A functional screen for sonic hedgehog regulatory elements across a 1 Mb interval identifies long-range ventral forebrain enhancers

Shh is expressed in the axial mesoderm (prechordal plate and notochord) and ventral midline (floor plate) of the overlying neural tube throughout most of the anteroposterior (AP) neuraxis(Echelard et al., 1993). It is from these sources that a morphogenic gradient of Shh is generated to promote distinct neuronal progenitor subtypes in the ventral neural tube. In the absence of Shh function, ventral CNS development is severely impaired because of neuronal patterning and pathfinding defects(Chiang et al., 1996; Charron et al., 2003). In humans, the ventral forebrain is particularly sensitive to the level of Shh expression, given that a 50% reduction in its normal levels causes holoprosencephaly (HPE), a structural brain malformation resulting in the fusion of the cerebral hemispheres, optic vesicles and other craniofacial anomalies (Roessler et al.,1996). In mice, loss of both Shh alleles is required to generate a similar HPE phenotype (Chiang et al., 1996). Over recent years, significant progress has been made in deciphering the molecular mechanisms used by Shh to pattern the ventral neural tube (Jessell, 2000; Patten and Placzek, 2000; Briscoe and Ericson, 2001; Ruiz i Altaba et al., 2003). By contrast, much less is known of the transcription factors acting upstream in the pathway that control Shh expression in key signaling centers,including the floor plate and axial mesoderm.

The etiology of HPE is complex, with multiple genetic and environmental factors contributing to the phenotype. To date, mutations in seven genes,several of which disrupt Shh or Nodal signaling, have been shown to account for nearly 20% of HPE cases in humans(Ming and Muenke, 2002). Additional causative loci have been identified in mice(Hayhurst and McConnell, 2003; Zoltewicz et al., 2004). Surprisingly, these genetic studies have yet to uncover the transcription factors directly responsible for regulating Shh expression in the rostral forebrain and prechordal plate. Although Shh expression is preferentially lost in the ventral forebrain of mouse embryos with attenuated Nodal signaling, the phenotype appears to result from a failure in the induction of prechordal plate mesoderm, rather than a direct effect on Shh transcription per se (Lowe et al., 2001; Dunn et al.,2004). This contrasts with the situation in zebrafish, where Nodal signaling is thought to directly regulate the expression of Shh in the ventral midline of the neural tube(Muller et al., 2000; Rohr et al., 2001).

Our efforts to identify the genes regulating Shh transcription in the ventral midline of the mouse CNS commenced with the analysis of 35 kb of genomic DNA surrounding Shh for regulatory potential in a transgenic mouse reporter assay (Epstein et al.,1999). Multiple enhancers were shown to activate lacZexpression at discrete positions along the AP axis of the neural tube(Epstein et al., 1999). Two Shh floor-plate enhancers, SFPE1 and SFPE2, were identified that regulated reporter activity in a Shh-like pattern in the ventral midline of the spinal cord and hindbrain. A third regulatory element, Shh brain enhancer 1 (SBE1), directed reporter activity to the ventral midbrain and caudal region of the diencephalon. The absence of reporter activity in rostral regions of the ventral forebrain in the 35 kb surveyed suggested that the enhancers controlling Shh expression in this domain must operate over greater distance. Two additional hints suggested that the Shh forebrain enhancers were long-range acting. First, a translocation breakpoint mapping 250 kb upstream of Shh was identified in an individual exhibiting a mild form of HPE, raising the possibility that the translocation separated a forebrain enhancer from the Shh transcription start site(Belloni et al., 1996). Second, Shh expression in the limb bud is regulated by sequences mapping∼1 Mb upstream of Shh(Lettice et al., 2003; Sagai et al., 2005).

In this study, we describe a functional genomic approach to the analysis of Shh regulatory sequences. Using modified Bac clones as reporter constructs, we employed an enhancer trap assay to systematically screen 1 Mb of DNA surrounding the Shh locus for the ability to target reporter gene expression to sites of Shh transcription in transgenic mouse embryos. Highly specific patterns of reporter expression were identified in the axial mesoderm, gut endoderm, limb bud and CNS, all known sites of Shh transcription. By coupling the Bac reporter assay with comparative sequence analysis, we identified three novel enhancers located over 400 kb from the Shh transcription start site that directed Shh-like expression to the ventral forebrain. This brings the total number of Shh CNS enhancers to six, the combined activity of which covers all regions of Shh transcription along the AP axis of the mouse neural tube. We also assessed the relative contribution of each CNS enhancer to the overall pattern of Shh expression by deleting individual elements in the context of the Bac reporter assay. Redundant mechanisms regulated Shh-like expression in the ventral spinal cord,hindbrain and subventricular zone of the medial ganglionic eminence (mge) in the telencephalon. Whereas Shh reporter activity in the ventral midbrain, much of the ventral diencephalon and the ventricular zone of the mge was found to be dependent on unique elements. The multitude of enhancers controlling Shh transcription in the CNS is further indication of the complex mechanisms operating to regionalize the ventral midline of the neural tube along the AP axis.

The eGFP reporter gene was introduced into Bacs RPCI-23 429M20 and RPCI-23 389P3 (CHORI, BacPac Resources) by homologous recombination in E. coli according to the method of Gong et al.(Gong et al., 2002). All references to nucleotide positions for mouse Chromosome 5 listed below were obtained from the mm5-May'04 assembly. Homology arms A and B flanking mouse Shh exon 1 (corresponding to the following nucleotide positions Chr5:26864937-26865255) were synthesized by PCR using the primer sets E169/E170 and E239/E243, respectively (see Table S1 in the supplementary material for the list of primer sequences described throughout this section). The fragments were cloned into the AscI and PacI restriction sites of the pLD53SCAEB shuttle vector (provided by N. Heintz, Rockefeller University). A detailed description of the Bac modification protocol is available(Jeong and Epstein, 2005). The correct positioning of eGFP within the individual Bacs was verified by probing XbaI-digested DNA with external probes on Southern blots.

Bac reporter constructs harboring deletions of SFPE1, SFPE2 and SBE1 in Bac 429M20eGFP were generated according to the same method described above. Homology arms A and B flanking the sequences to be deleted corresponding to mouse SFPE1 (Chr5: 26874573-26875234), SFPE2 (Chr5: 26857299-26858064) and SBE1 (Chr5: 26858065-26858607) were generated by PCR using the following primer sets: E260/E261, E262/E263 (SFPE1); E268/E269, E270/E271 (SFPE2); and E264/E265, E266/E267 (SBE1), and cloned into the pLD53SCAEB shuttle vector after removing the eGFP cassette. The construct carrying deletions of SFPE1 and SFPE2 was generated by deleting SFPE2 from Bac 429M20eGFP^ΔSFPE1. Bac reporter constructs harboring deletions of SBE2, SBE3 and SBE4 in Bac 447L17βlacZ were generated by the same approach. Homology arms A and B flanking the sequences to be deleted corresponding to mouse SBE2 (Chr5: 27274453-27275587), SBE3 (Chr5:27291667-27292041) and SBE4 (Chr5: 27184220-27185333) were generated by PCR using the following primer sets: E289/E290, E291/E292 (SBE2); E295/E296,E297/E298 (SBE3); and E360/E361, E362/E363 (SBE4), and cloned into the pLD53SCAEB shuttle vector after removal of the eGFP cassette. The constructs carrying combined deletions of SBE2 and SBE3, SBE2 and SBE4, SBE3 and SBE4,and SBE2, SBE3 and SBE4 were generated sequentially.

The Tn7βlacZ targeting vector (provided by D. Duboule, University of Geneva) containing Tn7 transposable elements surrounding the β-globin promoter, lacZ gene and SV40 poly(A) signal was introduced into overlapping Bacs obtained from the RPCI-23 (208K5, 447L17, 265M1, 214O17) and RPCI-24 (265M10) libraries (CHORI, BacPac Resources) according to the method of Spitz et al. (Spitz et al.,2003). Tn7βlacZ vector (40 ng) was mixed with 200 ng of Bac,GPS buffer and TnsABC (NEB). After incubation at 37°C for 10 minutes,Start Solution was added and the reaction was prolonged for 1 hour. After heat inactivation at 75°C for 10 minutes, electrocompetent DH10B cells were transformed with 1 μl of this incubate and grown overnight on LB plates containing 20 μg/ml Kanamycin and 12.5 μg/ml Chl. Positive colonies were identified by PCR using a primer pair corresponding to sequences from theβ-globin promoter and lacZ gene. Modified Bac clones were further analyzed by restriction enzyme digestion and pulsed-field gel electrophoresis to determine the number and location of integration sites. Only intact Bacs with a single Tn7βlacZ integration were used in the transgenic reporter assay.

All regulatory sequences were cloned into a reporter vector containing the Shh minimal promoter, lacZ gene and SV40 poly(A) signal. Each of the conserved regions (ECR1-ECR7) from the overlap between Bacs 447L17 and 265M1 was amplified by PCR (ECR1, Chr5: 27274453-27275587; ECR3, Chr5:27291667-27292041; and data not shown) using primer sets E278/E279 (ECR1) and E282/E283 (ECR3). SBE4 (Chr5: 27184220-27185333) was amplified by PCR using primers E337 and E338. Conserved SBE2 sequences from human (Chr7:155560539-155561314), chicken (Chr2: 7701492-7702259), frog (Scaffold57:2747939-2748753) and tetraodon (Chr6: 4066565-4067332) were amplified from genomic DNA by PCR using primer sets: E311/E313 (human); E309/E310 (chicken);E319/E320 (frog); E321/E322 (tetraodon). Conserved SBE4 sequences from chicken(Chr2: 7641242-7642395), frog (Scaffold57: 2689775-2690912) and tetraodon(Chr6: 4063308-4064467) were amplified from genomic DNA by PCR using primer sets: E372/E373 (chicken); E368/E369 (frog); E374/E375 (tetraodon). The frog DNA derived from X. tropicalis was kindly provided by Dr Frank Conlin(UNC, Chapel Hill, NC).

pCMV or pCMV-Nkx2.1 plasmids were transfected into NIH3T3 cells using FuGENE 6 transfection reagent (Roche). After 48 hours, whole-cell lysates were prepared in a buffer containing 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 1 mM DTT,protease inhibitor cocktail and 25% glycerol. For EMSA, 10 μg of protein from the cell lysates was incubated for 10 minutes at room temperature in a DNA-binding buffer containing 10 mM Tris-HCl (pH 7.4), 50 mM NaCl, 1 mM EDTA,1 mM DTT, 5% glycerol, 200 ng poly (dI-dC), 1 μg BSA in the presence or absence of competitor double stranded oligonucleotides. After 0.1 ng(5×10⁴ to 10×10⁴ cpm) of probe was added to the mixture, incubation was continued for an additional 20 minutes. Protein-bound DNA complex was separated from free probe on a 6.5% acrylamide gel run in 1×TBE (Tris-borate-EDTA) buffer. The sequence of probe and competitors were as follows: probe (also used as wild-type competitor),5′-ACAGAAATTGTTTTTTAAGTAGTTGCCCTCTGAAAATAT-3′; mutant competitor,5′-ACAGAAATTGTTTTTTGGAGCATTGCCCTCTGAAAATAT-3′. Underlined regions of the sequence indicate wild-type and mutant Nkx2-binding sites, respectively.

Conventional plasmid transgenes were prepared for microinjection as described (Epstein et al.,1999). Bac transgenes were linearized with the PI-SceI(NEB) restriction enzyme prior to injection(Jeong and Epstein, 2005). Transient transgenic embryos or mouse lines were generated by pronuclear injection into fertilized eggs derived from the (BL6×SJL) F1 mouse strain (Jackson Labs). The primers listed in parentheses were used for PCR genotyping of mice carrying: 429M20eGFP and 389P3eGFP (E128 and E129);208K5βlacZ, 447L17βlacZ, 265M1βlacZ, 214O17βlacZ and 265M10βlacZ (E276 and E277).

The assessment of β-galactosidase activity was performed as described(Epstein et al., 1999). Whole-mount RNA in situ hybridization was performed using a digoxigenin-UTP-labeled Shh riboprobe(Echelard et al., 1993). Stained embryos were photographed after dehydration in methanol and clearing in benzyl alcohol:benzyl benzoate (1:1). Representative embryos were rehydrated, immersed in 30% sucrose overnight, embedded and frozen in OCT and sectioned at 20 μm on a cryostat. Primary antibodies used for immunohistochemistry and dilutions were as follows: eGFP (Molecular Probes),1:1000; Shh (5E1, Developmental Studies Hybridoma Bank), 1:100. Detection of primary antibodies was achieved using Cy3- (Jackson ImmunoResearch Laboratories) or Alexa 488- (Molecular Probes) conjugated goat anti-mouse secondary antibodies.

To determine whether Shh forebrain enhancer(s) resided on sequences beyond the 35 kb previously analyzed, we isolated two Bac clones(429M20 and 389P3) from the RPCI-23 library (CHORI) that extended 180 kb upstream and 160 kb downstream of the Shh promoter, respectively. The first Shh exon in each Bac was replaced with an eGFP reporter gene by homologous recombination in bacteria according to the method of Gong et al.(Gong et al., 2002) (see Materials and methods for a description of the targeting strategy). Transgenic embryos were then generated with the modified Bacs and assessed for reporter activity. At 9.5 dpc, embryos carrying either of the 429M20eGFP or 389P3eGFP transgenes displayed eGFP expression in the ventral midline of the CNS in a similar pattern to that generated from the locally acting Shhregulatory elements (Fig. 1A,B; Table 1)(Epstein et al., 1999). In addition, Shh-like reporter activity was detected in the hindgut of embryos carrying 429M20eGFP but not 389P3eGFP, indicating that cis-acting sequences controlling Shh expression in the gut endoderm are located on Bac 429M20 (Fig. 1A,B).

It was clear from this experiment that regulatory sequences controlling Shh expression in the rostral forebrain were not located within 180 kb upstream or 160 kb downstream of the Shh promoter. However, as the previously identified local enhancers were active in embryos carrying 429M20eGFP, it provided us with an opportunity to assess their requirement in a genomic context more in keeping with the endogenous Shh locus. To address the contribution of SFPE1, SFPE2 and SBE1 in regulating Shh-like expression in the CNS, the sequences mediating their respective activities were deleted from Bac 429M20eGFP by homologous recombination in bacteria. Transgenic embryos carrying 429M20eGFP^ΔSFPE1 or 429M20eGFP^ΔSFPE2 showed no loss of eGFP expression in the floor plate of the hindbrain or spinal cord(compare Fig. 1B,I with 1C,K and 1D,L). By contrast, embryos carrying 429M20eGFP^{ΔSFPE1, ΔSFPE2}, a Bac transgene in which both floor plate enhancers were deleted, showed barely detectable levels of eGFP staining in the ventral midline of the hindbrain and spinal cord,despite normal expression in the ventral midbrain(Fig. 1E,M). Conversely,embryos carrying 429M20eGFP^ΔSBE1, a Bac with a deletion of SBE1, showed an absence of eGFP expression in the ventral midbrain and caudal diencephalon, despite normal staining in the hindbrain and spinal cord(compare Fig. 1B,I,J with F,N,O). These results suggest that Shh-like transcription in the floor plate of the hindbrain and spinal cord is dependent on the redundant activities of SFPE1 and SFPE2, and that Shh-like expression in the ventral midbrain and caudal diencephalon is solely dependent on SBE1.

The absence of reporter activity in the rostral forebrain of embryos carrying 429M20eGFP or 389P3eGFP suggested that the sequences regulating Shh expression in this tissue must operate over considerable distance. To further explore this possibility, we had to develop an efficient strategy of screening large amounts of DNA for regulatory activity. Comparative analysis of genomic DNA from divergent organisms is an effective method of identifying evolutionary conserved regions (ECR) that may function as regulatory elements (Nobrega et al.,2003; Thomas et al.,2003; Woolfe et al.,2005). The problem with applying this approach over large genomic intervals is that the number of ECRs is often too numerous to easily identify the ones of interest. For example, in the 800 kb interval between Shhand Rnf32, the next neighboring gene upstream, there are ∼170 ECRs (>70% identity over 100 bp) shared between human and mouse, of which 57 are also found in chicken (ECR browser: http://ecrbrowser.dcode.org/). Rather than testing each of the ECRs independently for Shh reporter activity, we implemented an enhancer-trap assay similar to the one designed by Spitz et al. (Spitz et al.,2003) in their study of the HoxD global control region. In this way, modified Bacs could be used as reporter constructs to first determine whether a given genomic interval possessed Shh enhancer activity. Five overlapping Bacs extending ∼1 Mb upstream of Shh were isolated (CHORI) and modified to incorporate a single copy reporter cassette containing a minimal β-globin promoter and lacZ gene(Fig. 2). The reporter cassette was randomly inserted into each of the Bacs by Tn7 transposition in E. coli (Spitz et al.,2003). Two independent transposition events were generated for each Bac and subsequently assayed for reporter activity in transgenic embryos.

Remarkably, of the five Bacs tested, four directed highly specific and reproducible patterns of reporter expression to the axial mesoderm, ventral forebrain and posterior limb bud, all known sites of Shhtranscription (Fig. 2). Of particular interest was the observation that X-gal staining in the ventral forebrain of embryos carrying either of two overlapping Bacs (447L17βlacZ and 265M1βlacZ) was similar to each other and highly reflective of the pattern of Shh expression in the ventral diencephalon and telencephalon (compare Fig. 2A with C,D).

In addition to the ventral forebrain, X-gal staining was observed in the notochord in embryos carrying any of three overlapping Bacs (208K5, 447L17,265M1), suggesting the existence of at least two long-range notochord enhancers (Fig. 2B,C,D). The early expression of Shh in the node, notochord, prechordal plate and ventral forebrain was also recapitulated in embryos carrying 447L17βlacZ(see Fig. S1 in the supplementary material). However, in contrast to the normal expression of Shh, X-gal staining was not maintained at rostral levels of the notochord in older embryos(Fig. 2B-D).

The most distant acting enhancer identified in our study mapped to Bac 265M10βlacZ and directed Shh-like reporter activity to the zone of polarizing activity (ZPA) in the posterior region of the fore- and hindlimb buds between 10.5 and 12.5 dpc (Fig. 2F; data not shown). This Shh ZPA enhancer is probably the same as the one reported by Lettice et al.(Lettice et al., 2003),located ∼850 kb upstream of Shh.

The similar pattern of X-gal staining in the forebrain of embryos carrying the 447L17βlacZ and 265M1βlacZ transgenes suggested that the cis-acting sequences responsible for this expression localized to the 70 kb region of overlap between the two clones. The pair-wise alignment (ECR browser: http://ecrbrowser.dcode.org/)of genomic sequence from this interval between mouse and human identified seven highly conserved regions of DNA showing at least 75% nucleotide identity that ranged in size from 0.37 to 1.4 kb(Fig. 3). To address their regulatory potential, each of the ECRs in this interval was amplified by PCR from mouse genomic DNA, cloned into a reporter cassette containing a minimal Shh promoter and lacZ gene and assayed for forebrain enhancer activity in transgenic embryos. Of the seven ECRs (ECR1-7) tested,embryos carrying ECR1 and ECR3 showed consistent X-gal staining in the ventral forebrain. The 1.1 kb ECR1 fragment, located 410 kb distal to Shh,directed lacZ expression to the rostral region of the ventral diencephalon in two lateral stripes that converged in the ventral midline at the level of the optic vesicles in a manner consistent with Shh and 447L17βlacZ reporter activity (compare Fig. 3D,J,P,V with 3A,G,M,S and 3B,H,N,T). Embryos carrying the 370 bp ECR3 element, which mapped 426 kb distal to Shh, directed reporter activity to only part of the Shh expression domain in the ventral telencephalon, encompassing the subventricular zone (svz) of the medial ganglionic eminence (mge) (Fig. 3E,K,Q,W). As the expression of Shh and 447L17βlacZ reporter activity was also detected in the ventricular zone (vz) of the mge,we presumed that a second telencephalic enhancer must regulate Shhtranscription at this site (Fig. 3S,T). None of the other ECR constructs directed reporter activity to Shh-expressing tissues, including the mge and axial mesoderm,suggesting that these enhancers must form on other sequences contained in Bac 447L17 (data not shown). In keeping with the nomenclature of previously identified Shh regulatory elements, we refer to the rostral diencephalic enhancer as Shh brain enhancer-2 (SBE2) and the telencephalic svz enhancer as SBE3 (Table 1).

The pattern of X-gal staining in the ventral telencephalon of embryos carrying 447L17βlacZ and 265M1βlacZ appeared similar when viewed as whole embryos (Fig. 3B,C);however, upon their sectioning, subtle yet significant differences were realized. 447L17βlacZ-containing embryos showed X-gal staining in both the vz and svz of the mge in a Shh-like pattern, whereas,265M1βlacZ containing embryos only stained positively in the svz of the mge (Fig. 3S-U). This indicated that the missing mge enhancer must reside somewhere on the 100 kb region of Bac 447L17 that did not overlap with Bac 265M1. In analyzing the whole genome alignment (ECR browser) of human, mouse and chicken covering this 100 kb interval, we identified only one highly conserved region (77% identity over 350 bp) positioned ∼319 kb upstream of the Shh promoter(Fig. 3). A 1.1 kb fragment overlapping this ECR was amplified from the mouse genome and tested for reporter activity. Transgenic embryos carrying the 1.1 kb fragment showed X-gal staining in both the vz and svz of the mge in a manner comparable with 447L17βlacZ and Shh expression(Fig. 3F,X). X-gal staining was also detected in the ventral diencephalon in the vicinity of prosomere 3(Fig. 3F,L). This enhancer was designated SBE4 (Table 1).

We next addressed whether the sequences mediating SBE2, SBE3 and SBE4 were required for the Shh-like pattern of expression in the ventral forebrain of embryos carrying 447L17βlacZ. We reasoned that if SBE2-4 were necessary then deleting their sequences from 447L17βlacZ would abrogate reporter activity in the ventral forebrain. However, if ventral forebrain expression persisted, in whole or in part, in the absence of SBE2-4,then additional regulatory elements would have to be implicated.

Transgenic embryos carrying 447L17βlacZ^ΔSBE2 were devoid of reporter activity throughout most of the rostral-ventral diencephalon with the exception of a small region corresponding to prosomere 3(p3) that showed weak bilateral patches of X-gal staining (compare Fig. 4A-C with Fig. 3B,H,N). Reporter activity in the ventral telencephalon and notochord were unaffected by the deletion of SBE2 (Fig. 4A,D). Interestingly, embryos carrying 447L17βlacZ^{ΔSBE2,ΔSBE3} showed a similar pattern of X-gal staining to those carrying 447L17βlacZ^ΔSBE2(Fig. 4E-H). This finding indicates that SBE3 is not required for svz expression in the ventral telencephalon and that another forebrain element in Bac 447L17 must compensate in its absence (Fig. 4D,H). As SBE3 and SBE4 were each capable of directing reporter expression to the svz of the ventral telencephalon, it is likely that the two enhancers function in a partially redundant manner. To test this hypothesis, we engineered deletions in Bac 447L17βlacZ of either SBE4 on its own or in combination with SBE3 and tested the respective transgenes (447L17βlacZ^ΔSBE4,447L17βlacZ^{ΔSBE3, ΔSBE4}) for reporter activity. Embryos carrying 447L17βlacZ^ΔSBE4 showed no X-gal staining in the vz of the ventral telencephalon, but did retain staining in the svz (compare Fig. 4L with Fig. 3T). Only when SBE3 and SBE4 were both deleted (447L17βlacZ^{ΔSBE3, ΔSBE4})was X-gal staining completely eliminated from the ventral telencephalon(Fig. 4M,P). These results confirm that the Shh-like expression derived from Bac 447L17βlacZ in the svz of the ventral telencephalon is dependent on both SBE3 and SBE4, and that the vz expression in the mge is solely dependent on SBE4.

Redundant mechanisms were also identified in the regulation of Shh-like reporter activity in the p3 region of the ventral diencephalon. Only when SBE2 and SBE4 were deleted from Bac 447L17βlacZ(447L17βlacZ^{ΔSBE2, ΔSBE4}) was reporter activity abrogated from the p3 domain (Fig. 4Q,R). Interestingly, transgenic embryos carrying 447L17βlacZ^{ΔSBE2, ΔSBE3, ΔSBE4}, in which all three forebrain enhancers were deleted, showed no X-gal staining anywhere in the ventral forebrain, despite persistent notochord staining, indicating that all Shh-like expression generated from Bac 447L17βlacZ is regulated by SBE2-4 (Fig. 4U-X).

The sequences mediating the activities of SBE2-SBE4 were identified based on their high degree of conservation between mouse and human (80% for SBE2 and SBE4 and 72% for SBE3). To determine the extent to which the conservation of sequence reflected the conservation of enhancer function, we performed blat and blast searches using the UCSC(http://www.genome.ucsc.edu/)and Ensemble(http://www.ensembl.org/)genome browsers, respectively, and identified enhancer sequences from a number of vertebrate species. SBE2 sequences were identified in most vertebrate phyla at various distances from the Shh promoter(Fig. 5A,B). However, SBE3 sequences were only found in closely related organisms, including chimp and rat, raising the possibility that the SBE3 element arose relatively recently in the clade containing rodents and primates (data not shown). Sequences mediating SBE4 were identified in most organisms surveyed; however, the degree of conservation was reduced in comparison with SBE2(Fig. 6A,B).

The SBE2 sequences from human, chicken and frog were each sufficient to drive lacZ expression into the ventral diencephalon of transgenic mice, in keeping with the significant preservation of nucleotide identity across much of this element (Fig. 5C-E). By contrast, the SBE2 sequence from puffer fish, which differed significantly from the other vertebrates, was incapable of regulating transcription in the ventral diencephalon of mouse embryos(Fig. 5F). Similarly, the SBE4 element from chicken showed conserved reporter activity in mice, but the more divergent frog and puffer fish SBE4 sequences were not active(Fig. 6C-E). Whether these findings indicate that puffer fish and frog use other cis-acting sequences to regulate distinct aspects of Shh expression in the forebrain, or that the sequences tested can only accommodate enhancer function in the puffer fish and frog brains remains to be determined.

To begin to understand the molecular mechanisms governing Shhforebrain enhancer activity, we surveyed the sequence of one of the enhancers,SBE3, for known transcription factor-binding sites and considered only those sites showing a correlation between the expression of their cognate DNA-binding protein and that of Shh. A binding site matching the consensus for Nkx2 proteins (T(T/C)AAGT(A/G)(G/C)TT)(Watada et al., 2000) was identified in the sequence of the SBE3 enhancer in a region that was 100%conserved between human and mouse (Fig. 7A). The expression of Nkx2.1 overlaps with that of Shh in the ventral forebrain including the svz of the mge(Shimamura et al., 1995). Moreover, the telencephalic expression of Shh is downregulated in Nkx2.1^-/- embryos(Sussel et al., 1999). However, as the mge fails to form in these mutants, it is unclear whether the regulation of Shh by Nkx2.1 is direct. To address this issue, we performed electromobility shift assays, the results of which suggested that Nkx2.1 protein was able to bind to its target recognition sequence in the SBE3 enhancer (Fig. 7B). We next determined the requirement of the Nkx2-binding site in the context of the SBE3 reporter assay. Transgenic embryos carrying an SBE3 reporter construct that lacked the core Nkx2 binding site (AAGTAG) failed to activate lacZexpression in the svz of the mge (Fig. 7D,F). Based on these findings, we conclude that Nkx2.1 is a direct regulator of SBE3 activity.

In this report, we describe a functional genomic approach to the analysis of Shh regulatory sequences. An enhancer trap assay was used to systematically screen 1 Mb of DNA surrounding the Shh locus for the ability to target reporter gene expression to ventral cells of the mouse neural tube. The strength of our technique in analyzing large amounts of DNA for regulatory potential relied on coupling a transgenic Bac reporter assay with comparative sequence analysis. By performing the Bac reporter assay first, a reasonable genomic interval could be defined in which to search for conserved regulatory elements. Other methods that rely solely on sequence conservation are often less efficient at identifying enhancers because the process of evaluating evolutionary conserved regions (ECRs) for regulatory activity is random (Nobrega et al.,2003; Woolfe et al.,2005). This is particularly problematic when the enhancers are spread out across a large genomic landscape, as with Shh, because the number of highly conserved sequences can be large. Using our methodology, we were able to rapidly identify several novel enhancers distributed over 400 kb from the Shh promoter that regulate Shh transcription in the embryonic forebrain. Together with our previous analysis of locally acting Shh regulatory sequences, our studies have uncovered a total of six enhancers that are capable of directing reporter activity to all sites of Shh transcription in the ventral neural tube of the mouse embryo(Fig. 8; Table 1).

One limitation for all approaches that use multi-species sequence comparison to uncover cis-acting regulatory elements is the potential to overlook enhancers whose sequence or position in the genome has not been conserved across phyla (Frazer et al.,2004). This may explain why we were unable to identify the precise location of the Shh notochord element on Bac 447L17. In such instances, deletion mapping may be a more suitable strategy to narrow down the location of the enhancer of interest.

Shh expression in the floor plate is induced by Shh signaling from the underlying notochord (reviewed by Placzek and Briscoe, 2005). According to our results, this homeogenetic activation of Shhtranscription within the floor plate of the spinal cord and hindbrain is dependent on the redundant activities of SFPE1 and SFPE2. Both floor plate enhancers are reliant on Gli2, as Shh expression and consequently floor-plate formation are absent in Gli2^-/- embryos(Ding et al., 1998; Matise et al., 1998). However,Gli2 is unlikely to be regulating SFPE1 and SFPE2 directly as Gli-binding sites were not identified in sequences mediating these enhancers(Epstein et al., 1999; Jeong and Epstein, 2003). Instead, Foxa2 and Nkx6 proteins are probably acting as transcriptional intermediaries downstream of Gli2 as conserved binding sites matching their consensus are required for SFPE2 activity(Jeong and Epstein, 2003). To determine whether SFPE1 and SFPE2 are under the influence of the same transcription factors, we screened the 300 bp segment mediating SFPE1 for Nkx6- and Foxa2-binding sites. However, no sequences even remotely matching their consensus binding sites were identified (data not shown). Taken together, these findings suggest that although SFPE1 and SFPE2 possess equivalent functions, the transcription factors responsible for their activation appear to be different.

Ventral midline cells of the forebrain exhibit distinct properties from floor-plate cells in more posterior regions of the CNS and are thought to emerge from a discrete pool of progenitors through slightly different inductive signaling events (Placzek and Briscoe, 2005). Lineage tracing studies performed in the chick indicate that ventral midline cells of the forebrain derive from a small domain, termed `area a', situated immediately rostral to Hensen's node(Garcia-Martinez et al., 2004; Patten et al., 2003). The prechordal plate that underlies `area a' cells is the source of Shh and Nodal, the combination of which is thought to convert`area a' progenitors into ventral midline cells(Patten et al., 2003). This contrasts with the origin of floor-plate precursors in more posterior regions of the embryo, which derive from the node and are dependent on notochord-derived Shh signaling for their formation(Placzek and Briscoe, 2005). Studies in zebrafish suggest that in addition to Nodal and Shh, suppression of Wnt signaling in the ventral midline of the rostral diencephalon plays a role in distinguishing the hypothalamic infundibulum from floor-plate cells that occupy more posterior regions of the neural tube(Kapsimali et al., 2004). It remains to be confirmed if `area a' cells and attenuated Wnt signaling are involved in ventral midline development in the mouse forebrain. Nevertheless,the distinct signaling properties of the prechordal plate versus the notochord are likely to play a significant part in distinguishing Shhexpression in the ventral forebrain from other regions of the CNS as evidenced by the separate enhancers controlling Shh transcription in rostral versus caudal region of the neural tube.

The finding that multiple enhancers control Shh expression in the forebrain argues that multiple transcription factors must be involved in this regulation. One likely candidate, the Nkx2.1 homeodomain protein, is expressed in the basal telencephalon and rostral region of the ventral diencephalon in a pattern that overlaps with Shh(Shimamura et al., 1995). Importantly, Shh transcripts were significantly reduced from these domains in Nkx2.1^-/- embryos(Sussel et al., 1999). Indication that at least one aspect of Shh regulation by Nkx2.1 is direct comes from our identification of a functional Nkx2.1-binding site in sequences mediating SBE3. In the absence of this binding site, SBE3 reporter activity was abrogated from the svz of the mge(Fig. 7). SBE2 and SBE4 also direct Shh-like reporter activity in a manner that overlaps with Nkx2.1, raising the possibility that Nkx2.1 binds to these regulatory sequences as well. However, as SBE2 and SBE4 control different aspects of Shh expression in the forebrain, Nkx2.1 would have to be acting in concert with other transcription factors.

Although the identities of the transcription factors that cooperate with Nkx2.1 to positively regulate Shh expression in the forebrain remain unclear, at least one negative regulator, Tbx2, has been proposed. Shh is not expressed in the ventral midline of the diencephalon corresponding to the hypothalamic infundibulum, despite the presence of Nkx2.1 in this domain. Instead, Shh transcripts are found in two bilateral stripes adjacent to the ventral midline. Recent data in the chick suggest that Tbx2 is activated specifically in the region of the hypothalamic infundibulum in response to Bmp signaling from the prechordal plate and is responsible for the repression of Shh in this domain (E. Manning and M. Placzek,personal communication). There is some evidence to suggest that the repression of Shh by Tbx2 is direct as the removal of a Tbx-binding site in sequences mediating SFPE2 caused ectopic lacZ expression in the ventral midline of the diencephalon (Jeong and Epstein, 2003). Analysis of the crucial sequences and cognate transcription factors required for the activities of SBE2 and SBE4 should provide additional insight to the mechanisms regulating Shhtranscription in the mouse forebrain.

HPE is a genetically heterogeneous condition caused predominantly by mutations in Shh (Ming and Muenke, 2002). The initial indication for Shh as an HPE candidate gene was based on the analysis of two individuals with HPE who were carrying cytogenetically detectable rearrangements involving chromosome 7q36,where Shh was shown to map(Belloni et al., 1996). Interestingly, the translocation breakpoints in these individuals did not disrupt Shh-coding regions but instead mapped 15 kb and 250 kb distal to the Shh gene. It is unlikely that these translocations interrupted other genes in the area as the interval between Shh and Rnf32, the next distal gene over, is separated by an 840 kb gene desert. Rather, our findings suggest that the cause of HPE in these translocation cases was due to the displacement of Shh enhancers from its promoter (Fig. 8). A similar explanation may also account for the HPE phenotype exhibited by the short digits (Dsh) mouse mutant, which contains an 11.7 Mb inversion on mouse chromosome 5 with a distal breakpoint 13 kb upstream of Shh(Niedermaier et al.,2005).

The positioning of each of the forebrain and axial mesoderm enhancers on the distal side of the translocation breakpoints in the individuals with HPE does little to resolve which elements in particular are required for proper forebrain development. In principle, a reduction in Shh transcription from any of the non-redundant enhancers could cause specific phenotypes associated with HPE. The expression of Shh in the prechordal plate and ventral forebrain is responsible for establishing patterns of growth and differentiation of neuronal and glial precursors in the rostral CNS, including bifurcation of the telencephalic vesicles and craniofacial structures(Chiang et al., 1996; Rallu et al., 2002; Fucillo et al., 2004). Gain- and loss-of-function studies in several organisms have implicated the source of Shh in the prechordal plate as requisite for ventral forebrain development (reviewed by Hayhurst and McConnell, 2003; Wilson and Houart, 2004). However, as the ventral forebrain is missing in mutants lacking a prechordal plate, the specific contribution of Shh signaling from the ventral forebrain itself remains unclear. Interestingly, in chick embryos, the severity of craniofacial defects resulting from a block in Shh signaling correlated with where and when the signal was disrupted(Marcucio et al., 2005). These studies showed that the expression of Shh in the ventral diencephalon and telencephalon had a significant impact on craniofacial morphogenesis. The inactivation of specific enhancers in the mouse that direct Shhexpression to the ventral diencephalon (SBE2) and ventral telencephalon (SBE4)will provide a unique opportunity to investigate more thoroughly the contribution of Shh from a given source with respect to its role in forebrain and craniofacial development.

The objective of the current study was to identify enhancers controlling Shh transcription in the mouse CNS. Coincidently, reporter activity from Bac transgenic embryos was also detected in the gut, axial mesoderm and posterior limb bud, all known sites of Shh expression. Previous studies have shown that the Shh limb enhancer is located ∼1 Mb away from Shh-coding sequences and is mutated in cases of preaxial polydactyly in humans, mice and chickens(Lettice et al., 2003; Sagai et al., 2004; Maas and Fallon,2005). The types of defects caused by alterations in Shhlimb enhancer activity are representative of the growing number of developmental anomalies that result from the disruption of cis-acting regulatory sequences, either by mutation or position effect resulting from chromosomal rearrangements (Kleinjan and van Heyningen, 2005; Lettice and Hill, 2005). Undoubtedly, this number will continue to rise with improved methods of identifying functional regulatory elements. Approaches such as ours will be useful in identifying other long-range enhancers that regulate tissue-specific expression of genes during development.

We thank Dr Jean Richa and his staff at the University of Pennsylvania Transgenic and Mouse Chimeric Facility for their assistance in transgenic mouse production, and Drs Haig Kazazian, Steve Liebhaber and Tom Kadesch for helpful comments on the manuscript. The authors are grateful to Drs Nat Heintz and Denis Duboule for providing the pLD53SCAEB and Tn7βLacZ plasmids,respectively. This work was supported by NIH grant R01 NS39421 from NINDS(D.J.E.), a Pew Scholar Award in the Biomedical Sciences (D.J.E.) and the Division of Intramural Research, National Human Genome Research Institute,National Institutes of Health (M.M.).