Region-specific requirement for cholesterol modification of sonic hedgehog in patterning the telencephalon and spinal cord

The appearance of distinct neuronal cell types at defined positions in the ventral neural tube is dependent on inductive signaling mediated by sonic hedgehog (Shh) (Agarwala et al.,2001; Dale et al.,1997; Fuccillo et al.,2006; Jessell,2000; Sur and Rubenstein,2005). In mice lacking Shh, most ventral neuronal cell types are lost along the rostral-caudal axis of the neural tube, whereas dorsal structures are expanded ventrally (Chiang et al., 1996; Rallu et al.,2002). It is thought that neurons generated in progressively more ventral regions in the spinal cord require correspondingly higher levels of Shh for their induction (Briscoe and Ericson, 2001; Jessell,2000). In the basal ganglion of the telencephalon, the medial ganglionic eminence (MGE) and lateral ganglionic eminence (LGE), which give rise to the globus pallidus and striatum, respectively, are generated sequentially in response to Hedgehog (Hh) signaling(Ericson et al., 1995; Kohtz et al., 1998; Shimamura and Rubenstein,1997).

Although Shh can elicit diverse cell fates in the neural tube, much less is known about the factors that regulate Shh range and activity gradient. Active Shh is dually lipid-modified (ShhNp), with palmitic acid at its N-terminus and cholesterol at its C-terminus (Mann and Beachy, 2004). The hydrophobic nature of lipids indicates that there might be a dedicated release mechanism for Shh. Disp1, a multi-pass transmembrane protein, appears to fulfil this mechanistic requirement(Caspary et al., 2002; Kawakami et al., 2002; Ma et al., 2002), as its function is necessary only for lipidated Shh secretion(Li et al., 2006; Tian et al., 2005). Lipid-modified Shh has also been shown to form a large multimeric complex in vitro (Chen et al., 2004a; Zeng et al., 2001). It has been suggested that the lipid moieties are embedded in the core of these complexes, analogous to micelles, thus facilitating Shh interaction with - and its spreading across - the extracellular matrix(Zeng et al., 2001). The biological function of lipid moieties in regulating Shh function has been investigated recently. Misexpression studies in rats indicated that palmitoylation is required for Shh to induce ectopic ventral cell fates in the dorsal telencephalon (Kohtz et al.,2001), suggesting that palmitoylation is required for Hh activity. Indeed, mice deficient in an enzyme that catalyzes palmitoylation, or carrying a variant of Shh that is incapable of being palmitoylated, showed developmental defects similar to those of loss of Shh function(Chen et al., 2004a).

In contrast to palmitoylation, lack of cholesterol modification (ShhN) does not appear to significantly affect the intrinsic potency of Shh. It has been shown that ShhN isolated from tissue culture cells can induce ectopic ventral cell types in neural explants with comparable or higher efficiency than ShhNp at similar concentration (Feng et al.,2004). Instead, the cholesterol moiety appears to affect the range of Hh spreading in the target field. Recent mosaic studies with Drosophila wing imaginal discs showed that Hh lacking cholesterol(Hh-N) has an extended range of spreading and for inducing ectopic low-threshold Hh target genes, although the expression of high-threshold target genes near the source was reduced(Callejo et al., 2006; Dawber et al., 2005; Gallet et al., 2006). However,another report indicated that wing disc clones that expressed Hh-N were unable to elicit low-threshold Hh target genes at a distance(Gallet et al., 2006). These differences might be associated with different levels of Shh that are expressed at the sites of synthesis(Guerrero and Chiang,2006).

Previous efforts to understand the role of cholesterol modification concluded that the cholesterol moiety is required for long-range spreading and signaling in the mouse limb buds (Lewis et al., 2001). Accordingly, it has been proposed that cholesterol modification is required for the formation of Shh multimers involved in long-range signaling and proper vertebrate forebrain patterning(Feng et al., 2004). However,a more recent study suggested that ShhN retained some paracrine activity because it was capable of inducing several ventral neuronal cell types in the spinal cord (Tian et al.,2005). Given this controversy, the role of the cholesterol moiety in Shh function has been re-assessed recently in the limb bud(Li et al., 2006). That study found that cholesterol modification is required to restrict the spread of Shh across the anteroposterior axis of the limb bud, similar to Hh-N function in Drosophila. The extended range of ShhN movement across the anteroposterior axis of the limb bud also resulted in an appreciable reduction of local ShhN levels, suggesting that the cholesterol moiety not only regulates the range, but also the shape of the Shh morphogen gradient.

In this study, we utilized mice exclusively expressing the ShhN form of Shh to address the function of the cholesterol moiety in patterning the spinal cord and telencephalon. We found that dorsoventral patterning of the telencephalon in embryos lacking cholesterol modification of Shh was severely compromised, as indicated by the lack of MGE progenitors and the appearance of widespread ectopic LGE progenitors in the dorsal telencephalon. By contrast,patterning of the ventral spinal cord was largely unaffected and no ectopic expression of ventral cell fates at dorsal sites was observed in Shh^N/- embryos. These findings suggest that patterning of the telencephalon is highly sensitive to alterations in Shh spreading behavior.

To generate conditional Shh^flox/+ mice, loxP sequences were inserted flanking the Shh processing domain, which is situated in exon 3 (Li et al., 2006). The resulting mice containing the Shh^flox (unrecombined)allele were maintained as heterozygotes, because homozygotes die soon after birth (Li et al., 2006). The Shh^flox allele has reduced capacity to generate lipid-modified Shh (ShhNp) owing to interference of the loxP site with the Shh autocatalytic processing reaction (Li et al., 2006) (data not shown), thus it should be considered hypomorphic. Embryos uniquely expressing Shh without cholesterol modification were generated by mating Shh^flox/+ animals with the Shh^+/-;Sox2-Cre deleter strain as previously described(Li et al., 2006). Shh^N/+;Disp1^-/- embryos were obtained from crosses between Shh^flox/+;Disp1^+/- and Sox2-Cre;Disp1^+/- mice. Disp1^+/- mice have been described (Ma et al.,2002). Sox2-Cre mice(Hayashi et al., 2002) and Ptch1-lacZ mice (Goodrich et al.,1997) were obtained from the Jackson Laboratory. All control embryos were either Shh^+/+or Shh^+/-;we have not observed any differences in the development or Shh staining patterns of these embryos (see Fig. S3 in the supplementary material).

Immunohistochemistry analyses were performed on tissue sections collected from OCT- or paraffin-embedded embryos as previously described(Li et al., 2006). The primary antibodies used were rabbit anti-Foxa2 (Li et al., 2007) (1:100), mouse anti-Nkx2.2 (DSHB, 1:2), mouse anti-Isl1/2 (DSHB, 1:10), sheep anti-Chx10 (Exalpha Biologicals, 1:50), mouse anti-En1 (DSHB, 1:5), mouse anti-Pax7 (DSHB, 1:2), mouse anti-Pax6, (DSHB,1:1), mouse anti-Nkx2.1 (Neomarkers, 1:100), rabbit anti-Nkx6.1 (1:3000, gift of Christer Betsholz, Karolinska Institute, Stockholm, Sweden), rabbit anti-Olig2 (1:3000, gift of Hirohide Takebayashi, Kyoto University, Yoshida,Japan), mouse anti-Mash1 (1:100, gift of Jane Johnson, University of Texas Southwestern Medical Center, Dallas, TX, USA), rabbit anti-Gsh2 (1:500, gift of Kenneth Campbell, University of Cincinnati, Cincinnati, OH, USA), mouse anti-Tuj1 (Sigma, 1:400), goat anti-Shh (Santa Cruz Biotechnology, 1:1000). For western analysis, protein lysate samples (200 μg each), collected from E15.5 whole brains, were resolved on 6% SDS-polyacrylamide gels. Gli3-190 and Gli3R species were detected using a Gli3 N-terminal-specific antibody as previously described (Litingtung et al.,2002).

5-Bromodeoxyuridine (BrdU) in vivo labeling and TUNEL analysis were performed as previously described(Litingtung et al., 1998). Proliferative and apoptotic cell countings were carried out with at least five sections from three different embryos for each genotype. To quantify BrdU- or TUNEL-positive cells, coronal sections of the telencephalic neuroepithelium were subdivided into three regions: ventral (v), dorsal (d) and dorsal midline(dm). v is defined as the region of neuroepithelial cells ventral to the dorsal edge of the morphologically distinct LGE (see Fig. 2, arrowhead). The dm region consists of 200 neuroepithelial cells situated at the center of the dorsal midline and its immediate neighbors. The d region is located between regions v and dm. The percentage of BrdU-positive cells was determined by counting these in segments of 200 cells within the designated region of the telencephalon and dividing by the total number of cells in the segment. All sections were counterstained with the nuclear dye DAPI to highlight the total number of cells. At least five segments in each region were counted to generate data for statistical comparison.

X-Gal staining was performed according to standard protocols(Hogan et al., 1994). Whole-mount and section in situ hybridizations were performed as described(Li et al., 2004). The following cDNAs were used as templates for synthesizing digoxygenin-labeled riboprobes: Shh (Li et al.,2006), Ptch1(Goodrich et al., 1996), Gli3 (Li et al.,2004) and Nkx2.1(Lazzaro et al., 1991).

To assess differences among groups, statistical analyses were performed using a one-way analysis of variance (ANOVA) with Excel (Microsoft) and significance accepted at P<0.05. Results are presented as mean±s.e.m.

To investigate the effect of ShhN on forebrain development, sections of control and Shh^N/- mutant embryos in the telencephalic region were stained with Hematoxylin and Eosin and analyzed for morphological differences. At E9.5, we did not detect any apparent external morphological abnormalities in the brains of Shh^N/- and control embryos(data not shown). At E10.5, however, Shh^N/- mutant forebrain exhibited several distinct malformations, including enlarged brain ventricles and narrowed frontal nasal processes(Fig. 1, compare C,D with the control in A,B). At E10.5, the MGE is evident as a neuroepithelial eminence protruding into the lateral ventricles of control embryo telencephalon; the MGE is also a proliferative zone that later gives rise to the globus pallidus(Fig. 1B,H). Strikingly, the Shh^N/- telencephalon failed to form MGE, but instead developed a flattened neuroepithelium (Fig. 1D,J). The LGE, a second morphologically distinct ventral structure, develops between E11.5 and E12.5 in control embryos and is situated between the cortex and MGE (Fig. 1H). No LGE bulge was observed in the E12.5 Shh^N/- telencephalon(Fig. 1J). The loss of morphologically distinct ganglionic eminences is indicative of impaired ventral patterning in the Shh^N/- telencephalon. Notably,the Shh^-/- telencephalon also did not display any ventral structures (Fig. 1E,F,K,L). In addition to ventral defects, dorsal telencephalic midline structures such as the hippocampus primodium (HP) and choroid plexus (CP) were also malformed in the Shh^N/- mutant (Fig. 1, compare B,H with D,J). Interestingly, the widely separated telencephalic ventricles in Shh^N/- clearly differed from that of the Shh^-/- embryo, in which the telencephalic ventricles were fused (Fig. 1, compare C with E), suggestive of an altered Shh signaling gradient in Shh^N/- forebrain rather than a loss of signaling activity. Analysis of E18.5 Shh^N/- embryos revealed several characteristics of the holoprosencephaly (HPE)-like phenotype including defective eye formation, hypoplastic mid-facial structures, a single centrally located nostril and the absence of olfactory bulb in the forebrain(Fig. 1M-R).

To determine whether altered cell apoptosis and/or proliferation might contribute to the defective development of ganglionic eminences in Shh^N/- mutants, we performed the TUNEL cell death and BrdU proliferation assays. Analysis of E11.5 embryos indicated that cell death in the Mash1-positive (Ascl1 - Mouse Genome Informatics) LGE domain and the cortex of the neuroepithelial layer and its surrounding mesenchyme was increased in the Shh^N/- telencephalon compared with the control (Fig. 2A,B,D,E). Intriguingly, the percentage of BrdU-positive cells in the proliferative zones of the LGE (Fig. 2H,K;42±3.7 versus 58±1.8, P<0.001) and the cortex(Fig. 2I,L; 50±1.8 versus 66±4.7, P<0.001) were moderately increased in the Shh^N/- telencephalon, raising the possibility that this increase in cell proliferation may be attributed to exposure of neural progenitors to low-level, long-range Shh signaling.

We next determined whether the capacity for neural progenitor cells to proceed with differentiation is altered in Shh^N/- mutant telencephalon. We stained Shh^N/- and control sections with an antibody against neuronal class III tubulin (Tuj1; Tubb3 - Mouse Genome Informatics) that marks all differentiated neurons. At E11.5, Tuj1-positive cells normally consisted of a thick layer of differentiated neurons within the subventricular zone and mantle zone of MGE and LGE(Fig. 2O). However, in Shh^N/- mutants, we observed significant reduction in the population of differentiating neurons within the presumptive LGE domain(Fig. 2T). There was no obvious difference in differentiation in the cortex and dorsal midline region between control and Shh^N/- embryos at this stage(Fig. 2P,Q,U,V). Similarly,Isl1, a ventral-specific differentiation marker, also showed significant reduction in the differentiation capacity of Shh^N/-ventral neural epithelium (Fig. 2R,S,W,X). Collectively, these observations suggest that increased apoptosis and reduced neuronal differentiation contribute to the defective development of LGE in Shh^N/- mutants.

In addition to basal ganglions, dorsal midline derivatives such as the secretory epithelial cells of the choroid plexus and hippocampus primodium were also affected in Shh^N/-(Fig. 1H,J). The rapid invagination and remodeling of the dorsal midline neuroepithelium is normally accompanied by reduced proliferation and increased cell death(Furuta et al., 1997; Nagai et al., 2000). Although cell death still occurred in the Shh^N/- dorsal roof(Fig. 2C,F; 41±5.7 versus 44±6.9, P=0.48), we observed a near 100% increase in cell proliferation in the roof and medial wall of the dorsal midline(Fig. 2J,M, arrowheads;34.4±1 versus 72±3, P<0.001), suggesting that ShhN might have long-range effects on dorsal midline development (see below). Taken together, these results indicate that the cholesterol moiety of Shh is required for normal development of both dorsal and ventral telencephalic structures.

To gain further insight into the effect of ShhN on telencephalic patterning, we performed molecular characterization by examining markers that are regulated by Shh signaling. Nkx2.1 (Titf1 - Mouse Genome Informatics) is a homeobox gene whose expression in the mouse is first detectable in the basal telencephalic neuroepithelium at the three-somite stage, and subsequently localizes to a morphologically distinct region, the MGE (Shimamura et al., 1995). At E12.5, Nkx2.1 was uniformly expressed in the MGE of the control embryo,whereas no Nkx2.1 expression could be detected in the Shh^N/- mutant telencephalon, indicating severe loss of MGE identity (Fig. 3A,B). Mash1, which encodes a neural basic helix-loop-helix transcription factor, is expressed in the pan-ventral telencephalic region, including the MGE and LGE (Casarosa et al.,1999; Horton et al.,1999; Lo et al.,1991). Mash1 expression in these regions is regulated by Hh signaling and marks a transient population of committed neural precursors before differentiation (Fuccillo et al.,2004). Although Mash1 expression was present in both MGE and LGE of control embryos, Mash1-positive cells encompassed a relatively large ventral portion of the Shh^N/- mutant telencephalon(Fig. 3E,F), indicating that LGE progenitor cells were present despite the lack of development of a morphologically distinct LGE structure. Pax6 expression is normally confined to the dorsal cortical region abutting the LGE boundary(Fig. 3I). We observed prominent expansion of Pax6 expression in the ventral neuroepithelium, with reduced levels in the presumptive LGE domain(Fig. 3J, arrowheads). The expanded Pax6 expression domain was observed in the Shh^-/-mutant telencephalon, which often collapsed at this stage with no expression of ventral markers such as Mash1 and Nkx2.1(Fig. 3D,H,L). Although these results point to a requirement for the cholesterol moiety of Shh in patterning the ventral telencephalon, the possibility remains that the expression level of the ShhN allele is not as high as the wild-type allele when it is expressed alone in Shh^N/- embryos, thus generating haploinsufficiency. Because it is not possible to generate ShhNhomozygous mutants (see Materials and methods), we compared the patterning capacity of the Shh^flox allele, in which the ShhNallele is generated upon Cre-mediated deletion of the sequence encoding the C-terminal autoprocessing domain. With the exception of the cholesterol moiety, the proteins generated from these two alleles are the same. However,the Shh^flox allele is hypomorphic in that it generates a lower level of bioactive lipid-modified ShhNp owing to loxP sequence interference with autoprocessing of the Shh precursor(Li et al., 2006) (data not shown). Thus, Shh^flox/- embryos died soon after birth with multiple defects, including cleft secondary palate and distorted eye formation(see Fig. S1A-E in the supplementary material), characteristics of impaired Shh signaling (Rice et al.,2004; Zhang and Yang,2001). However, dorsoventral patterning of the telencephalon was maintained in hypomorphic Shh^flox/- mutants, although a reduction of the LGE domain was evident(Fig. 3C,G,K). Therefore, the severe patterning defect in Shh^N/- is primarily attributed to failure of cholesterol modification rather than haploinsufficiency, which is also consistent with the ectopic activation of Shh pathway activity observed in the dorsal telencephalon (see below).

As noted above, we observed increased cell proliferation in the Shh^N/- telencephalon at E11.5, suggestive of long-range spreading of ShhN. Indeed, we detected ectopic expression of Shh-responsive genes that mark LGE neurons in the dorsal Shh^N/-telencephalon beginning at E13.5. Specifically, we found that Gsh2, a homeobox protein, was ectopically expressed in the dorsal midline region at E13.5 in the Shh^N/- telencephalon(Fig. 4C,Ca), as compared with its normally restricted expression in the ventricular zone of control MGE and LGE (Fig. 4A,Aa). Similarly,Mash1 was also ectopically expressed at the dorsal midline(Fig. 4B,D). By E15.5,expression of Gsh2 and Mash1 was detected in the ventricular zone of the striatum, a derivative of the LGE, but not in the cortical region(Fig. 4E-F). However, in Shh^N/- their expression was prominent along the dorsoventral axis of the telencephalic neuroepithelium(Fig. 4G-Hc). We found that the expression of Gli3, which encodes a zinc-finger transcriptional repressor in the absence of Shh signaling, was selectively excluded from the dorsal midline in Shh^N/- as well as control telencephalon(Fig. 4I-Ja). This observation is consistent with the finding that dorsal midline cells were more sensitive to the long-range patterning effect of ShhN than the lateral cortex where Gli3 is expressed.

The emergence of ectopic LGE marker expression and defective midline dorsal structures at late gestation in Shh^N/- are consistent with the notion that the dorsal telencephalon might be exposed to low levels of ectopic ShhN signal. Indeed, we detected low but consistent Shh pathway activation in the dorsal telencephalon of Shh^N/- mutants as indicated by sensitive Ptch1-lacZ reporter expression(Fig. 5B), whereas no dorsal Shh signaling was found in a comparable region of the control embryo(Fig. 5A). Because Shh pathway activation leads to the accumulation of full-length Gli3 (Gli3-190) through the inhibition of Gli3 repressor (Gli3R) formation(Litingtung et al., 2002; Wang et al., 2000), we determined the relative amount of Gli3R to Gli3-190 as another readout of Shh signaling. Consistent with widespread Ptch1 reporter expression, we observed as much as a 50% reduction in the Gli3R/Gli3-190 ratio in E15.5 whole-brain extracts of Shh^N/- mutants as compared with controls(Fig. 5C,D). Ectopic Shh signaling is not the result of aberrant ectopic Shh expression, as we did not observe Shh-expressing cells or their descendents in the dorsal telencephalon using the Shh-Cre transgenic line that we described previously(Li et al., 2006) (data not shown). Thus, it appears that the cumulative effect of long-range ectopic ShhN signaling resulted in the induction of ventral marker expression in the dorsal neural progenitor population.

In order to determine whether specification of ventral cell types was similarly compromised in Shh^N/- mutant spinal cord, as in the telencephalon, we analyzed the expression patterns of several homeodomain transcription factors that identify different cell types including neuronal progenitors in the spinal cord (Briscoe and Ericson, 2001; Jessell,2000). Pax7 is a class II factor that is repressed by Shh and is expressed in all dorsal progenitor domains, whereas Nkx6.1 (Nkx6-1 - Mouse Genome Informatics) is a class I factor that is induced by Shh and expressed in the broad ventral progenitor domains of V2 (p2), V3 (p3) interneurons and motor neurons (pMN). Similarly, Olig2 and Nkx2.2 (Nkx2-2 - Mouse Genome Informatics) are class I factors that are selectively expressed in the pMN and p3 progenitors, respectively. In Shh^N/- mutants, all ventral cell types were generated at defined locations(Fig. 6), in agreement with previously reported ShhN function in the spinal cord(Tian et al., 2005). The number of floorplate cells labeled by Foxa2 was comparable between control and Shh^N/- embryos (Fig. 6C,D,N). However, there was an overall reduction in the total number of ventral neuronal progenitor domains and corresponding Chx10⁺ (V2, P=0.001), En1⁺ (V1, P=0.001) and Isl1⁺ (MN, P=0.007) neuronal subtypes that correlated with the reduced neural tube size in Shh^N/- embryos (Fig. 6N,O). Taking into account the difference in neural tube size at E10.5 (the control spinal cord was about 44% larger than that of Shh^N/-; see Fig. S2J in the supplementary material), the Nkx6.1⁺ (p3, pMN and p2) progenitor domains in Shh^N/- were slightly reduced as compared with the control(Fig. 6M, 35±3% versus 31±3%, P=0.05). Notably, there was an apparent expansion of the Pax7⁺ progenitor domain(Fig. 6M, 49±2% versus 63±2%, P<0.001) and concomitant reduction of p0+p1 (V0 and V1 progenitors) progenitor domains (Fig. 6M, 16±2% versus 6.2±1%, P<0.001). The expansion of the Pax7⁺ domain is not a consequence of maintaining dorsal tissue mass as a result of reduced neural tube size, as ventral expansion of dorsal progenitors was still observed at E9.5 when Shh^N/- spinal cord exhibited a size comparable to the control (see Fig. S2A-H in the supplementary material). These results indicate that although neuronal subtypes are specified and patterned in Shh^N/- spinal cord, the cholesterol moiety of Shh is,nevertheless, required for normal expansion of ventral progenitors,particularly p0+p1 progenitor populations.

We next examined ShhN protein distribution in the telencephalon and spinal cord at stages when ventral neural progenitors were being specified. In the developing forebrain, Shh expression is first detected in the dorsal foregut endoderm (also referred to as the prechordal plate) underlying the future telencephalic neuroepithelium, and the ventral neuroepithelial expression of Shh does not occur until the eight-somite stage(Echelard et al., 1993; Shimamura and Rubenstein,1997) (Fig. 7A,C). In Shh^N/- embryos, Shh RNA expression in the dorsal foregut was maintained (Fig. 7E, arrow), but the level of ShhN protein was significantly reduced (Fig. 7G, arrow). Accordingly, Shh RNA expression in the overlying neuroepithelium was almost undetectable (Fig. 7E,G,arrowheads). By E9.5, Shh expression was completely abolished,although its expression in other regions of the brain was maintained(Fig. 7B,F). Similarly, we failed to detect ShhN protein in Shh^N/- ventral telencephalon, whereas Shh protein was always observed in the midline of control embryos (Fig. 7D,H). As expected, Nkx2.1 expression, which overlapped with Shh expression in the ventral telencephalon, was not observed(Fig. 7I,M). The expression of Ptch1, a downstream target of Shh signaling(Goodrich et al., 1996), was also severely compromised in this region(Fig. 7J,N). In contrast to the telencephalon, Shh transcript and ShhN protein expression in the floorplate and notochord at E10.5 were clearly detectable, but the ShhN protein level was significantly reduced(Fig. 7K,L,O,P). Interestingly,ShhN protein predominantly localized to the apical membrane of floorplate cells (arrowheads in Fig. 7P),whereas ShhNp protein localized to both apical and basolateral membranes. This is in agreement with a recent study in Drosophila in which Hh-N accumulated only at the apical membrane(Callejo et al., 2006). Intriguingly, we found that ShhNp protein was distributed along the apical cell membrane of ependymal cells (Fig. 7L, arrows), which line the central canal of the spinal cord. By contrast, we never observed ependymal localization of ShhN in the mutant spinal cord. It is possible that this ependymal cell layer represents an active site where ShhNp is normally concentrated and functions to properly expand ventral progenitor populations. Thus, the results suggest that ShhN has a propensity to spread away from its site of synthesis at the expense of its local concentration, similar to the situation observed in the limb bud(Li et al., 2006). As Shh expression in the ventral telencephalon is induced by the underlying prechordal plate (Ericson et al.,1995; Shimamura and Rubenstein, 1997), the ventral patterning defects observed in the Shh^N/- telencephalon are likely to be attributable to the inability of telencephalic neuroepithelium to establish Shhexpression in response to altered Shh spreading behavior in the prechordal plate.

To confirm that the differential effects of ShhN on the development of the telencephalon and spinal cord were primarily mediated by paracrine signaling,we examined the patterning effects of ShhN in a Disp1^-/-background, in which all Shh paracrine activities are elicited by ShhN because ShhNp is retained in Shh-producing cells. Disp1^-/- single mutants normally die at or prior to E9.5 with a phenotype resembling Smo^-/- embryos (Caspary et al., 2002; Kawakami et al.,2002; Ma et al.,2002). However, the gross morphology of Disp1^-/-;Shh^N/+ embryos largely resembled that of Shh^N/- embryos, with enlarged brain ventricles, closely positioned frontal nasal process and defective eye formation(Fig. 8B, data not shown). As expected, we observed unperturbed dorsoventral neuronal patterning in E10.5 Disp1^-/-;Shh^N/+ spinal cord, as demonstrated by well-defined dorsal Pax7 and ventral Nkx6.1 marker expression(Fig. 8G). Notably, we found an expanded pool of floorplate cells in Disp1^-/-;Shh^N/+ spinal cord that often co-expressed the V3 interneuron marker Nkx2.2 and the floorplate marker Foxa2(Fig. 8H-J). This is consistent with the notion that ShhN protein is capable of inducing floorplate formation and ShhNp cannot travel beyond the floorplate in the absence of Disp1 function, resulting in an elevated local concentration of Shh in the floorplate region (Fig. 8F,note how the robust floorplate Shh staining and the absence of ependymal Shh staining differ from that of control and Shh^N/-embryos).

By contrast, the Disp1^-/-;Shh^N/+telencephalon showed patterning defects similar to that of Shh^N/- embryos at E12.5, with the absence of MGE neurons and ventral expansion of the dorsal marker Pax6(Fig. 8C-E), commensurate with the absence of Shh protein in the telencephalon(Fig. 8A). The above results support the notion that there is a more stringent requirement for cholesterol-modified Shh in patterning the ventral telencephalon than the spinal cord.

To our knowledge, Shh is the only metazoan protein that has been shown to be covalently modified by a cholesterol moiety. To fully appreciate the biological function of this post-translational event, it is important to understand not only the role of cholesterol in regulating Shh protein trafficking, but also the tissue-specific requirement for this cholesterol adduct. This study indicates that cholesterol modification of Shh is required for patterning the telencephalon. In Shh^N/- telencephalon,induction of ventral and dorsal anatomical structures and their corresponding cellular and molecular identities were significantly impaired. By contrast,Shh activity is not required for the generation of different ventral neuronal cell types in the spinal cord, although it is required for the full expansion of ventral progenitor domains. Thus, alterations in Shh local concentration and spreading behavior had more profound effects on the development of the telencephalon than of the spinal cord.

One of the characteristic features in early development of the vertebrate brain is the appearance of brain ventricles. Generating a uniform and coordinated outgrowth of neuroepithelium from the neural tube involves balanced cell proliferation, apoptosis and differentiation. Previous studies provided evidence that Shh is an essential player to initiate and maintain ventricle expansion (Britto et al.,2002). Ablating Shh signaling by removing notochord resulted in elevated apoptosis and significantly reduced cell proliferation throughout the neuroepithelium of the developing avian midbrain and forebrain(Britto et al., 2002). This growth retardation resulted in a collapse of the telencephalon and metencephalon ventricles. Furthermore, gradual blockade of Shh signaling by increasing amounts of cyclopamine led to progressive brain collapse associated with defective cell proliferation and massive apoptosis(Cooper et al., 1998). Shh^-/- forebrain development was also severely affected and it developed into one fused ventricle(Chiang et al., 1996). In contrast to reduced or collapsed brain ventricles associated with a reduction or absence of Shh signaling, Shh^N/- embryos showed enlarged forebrain ventricles, which were more evident at later stages of development when ectopic activation of Shh signaling was detectable. This observation suggested that the cholesterol moiety of Shh plays a key role in maintaining forebrain ventricle expansion by regulating the range of Shh spreading.

The ability of ShhN to enhance dorsal midline cell proliferation and elicit ectopic Shh-responsive gene expression in the dorsal telencephalon is consistent with the capacity of ShhN to spread over a long range. However, the induction of these genes can only be detected beginning at E13.5, suggesting that either the dorsal telencephalon is incompetent to receive the ShhN signal, or the long-range ShhN concentration is below the threshold for target gene expression at earlier stages. We favor the latter possibility, as previous neural explant studies showed that the competence to receive Shh signal in the dorsal telencephalon was diminished by E11.5(Kohtz et al., 1998). Interpretation of a morphogen signal by responding cells is achieved by the signaling intensity, which is conveyed by the concentration of ligand and the duration of exposure (reviewed by Gurdon and Bourillot, 2001). Long-lasting residence of ligand-receptor signaling complex has been shown to lead to prolonged signaling events in activin-dependent TGF-β signaling(Jullien and Gurdon, 2005). Interestingly, the activated form of smoothened (Smo), a seven-pass transmembrane receptor that activates downstream effectors of the Shh pathway,is internalized in a Shh-dependent manner(Chen et al., 2004b). Recent time-course analysis suggested that spinal cord progenitor cells could integrate the level of signaling over time, providing evidence that the signaling duration is an important parameter, in addition to signaling strength, to control dorsoventral patterning of the spinal cord(Stamataki et al., 2005). Therefore, we speculate that prolonged exposure of the telencephalic neuroepithelium to ShhN has a cumulative influence on the mediation of the gradual deposition of activated Smo intracellularly, which might in turn lead to widespread expression of Shh-responsive genes such as Gsh2 and Mash1 in Shh^N/- telencephalic neuroepithelium at later stages.

Although the mechanism of Shh spreading is not well understood,lipid-dependent Hh multimers have been proposed to regulate Hh spreading by interacting with extracellular matrix molecules such as heparan sulfate proteoglycans (HSPGs) (Callejo et al.,2006). Furthermore, hedgehog-interacting protein (Hhip), a membrane glycoprotein, can bind to Shh, thereby inhibiting its spread(Chuang et al., 2003; Chuang and McMahon, 1999). A recent mathematical model for Shh signaling dynamics predicted that the restrictive effect of these Shh-interacting molecules lowers the diffusion constant and concentrates the ligand near its secreting source(Saha and Schaffer, 2006). It is therefore possible that ShhN protein can escape the synergistic regulatory mechanisms exerted by Shh-interacting molecules, generating a high diffusion rate as predicted by the model (Saha and Schaffer, 2006). Given that ShhN transcript is undetectable in the ventral telencephalon and that ShhN would have to traverse across the neural epithelium - a relatively long distance - to activate Shh-responsive gene expression extending to the dorsal midline, we favor the idea that ShhN could spread into the subjacent cerebrospinal fluid (CSF) from its site of synthesis in the neural tube. This proposition is analogous to the situation in Drosophila, where Hh-N is expressed from peripodial cells of the wing imaginal disc and is proposed to be secreted into the lumen to activate low-threshold target genes as it accumulates near the apical surface (Gallet et al.,2006).

Diffusive molecules and proteins of the CSF can play crucial roles in regulating neuroepithelial cell behavior and it has been shown that CSF promotes neuroepithelial cell survival, proliferation and neurogenesis in mesencephalic explants (Gato et al.,2005; Parada et al.,2005). Recent studies have suggested that the Slit family of chemorepulsive proteins may be secreted into the CSF to generate a cilia-dependent concentration gradient necessary for vectorial migration of neuroblasts in the brain (Sawamoto et al.,2006). In this context, it is interesting to note that Shh-expressing Cos cells implanted within the lumen of the midbrain can rescue midbrain expansion defects generated by extirpation of the notochord,suggesting that the precise source of Shh signaling activity is not crucial for normal morphogenesis of early brain ventricles(Britto et al., 2002). Furthermore, we found that ventral spinal cord ependymal cells might be an important site at which ShhNp protein concentrates and/or through which it travels. Intriguingly, we found by immunostaining that ShhNp protein can be visualized along the apical side of the innermost ventricular zone neuroepithelial cells as early as E9.5. Although the significance of apical surface neuroepithelial expression is not clear, the fact that this pattern is absent in the ShhN mutant suggests that it might be required for expansion of ventral progenitor domains. Recent studies in the chick spinal cord have also reported similar apical surface neuroepithelial expression at the onset of oligodendrogenesis, suggesting a role of ShhNp in the expansion of oligodendrocyte precursors (Danesin et al., 2006). We envision two possible explanations for how Shh is localized to this region. First, a significant amount of Shh protein could travel in a planar direction through the innermost neuroepithelial cell layer,and subsequently spreads laterally into the ventricular zone to establish different neuronal progenitor identities. Alternatively, it is possible that ShhNp protein from the floorplate is secreted into the CSF, and ShhNp receptors that reside on the apical membrane of neuroepithelial cells function to concentrate and endocytose this ShhNp from the CSF. These two tentative explanations are not mutually exclusive. Megalin (Lrp2) is a member of the low-density lipoprotein (LDL) receptor-related protein family. Megalin function is essential for forebrain formation and its abundant expression at the apical surface of the CNS neuroepithelium at midgestation is very similar to ShhNp expression, notably at the apical surface of neuroepithelial cells(Assemat et al., 2005; Willnow et al., 1996). A mechanism has been proposed in which megalin is required to supply sufficient cholesterol to the rapidly dividing neuroepithelium before neural tube closure, possibly through the endocytosis of cholesterol-containing lipoprotein (Herz et al.,1997). Since it has been shown in tissue culture studies that megalin could function as an endocytic Shh receptor(McCarthy et al., 2002), it is possible that megalin has a direct role in propagating Shh signaling. Another endocytic receptor, cubilin (Cubn - Mouse Genome Informatics), an EGF-CUB protein, has a nearly identical apical surface neuroepithelial expression pattern to megalin (Assemat et al.,2005). Interestingly, a cubulin-related protein encoded by the zebrafish scube2 gene has been shown to mediate Shh signaling in embryos (Hollway et al., 2006; Woods and Talbot, 2005). Therefore, it is possible that these two proteins act synergistically in promoting Shh signaling during mammalian development. It would be interesting to investigate the functional relationship of Shh, megalin and cubulin in ependymal cells and the existence of secreted Shh protein in the CSF.

Our study suggests that alteration of Shh spreading behavior has differential effects in patterning the telencephalon versus spinal cord. This is most likely to be attributable to regional differences in the generation and maintenance of Shh expression in the neuroepithelium. In the spinal cord, the development of ventral progenitor cells is mediated by Shh secreted from the notochord and floorplate(Jessell, 2000). Shhexpression in the floorplate is induced and maintained by the notochord, which is in close contact with the overlying spinal cord for an extended period of time. In the telencephalon, the development of basal ganglia appears to be mediated initially by Shh secreted from the prechordal plate (the dorsal foregut) and then by the ventral neuroepithelium(Ericson et al., 1995; Shimamura and Rubenstein,1997). In contrast to the floorplate, Shh expression in the ventral telencephalic neuroepithelium becomes quickly independent of the influence of the underlying prechordal plate. Thus, a reduction in local Shh concentration in the prechordal plate and notochord, as shown in Shh^N/- embryos, would have more profound effects on Shh expression and maintenance in the telencephalon than in the spinal cord.

An unexpected finding is that ShhN exerted no effect on dorsal progenitors,as there was no evidence of expansion or ectopic ventral progenitor cell fates in the dorsal region of the spinal cord. This observation is in contrast to findings in the telencephalon and limb buds(Li et al., 2006), where ectopic Shh pathway activation was detected far from the source of ShhN synthesis. There are at least two, non-mutually exclusive explanations for this difference. First, temporal responsiveness to ectopic ShhN might be different between dorsal telencephalon and dorsal spinal cord. Neural explant studies in rats indicated that E11.5 dorsal telencephalon (∼E10.5 in mouse) remained competent to receive exogenous Shh stimulation(Kohtz et al., 1998), whereas E11 (∼E10 in mouse) spinal cord did not demonstrate such competence(Charron et al., 2003). Likewise, the spinal cord might have lost competence to respond to ectopic ShhN by the time the signal reaches a threshold concentration necessary for cell fate specification. Alternatively, the concentration threshold required for ShhN to elicit ectopic patterning effects in the dorsal spinal cord is higher than for the dorsal telencephalon.

We thank Seema Agarwala and Michael Cooper for critical reading of the manuscript; Yina Li for her advice on embedding early-stage embryos; and Christer Betsholz, Kenneth Campbell, Jane Johnson and Hirohide Takebayashi for reagents. This work was supported by a grant from the National Institutes of Health (NS42205).