Distribution of Sonic hedgehog peptides in the developing chick and mouse embryo

Regulation of axial polarity in the vertebrate central nervous system (CNS), somite and limb appears to depend on intercellular signals. For example, in the CNS, contact-mediated signals emanating from the underlying notochord are thought to initiate the development of a specialized population of ventral midline cells, the floor plate, in overlying neural ectoderm (Yamada et al., 1991; Placzek et al., 1993). Diffusible signals from the floor plate and notochord then induce the formation of motor neurons on either side of the floor plate (Yamada et al., 1991, 1993). Similarly, the notochord and floor plate appear to be the source of signals responsible for sclero-tomal development in the somite (Porquie et al., 1993; BrandSaberi et al., 1993). Thus, ventral polarity in the CNS and somite is regulated by these two signaling centres. In the limb, a signaling centre localized in posterior mesenchyme, the zone of polarizing activity (ZPA), controls anterior-posterior polarity (Saunders and Gasseling, 1968). Several lines of evidence suggest that the ZPA is the source of a diffusible factor or factors whose concentration determines polarity: high concentrations for posterior structures, low concentrations for more anterior limb elements (Tickle, 1981). Interestingly, ZPA activity is found in the notochord and floor plate (Wagner et al., 1990), suggesting that common signals may regulate patterning of the CNS, somites and limbs.

Recent studies have identified a family of vertebrate genes encoding putative signaling molecules related to the Drosophila segment polarity gene, hedgehog (hh). One member of this family, Sonic hedgehog (Shh), is expressed in the notochord, floor plate and ZPA of the limb (Echelard et al., 1993; Krauss et al., 1993; Riddle et al., 1993; Chang et al., 1994; Roelink et al., 1994). Moreover, functional studies in zebrafish, chick, rats and mice indicate that Shh may encode a common signal regulating polarity in the CNS, somite and limbs. Ectopic expression of Shh in the CNS in vivo, or application of Shh-expressing cells or purified peptides to neural explants in vitro, leads to induction of floor plate and motor neurons (Echelard et al., 1993, Krauss et al., 1993; Roelink et al., 1994; Martí et al., 1995). These results indicate that Shh peptides may mediate both short-(floor plate) and long-(motor neuron) range induction in the vertebrate CNS. In the somite, similar approaches have shown that Shh can induce the expression of Pax-1, a sclerotomal marker (Fan and Tessier-Lavigne, 1994; Johnson et al., 1994), whilst in the limb, Shh-expressing cells placed at the anterior margin result in mirror-image duplications of digits, similar to those produced in response to grafts of ZPA cells (Riddle et al., 1993; Chang et al., 1994).

These experimental studies are consistent with Shh playing a major role in patterning, acting as a short -(floor plate) and long-(motor neuron, limb and sclerotome) range signal. Similarly, Drosophila Hh has been proposed to act locally, in limb development, as well as at a distance, in cellular patterning of the epidermis and in driving morphogenesis of the eye (Ingham, 1993; Basler and Struhl, 1994; Heemskerk and DiNardo, 1994; Tabata and Kornberg, 1994). These results are supported by antibody studies which indicate that Hh is secreted and can be detected a few cell diameters beyond cells transcribing the hh gene (Taylor et al. 1993; Lee et al. 1994; Tabata and Kornberg, 1994).

In the vertebrate, processing and secretion of Shh has been addressed in a variety of cultured cells. The primary Shh translation product enters the secretory pathway, undergoes signal peptide cleavage and glycosylation at a single conserved N-linked glycosylation site and is cleaved into two peptides: an amino 19×10³M_r species and a carboxyl 27×10³M_r glycosylated form (Chang et al., 1994; Lee et al., 1994; Bumcrot et al., 1995). Interestingly, processing is autocatalysed, the catalytic site mapping to the carboxyl peptide (Lee et al., 1994). Both peptides are secreted, although only the larger form is readily detected in the supernatant when the mature protein is processed (Lee et al., 1994; Bumcrot et al., 1995). Initially it was suggested that the two secreted peptides may independently mediate short-or long-range signaling functions (Lee et al., 1994). However, recent results indicate that all signaling activity resides in the 19×10³M_r peptide in Drosophila Hh (Fietz et al., 1995; Porter et al., 1995) and mouse Shh (Martí et al., 1995).

We have generated peptide-specific polyclonal antibodies and have used these to investigate the expression of Shh in the notochord, CNS and limb of mouse and chick embryos. In general, the temporal expression of Shh is consistent with its postulated signaling functions. However, the 19×10³M_r peptide is detected in close association with cells transcribing the Shh gene. Thus, we are unable to detect significant diffusion of the Shh signal in the vertebrate embryo.

Whole-mount in situ hybridization was performed according to Parr et al. (1993). Digoxigenin-labelled RNA probes were prepared according to Wilkinson (1992). Mouse and chick Sonic hedgehog probes have been described in Echelard et al. (1993) and Riddle et al. (1993).

Generation of Ab80 has been described elsewhere (Bumcrot et al., 1995). Ab79 was raised against a bacterially expressed, hexa-histidine tagged protein corresponding to approximately the amino terminal two thirds of mouse Shh. The antiserum was affinity purified as described elsewhere (Bumcrot et al., 1995). Ab77 was raised against a glutathione-S-transferase fusion protein containing amino acids 369 to 425 of chick Shh (Riddle et al., 1993). For all three antisera, inoculation of New Zealand White rabbits, as well as test and production bleeding, was carried out at Hazleton Products, Inc.

COS cells were transfected with mouse and chick Shh expression constructs as described (Bumcrot et al., 1995). Cell lysates were prepared and analyzed by immunoblotting as described (Bumcrot et al., 1995). Affinity-purified Ab79 and Ab80 were each used at a dilution of 1:200.

Embryos were fixed in 4% paraformaldehyde in PBS (pH 7.4) for 16 hours at 4ºC, rinsed in PBS pH 7.4, dehydrated in an ethanol series and stored in 100% ethanol at −20ºC. Before immunostaining, embryos were rehydrated through an ethanol series. Endogenous peroxidase activity was inactivated with 1% H₂O₂ in PBS. Embryos were then rinsed in PBT (PBS containing 0.5% Triton X-100) and blocked for 2 hours at room temperature, using 10% heat-inactivated sheep serum and 1% bovine serum albumin (BSA) in PBT. Embryos were incubated overnight at 4ºC with antisera diluted 1:500 in PBS containing 1% sheep serum and 0.1% BSA and 0.5% Triton X-100. Shh antibodies were detected with goat anti-rabbit-HRP (1:250) (Jackson Immunochemicals), followed by DAB color detection, or with biotinylated goat anti-rabbit antibodies (1:500) (Jackson Immunochemicals) and a streptavidin-conjugated β-gal (1:500) (Boehringer) followed by β-galactosidase color detection.

Embryos were mounted in 85% glycerol in PBT, or were cleared in benzyl alcohol-benzyl benzoate (1:2) before photographing with an Olympus SZH stereomicroscope.

Antibody staining on cryostat (10 μm) or paraffin (10 μm) sections gave similar results and thus were used interchangeably. Sonic hedgehog antibodies were used at a dilution of 1:500.

A rabbit polyclonal antibody to the LIM homeodomain protein islet-1 (kindly provided by T. Edlund) was used in order to detect early differentiating motor neurons. A mouse monoclonal antibody to the IgG-like glycoprotein SC-1/BEN (Developmental Studies Hybridoma Bank) was used in order to stain floor plate and early differentiating motor neurons. HNF-3β was detected using a rabbit polyclonal anti-HNF-3β antibody (H. Sasaki, unpublished data). Comparison of immunostaining in mouse embryos with previous reports of expression of related family members, including HNF-3α, together with western blot analysis of cross reactivity, suggest that this antibody is HNF-3β specific. Rabbit primary antibodies were detected with either HRP-conjugated goat anti-rabbit or AP-conjugated goat anti-rabbit. Mouse antibody was detected with HRP-conjugated goat anti-mouse.

Double labelling for RNA and protein was performed following the whole-mount in situ hybridization (Parr et al., 1993) with several modifications. The proteinase K and RNAse treatments were eliminated and postfixation was performed in 4% paraformaldehyde. After postfixation, embryos were photographed, dehydrated and paraffin embedded. Antibody staining was performed on 10 μm paraffin sections as described above.

Double labelling for two different antigens was performed simultaneously on either cryostat or paraffin sections. Antibody detection was performed by incubating with the relevant secondary antibodies (anti-rabbit AP, and anti-mouse HRP), followed by sequential color detection.

Polyclonal antibodies were raised in rabbits against mouse and chick Shh proteins (Fig. 1A). A summary of these activities is presented (Fig. 1B). Ab77 was raised against a C-terminal region of the 26×10³M_r carboxyl peptide (C-peptide) of chick Shh. We have been unable to affinity purify this antiserum effectively due to the insolubility of the immunogen, thus western blot analysis of the antibody specificity has not been performed. However, on the basis of immunostaining of COS cells transfected with full-length and truncated mouse and chick Shh proteins, it is clear that this antiserum only cross-reacts with the unprocessed form of chick Shh and its C-peptide cleavage product. There is no cross reactivity with mouse Shh which is quite divergent in this region (Echelard et al. 1993). Ab79 was raised against a 28×10³M_r mouse Shh peptide which encompasses the entire N-peptide and a short region of the C-peptide. Western analysis of COS cells expressing chick and mouse Shh peptides indicates that Ab79 recognizes the fulllength and amino 19×10³M_r Shh peptides of chick and mouse Shh (Fig. 1C), as well as their zebrafish and human counterparts (data not shown). In addition, there is weak cross reactivity against both the carboxyl 26×10³M_r chick and 27×10³M_r mouse peptides (Fig. 1C). Ab80, which was raised against a small region of the N-peptide, has been described in detail elsewhere (Bumcrot et al., 1995). This antibody only recognizes the full-length and amino 19×10³M_r peptide of chick and mouse Shh (Fig. 1C). Cell culture and embryonic studies indicate that the vast majority of Shh protein is rapidly processed to generate the amino and carboxyl peptides (Bumcrot et al., 1995). Thus, it is likely that the various antisera only detect the terminal cleavage products in this study. Moreover, despite the extensive similarity amongst vertebrate Hh proteins in the N-peptide (Echelard et al., 1993), it is likely that our studies only detect Shh peptides. A detailed analysis of the expression of Shh, Dhh and Ihh shows no overlap in their respective expression domains in the notochord, CNS or limbs (M. Bitgood and A. P. M., unpublished data)

Whole-mount immunohistochemistry was performed to mouse and chick embryos from early head fold stages. Identical results were obtained with Ab79 and Ab80 in the mouse and with all three antisera in the chick. Results are presented for Ab80 (mouse) and Ab79 (chick) (Figs 2-4). At early head fold stages in the mouse (8.0 days postcoitum [dpc]) prior to formation of the 1st somite and in the stage 5-6 chick (Hamburger and Hamilton, 1951), Shh-expressing cells extend from the node into the node-derived cells of the head process which underlie the neural plate (Fig. 2A,B; data not shown). Expression in the head process is mosaic, with Shh protein outlining single cells at the midline (Fig. 2B). By early somite stages, Shh-expressing cells extend more anteriorly, into the prechordal plate mesoderm underlying the forebrain of the chick (Fig. 3A,B), and caudally along the entire length of the head process and notochord into the node in both the chick and mouse (Figs 2C, 3A-E). Expression is lost in the prechordal plate of the chick by stage 15 (Fig. 4A,B), so that the chick and mouse at this stage (mouse 9.5 dpc; 20-25 somites) show similar mesodermal expression, along the entire length of the head process and notochord (Figs 2D, 4A-D). In the chick at stage 18 and in the mouse at 11.5 dpc, Shh expression becomes restricted to the notochord caudal to the forelimb buds (data not shown).

Shh is detected in the CNS of the mouse in the ventral midbrain at the 8-somite stage (Fig. 2C) and slightly earlier in the chick. Expression then extends rostrally and caudally such that, by stage 10 in the chick, Shh peptides are present in the ventral CNS from the forebrain to the spinal cord up to the level of the fifth somite (Fig. 3A-E). By 9.5 dpc in the mouse, approximately stage 15 in the chick, Shh is restricted to the ventral midline in most of the brain, with the exception of the rostral diencephalon where expression skirts the presumptive hypothalamic region, the paramedian plate (Figs 2D, 4B). In the spinal cord, ventral midline expression extends caudally to the position of the 4th and 5th most recently formed somites (Figs 2D, 4A,D). Thus, there is a continuous ventral strip of Shh expression at, or close to, the ventral midline of the CNS extending from the rostral limit of the CNS in the forebrain to caudal regions of the spinal cord. This expression persists in the mouse until at least 15.5 dpc (data not shown). In addition, there is a dorsal expansion of Shh expression at the zona limitans intrathalamica, a major pathway of axonal migration that lies at the boundary between prosomeres 2 and 3 in the diencephalon (Figdor and Stern, 1993), which is visible in the stage 20 chick (Fig. 4E) and 10.5 dpc mouse brain (data not shown).

Whole-mount analysis of Shh expression at the midline suggests that Shh peptides are restricted to a domain very similar to that previously reported for Shh transcripts (Echelard et al., 1993; Riddle et al., 1993). To examine the distribution of Shh peptides more closely, particularly with relation to ventral patterning of the CNS, we performed a number of single-and double-labelling studies on sections of chick and mouse embryos.

As with earlier studies, similar results were obtained with the various antisera, with one exception. Ab77 and Ab79, but not Ab80, detect a halo of Shh peptide surrounding the notochord of the stage 15 chick embryo at the brachial level (Fig. 5A-C), presumably reflecting the distribution of the carboxyl 26×10³M_r peptide, which has been shown to be more freely diffusible in cell culture (Bumcrot et al., 1995). In all other respects immunostaining is identical and similar to that obtained at brachial levels of the mouse using either Ab79 or Ab80 (Fig. 5D,E). In both the notochord and ventral CNS, Shh peptides are pericellular, indicating secretion of Shh protein. Moreover, in the ventral spinal cord, most of the protein accumulates at the luminal cell surface in a region corresponding, approximately, to the limits of the morphologically defined floor plate (Kingsbury, 1930).

Double labelling for expression of Shh RNA and protein in the mouse (Fig. 5E) and chick (data not shown) indicates that peptides are only detected on, or very close to, the surface of cells transcribing the Shh gene. Thus, we are unable to detect any diffusion of the 19×10³M_r signaling peptide. At a later stage, in the spinal cord of the 9.5 dpc mouse embryo at the brachial level (Fig. 5F,G), we observe a similar tight relationship between the distribution of Shh protein and Shh-transcribing cells. However, both now appear to extend ventrolateral to the morphologically defined floor plate. This expansion is also apparent in stage 18 chick embryos (data not shown). Interestingly, at 10.5 dpc, when commissural axon tracts project contralaterally across the floor plate, Shh peptides become widely dispersed beyond the floor plate cells that are transcribing the Shh gene (Fig. 5I), suggesting that the 19×10³M_r peptide may adhere to the surface of these migrating axons. It is clear that this, and all other aspects of antibody immunoreactivity at each of the stages examined, are specific since all immunostaining is blocked by preincubation of antisera in recombinant Shh peptide (Fig. 5H and data not shown).

It has been suggested that Shh expression at the midline might be activated by HNF-3β, a member of the winged-helix class of transcriptional regulators, and that HNF-3β and Shh may subsequently regulate each other’s expression by a positive feedback mechanism (Echelard et al., 1993). To explore the relationship between these two factors, mouse embryo sections were double-stained for Shh RNA, which is spatially and temporally equivalent to the domain of Shh protein expression (see above), and for HNF-3β protein, using an HNF-3β-specific antibody (H. Sasaki, unpublished data). As expected from earlier studies (Echelard et al., 1993), HNF-3β expression in the ventral CNS precedes that of Shh (data not shown), consistent with HNF-3β playing a role in the activation of Shh. By 9.5 dpc, Shh expression is also detected in the ventral spinal cord at brachial levels. Interestingly, there is a graded distribution of HNF-3β expression; highest ventrally in the floor plate cells that also express Shh, and weaker laterally in cells outside of the Shh expression domain (Fig. 6A). Thus, the subsequent maintenance of high levels of HNF-3β correlates with the presence of high levels of Shh peptides. These results raise the interesting possibility that cells expressing high versus low levels of HNF-3β may have different fates, floor plate and motor neurons, respectively. That there are dosage requirements for HNF-3β for normal development of the nervous system is suggested by the neurological phenotype observed in mice lacking a single copy of the gene (Ang and Rossant, 1994; Weinstein et al., 1994).

In vitro studies have demonstrated that the 19×10³M_r Shh peptide is both necessary and sufficient for the induction of motor neurons in vitro (Martí et al., 1995), as judged by the activation of an early motor neuron marker, Islet-1, a LIM domain transcription factor (Thor et al., 1991; Yamada et al., 1993). Thus, Shh produced by the notochord and floor plate may play a role in motor neuron induction in vivo. This conclusion is supported by double labelling studies in stage 18 chick embryos. Clearly, Islet-1 and Shh-expressing cells partially overlap in the ventral CNS (Fig. 6B). Thus at least some motor neuron precursors appear to arise from, or are in contact with, Shh-expressing cells. Shh expression was also compared to that of the IgG-related glycoprotein SC-1/BEN (Tanaka and Obata, 1984; Yamada et al., 1991), a floor plate and motor neuron marker. At the ventral midline, SC-1/BEN is restricted to floor plate cells, whilst Shh expression extends ventrolaterally, abutting the most ventral SC-1/BEN-expressing motor neurons (Fig. 6C). Thus, even in the absence of appreciable diffusion of Shh protein, Shh expression is consistent with the direct induction of at least the ventralmost motor neurons.

Normal expression of Shh in the ZPA (Riddle et al., 1993; Chang et al., 1994), together with in ovo studies in which ectopic expression in the anterior mesenchyme of the limb leads to limb duplications (Riddle et al., 1993), support the hypothesis that Shh mediates the function of the ZPA. As the ZPA has been hypothesized to be the source of a long-range morphogen (Tickle et al., 1975; Tickle, 1981), the distribution of Shh peptides in the limb mesenchyme is of considerable interest.

Chick and mouse limbs express Shh peptides throughout the period in which Shh is transcribed in a region approximating to the operationally defined ZPA (Riddle et al., 1993). Similar results are obtained with antisera recognizing either Shh peptide (Fig. 7A,B). Immunoreactivity remains tightly localized to the posterior-distal limb mesenchyme, extending approximately 50 to 75 μm across the anterior-posterior axis between stages 20 to 25 (Fig. 7A-C). Expression of Shh RNA and protein in the posterior mesenchyme is not uniform. The highest levels of both are observed in cells immediately beneath the AER (Fig. 7C), although no immunoreactivity is seen in the AER itself. Further, isolated patches of small groups of posterior mesenchyme cells (3 to 5) are observed which clearly do not immunostain (arrows, Fig. 7C) despite the fact that these clusters are surrounded by cells transcribing Shh, which have Shh protein on their surface. Thus, although we can see differences in Shh expression in the posterior limb bud mesenchyme, we cannot detect any long-range diffusion of Shh protein.

Our results provide evidence that Shh peptides are expressed in midline cell types, from gastrulation stages, in the chick and mouse embryo. In the mesoderm, Shh expression extends from the most rostral midline mesoderm of the prechordal plate, underling presumptive forebrain regions, to the notochord at spinal cord levels. By early somite stages Shh peptides are present at the ventral midline of the presumptive CNS, extending rostrally and caudally, to occupy most of the ventral midline. Later an exception is seen in the rostral diencephalon where Shh is not expressed in the midline of the presumptive hypothalamic region. Whilst Shh expression overlaps the classically defined wedge-shaped region termed the floor plate, it is clear that Shh-expressing cells also extend laterally, beyond the floor plate. Thus, at spinal cord levels, the domain of Shh expression is considerably broader than that of the floor plate marker, SC-1/BEN. In the most rostral regions of the brain, which do not exhibit a floor plate morphology, Shh is also expressed in midline cells. Therefore, expression of Shh does not demarcate a particular cell type, but rather a position within the ventral CNS.

Several processes have been associated with the signaling properties of the notochord and floor plate region; the induction or self induction of floor plate, the specification of neuronal identity (motor neurons) and the induction of sclerotome development in the ventral somite. Recent studies, both in vitro and in vivo, indicate that all these activities may be attributed to Shh. Thus, either ectopic expression of Shh in the embryo (Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994) or application of Shh-expressing cells (Roelink et al., 1994) or Shh peptides (Martí et al., 1995) to neural explants in vitro leads to floor plate induction. Moreover, purified Shh at the appropriate concentration can induce motor neurons in the absence of floor plate induction (Martí et al., 1995). Finally, Shh can activate the ventral sclerotomal marker Pax-1 (Fan and Tessier-Lavigne, 1994; Johnson et al., 1994). These inductive events, which apparently involve both short-and long-range interactions, all appear to be mediated by the 19×10³M_r Shh peptide (Martí et al., 1995; unpublished data). Similarly, all Drosophila Hh signaling can be ascribed to a homologous 19×10³M_r peptide (Fietz et al., 1995; Porter et al., 1995). Thus, a key question is how does the expression and distribution of this peptide correlate with these signaling activities?

Detailed analysis of the timing of floor-plate-and motorneuron-inducing activity by the notochord and floor plate has come from the studies of Placzek et al. (1993) and Yamada et al. (1993), which are summarized and compared to the expression of Shh peptides (Table 1). Data are based on the inducing properties of notochord and floor plate (excised at particular axial levels) on chick and rat neural explants in vitro. Shh expression in the stage 6 chick notochord is consistent with the floor-plate-inducing properties of the notochord at this time. However, strong Shh expression is also detected at stage 10, when floor-plate-inducing activity is apparently already lost from the notochord. In these experiments, floor plate induction is assayed after 24-48 hours of culture, and it is not clear for what proportion of this period the neural explant must receive a floor-plate-inducing signal to initiate floor plate development. As our data indicates that Shh expression in the notochord decreases after stage 15, there may be insufficient expression, either quantitatively or temporally, to effect floor plate induction. In contrast, both floor-plate-inducing activity and Shh expression are present at late stages (Table 1).

However, floor-plate-inducing activity appears to be present in the ventral neural plate prior to expression of Shh, although it is possible that explanted floor plate tissue may activate Shh during the course of the culture period, as would occur in vivo. In summary, our study is consistent with the conclusion that the 19×10³M_r Shh peptide may mediate the local induction of floor plate by the notochord and/or floor plate in vivo. The inconsistencies can be accounted for by the experimental approaches which do not assay inducing activity at the time of explant, but rather at some indeterminate time during the culture period.

A similar analysis of motor neuron induction indicates a striking correlation with the expression of Shh in the inducing tissues (Table 1). Interestingly, notochord of stage 15/20 embryos is unable to induce floor plate but does induce motor neurons. As Shh levels in the notochord are decreasing after stage 15, the changing inductive capacity of the notochord may reflect decreasing levels of Shh activity (see later). A detailed analysis of sclerotomal-inducing activity in the notochord and floor plate has not been performed, although it is clear that Shh is expressed in the inducing tissue at the time of grafting (Fan and Tessier-Lavigne, 1994). Thus, the temporal expression of Shh in midline tissues is generally consistent with its mediating inductive events in the spinal cord and somites.

In contrast to floor plate induction, motor neuron and sclerotome induction do not appear to require contact with the inducing tissue, suggesting that long-range inductions are mediated by a diffusible signal. However, we have been unable to detect appreciable diffusion of the 19×10³M_r Shh peptide in vivo. There are several possible explanations that may reconcile the various data.

For example, whereas in vitro a diffusible signal from midline tissues may induce motor neurons and sclerotome, this may not be the case in vivo. Our results indicate that some ventral motor neurons may develop in direct contact with cell bound Shh; however, this is unlikely to be true of all motor neurons. Alternatively, Shh may induce the expression of a second diffusible signal which mediates long-range signaling. This does not appear to be the case in the somite at least (Fan and Tessier-Lavigne, 1994). The simplest explanation, and the one we favor, is that long-range signaling is mediated by a diffusible form of the 19×10³M_r peptide, but this cannot be detected using current procedures. If correct this model would predict that floor plate induction may require high concentrations or a cell surface bound form (or both), whilst motor neuron and sclerotome induction may be initiated in response to low amounts of diffusing Shh protein. Indeed, in vitro studies indicate that motor neuron induction can occur over a 100-fold concentration range of the 19×10³M_r peptide, which is unable to induce floor plate (Martí et al., 1995).

Interestingly, our studies demonstrate a distinct polarity to the distribution of Shh peptides on the surface of ventral midline cells in the CNS. Shh accumulates towards the luminal, and to a lesser extent, the basal surface, of the cell. It is tempting to speculate that if there is diffusion of the 19×10³M_r peptide in the spinal cord, it is through the lumen acting at the luminal surfaces of responding cells.

It is clear from the work reported here that Shh peptides are present in the midline mesoderm and ventral CNS at all axial levels. Interestingly, grafts of trunk notochord induce serotoninergic cells in hindbrain explants (Yamada et al., 1991), whilst floor plate from the spinal cord induces dopaminergic neurons in midbrain explants (Hynes et al., 1995). The induction of the appropriate ventral cell type for each axial level suggests that different responses may be mediated by common midline signals, the specificity of the response being determined by the receiving tissue. Indeed, studies on the zebrafish embryo indicate that the ventral midline of the CNS has common properties along its entire length (Hatta et al., 1991, 1994). The expression of Shh makes it an attractive candidate for this common ventralizing activity. Thus, the response to Shh signaling in the CNS may be governed by both the distance from the source of Shh-expressing cells and the position of responding cells along the anterior-posterior axis.

Although there is evidence that Shh may play an inductive role in the early CNS, its expression at later stages suggests additional functions. At the ventral midline, it appears that Shh peptides move laterally on the surface of migrating axons. In the brain, we observe strong expression of both Shh peptides at the zona limitans intrathalamica, separating prosomeres (P) 2 and 3 in the rostral diencephalon (Figdor and Stern, 1993; Puelles and Rubinstein, 1993). Expression here is associated with the axon tracts of a major dorsal projection of ventral neurons, the mamillothalamic tract. Thus, there is some suggestion that Shh may play a role in axonal outgrowth or guidance. Alternatively, such boundaries may initiate or maintain an underlying segmental organization in the forebrain (Puelles and Rubinsten, 1993). Interestingly, Drosophila Hh participates in a reciprocal signaling pathway, which maintains the parasegmental boundary. A further parallel with the Drosophila embryo can be drawn from the fact that this strip of Shh-expressing cells in the diencephalon separates cells in P2 that express Wnt-3 from those in P3 expressing Dlx-1/2. (Puelles and Rubinstein, 1993). Drosophila counterparts of these genes, wingless and distalless, respectively, are regulated by Hh (Ingham 1993; Diaz-Benjumea et al., 1994).

The results of elegant in vivo surgeries in the chick limb suggest that the cells of the ZPA produce a signal which acts over 200 μm, or 20 cell diameters, to pattern the anteriorposterior axis of the limb (Honig, 1981). High levels of this signal are thought to be required for the specification of posterior limb elements, whilst lower levels are sufficient for anterior structures (Tickle et al., 1975; Tickle, 1981). Thus, the limb offers one of the best examples in vertebrate embryogenesis where a long-range morphogen may play a significant role. Shh is expressed in, and mediates the activity of the ZPA of the limb (Riddle et al., 1993; Chang et al., 1994), and is thus a good candidate for this postulated morphogen. We show that Shh peptides are expressed in the posterior distal limb bud mesenchyme at the appropriate stages. There is a graded distribution of both RNA and protein, which are highest in the vicinity of the AER, but detectable peptide extends no more than 50 to 75 μm along the anterior-posterior axis. Thus, we are unable to demonstrate any long-range diffusion of Shh in the limb. As argued in the previous section, this most likely reflects the technical difficulty of detecting low amounts of protein, although the alternative explanation of a second signal has to be treated seriously.

Recently, BMP-2, a member of the TGF-β family of secreted peptide factors, has been shown to be expressed in the ZPA and appears to extend anterior to the domain of Shh expression (Vogel and Tickle, 1993). Interestingly, Dpp, a close relative of BMP-2, is activated in response to Hh signaling in Drosophila (Heberlin et al., 1993; Ma et al., 1993). Moreover, ectopic activation of Dpp, in the absence of Hh, gives similar pattern alterations to ectopic expression of Hh (Capdevila and Guerrero, 1994; Ingham and Fietz, 1995). Thus, Dpp may have long-range activity and may mediate the dose-dependent effects of Hh. Whether BMP-2, or another related factor, mediates the apparent long-range action of Shh in the vertebrate limb remains to be determined. Experiments implanting BMP-2 containing beads into the anterior margin of the chick limb have not resulted in alteration of limb pattern (Vogel and Tickle, 1993).

This work was supported by grants from the HFSP and NIH to A. P. M., and by the Spanish Ministry of Education (E. M.) and the Muscular Dystrophy Association (D. A. B.). We wish to thank Bianca Klumpar for her technical assistance. We would like to thank Dr Thomas Edlund for the gift of the anti-islet-1 antiserum and Dr Cliff Tabin for the chick Shh cDNA clone.