Cis-acting regulatory sequences governing Wnt-1 expression in the developing mouse CNS

The Wnt-1 protooncogene was identified in several independently arising mammary tumors in which proviral insertion of Mouse Mammary Tumor Virus (MMTV) resulted in activation and ectopic expression of an apparently normal Wnt-1 gene product (Nusse and Varmus, 1982; Nusse et al., 1984; van Ooyen and Nusse, 1984). Cell culture (Brown et al., 1986; Rijsewijk et al., 1987; Jue et al., 1992) and transgenic mouse (Tsukamoto et al., 1988; Kwan et al., 1992) studies have clearly demonstrated that ectopic expression of Wnt-1 in mammary epithelium leads to morphological transformation in vitro and dramatic hyperplasia in vivo. Thus, a causal relationship between ectopic expression of Wnt-1 and the development of mammary adenocarcinomas has been established (reviewed by Nusse and Varmus, 1992). More recently, studies in a number of developmental systems, most notably Drosophila and the mouse, have started to address the normal role of Wnt-1.

Wnt-1 in the mouse, and its Drosophila counterpart the segment polarity gene wingless (wg), encode cysteine-rich, secreted glycoproteins (Papkoff et al., 1987; Brown et al., 1987; van den Heuvel et al., 1989; Gonzalez et al., 1991) which are members of a large family of putative signaling molecules (reviewed in McMahon, 1992; Nusse and Varmus, 1992). In Drosophila, analysis of multiple alleles indicates that wg is required throughout embryonic and larval development (Baker, 1988). The best studied aspect of wg-mediated regulation lies in patterning of the embryonic segment where wg is necessary to maintain expression of the homeobox-containing transcription factor, engrailed, in posterior cells (DiNardo et al., 1988; Martinez-Arias et al., 1988; Heemskerk et al., 1991; Gonzalez et al., 1991; Bejsovec and Martinez-Arias, 1991) and in the subsequent regulation of cell fate choices adopted by these cells (Dougan et al., 1991). In the absence of correct wg signaling, posterior pattern elements in each of the segments are lost, partly by cell death (Perrimon and Mahowald, 1987), and partly by cells adopting more anterior fates (NüssleinVolhard and Wieschaus, 1980; Dougan et al., 1991). Although later functions are less well understood, wg appears to be essential for development and/or patterning of the limb (Cohen, 1990; Couso et al., 1993; Struhl and Basler, 1993), wing (Couso et al., 1993; Williams et al., 1993) and Malpighian tubules (Skaer and Martinez-Arias, 1992). Thus, wg is used in many different contexts during insect development.

In contrast, in the mouse and all other vertebrate embryos studied to date, Wnt-1 is only expressed during development in the central nervous system (CNS; Wilkinson et al., 1987; Davis et al., 1988; Molven et al., 1991; McMahon et al., 1992; Wolda et al., 1993). Wnt-1 expression is initiated at neural plate stages throughout the presumptive midbrain, then becomes rapidly restricted to a tight circle lying just anterior of the mid/hindbrain isthmus, by neural tube closure. In addition, Wnt-1 expression persists at the dorsal midline of the midbrain and extends anteriorly at the dorsal midline into the diencephalon and continues down the dorsal midline of the caudal hindbrain (myelencephalon) and spinal cord. Transient ventral expression is also seen in the diencephalon and midbrain shortly after neural tube closure. Thus, the earliest Wnt-1 expression demarcates the presumptive midbrain suggesting that Wnt-1 signaling may regulate the development of this brain region and later expression is consistent with a potential role in establishing dorsal-ventral pattern in much of the CNS. Analysis of mutant alleles of Wnt-1 generated either spontaneously (Thomas et al., 1991) or by gene targeting (Thomas and Capecchi, 1990; McMahon and Bradley, 1990) has clearly demonstrated that Wnt-1 is primarily required for development of the midbrain (Thomas and Capecchi, 1990; McMahon and Bradley, 1990; Thomas et al., 1991; McMahon et al., 1992); the loss of the midbrain leading to a secondary loss of anterior hindbrain regions (McMahon et al., 1992). Although it has not been determined if Wnt-1 controls cell fate decision in the midbrain or some other process such as cell growth or survival (for review see McMahon, 1992), these studies demonstrate that activation of Wnt-1 is essential for normal brain patterning. Moreover, the establishment of a specific domain of Wnt-1 expression as early as neural plate stages strongly suggests that activation of Wnt-1 may be regulated by regional patterning signals that are thought to be initiated by the midline mesoderm (for review see Slack and Tannahill, 1992; McMahon, 1993). Interestingly, although Wnt-1 shows dorsally restricted expression in much of the CNS by neural tube closure, there is no apparent phenotype in null mutant embryos. A second Wnt-1 gene, Wnt-3a appears to be coactivated with Wnt-1 in these regions (Roelink and Nusse, 1991; McMahon et al., 1992) and it has been suggested that Wnt-3a signaling may be functionally redundant with that of Wnt-1 (McMahon et al., 1992).

To address the question of how regional expression of Wnt-1 is initiated in the mammalian CNS, we have studied the cis-acting regulatory sequences governing embryonic Wnt-1 expression in transgenic mice. We demonstrate that temporal and spatial expression of Wnt-1 in the presumptive midbrain and at the dorsal midline of the CNS is controlled by a 3′ enhancer. Finally, our experiments highlight a possible link between Wnt-1 expression and the neural crest cell lineage in the mouse.

Genomic Wnt-1 fragments were obtained by screening a λ-GEM12 (Promega) 129/Sv mouse genomic library with a 375 bp MluI-BglII fragment derived from the fourth exon of the murine Wnt-1 gene. Two overlapping clones W1-5.1 and W1-15.1 (Fig. 1) were used in this study. The genomic inserts were excised by NotI digestion and subcloned into pGEM-5Zf⁻ (Promega) generating pW1-5.1 and pW1-15.1. Throughout, we distinguish plasmid-containing genomic inserts from the isolated genomic inserts by the prefix ‘p’for the former.

As an initial step towards the generation of fusion constructs between Wnt-1 genomic sequences and the E. coli lacZ gene, a polylinker was inserted upstream of a lacZ-SV40pA cassette contained in the plasmid pβ-actin PSDKlacZpA (J. Rossant, unpub-lished data). This plasmid was completely digested with HindIII, excising the human β-actin promoter fragment. The oligonucleotides lac1 and lac2 were annealed in 1× ligation buffer (Maniatis et al., 1982), then ligated into the HindIII cut vector to generate pPSD-KlacZpA1, which contains the following restriction sites: SalI, NotI, SnaBI, BstEII, AatII, HindIII, and BglII just upstream of the NcoI site located at the translation initiation site of the lacZ gene.

pWZT7 was generated by ligating a 1.3 kb KpnI-NcoI fragment containing 5′ untranscribed sequence including the minimal Wnt-1 promoter (Nusse et al. 1990) and 5′ transcribed but untranslated sequence up to and including the initiation methionine codon (Fig. 1) into BglII-NcoI-cut pPSDKlacZpA1. To generate other Wnt-1-lacZ fusion constructs, pWZT7 was digested with BamHI, which cuts 3′ of the SV40 polyadenylation cassette, and oligos lac3 and lac4 annealed and ligated into this vector. This adaptor introduced XhoI, BglII, AflII, NruI, RsrII, PacI and SalI restriction sites 3′ of the polyadenylation signal generating pWZT7*. pWZT9B and pWZT14B were obtained by isolating a 3′ proximal 5.5 kb BglII fragment from pW1-5.1, located just downstream of the Wnt-1 polyadenylation sequence and ligating this into Bg1II-cut pWZT7*.

pWZT73 and pWZThsp5 were obtained by the addition of NotI linkers on to the 5.5 kb proximal 3′ BglII fragment and ligation into either NotI-cut pWZT7* to generate pWZT73 or into NotI-cut phspPTlacZpA (Kothary et al. 1989) to place this 3′ Wnt-1 genomic sequence upstream of a 600 bp hsp68 promoter generating pWZThsp5.

pWZT10AC was constructed by a multistep approach. The pW1-15.1 subclone was completely digested with restriction enzymes AatII and ClaI, and the 2,774 bp Wnt-1 AatII-ClaI fragment isolated (Fig. 1). This fragment was ligated into the pGEM-7Zf⁻ vector (Promega) digested with AatII and ClaI, generating pW1-18. This plasmid was digested at the unique NcoI site located at the Wnt-1 translation initiation codon, and the annealed oligonucleotides NB1 and NB2 were inserted, regenerating only the 5′ NcoI site and introducing a new 3′ BglII site. This plasmid, pW1-33, was cut with NcoI-BglII and ligated into the NcoI-BglII lacZ-SV40pA expression cassette from pWZT7* to generate pWZT17NB. Finally, pWZT17NB was partially digested with AatII and ClaI (AatII and ClaI both cut within the lacZ gene) and a 6.5 kb partial fragment isolated containing 5′ and 3′ Wnt-1 sequences surrounding the lacZ-SV40 expression cassette. This fragment was subcloned into AatII-ClaI digested pW1-15.1 to generate pWZT10AC. pWZT8C was obtained by ligating a 21.5 kb ClaI-NotI Wnt-1 3′ fragment isolated from pW1-5.1 and a 13 kb NotI (complete digest)-ClaI (partial) Wnt-1-lacZ-SV40pA-Wnt-1 fragment isolated from pWZT10AC into NotI-digested SuperCos 1 (Strata-gene).

Oligonucleotide sequences are as follows:

lac15′-AGCTGTCGACGCGGCCGCTACGTAGGTTACCGACGTCAAGCTTAGATCTC-3′

lac25′-AGCTGAGATCTAAGCTTGACGTCGGTAACCTACGTAGCGGCCGCGTCGAC-3′λ

lac3 5′-GATCTATCGATCTCGAGAGATCTCTTAAGTCGCGACGGACCGTTAATTAAGTCGAC-3′

lac45′-GATCGTCGACTTAATTAACGGTCCGTCGCGACTTAAGAGATCTCTCGAGATCGATA-3′

NB15′-CATGGCTACGTCACTTAGATCTATGTTACTCA-3′λ

NB2 5′-CATGTGAGTAACATAGATCTAAGTGACGTAGC-3′λ

For mapping Wnt-1 regulatory elements, transgenic mice were generated by microinjection of linear DNA fragments, separated from plasmid vector sequences into the male pronucleus of B6CBAF1/J (C57BL/6J × CBA/J) zygotes as described by Hogan et al. (1986). B6CBAF₁/J females were also used as recipients for injected embryos and for all breeding studies.

Wnt-1-lacZ transgenic embryos were identified by PCR analysis of proteinase-K digests of yolk sacs. Briefly, yolk sacs were carefully dissected free from maternal and embryonic tissues, avoiding crosscontamination between littermates, then washed once in PBS. After overnight incubation at 55°C in 50 μl of PCR proteinase-K digestion buffer (McMahon and Bradley, 1990), 1 μl of heat-inactivated digest was subjected to lacZ-specific polymerase chain reaction (PCR), 94°C/30 minutes λ 55°C/30 minutes λ 72°C/1 minute at 40 cycles in a 20 μl total reaction volume. The oligonucleotides, 5 ′ -TAC-CACAGCGGATGGTTCGG - 3 ′ and 5 ′ - GTGGTGGT- TATGCCGATCGC - 3 ′ amplify a 352 bp lacZ-specific product. As an internal control for the presence of mouse genomic DNA in the yolk sac digests, ZP3 primers (Lira et al. 1990), which amplify a 511 bp fragment, were used in the same reactions.

Embryos from transgenic experiments at 7.5 to 12.5 days post coitum (dpc) were stained with X-Gal as described in Whiting et al. (1991). The embryos were stained for periods ranging from 30 minutes to 12 hours according to the strength of Wnt-1-lacZ transgene expression. The reaction was stopped by washing the embryos in PBS, and immediately postfixing in 4% paraformaldehyde (Sigma). Whole embryos were photographed on an Olympus SZH microscope with an Olympus C-35AD-4 camera, using Kodak Ektachrome 64T film. For histological analysis, embryos were embedded in paraffin wax and 6 μm sections cut, rehydrated and mounted in 80% glycerol under a coverslip. Photomicrographs were obtained on a Leitz Orthoplan 2 microscope under Nomarski optics using Fujichrome 64T film.

Whole-mount in situ hybridization to 9.5 dpc embryos was performed essentially as described by Wilkinson (1992) using single-strand digoxigenin-UTP labeled RNA probes. The lacZ probe was a 495 nt EcoRI-MluI fragment at the 3′ end of the lacZ gene (Kalnins et al., 1983). The Wnt-1 probe contains a 1.0 kb region from the start of the 3′ untranslated sequence to the polyadenylation site of the Wnt-1 gene (van Ooyen and Nusse, 1984).

As a first step towards understanding how regional gene expression is initiated in the early mammalian CNS, we have started to dissect the cis-acting regulatory sequences that govern expression of Wnt-1. Two overlapping genomic fragments covering a 30 kb area encompassing the Wnt-1 gene (Fig. 1) were used to generate all the Wnt-1-lacZ fusion constructs analyzed in our transgenic studies. The first fusion construct, WZT8C, contained 7.0 kb of genomic sequence upstream of the Wnt-1 transcription start sites fused directly to an E. coli lacZ-SV40pA reporter gene cassette. All Wnt-1 exons and introns (4 kb) and approximately 17.5 kb of non-transcribed 3′ flanking region were placed downstream of the reporter. WZT8C was injected into the pronucleus of fertilized eggs and generation 0 (G₀) embryos, examined at 9.5 dpc to 11.5 dpc for presence of the transgene and expression of β-galactosidase. A total of 126 generation 0 (G₀) embryos were analyzed. Of these, only two were transgenic (Table 1), an unusually low frequency. However, in both cases, transgene expression was similar to the previously reported distribution of Wnt-1 RNA (Wilkinson et al., 1987; Parr et al., 1993).

The lacZ reporter gene was expressed at the dorsal midline of the diencephalon, midbrain and spinal cord; in a circle just anterior to the mid/hindbrain border and in the ventral midbrain close to, but excluding, the floor plate (data not shown). Moreover, a characteristic gap of expression occurred in the metencephalon, as previously reported (Wilkinson et al., 1987; McMahon et al., 1992; Parr et al., 1993). Thus, it is clear that Wnt-1 regulatory regions are contained on this large reporter construct. Interestingly, these two transgenic embryos both showed ectopic β-galactosidase activity in newly emerging neural crest derivatives along the trunk and in the branchial arches. In addition, ectopic activity was detected in a small region of the ventral telencephalon.

The poor transgenic frequency precluded our establishing stable transgenic lines with this construct. Thus we decided to test subfragments for cis-acting regulatory sequences. 26 G₀ transgenics containing a reporter construct, WZT10AC (Fig. 1), lacking 3′ flanking sequence but retaining 7.5 kb of upstream sequence and the Wnt-1 transcription unit, were examined. Only two of these showed β-galactosidase expression and, in both instances, expression was weak and did not occur within the normal Wnt-1 expression domain (Table 1). These results support the conclusion that regulatory sequences lie 3′ of the Wnt-1 polyadenylation site. To test this hypothesis directly, transgenic G₀ embryos were generated with a third construct, pWZT9B (Fig. 1), which contains 1.3 kb upstream of the Wnt-1 translation initiation codon, including the minimal Wnt-1 promoter, together with 5.5 kb of 3′ Wnt-1 flanking sequences lying just downstream of the Wnt-1 polyadenylation site. Eleven transgenic mice were generated using the WZT9B construct; among these, five were established as stable lines, and the remainder were examined directly as G₀ embryos. Of these WZT9B transgenics, eight expressed the lacZ reporter gene with a pattern similar to that of the endogenous Wnt-1 gene (Table 1) and WZT8C transgenic embryos. Transgenics containing just the 1.3 kb Wnt-1 promoter region showed no lacZ expression as expected (WZT7 in Fig. 1, Table 1). Moreover, the 3′ regulatory region conferred correct expression on a heterologous heat-shock promoter (WZThsp5 in Fig. 1 and Table 1). These results demonstrate that sequences that are sufficient for spatial regulation of the Wnt-1 gene at 9.5 to 10.5 dpc are located within a 5.5 kb region just downstream of the Wnt-1 polyadenylation signal.

In order to determine whether the spatial restriction conferred by the 5.5 kb BglII fragment on the lacZ-SV40pA reporter gene is position- or orientation-dependent, this element was inverted (WZT14B) or placed upstream of the Wnt-1 promoter (WZT73; Fig. 1). Five of sixteen WZT14B and two of seven WZT73 G₀ transgenic embryos showed correct expression of β-galactosidase in the Wnt-1 expression domain (Table 1). Thus, although altering the position and orientation of the 3′ regulatory element resulted in a decrease in the percentage of transgenics expressing lacZ (72% to 29-36%), correct spatial expression was maintained. In summary, the relative insensitivity of the BglII fragment to position, orientation and promoter sequences indicates that this element contains a cis-acting Wnt-1 enhancer.

The establishment of stable transgenic lines with the WZT9B construct allowed us to examine temporal and spatial expression of the transgene in detail. Five independent transgenic males (lines 97, 135, 188, 195, 206) were generated, and several litters from each line analyzed from 7.5 dpc to 11.5 dpc. F₁ embryos from each of the WZT9B stable lines showed similar reporter gene expression with minimal ectopic expression. Strong histochemical staining was obtained in as little as 5 minutes (line 135) at 37°C, although most lines required 15 to 30 minutes to reveal the normal Wnt-1 expression domain. Differences between lines may reflect position effects rather than transgene copy number as the weakest expression was obtained in offspring from the line with the highest transgene copy number (line 188).

In situ hybridization studies on sections (McMahon et al., 1992) and in whole mounts (Parr et al., 1993) of neural platestage mouse embryos indicate that Wnt-1 transcripts are first detected in the presumptive midbrain at the 1-somite stage. Analysis of transgenic lines first detected β-galactosidase activity restricted to the presumptive midbrain prior to formation of the first somite, presumably reflecting the high sensitivity of histochemical staining (Fig. 3A,B). Thus, activation of Wnt-1 in the midbrain is one of the first signs of regional specialization within the brain. Whilst expression does not occur in all cells at this time, expressing cells are widely distributed across the lateral-medial (future dorsal-ventral) axis as documented for Wnt-1 itself (McMahon et al., 1992; Parr et al., 1993). Thus, activation of the reporter construct under the control of the 3′ Wnt-1 enhancer mimics in vivo activation.

β-galactosidase expression extends progressively more caudal along the dorsal aspects of the hindbrain from three somites, whilst broad midbrain expression is maintained (Fig. 2C-F). Coupled with the progressive caudal extension of Wnt-1 expression at the dorsal midline, there is an anterior extension into the diencephalon and a loss of broad midbrain expression. By 9.5 dpc midbrain expression is limited to the dorsal midline (continuous with the diencephalon), a circle anterior of mid/hindbrain isthmus and the ventral midline either side of the floor plate. This final distribution of Wnt-1 transcripts, which is established by 9.5 dpc, is maintained for several days (Wilkinson et al., 1987). As expected, lacZ expression in the WZT9B lines also shows correct temporal and spatial modifications by 9.5 dpc (Fig. 3A,B) and is maintained until at least 11.5 dpc (Fig. 3C,D). Thus, the cis-acting Wnt-1 regulatory element confers all aspects of the dynamic regulation of Wnt-1 gene expression on the reporter gene construct.

Although we observed a strong concordance between the reported distribution of Wnt-1 transcripts and the observed expression of β-galactosidase in WZT9B transgenic offspring at each developmental stage, some differences were observed. For example, all lines showed strong ectopic lacZ expression in the ventral telencephalon at 9.5 and 10.5 dpc (Figs 3A,C, 5D, 6B). Ectopic expression was also seen in WZT8C transgenics, suggesting that some region not present in 30 kb of genomic DNA, which includes the Wnt-1 gene and 26 kb of flanking sequences, is necessary for telencephalic repression of Wnt-1 expression. More interestingly, we observed widespread expression of lacZ in neural crest derivatives throughout development, both in WZT9B and WZT8C transgenic embryos. This is well illustrated by examining WZT9B transgenic embryos at 9.5 dpc.

Littermates (from line 206) were removed and incubated in the β-galactosidase substrate X-Gal for varying lengths of time. Short incubations resulted in histochemical staining restricted to the normal Wnt-1 expression domains except in more caudal regions of the spinal cord where weakly stained cells were visible with a segmental arrangement (Fig. 4, 30′). With progressively longer incubation, broad expression of Wnt-1 was detected in cells of the emerging cranial and trunk neural crest (Fig. 4, 60-120′). These include neural crest cells around the eye, in branchial arches, cranial ganglia, dorsal root ganglia and developing enteric nervous system (Figs 4, 5). β-galactosidase activity was highest in the most recently formed neural crest. Thus, in contrast to normal Wnt-1 transcription, which is not seen in migrating neural crest, βgalactosidase activity is apparent in several neural crest derivatives. The midbrain also showed more extensive βgalactosidase activity on long incubation (Fig. 4, 120′) reflecting the previously widespread distribution of lacZ expression at earlier stages.

The distribution of labeled neural crest cells was examined in more detail in sectioned material from WTZ9B transgenics (Fig. 5). In the midbrain, neural crest is generated from 4-to 7-somite stages (see Discussion). In midbrain sections at 3 somites, broad β-galactosidase activity is apparent in the neural plate; no activity is visible elsewhere (Fig. 5A). In contrast, β-galac-tosidase-expressing midbrain neural crest cells can be seen in the head mesenchyme by the lateral (future dorsal) aspects of the midbrain at the 5-somite stage (Fig. 5B). By 9.0 dpc, cranial neural crest is found in anterior head regions in both the first and second branchial arch and surround-ing the rostral spinal cord (Fig. 5C). At 9.5 dpc, cranial crest cells still expressing β-galactosidase are condensing to form cranial sensory ganglia (Fig. 5E). By 10.5 dpc, new neural crest cells are only generated in trunk regions of the CNS. In caudalmost areas of the spinal cord, β-galactosidase expression is restricted to the dorsal midline in a distribution indistinguishable from endogenous Wnt-1 RNA expression (Fig. 5F). The trunk crest, which emerges (Fig. 5G) and migrates ventrally (Fig. 5H) at progressively more rostral levels in the spinal cord, continues to express β-galactosidase.

Previous cell lineage studies in the chick (BronnerFraser and Fraser, 1988) injecting single cells at the dorsal midline of the spinal cord have shown that these cells will give rise to progenitors which form the dorsal midline of the CNS (roof plate) and neural crest derivatives, suggesting that these cells share a common lineage. Wnt-1 expression at the dorsal midline, which precedes emigration of neural crest cells, is confined, at the RNA level, to the CNS. The observed expression of the lacZ reporter gene may simply reflect abnormal ectopic activity of the regulatory elements in the neural crest. However, given the normal location of Wnt-1 expression and the cell lineage studies in the chick, it seems as likely that neural crest cells and the roof plate share a Wnt-1-expressing precursor cell. Perdurance of reporter gene RNA or residual enzyme activity thereby would mark the neural crest cells arising from these dorsal midline precursor cells. As would be predicted if this explanation is correct, most recently emerging neural crest cells stain strongest and little or no neural crest cell staining is visible by 11.5 dpc except in the most caudal region of the trunk where the newest crest cells have recently emerged (Fig. 3D). This explanation gains support from wholemount in situ hybridization analysis of Wnt-1 and lacZ RNA expression in transgenic and non-transgenic embryos (Fig. 6). Transcripts show similar spatial distribution within the CNS although there is strong ectopic expression of lacZ RNA in the ventral telencephalon (Fig. 6B), and possibly stronger expression of lacZ RNA than expected in the ventral midbrain (compare Fig. 6A with 6B). However, even after extensive incubation, lacZ RNA was only detected in the most recently arising neural crest cells of the trunk as they migrate through the sclerotomal component of the somite. In contrast, long periods of staining for lacZ expression detected β-galactosidase activity in forming cranial and dorsal root ganglia and in crest derivatives within the branchial arches (Fig. 4).

Our results demonstrate that correct temporal and spatial expression of the mouse Wnt-1 gene in the CNS depends upon a cis-acting regulatory region lying 3′ of the Wnt-1 polyadenylation site. Sequences contained within a 5.5 kb fragment can act independent of orientation, and on a heterologous promoter, to activate, modify and maintain expression of a reporter gene construct in the normal Wnt-1 expression domain. In this study, it is impossible to quantitate rigorously the effects of position and direction on the Wnt-1 enhancer, which would require the generation of many stable trangenic lines. However, constructs in which the enhancer is displaced give rise to a smaller proportion of G₀ embryos, which express correctly the reporter, indicating an increased sensitivity to position effect. Whilst reporter gene expression largely mimics the normal Wnt-1 expression domain, consistent ectopic expression in the ventral telencephalon was observed suggesting that negative regulatory elements responsible for repression of Wnt-1 expression in the telencephalon are missing from the test constructs. However, it is clear that expression is substantially normal. Although the Wnt-1 gene has been cloned from a number of species including zebrafish (Molven et al., 1991), Xenopus (Noordemeer et al., 1989), man (van Ooyen et al., 1985) and Drosophila (Baker, 1987; Risjewijk et al., 1987; Cabrera et al., 1987; Uzvölgyi et al., 1988), our results constitute the first identification of cis-acting regulatory regions governing normal embryonic expression.

Currently, there is considerable interest in the role of Wnt-1 in vertebrate and invertebrate development where Wnt-1 signaling plays key regulatory roles. In vertebrates, Wnt-1 expression is initiated at early neural plate stages in the presumptive midbrain and is essential for subsequent development of this and the surrounding metencephalic region (Thomas and Capecchi, 1990; McMahon and Bradley, 1990; McMahon et al., 1992; Thomas et al. 1991). Thus, activation of Wnt-1 is one of the earliest signs of regional organization of the neural plate. A number of studies (reviewed in Slack and Tannahill, 1992; McMahon, 1993) have led to the conclusion that regional gene expression in the anterior CNS is initiated in response to mesodermally derived signals from the dorsal mesoderm operating vertically and/or in the plane of the neural plate. Thus transcriptional activation of the Wnt-1 gene, which encodes a putative signaling molecule, may be a direct response to reception of mesodermally initiated signal or signals that regulate anterior-posterior patterning of the vertebrate brain. Elsewhere in caudal regions of the hindbrain and spinal cord, Wnt-1 expression at the dorsal midline is one of the first examples of dorsal-ventral polarity in the CNS (Wilkinson et al., 1987; McMahon et al., 1992) suggesting that Wnt-1 signaling may potentially regulate some aspect of dorsal neural tube development. Although there is no apparent phenotype in these regions in Wnt-1 mutant embryos (Thomas and Capecchi, 1990; McMahon and Bradley, 1990) a second Wnt-gene, Wnt-3a, is coactivated in this domain suggesting that these members may be functionally redundant (McMahon et al., 1992), a conclusion supported by biological assays of Wnt-properties (Wolda et al., 1993; G. Wong, B. Gavin and A.P. McMahon, unpublished data). Thus, dissection of the regulatory pathway that controls Wnt-1 expression in the vertebrate CNS will lead to a better understanding of how CNS pattern is regulated and may potentially identify aspects of these pathways which regulated other Wnt genes such as Wnt-3a. Recent experiments indicate that establishment of dorsal expression of Wnt-1 may depend on signaling from the non neural ectoderm (Takada et al., 1994) It will be interesting to determine where these ectoderm-responsive regulatory regions lie in the Wnt-1 enhancer.

At present, there are few clues as to the nature of cis-acting regulatory factors. Several studies (St-Arnaud et al., 1989; Schuuring et al., 1989) have demonstrated that P19 embryonal carcinoma cells when induced to differentiate along neural pathways by application of retinoic acid (Jones-Villeneuve et al., 1982; McBurney et al., 1982) activate endogenous Wnt-1 expression. Activation of reporter gene fusion constructs transfected into P19 cells has identified an activity that binds to a GC-rich site (−206 to −186) (St-Arnaud and Moir 1993) in the Wnt-1 minimal promoter. This factor, WiF-1, is only present following retinoic acid treatment and its appearance correlates well with activation of reporter constructs as well as the endogenous Wnt-1 gene. However, our experiments indicate that Wnt-1 promoter sequences are not essential for normal embryonic expression. Thus, it seems unlikely that WiF-1-mediated regulation plays a significant role in vivo.

Much of the skeletal structures of the head, peripheral nervous system of the head and trunk, and pigment of the body is derived from the neural crest (for review, see LeDouarin 1982). These cells, which have their origin in the dorsal aspects of the neural plate in all vertebrates, are formed in a rostral-caudal gradient during development. The first crest cells arise in the midbrain/hindbrain regions at 4-somites in the mouse (Nichols, 1981; Chan and Tam, 1988; Serbedzija et al., 1992) and emigration is completed by the 7- (Chan and Tam, 1988) to 14- (Serbedzija et al., 1992) somite stage. In the trunk, crest cells start to emerge from the neural tube 2 to 4 somites rostral to the most recently formed somite and continue to emigrate from this region for 36 to 48 hours (Rawles, 1947; Erickson and Weston, 1983; Serbedzija et al., 1990). The pathway of crest migration and the contribution of crest cells to specific derivatives has been extensively studied in avian species and to a lesser extent in the mouse (for review, see Bronner-Fraser, 1993). These studies indicate that there is a ventral-to-dorsal progression in migratory pathways and that some, if not all, crest cells are multipotent during this migratory phase. Stemple and Anderson (1992) have isolated neural crest cells and demonstrated in vitro, that single cells can generate neural and non-neural crest derivatives. Thus, the diversity of cell types formed in the trunk region may be controlled by environmental factors encountered along the migratory pathway or at the end points of migration.

Although the migratory and terminal differentiation phases of neural crest development have been extensively investigated, relatively little is known of the lineage relationships and mechanisms by which these cells are segregated from the neural tube. The observed expression of active β-galactosidase enzyme within migrating and differentiating neural crest cells in our studies suggests that Wnt-1-expressing cells generate neural crest cells and the roof plate in both cranial and caudal regions. Alternatively, lacZ expression in the neural crest may result, in transgenic mice, from the absence of a repressor element which keeps the gene shut off in the neural crest lineage, or from de novo activation. There is some evidence in support of the former explanation. Firstly, high levels of lacZ RNA are only detected in the normal Wnt-1 expression domain. Low levels of lacZ RNA are restricted to the newly emerging neural crest. Even after long periods of incubation, transcripts are not detected in more mature crest population. Thus expression is transient, consistent with the gradual decay of a more stable lacZ RNA. At the protein level, β-galactosidase activity is only strong in the normal Wnt-1 expression domain and in the newly emerging crest. Staining in earlier formed crest lineages shows a strict temporal-spatial relationship, strongest in the most caudal and recently formed neural crest, weakest in the most rostral and earliest crest cells. Thus the data appears most consistent with the gradual loss of lacZ RNA and β-galactosidase enzyme activity in the neural crest rather than an extremely transient de novo activation. Given the previous results from clonal analysis of cell lineage in the dorsal neural plate, which demonstrated that the roof plate and neural crest shared a CNS precursor (Bronner-Fraser and Fraser, 1988), the simplest explanation of our results is that this precursor normally expresses Wnt-1. If this is the case, down regulation of Wnt-1 is presumably an early step in the genesis of neural crest cells. Thus, the perdurance of high levels of lacZ RNA and β-galactosidase enzyme activity in transgenic mice may act as in vivo lineage tracer identifying recent ancestral relationships between the CNS and neural crest in cranial and trunk region. This issue may be best addressed by the introduction of the lacZ gene into the normal context of the Wnt-1 locus. However, this data is not currently available.

There are several significant aspects to this observation. By isolating populations of β-galactosidase-expressing cells using fluorescent sorting techniques, premigratory crest precursors may be purified and their differentiation program studied in vitro. Simple tracing of β-galactosidase-expressing cells will also facilitate the analysis of normal neural crest cell migration. By tracing the distribution of β-galactosidase-expressing, migrating neural crest cells, it should be possible to determine whether all crest cells are formed from the postulated Wnt-1-expressing precursor or if there are populations that do not express Wnt-1. Initial results indicate that cranial and spinal ganglia, and skeletogenic neural crest cells in the branchial arches, are all derived from Wnt-1-expressing precursor cells. In addition, by crossing our transgenic stocks with mouse mutants showing neural crest phenotypes, e.g., Splotch mice, which carry mutation within the Pax-3 gene (Epstein et al., 1991), we may be better able to understand the nature of the neural crest deficiencies in such mutants. Finally, these results may point to new phenotypes, which we may expect in compound Wnt mutants. As discussed earlier, Wnt-1 and Wnt-3a, which are coactivated at the dorsal midline of much of the CNS, may share equivalent biological properties. Like Wnt-1 mutants, Wnt-3a null mutant embryos show no obvious phenotype in the dorsal CNS or PNS derivatives (Takada et al. 1994). Thus, the generation of compound Wnt-1/Wnt-3a mutant embryos may be required to reveal complete phenotypes for these members. If Wnt-1/Wnt-3a regulate the development of a roof plate and neural crest precursor cell, loss of both of these genes may lead to an additional neural crest phenotype.

We thank Dr Anton Berns for the gift of 129/Sv genomic library, Dr Janet Rossant for pApB-actin PSDKlacZ pA, Dr Paul Wassarman, Dr Colin Stewart for critical reading, and Alice O’Connor for typing the manuscript. Y. E. wishes to thank Dr Gwen Wong, Dr Mark Labow, and Dr Kevin Stark for helpful discussions and technical advices. Y. E. was a recipient of Canadian Medical Research Council and Human Frontier Science Program Organization fellowships.