The generation of anterior-posterior polarity in the vertebrate brain requires the establishment of regional domains of gene expression at early somite stages. Wnt-1 encodes a signal that is expressed in the developing midbrain and is essential for midbrain and anterior hindbrain development. Previous work identified a 5.5 kilobase region located downstream of the Wnt-1 coding sequence which is necessary and sufficient for Wnt-1 expression in vivo. Using a transgenic mouse reporter assay, we have now identified a 110 base pair regulatory sequence within the 5.5 kilobase enhancer, which is sufficient for expression of a lacZ reporter in the approximate Wnt-1 pattern at neural plate stages. Multimers of this element driving Wnt-1 expression can partially rescue the midbrain-hindbrain phenotype of Wnt-1−/− embryos. The possibility that this region represents an evolutionarily conserved regulatory module is suggested by the identification of a highly homologous region located downstream of the wnt-1 gene in the pufferfish (Fugu rubripes). These sequences are capable of appropriate temporal and spatial activation of a reporter gene in the embryonic mouse midbrain; although, later aspects of the Wnt-1 expression pattern are absent. Genetic evidence has implicated Pax transcription factors in the regulation of Wnt-1. Although Pax-2 binds to the 110 base pair murine regulatory element in vitro, the location of the binding sites could not be precisely established and mutation of two putative low affinity sites did not abolish activation of a Wnt-1 reporter transgene in vivo. Thus, it is unlikely that Pax proteins regulate Wnt-1 by direct interactions with this cis-acting regulatory region. Our analysis of the 110 base pair minimal regulatory element suggests that Wnt-1 regulation is complex, involving different regulatory interactions for activation and the later maintenance of transgene expression in the dorsal midbrain and ventral diencephalon, and at the midbrain-hindbrain junction.
Several lines of evidence indicate that the vertebrate brain develops on a segmental template. The forebrain is thought to be composed of six segments or neuromeres, the midbrain one, and the hindbrain eight (Puelles and Rubenstein, 1993; Rubenstein et al., 1994). While the process of induction and regional specification of the embryonic brain is poorly understood, both vertical signals, arising from underlying endodermal or mesodermal tissues, as well as planar signals, operating within the epithelial layer of the neural plate itself, may be involved (Ruiz i Altaba, 1994; Kelly and Melton, 1995). The role of transcription factors as components of such signaling pathways has been the subject of extensive analysis in the developing hindbrain, where certain Hox genes have essential roles in determination of rhombomeric identity (Krumlauf, 1994; Lumsden and Krumlauf, 1996). Identification of cis-acting regulatory sequences for these factors has facilitated an understanding of upstream mechanisms that determine segmental pattern. For example, Hox-b3 has recently been shown to be a direct target of kreisler (Manzanares et al., 1997).
To further investigate the genetic pathways that control the establishment and elaboration of neuromeric development, we have focused our attention on the embryonic midbrain. Wnt-1 encodes a secreted glycoprotein which is first expressed in the presumptive midbrain at early somite stages in a pattern that is conserved in all vertebrates examined thus far (Wilkinson et al., 1987; Molven et al., 1991; Wolda et al., 1993; Hollyday et al., 1995; McMahon, 1992). Wnt-1 signaling is essential for development of both the midbrain and adjacent rhombomere 1 of the hindbrain (McMahon and Bradley, 1990; Thomas and Capecchi, 1990; McMahon et al., 1992; Mastick et al., 1996; Serbedzjia et al., 1996). An understanding of Wnt-1 regulation would, therefore, shed light on transcriptional mechanisms that underlay regional organization of the vertebrate CNS.
The expression pattern of Wnt-1 in the embryonic midbrain is dynamic and can, in principal, be divided into activation and refinement phases at neural plate stages, as well as a maintenance phase, which commences approximately at the time of neural tube closure and persists throughout embryogenesis and into adulthood (Wilkinson et al., 1987; McMahon et al., 1992; Echelard et al., 1994; D. Rowitch, N. Zec and A. McMahon, unpublished observations). At the condensation of the first somite, Wnt-1 mRNA transcripts are initially detected in a broad segmental domain that demarcates the presumptive midbrain. In the subsequent refinement phase of expression, Wnt-1 transcripts are restricted to the dorsal and ventral midline of the midbrain and caudal diencephalon, a narrow ring rostral to the midbrain-hindbrain junction and the roof plate of the spinal cord. The rostral limit of Wnt-1 expression coincides with the prosomere 2/3 boundary of the diencephalon, and Wnt-1 is expressed in the roofplate of the spinal cord caudal to rhombomere 2.
Regulation of Wnt-1 is poorly understood. We have focused on the activation phase of Wnt-1 expression with the long-term goal of understanding how expression is initiated in a segmentally restricted pattern. Previous work identified a 5.5 kilobase (kb) enhancer 3′ of the Wnt-1 coding region which was capable of driving a lacZ reporter in the correct spatiotemporal pattern (Echelard et al., 1994). Furthermore, the 5.5 kb region has recently been deleted from the Wnt-1 locus by gene targeting to assess its function in vivo. This analysis demonstrated that the 5.5 kb regulatory element is necessary for early activation of Wnt-1 expression (Danielian et al., 1997). Herein we report the identification of an evolutionarily conserved 110 base pair (bp) minimal regulatory region within this enhancer which is sufficient for activation of a lacZ reporter in the Wnt-1 pattern at neural plate stages. Moreover, we show that expression of a Wnt-1 transgene under control of the minimal regulatory element is able to partially rescue the midbrain-hindbrain phenotype of Wnt-1 null mutants, verifying that it contains biologically relevant transcription factor binding sites.
MATERIALS AND METHODS
DNA constructs and sequencing
The reporter transgene vector, XB3 (p1048), is a modification of pWZT7 (Echelard et al., 1994) and comprises the Wnt-1 promoter, lacZ gene, SV40 large T antigen poly(A) site and a downstream Eco72I restriction endonuclease site, which allows blunt-ended enhancer test fragments to be cloned in an arrangement that mimics the positions of gene and enhancer at the Wnt-1 locus (Fig. 1A). DNA sequences tested in reporter transgenes (constructs 1-16; Fig. 1B,C) were removed from the 5.5 kb Wnt-1 enhancer, end-filled using either Klenow fragment of DNA polymerase I (5′ overhang) or bacteriophage T4 DNA polymerase (3′ overhang), and subcloned in to XB3 oriented 5′ to 3′ with respect to lacZ. Restriction endonuclease cleavage sites used in plasmid construction are listed according to their nucleotide position within the 5.5 kb enhancer (GenBank accession number AFO39680). The following construct numbers listed refer to Fig. 1B: 1, BglII (1) to HindIII (720); 2, BglII (1) to FspI (3437); 3, BglII (1) to EagI (4210); 4, BamHI (2309) to BglII (5441); 5, FspI(3437) to EagI (4210); 6, ScaI (3673) to BglI (4790). The following construct numbers listed refer to Fig. 1C: 7, ScaI (3673) to StyI (4600); 8, ScaI (3673) to ApaI (4394); 9, ScaI (3673) to ApaI (4394) + ApaI (4616) to BglI (4790); 10, ScaI (3673) to StuI (4197); 11, StuI (4197) to BglI (4790); 12, StuI (4197) to ApaI (4394); 13, ApaI (4394) to ApaI (4616); 14, StyI (4600) to BglI (4790); 15, Tsp509I (4368) to Tsp509I (4477); 16, three copies of Tsp509I (4368) to Tsp509I (4477). For all reporter constructs listed above, the restriction endonuclease, SalI, was used to purify the transgene from the vector sequences prior to microinjection.
In order to test the pufferfish regulatory element, two copies of a 102 bp fragment, generated by digestion of p38-4e1 with the restriction endonuclease Tsp509I, were first cloned into the EcoRI site of Bluescript KSII+ in the same orientation, and then into the reporter construct, XB3, yielding XB3-Fugu 102, which comprises a total of 4 copies of the 102 bp fragment oriented 5′ to 3′ with respect to lacZ. The restriction endonuclease, SalI, was used to purify the transgene from the vector sequences prior to microinjection.
To test putative Pax DNA-binding sites within the 110 bp regulatory element, a synthetic oligonucleotide pair that spanned sequences from the Tsp509I and AseI sites (Fig. 4A) containing mutations of 5′ and 3′ ‘Pax’ sites (Fig. 4B), was used to replace wild-type sequences. PCR methodology was used to create deletions of the 5′ and 3′ sites (Fig. 4B, details available). These constructs were then multimerized and tested in transgenic embryos essentially as described above. In order to generate recombinant Pax-2 protein, an end-filled NotI-BamHI fragment from pCMV Pax-2b (Dressler et al., 1990) was inserted into the BamHI site of pGEX-2T (Pharmacia Biotech) creating an in-frame fusion between Glutathione S-transferase (GST) and sequences encoding the DNA-binding domain of Pax-2b. Recombinant protein was generated and purified using standard protocols.
The plasmid Wnt-1 R110, which was used to partially rescue the null phenotype of a Wnt-1 mutant generated by gene targeting (McMahon and Bradley, 1990), was constructed as follows. A ClaI-BglII fragment (positions 2344 and 4432, respectively, van Ooyen and Nusse, 1984) was subcloned to create p1412, which contains a unique MluI site at position 4062 in the 3′ untranslated region of Wnt-1. Plasmid 1412 was digested with MluI, end-filled using the Klenow fragment of DNA polymerase I and then treated with calf intestine alkaline phosphatase. Subsequently, an EcoRV-SacI fragment of lacZ was end-filled and cloned 5′-3′ into the above vector to create pWnt-o-tag. A ClaI-BglII fragment was excised from pWRES4 (Danielian et al., 1997) and replaced by similar fragment generated from pWnt-o-tag to create p1424. The 5.5 kb enhancer region of p1424 was then replaced by an oligonucleotide polylinker (which contains a MunI restriction endonuclease cleavage site) into which the trimer array of the 110 bp regulatory element from construct 16 (Fig. 1C) was cloned to create the final transgenic construct, Wnt-1 R110 (p1428) (Fig. 3a). An AatII fragment of this plasmid was purified for pronuclear injection.
DNA sequencing of both strands of the 5.5 kb murine Wnt-1 enhancer was carried out using both ABI dye terminator and di-deoxy chain termination methodologies. Contigs were sorted using IntelliGenetics software. The DNA sequence has been deposited in the GenBank database with the accession number AFO39680.
Cloning the pufferfish wnt-1 locus
Degenerate PCR primers wnt1f (CAA/G GAA/G TGT/C AAA/G TGT/C GG) and wnt1r (ACA/G TGA/G CAA/G CAC CAA/G CAC CAA/G TGA/G AA) were used to amplify and clone a 400 bp fragment of Fugu wnt-1 exon 4 from Fugu genomic DNA. The fragment was used to probe a gridded Fugu genomic cosmid library (‘Elgar library; Baxendale et al., 1995). Cosmid clone 138E3 of the Elgar library was isolated. This was sonicated and subcloned for analysis with ABI dye terminator chemistry on an ABI 373 or 377 automated DNA sequencer. Sequences were assembled with the program ‘seqman’ from the lasargene package (DNASTAR, Inc.). A 33 kb Fugu genomic sequence has been deposited in the GenBank database with the accession number U82608. The Fugu and mouse noncoding sequences were aligned with the programs ‘bestfit’ of the GCG package and ‘clustal W’ (Thompson et al., 1994).
Production and genotyping of transgenic mice
For mapping of Wnt-1 regulatory elements, transgenic mice were generated by microinjection of linear DNA fragments, separated from plasmid vector sequences into pronuclei of BL6CBAF1/J (C57BL/6J ×CBA/J) zygotes as described (Echelard et al., 1994). For testing the Wnt-1 rescue transgene, zygotes for pronuclear injection were obtained from intercrosses of males heterozygous for the Wnt-1 null allele (Wnt-1+/−; McMahon and Bradley, 1990) and BL6CBAF1/J females. Mutant embryos were subsequently generated by intercrossing males hemizygous for the transgene and heterozygous for the Wnt-1 null allele (Tg+,Wnt-1+/−) with heterozygous females (Wnt-1+/−).
Genotyping of transgenic mice or embryos injected with lacZ reporter constructs by PCR was carried out as described in Echelard et al. (1994). Otherwise, lacZ-containing transgene arrays in mouse lines were identified by Southern blot analysis of products of genomic DNA (Laird et al., 1991) digestion with the restriction endonuclease, EcoRI, probed with a labeled DNA fragment comprising nucleotides 1125 (EcoRV) to 1950 (SacI) of the lacZ sequence. Transgenic mice injected with Wnt-1 R110 were genotyped by Southern blot analysis of genomic DNA digested with the restriction endonucleases, SacI and MunI, using a labeled DNA fragment comprising nucleotides 2344 (ClaI) to 4062 (MluI) of the Wnt-1 genomic sequence (van Ooyen and Nusse, 1984). This generated the following diagnostic fragments: 5.5 kb and 4.0 kb (Wnt-1 neo-targeted allele and Wnt-1 wild-type allele, respectively; McMahon and Bradley, 1990) and 3.2 kb (Wnt-1 R110-specific), from which mice or embryos could be assigned with respect to the Wnt-1 and transgene loci. Wnt-1 R110 Tg+ animals were confirmed by PCR using the primers: AACTCTTGAAGGTGTTGCGG (#1966) and TACCACAGCGGAT-GGTTCGG (#137) (Fig. 3A) under conditions described in Echelard et al. (1994).
Whole-mount β-galactosidase staining
14.5 d.p.c. mouse embryos were fixed in Bouin’s fixative for at least 24 hours. They were then dehydrated, embedded and sectioned at 6 μm according to standard protocols. Sections were stained with haematoxylin-eosin and photographed (Kodak 64T film) on a Leitz Aristoplan DMZ compound microscope.
Band-shift reactions (15 μl) were performed in the following solution: 50 mM KCl, 20 mM Hepes pH 7.4, 1 mM DTT, 10% glycerol, containing 1.25 μg of poly(dIdC), 70 μg of bovine serum albumin and 0.1 μg GST or GST-Pax-2 protein. These reactions were incubated for 10 minutes before the addition of 1 ng of radiolabelled probe and then incubated for a further 20 minutes with or without the addition of annealed pairs of oligonucleotides as follows: (1) Pax-2 consensus oligonucleotide pair: 5′-CTAGAGGAATTCAGGAAATTGTCACGCATGAGTGGTTA-GCTCGAT-3′ 5′-CTAGATCGAGCTAACCACTCATGCGTGACAATTTCCTG-AATTCCT-3′ (2) 5′ Site oligonucleotide pair: 5′-CTAGATTAATGAATTGTCCATCACGCCTTTCAGGGCCC-GCCCGT-3′ 5′-CTAGACGGGCGGGCCCTGAAAGGCGTGATGGACAATT-CATTAAT-3′ (3) 3′ Site oligonucleotide pair: 5′-CTAGATTCAGGGCCCGCCCGTCAGCCTGGATTAATCTT-CGGAGC-3′ 5′-CTAGATTAATGAATTGTCCAGACCGCCGCTCAGGGCCC-GCCCGT-3′ (4) HBS oligonucleotide pair: 5′-CACTCTAGACATTGTCACTAATTGAGGTAATTATCTGTG-ATATCTAGACCGGTAC-3′ 5′-CGGTCTAGATATCACAGATAATTACCTCAATTAGTGACA-ATGTCTAGAGTGGGCC-3′ Reactions were then applied to a pre-run 5% polyacrylamide gel and subjected to electrophoresis at 100 V (4°C). Gels were then dried and bands visualized by autoradiography. Radiolabelled probes were generated by end labeling with [α-32P] dCTP and Klenow enzyme either before or after gel-purification or were generated by PCR using T4 polynucleotide kinase 32P-labeled oligonucleotide primers.
Deletion analysis of the 5.5 kb Wnt-1 enhancer in transgenic mice
Previous studies have documented that an enhancer sequence lying within a 5.5 kb region 3′ of the Wnt-1 coding sequence is both necessary and sufficient for transcriptional activation of Wnt-1 . To further define the cis-acting regulatory sequences, we performed a detailed dissection of the 5.5 kb region (Fig. 1). Reporter transgenes were tested in a vector comprising the Wnt-1 promoter, lacZ and SV40 large T antigen poly(A) sequences. Various extents of the 5.5 kb regulatory region were placed downstream of lacZ in keeping with their normal position relative to the Wnt-1 gene and promoter (Fig. 1A ; van Ooyen and Nusse, 1984; Echelard et al., 1994). Reporter constructs were tested for activity directly in injected (Go) founder embryos initially at 10.5 d.p.c. to determine whether expression was appropriately regulated in the brain. Injections of 1-cell mouse embryos with Wnt-1 reporter transgenes routinely yielded 10-40% transgenic embryos at 10.5 d.p.c. Because of copy number and integration site influences on transgene expression (Dobie et al., 1997), as well as mosaicism (Echelard, 1996), approximately half of transgenic embryos express lacZ in the Wnt-1 domain. Consequently, at least eight Go transgenics were generated per construct tested and were only scored as positive for expression when at least 4 embryos showed a similar pattern of β-galactosidase staining. Transgenic embryos that stained in less than 30 minutes were considered to have relatively ‘strong’ acting regulatory elements directing expression of lacZ (black colored boxes, Fig. 1B,C), whereas those taking 1 to 24 hours to stain were judged as relatively ‘weak’ (gray colored boxes, Fig. 1B,C). While these designations reflect consistent staining characteristics in at least 4 embryos, the strength of a regulatory element obviously cannot be strictly quantified in the transgenic reporter where position-dependent effects and copy number vary. Nevertheless, they do offer some relative indication of function.
As indicated in Fig. 1B, a 1.1 kb fragment of DNA (construct 6), located approximately 4 kb 3′ of the Wnt-1 poly(A), was identified which possessed activity similar to the entire 5.5 kb enhancer fragment (Fig. 2A). Follow-up studies in two stable transgenic lines indicated that this element was able to activate Wnt-1 reporter gene expression appropriately at the 1-somite stage (data not shown). Whereas other regions of the enhancer have activity, only this region was responsible for strong activation. It is noteworthy that construct 1 (Fig. 1B) generated the full Wnt-1 pattern in 1 of 19 transgenic embryos tested. It is possible, therefore, that sequences within construct 1 possess weak activity in an integration site-dependent manner. Truncation at the 3′ end of the 5.5 kb enhancer resulted in weaker midbrain expression and 3. ectopic expression in the forebrain (constructs 3, 5, Fig. 1B). Subsequently, we performed a more extensive deletion analysis of the 1.1 kb region.
Several interesting observations were made from this series of constructs (Fig. 1C). Firstly, deletion of an 150 bp fragment within the 1.1 kb enhancer (construct 9) resulted in ectopic expression of the lacZ transgene in the dorsal telencephalon (e.g., Fig. 2B). Thus, sequences within this region (construct 13) are responsible for repression of the Wnt-1 reporter transgene in the forebrain. Constructs comprising proximal or distal halves of the 1.1 kb enhancer both yielded expression in the Wnt-1 pattern (constructs 10 and 11 and Fig. 2B,C). Thus, it is likely that cis-acting regulatory sequences responsible for Wnt-1 expression are repeated. Further dissection of the 3′ half of the 1.1 kb fragment (construct 11) into three rather similar sized pieces (constructs 12, 13, 14) failed to demonstrate any activity, suggesting that sequences present in two or more of the fragments were necessary. Given that none of the subfragments supported activity in transgenic embryos, we performed a scan for direct repeat nucleotide sequences shared by constructs 10 and 11 and this identified a region that spanned two of the subfragments (constructs 12, 13; see Fig. 4A). An 110 bp DNA fragment from this region (construct 15), was excised and tested for regulatory function. One of sixteen transgenic embryos demonstrated weak activity in the dorsal midbrain (data not shown), indicating that this region may include some of the relevant cis-acting regulatory sequences. The low frequency of expression in transgenic embryos may result from position effects often observed with small regulatory DNA fragments (Dobie et al., 1997).
To determine if multimers of the 110 bp fragment would more efficiently drive reporter gene expression, three copies were cloned into the vector, XB3 (construct 16, Fig. 1C). Now, in 31% of transgenics, the reporter gene was expressed in the midbrain portion of the Wnt-1 pattern (Fig. 2D). In contrast, expression in the hindbrain and spinal cord was largely missing, indicating that the regulatory region for more caudal aspects of transgene expression in the neural tube lie elsewhere. To verify that sequences within this 110 bp element were sufficient to direct activation of the lacZ reporter with a temporal and spatial pattern identical to Wnt-1 itself, we generated a permanent transgenic mouse line (Wnt-1 110 lacZ). Embryos were collected at the 1-to 2-somite stage, when Wnt-1 mRNA transcripts are first detected (Wilkinson et al., 1987; McMahon et al., 1992), and β-galactosidase activity was compared with similar stage embryos generated from a previously described line of transgenic reporter mice that contain the full 5.5 kb enhancer (Echelard et al., 1994; Fig. 2E,F). These results demonstrate that the 110 bp minimal enhancer element was sufficient to drive lacZ expression in the midbrain as early as the 1-somite stage (Fig. 2E). However, the 110 bp enhancer was clearly weaker-acting than the full 5.5 kb enhancer, indicating that sequences important for normal levels of expression lie outside this region. Additionally, only 31% of transgenic embryos injected with construct 16 expressed lacZ, compared with approximately 50% of transgenics generated with the full-length 5.5 kb enhancer. This suggests that the 110 bp enhancer lacks sequences that confer relative position independence and/or bolster levels of expression to detectable levels.
Activation of Wnt-1 in the developing brain
To further characterize the capability of the minimal enhancer to drive appropriate expression of Wnt-1 in the neural plate, we determined whether the 110 bp element was sufficient to rescue the Wnt-1 null phenotype. The transgene used in these studies contained the Wnt-1 promoter and transcription unit with a tag inserted into the 3′ untranslated region and, finally, three copies of the 110 bp enhancer, placed 3′ of the gene (Fig. 3A). Of 15 founder transgenic animals, 8 transmitted the Wnt-1 110R transgene through the germline and whole-mount in situ hybridization demonstrated that 3 of these lines expressed the rescue transgene in the expected pattern (data not shown). Thus, 27% of founder transgenics expressed detectable levels of the transgene-derived Wnt-1 transcript, in accordance with the statistics derived from the lacZ reporter transgene studies (construct 16, Fig. 1C). One of these transgenic lines (Wnt-1 110R-18) was chosen for further study based on its relatively high level of transgene expression.
The Tg+, Wnt-1+/− founder male was crossed to Wnt-1+/−?females and 92 weanlings were genotyped (Fig. 3B). Since?Wnt-1 null mutant pups die at birth (McMahon and Bradley,?1990), complete rescue would be indicated by the generation of Tg+, Wnt-1−/− survivors. However, no Wnt-1−/− pups survived to weaning, indicating that expression in the Wnt-1 R110-18 transgenic line was not sufficient for complete rescue of Wnt-1 null mutants. The Wnt-1 null phenotype comprises failure of mesencephalic and metencephalic development during embryogenesis by 9.75 d.p.c. (McMahon and Bradley, 1990; Thomas and Capecchi, 1990). Thus, fourteen Tg+, Wnt-1−/− embryos were also analyzed at 14.5 d.p.c. Of these, 35% had the Wnt-1 null phenotype, which is easily distinguished from wild-type based on the cephalic contour in the parieto-occipital region. The remaining 65% of Tg+, Wnt-1−/− embryos were scored as wild-type before the genotype was determined by Southern blot. Histological analysis was then performed on five such embryos and three Tg+, Wnt-1−/− embryos with null phenotype and these were compared with wild-type controls (Fig. 3). The midbrain tectum and cerebellar anlage has failed to develop in embryos lacking Wnt-1 (compare Figs 3C,D). However, in marked contrast, in five Wnt-1 null mutants that carry the Wnt-1 R110 rescue transgene (Tg+, Wnt-1−/−), the tectum and cerebellar anlage are identifiable (Fig. 3E,F). In some cases, these structures appear to be developing quite normally (compare Fig. 3C,F). In three Tg+, Wnt-1−/− embryos, histological analysis confirmed the null phenotype, demonstrating that the Wnt-1 R110 transgene in these examples is not sufficient to maintain development of mesencephalic and metencephalic structures when analyzed at?14.5 d.p.c. (data not shown). No other abnormalities were observed in the CNS of wild-type embryos that carried the Wnt-1 R110 transgene. In summary, Wnt-1−/− embryos lack mesencephalic and metencephalic regions from 9.75 d.p.c. (McMahon and Bradley, 1990; Thomas and Capecchi, 1990; Mastick et al., 1996). However, expression of Wnt-1 under control of the 110 bp regulatory region is sufficient to support apparently normal development of the midbrain and cerebellum up until 14.5 d.p.c., confirming that it contains transcription factor binding sites with relevance in a biological context. The variability of rescue most likely reflects inconsistency in transgene expression on a mixed genetic background.
Conservation of a putative regulatory region in the pufferfish wnt-1 locus
Homologues of Wnt-1 have been found in all vertebrates studied to date and the conserved pattern of expression in the midbrain is in keeping with evolutionary conservation in the regulation of Wnt-1 expression in the developing vertebrate CNS (McMahon, 1992). Cross-species comparison of such enhancer regions offers a powerful strategy for identifying relevant cis-acting regulatory sequences. To this end, we screened DNA sequences flanking the wnt-1 locus in the pufferfish (Fugu rubripes) for homology to the murine Wnt-1 enhancer sequences. The pufferfish is a vertebrate with a compact genome (1/8 the size of the mouse; Brenner et al., 1993), which may facilitate analysis of complex regulatory regions (Marshall et al., 1994; Aparicio et al., 1995). Alignment of murine and pufferfish sequences flanking Wnt-1 identified a region located approximately 1.8 kb downstream of the stop codon of Fugu wnt-1 covering about 200 bp with a nucleotide identity of about 70% with the mouse Wnt-1 enhancer (but inversely oriented with respect to the murine sequences). Within this element a 102 bp region, which is 69% homologous to the murine 110 bp Wnt-1 minimal regulatory region, was identified (Fig. 4). The Fugu-derived 102 bp fragment was multimerized to four copies and tested in the mouse in vivo reporter assay. As shown in Fig. 5, this construct directed expression of lacZ broadly within the midbrain at neural plates stages (7-somites, Fig. 5A,B), in a pattern very similar to the mouse 110 bp enhancer reporter line (Fig. 5C). However, after neural tube closure (9.5-10.5 d.p.c.; 25-32-somites), the Fugu-derived 102 bp sequences directed lacZ expression only in the dorsal midbrain and caudal diencephalon (Fig. 5D). Expression at the midbrain-hindbrain junction and in the ventral midbrain was absent (compare Figs 2D and 5D). Thus, sequences conserved between the Fugu and mouse are likely to represent regions associated with correct activation of Wnt-1, while differences reflect divergence in regions responsible in some measure for refinement and/or maintenance of the Wnt-1 expression pattern.
To gain further perspective on such conserved and non-conserved sequences, a distal subfragment of the mouse minimal regulatory element generated by digestion with the restriction endonucleases AseI and BsmI (see Fig. 4) was multimerized and tested in the in vivo reporter assay. This resulted in β-galactosidase activity solely in the region of the ventral diencephalon both at neural plate stages and after neural tube closure (data not shown). On the contrary, when the proximal fragment of the minimal regulatory element generated by digestion with the restriction endonucleases Tsp509I and AseI (see Fig. 4) was tested, only 1 of 28 transgenic embryos demonstrated staining of β-galactosidase in the midbrain at neural plate stages (data not shown). Thus, it would appear that both proximal and distal halves of the 110 bp regulatory element cooperate to generate the full pattern of activity in the early midbrain. Alternatively, it is possible that cleavage with the restriction endonuclease, AseI destroys an important regulatory site. We consider this unlikely, however, because the AseI site is not conserved in the Fugu element.
Genetic evidence has indicated that activation of the Wnt-1 orthologue, wingless (wg), during Drosophila segmentation requires paired, a Pax transcription factor (Ingham, 1991; Ingham and Hidalgo, 1993), and that the nature of such interactions is likely to be complex (Mullen and DiNardo, 1995). In vertebrates, Pax-2 and Pax-5 are both expressed in the presumptive midbrain (Nornes et al., 1990; Krauss et al., 1991; Puschel et al., 1992; Adams et al., 1992; Asano and Gruss, 1992; Urbanek et al., 1994; Rowitch and McMahon, 1995). The finding that Pax-2 expression precedes activation of Wnt-1 (Rowitch and McMahon, 1995; McMahon et al., 1992) was in keeping with the possibility that its function was conserved in vertebrates, a proposal that has also been made with regard to regulation of Engrailed genes (Joyner and Martin, 1987; Davis and Joyner, 1988) by Pax transcription factors (Song et al., 1996, Joyner, 1996). Consistent with this model, loss-of-function analyses of the zebrafish Pax-2 homologue leads to a failure of Wnt-1 activation and a severe phenotype with deletion of midbrain and cerebellar structures (Krauss et al., 1992; Brand et al., 1996). Murine loss-of-function alleles for Pax-2 (Keller et al., 1994; Favor et al., 1996; Torres et al., 1996) and Pax-5 (Urbanek et al., 1994) have been described and, while these result in abnormalities in the mid-hindbrain region, none are as severe as that of Wnt-1 mutant, implying that Wnt-1 expression is not completely lost. However, given the recent evidence for functional overlap of Pax-2 and Pax-5 in CNS development (Urbanek et al., 1997; Schwartz et al., 1997), final interpretation of this issue awaits analysis of Wnt-1 expression at the 1- to 2-somite stage in compound Pax-2−/−; Pax-5−/− mutant embryos.
Based on similarity to Pax consensus sequences (Czerny et al., 1993; Epstein et al., 1994; Phelps and Dressler, 1996; Fig. 4B), we identified two potential Pax-2-binding sites in the mouse 110 bp element. To assess if Pax-2 was able to bind to the Wnt-1 110 bp regulatory element in vitro, band shift assays were performed using a GST-Pax-2 DNA-binding domain fusion protein (Fig. 6). This indicated that the fusion protein was able to bind to the 110 bp element and that such binding was competed by a consensus Pax-2-binding site. In order to locate potential Pax-2-binding sites, annealed oligonucleotides of approximately 40 bp, comprising overlapping sequences from the 110 bp regulatory element, were used in competition experiments. This analysis identified oligonucleotides representing the putative 5′ and 3′ Pax sites (Fig. 4B) as the only regions able to compete. This competition was rather poor even at high molar excesses suggesting the possibility of other Pax-binding sites (Fig. 6). In keeping with this, deletion or replacement of the core regions of the putative 5′ and 3′ Pax sites with unrelated sequence failed to disrupt Pax binding to the 110 bp element in vitro (data not shown). However, additional deletion analysis of the 110 bp element indicated that binding of Pax proteins was indeed strongest to the region encompassing the putative sites described above and only weak binding to other areas was detected (data not shown). Identical results were found with full-length Pax-2 protein generated in reticulocyte lysate and transiently transfected COS cells, using larger DNA sequences and different conditions (data not shown). To determine the exact location of the Pax-binding sites, we performed DNase I footprinting with Pax-2 and Pax-5 but this failed to consistently identify a discreet binding site. Since the only possible Pax-2-binding sites identified in our analyses were the putative 5′ and 3′ sites, the effect of mutation of these sequences in transgenic mice was assessed despite the fact that such mutations did not disrupt binding in vitro. This was done because our gel shift conditions may have lacked factors (e.g., cooperating proteins; Fitzsimmons et al., 1996) necessary for strong binding to the putative Pax sites. As shown in Fig. 5E-G, such reporter constructs generated appropriate lacZ expression in the early midbrain and at 9.5 d.p.c. Deletion of these sequences from the 110 bp regulatory element (Fig. 4B) resulted in reporter gene expression solely in the ventral diencephalon after neural tube closure (Fig. 5H) but did not abolish early expression of lacZ in the neural plate (data not shown). From our results, we are unable to rule out a role for Pax proteins in the direct regulation of early Wnt-1 expression, however, the failure to identify binding sites argues against this possibility.
Wnt-1 is expressed in the primordial midbrain and targeted disruption of Wnt-1 results in failure of development of the midbrain and subsequent loss of the anterior hindbrain (McMahon and Bradley, 1990; Thomas and Capecchi, 1990; McMahon et al., 1992; Serbedzjia et al., 1996; Mastick et al., 1996). Establishing the molecular interactions that initiate and maintain Wnt-1 expression would further our understanding of mechanisms that determine pattern in the neural plate and subsequent regional development of the mammalian brain. Our approach has been to precisely map cis-acting regulatory sequences that govern Wnt-1 expression in order to investigate such upstream mechanisms.
Initiation of Wnt-1 expression in the embryonic brain
In previous work, a 5.5 kb cis-acting regulatory region that lies 3′ of the Wnt-1 gene was shown to be sufficient to direct expression of a lacZ reporter transgene in a pattern that closely resembled that of endogenous Wnt-1 (Echelard et al., 1994). Significantly, deletion of the 5.5 kb enhancer by homologous recombination results in no early expression of Wnt-1 and a phenotype that is identical to that of the Wnt-1 null mutant embryo (Danielian et al., 1997). This confirms the essential nature of the 5.5 kb enhancer. To further our understanding of Wnt-1 regulation, we have conducted an extensive dissection of the 5.5 kb enhancer.
Because there are no cell lines available that can model the dynamic pattern of Wnt-1 expression during midbrain development, our analysis was carried out entirely in mice. P19 cells have been shown to express Wnt-1 when induced to differentiate along a neural pathway by treatment with retinoic acid (St-Arnaud et al., 1989). Previously it was demonstrated that the DNA-binding site for WiF-1, which lies upstream of the Wnt-1 gene and mediates activity of a reporter construct in vitro (St-Arnaud and Moir, 1993), was neither required or sufficient for expression of a reporter gene in transgenic mice (Echelard et al., 1994). Subsequently, sequences located within the 5.5 kb enhancer that conferred reporter gene activity after RA treatment were mapped extensively in P19 cells to the stippled region depicted in construct 14 (R. St.-Arnaud, personal communication). However, such identified sequences were neither required (constructs 7, 15) nor sufficient (construct 14) for activity in the transgenic mouse. Thus, we conclude that RA-induced P19 cells are not an accurate model of early midbrain regulation of Wnt-1. Whether there is any relevance at all to in vivo regulation is doubtful. By the same token, murine sequences with homology to a functional promoter at the Xenopus wnt-1 locus (Gao et al., 1994) do not appear to be required for any aspect of reporter transgene expression in mice (Echelard et al., 1994).
Robust activity of reporter transgene expression maps to a 1.1 kb region at the 3′ side of the 5.5 kb enhancer (construct 6). Both the proximal and distal halves of this fragment possessed sequences sufficient for activity at 10. 5 d.p.c. in transgenic embryos (constructs 10 and 11). However, attempts to further subdivide the distal half of the 1.1 kb element (construct 10) failed to yield activity (constructs 11, 12, 13). Subsequently, we were successful in identifying a sequence, repeated in constructs 10 and 11, which, moreover, spanned one of the subdivisions. Multimers of this element demonstrated that this sequence was sufficient for temporal and spatial activation of reporter genes in the Wnt-1 pattern both at neural plate stages and after neural tube closure.
We have further investigated sequences in the 110 bp element by using the minimal element to partially rescue the midbrain phenotype of Wnt-1 null mutants. In particular, we have confirmed that the 110 bp regulatory element is sufficient to express Wnt-1 in a physiologically significant pattern at early stages of development because in 65% of Tg+, Wnt-1−/− embryos there is substantial rescue of the midbrain structures and cerebellar anlage at 14.5 d.p.c. However, since no Tg+, Wnt-1−/− pups survived to weaning there is some quantitative or qualitative deficiency in Wnt-1 expression from this element.
Regulatory sequences involved in activation and maintenance patterns of Wnt-1 expression
Comparative vertebrate genome sequencing with Fugu has been successfully used to identify enhancer elements for Hoxb-4 (Arparcio et al., 1995) and Hoxb-1 (Marshall et al., 1994). The large evolutionary distance separating the pufferfish and mammals (approx. 400 million years) ensures that most sequences have diverged, except where there is some functional importance, e.g., coding and regulatory elements. Comparison of the mouse 110 bp regulatory element to pufferfish DNA sequences revealed a region of high homology. In fact, 69% of the pufferfish genomic DNA sequence was identical to that of the mouse 110 bp regulatory region. As the pufferfish 102 bp regulatory element is capable of driving β-galactosidase activity in the mouse neural plate in a pattern similar to that generated with the murine sequences, the conserved regions are most likely involved in the normal activation of the endogenous Wnt-1 gene. Preliminary analysis of the putative human WNT-1 regulatory region indicates 100% identity in the region of the murine 110 bp regulatory element (D. Rowitch, E. Engle and A. Beggs, unpublished observations), suggesting that mechanisms involved in the regulation of WNT-1 are conserved, as has been suggested previously for the ENGRAILED genes (Song et al., 1996; Zec et al., 1997).
We have focused on conserved regions of the 110 bp regulatory element but have been unable to precisely define the sequences responsible for activation of transgene expression. The 3′ half of the 110 bp element directs expression solely in the ventral diencephalon, whilst neural plate activity of the reporter with the 5′ half has been demonstrated in only 1 of 28 transgenics tested (D. Rowitch, P. Danielian and A. McMahon, unpublished data). These results are most consistent with a model in which both the 5′ and 3′ halves of the 110 bp element cooperate to promote early midbrain activation of Wnt-1. In future work, processive mutations of the element will be tested in order to define cis-acting regulatory sequences within the 110 bp element necessary for transgene activation.
Transcriptional regulation in the mid-hindbrain region during development
The earliest known gene expressed in the midbrain-hindbrain region of the mouse is Pax-2 (Nornes et al., 1990, Puschel et al., 1992; Rowitch and McMahon, 1995), a paired domaincontaining transcription factor. Genetic evidence from Drosophila (Ingham and Hidalgo, 1993) and zebrafish studies (Brand et al., 1996) has raised the possibility that Pax transcription factors are required for activation of Wnt-1 in an overlapping region at the midbrain-hindbrain junction. Although we could demonstrate weak Pax-2 binding to the mouse 110 bp regulatory region in band shift assay, we could not determine exactly the position of binding to allow us to test definitively the hypothesis that Pax proteins are direct regulators of Wnt-1 transcription. The failure to identify specific binding sites may suggest that Pax proteins are not direct regulators of Wnt-1 transcription, but our findings cannot rule out the possibilities that Pax transcription factors bind the Wnt-1 enhancer with low affinity, promiscuously, at a variety of sites, or cooperatively, as has been described for Pax-5 DNA-protein interactions in mb-1 regulation (Fitzsimmons et al., 1996).
The homeobox transcription factor Otx-2 is expressed rostral to the midbrain-hindbrain junction and has essential roles in the formation of the anterior CNS (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996) and establishment of midbrain identity (Acampora et al., 1997). We have identified a candidate homeodomain consensus binding site, HBS1 (Shang et al., 1994; Iler et al., 1995, Fig. 4). However, whilst this site participates in aspects of Wnt-1 regulation (see next section), deletion of HBS1 does not prevent transgene activation at the 1- to 2-somite stage (D. Rowitch, Y. Echelard and A. McMahon, unpublished data). In addition to HBS1, other homeodomain core binding sites are present in the 110 bp regulatory region which have yet to be tested. The preliminary evidence at hand, however, does not support involvement of homeobox genes in Wnt-1 transgene activation. Indeed, identifying the factors responsible for this process will be a challenging undertaking.
Regulation of the Wnt-1 maintenance pattern of expression in the developing brain
Expression of Wnt-1 in its normal pattern during embryogenesis is essential for CNS development, and one component of this is the repression of Wnt-1 in the dorsal telencephalon. In previous work, HBS1, a homeodomain core binding sequence conserved with respect to the pufferfish regulatory sequences, was shown to govern repression of transgene expression in the dorsal telencephalon (Iler et al., 1995). The expression pattern of Emx-2 is consistent with a role in mediating repression of Wnt-1 during normal development (Boncinelli et al., 1993), and Emx-2 has been found to bind HBS1 in vitro (Iler et al., 1995). Moreover, Emx-2 null embryos ectopically express Wnt-1 in the dorsal telencephalon (Yoshida et al., 1997). Taken together, these experiments are consistent with a model in which Emx-2 normally functions to repress Wnt-1 expression in the developing telencephalon. It is not obvious why there should be a mechanism for repression of Wnt-1 in the forebrain, given that five other Wnt genes are expressed in that region (Parr et al., 1993; S. Lee and A. McMahon, unpublished observations). Perhaps it is important to regulate the total dosage of the signal; alternatively, Wnt-1 could have specific effects on cell fate. In any case, preliminary analysis indicates that failure to maintain repression of Wnt-1 may lead to a severe forebrain phenotype (D. Rowitch, S. Lee, Y. Echelard and A. McMahon, unpublished observations).
Interestingly, in null mutants of the homeobox gene, Gbx-2, ectopic expression of Wnt-1 has been observed in an abnormal anterior hindbrain region. This raises the possibility that Gbx-2 may function in part to restrict Wnt-1 expression to the midbrain (Wassarman et al., 1997). Our results are indicative of a complex regulatory mechanism at work in determining the pattern of Wnt-1 expression after neural tube closure. DNA-protein interactions responsible for the activation and maintenance phases of expression from the Wnt-1 reporter transgene appear to be different since the Fugu-derived 102 bp element governs activation but only the dorsal mesencephalic portion of the mature Wnt-1 pattern. Thus, it is of interest that deletion of the non-conserved putative ‘Pax’ sites from the 110 bp regulatory element drives reporter gene expression solely in the ventral diencephalon after neural tube closure, indicating that the deleted region is involved in regulating expression in the dorsal mesencephalon and in the ring rostral to the mes-metencephalic border. It has been suggested that a complex feedback loop may govern the later expression of Wnt-1 and Fgf-8 in adjacent cell populations at the mes-metencephalic border, and that signals from this region comprise an organizing activity controlling aspects of midbrain development (for review see Joyner, 1996). Our results point to a small regulatory region which may mediate the transcriptional response to these cellular interactions. In addition to its early function in the maintenance of Engrailed expression (Danielian and McMahon, 1996), later roles for Wnt-1 in the development of the midbrain have been proposed by several investigators (Bally-Cuif et al., 1995; Fritzsch et al., 1995; Porter and Baker, 1997; Rowitch et al., 1997). Further investigation is likely to shed light on mechanisms responsible for refinement and long-term maintenance of the Wnt-1 expression pattern providing new insights into patterning of the vertebrate brain.
D. H. R. acknowledges a Physician Postdoctoral Fellowship from the Howard Hughes Medical Institute and the NIH (grant HD01182) for support. P. S. D. was a recipient of a postdoctoral fellowship from the Human Frontiers Science Program. We thank Rene St-Arnaud for sharing unpublished results on Wnt-1 regulation in P19 cells, M. Busslingler for providing Pax-2 and Pax-5 cDNA clones, Greg Dressler for a Pax-2b clone and B. Klumpar, J. Williams and B. Wilburn for technical help. These studies were supported by a grants NS32691 and HD30249 from the NIH (to A. P. M.).