodermal derivatives were also present in the body wall,but the definition and organization of muscle fibers was less clear than in controls. This latter observation correlated with the fact that the germinal plate was very fragile and difficult to dissect in cyclopamine-treated embryos and that the embryos failed to flatten in the DV axis (data not shown).

Cyclopamine also caused defects in the dm' micromere lineage, which normally contributes presumptive circumferential muscles to the proboscis and esophagus (Huang et al.,2002). In normal development, these hoop-like fibers are found from the anterior tip of the proboscis to the anterior end of the crop(Fig. 9A,B). In embryos treated with 5 μM cyclopamine, the circumferential muscles are present in the anterior portion of the proboscis, but absent in the esophagus(Fig. 9D,E). In embryos treated with 10 μM cyclopamine, they are missing from both proboscis and esophagus(Fig. 9G,H). These observations were confirmed by the examination of sectioned embryos(Fig. 8L). The dm′micromere also contributes a network of fibers that extend throughout the body wall and ramify in the posterior sucker(Huang et al., 2002); these fibers were present at both concentrations tested(Fig. 9C,F,I).

In addition to the foregoing defects, cyclopamine-treated embryos were more transparent than controls. In normal development, most of the coelom of the leech embryo becomes filled with mesenchyme and other mesodermal derivatives,so that the remaining space is reduced to a relatively narrow system of interconnecting channels (reviewed bySawyer, 1986)(Fig. 8C-E); this process failed to occur in embryos treated with cyclopamine(Fig. 8M-O). Moreover, in sectioned, cyclopamine-treated embryos, we were unable to recognize any morphogically defined gonads, which appear at this stage in normal embryos as U-shaped structures connected to the ventral ectoderm (seeFig. 6E). In contrast, the ventral blood vessel was still present(Fig. 8E′,O′),which is further evidence for the specificity of the cyclopamine effects.

hh-class genes are involved in various aspects of patterning in diverse species. Functional studies have been largely confined toDrosophila and vertebrates, which represent two of the three main clades of bilaterally symmetric animals, Ecdysozoa and Deuterostomia,respectively. In the work presented here, we characterized Hro-hh, ahh ortholog from Helobdella robusta, a glossiphoniid leech and a member of the third bilaterian clade, Lophotrochozoa. As in other animals that have been examined, Hro-hh is expressed zygotically in a variety of cells and tissues, including subsets of body wall, gonad, nervous system, muscle and gut. We find that Hro-hh is expressed by a subset of the mesodermal derivatives of micromere lineages in the foregut and also by mesodermal and possibly endodermal tissues in the midgut; analysis of cyclopamine-treated embryos indicates that Hro-hh signaling is required for normal development of the proboscis and anterior midgut, and for mesenchymal tissues that normally come to fill much of the coelom in leeches.

Based on these results, we can try to interpret which elements of theHro-hh expression pattern are responsible for the developmental defects produced by cyclopamine. For example, it seems likely that the expression of Hro-hh by the radial and/or longitudinal muscle fibers of the proboscis is required for the normal development of the circular muscle fibers derived from micromere dm′. The source of the Hro-hh signals that regulate crop morphogenesis and mesenchyme formation is less obvious, but they may originate in the faint stripes of germinal plate expression seen at stage 9 (Fig. 5H).

Though expression of hh and en are tightly coupled during segmentation in insects, we found no evidence for patterned expression of a leech hh gene in the germinal bands or early germinal plate at the time and place where the segmental pattern of engrailed expression is first manifest (Wedeen and Weisblat,1991; Lans et al.,1993), nor is there a clear correlation between the patterns ofhh- and en-class gene expression in the ventral ganglia. Subject to the caveats necessary for any negative result, the apparent dissociation of en and hh gene expression during leech segmentation is consistent with other studies suggesting that the segmentation processes in Helobdella and Drosophila are not homologous at the molecular level (Iwasa et al.,2000; Pilon and Weisblat,1997; Savage and Shankland,1996; Seaver and Shankland,2000; Seaver and Shankland,2001; Shain et al.,1998; Song et al.,2002).

Given the diverse functions of hh-class genes in various organisms, what can we conclude regarding the function of thehedgehog gene in the common ancestor of the three main bilaterian clades? The eponymous hh was identified as a segment polarity gene inDrosophila, and vertebrate hh-class genes are critical for patterning limbs and somites. The question is still open(De Robertis, 1997;Dewel, 2000;Collins and Valentine, 2001;Valentine and Collins, 2000),but it seems unlikely to us that the ancestral bilaterian had these particular features, in which case the ancestral hh gene could not have functioned in any of those roles.

The first description of a hh homolog in Lophotrochozoa has recently been published for the limpet Patella vulgata, a gastropod mollusc; its expression in anterior and ventral midline ectoderm of the trochophore larva of that species was interpreted as supporting the dorsoventral axis inversion theory(Nederbragt et al., 2002). Our results, that Hro-hes is expressed at low levels throughout the epidermis of the germinal plate of the leech, but with no particular relationship to the developing nerve cord, does not particularly support this interpretation. Of course, Helobdella is a derived annelid, so the expression and function of Hro-hh may bear little resemblance to those of the ancestral hh gene. But the same argument holds for all extant species, including those whose external morphology resembles ancestral forms. Thus, whether the pattern of hh-class gene expression seen in the Patella trochophore reflects its expression in the urbilateria or an adaptation associated with torsion of the molluscan body plan during gastropod evolution, for example, remains to be seen. In any event, the expression of hh-class genes in the nervous system of vertebrates,insects, annelids and (presumably) mollusks suggests that the ancestralhh gene(s) may have been involved in some aspect of neural differentiation.

Another candidate function of the ancestral hh gene is in gut formation (Pankratz and Hoch,1995). That the gut is a plesiomorphic trait of bilaterian animals seems beyond dispute, and it has been suggested that signaling between invaginating gut and ectoderm, leading to the formation of stomodeum/foregut and/or proctodeum/hindgut, was among the first inductive interactions to have evolved (Wolpert, 1994). In chick, sonic hedgehog (shh, one of 3 hh-class genes known in vertebrates) is expressed in the endodermal epithelium throughout much of the gut (Echelard et al.,1993; Roberts et al.,1995; Roberts et al.,1998; Sukegawa et al.,2000). In Amphioxus, which arose from a basal branch of the chordate lineage, the one known hh-class gene is also expressed in endoderm, among other tissues (Shimeld,1999). In Drosophila, epithelial expression ofhh is an essential aspect of foregut and midgut formation, but in contrast to chick and Amphioxus, the source of the signal is ectodermal rather than endodermal (Hoch and Pankratz, 1996; Pankratz and Hoch, 1995). And now in leech, we find hh-class gene expression primarily in mesoderm, the third germ layer, during gut formation.

Similarities between hedgehog signaling pathways may also extend to the level of target genes. The inductive Shh signal regulates the concentric patterning of the surrounding mesenchyme in chick, negatively regulating the differentiation of smooth muscle, in part by activation of BMP4, a member of the TGFβ superfamily, in those cells(Roberts et al., 1995;Roberts et al., 1998;Sukegawa et al., 2000). Members of the multi-gene NK-2 family of homeobox genes are also activated by Shh signaling in vertebrates(Barth and Wilson, 1995;Briscoe et al., 1999). InAmphioxus, AmphiBMP2/4 is expressed in hypoblast and endodermal derivatives (Panopoulou et al.,1998), and AmphiNk2-2 is expressed in anterior nervous system and endoderm (Holland et al.,1998). In Drosophila, as in vertebrates, hhactivates the expression of TGFβ-related genes (dpp and60A) in adjacent visceral mesoderm. Moreover, an NK2-class gene (tinman) is also a downstream target of hh signaling inDrosophila heart differentiation(Yin and Frasch, 1998). In leech, Lox10, an NK2-class gene, is known to be expressed in the crop (Nardelli-Haefliger and Shankland, 1993); a TGFβ-related gene has been identified, but its expression remains unknown (Isaksen, 1992). Thus,comparison of hh-class gene expression and function in a vertebrate(chick), fly and leech reveals some parallels, supporting the hypothesis that signaling in gut formation was an ancestral role for hh-class gene in ancient bilaterians, but there are also significant differences. Further work should reveal whether this situation reflects divergence or convergence.

This work was supported by NSF grant IBN 9723114 and NIH grant RO1 GM 60240 to D.A.W., by a fellowship from the Korean Research Foundation to D.K. and by NASA grants NAG2-1349 to S.M.S. and NAG2-1359 to D.A.W. We thank Foster C. Gonsalves for assistance in preparing Fig. 3, and members of the Weisblat and Shankland labs for helpful discussions.