Characterization of the Complete Genomic Structure of the Human WNT-5A Gene, Functional Analysis of its Promoter, Chromosomal Mapping, and Expression in Early Human Embryogenesis*

We report the complete genomic organization of the human WNT-5A gene, which encodes a cysteine-rich growth factor involved in cell-cell signaling during growth and differentiation. The gene comprises five exons with the terminal exon coding for a large 3 (cid:42) -untrans-lated region of (cid:39) 6.5 kilobase pairs and utilizes multiple polyadenylation signals to generate at least four dis- crete transcripts. We discovered a new leader exon interrupted by a 411-base pair intron that was retained in our original cDNA cloning. The promoter region was located in a GpC-rich island and harbored numerous cis -acting elements including several GC boxes and Sp1, AP1, and AP2 binding motifs. It lacked TATA or CAAT boxes typical of housekeeping and growth factor genes. In support of this, primer extension revealed two tran- scription start sites. Transient cell transfection assays showed functional promoter activity for the 3.9-kilobase pair 5 (cid:42) -flanking region. Interestingly, internal and 5 (cid:42) deletions revealed that the distal promoter was not re- quired for full transcriptional activity and that the first 631 base pairs of WNT-5A harbored the strongest pro- moter activity. Using a panel of rodent-human hybrid DNAs carrying portions of chromosome 3p, we mapped the gene to 3p14.2-p21.1, Fetal Material, Synthesis of Riboprobes, and In Situ Hybridization Analysis —Human embryos were obtained from elective thera- peutic terminations of pregnancy. Approval was obtained from the Ethical Committee of the Medical Faculty, the University of Hong Kong for work with fresh embryonic materials and archival specimens. The gestational ages of archival specimens was determined by assessing the centers of ossification and the histological characteristics of the stained sections. The crown-rump length was measured for intact fresh speci- mens. Five embryos, two at 28 days and one each at 33, 38, and 42 days, were studied. Archival samples of human embryos were previously fixed in 10% (v/v) neutral buffered formalin. Fresh embryos were fixed in 4% paraformaldehyde. Both sources of human embryonic tissue were analyzed by in situ hybridization. Sense and antisense riboprobes for in situ hybridization were generated from a recombinant plasmid contain- ing a 498-bp Dra II– Sac I fragment of human WNT-5A cDNA (951–1448) subcloned into pBluescript II KS (Stratagene). Synthesis of [ 35 S]UTP- labeled riboprobes, in situ hybridization, autoradiography, and histological staining were carried out as described (36, 37). Single-stranded and antisense riboprobes labeled with [ (cid:97) 35 S]UTP were generated using T3 and T7 RNA polymerase, respectively. As controls, hybridiza- tions were riboprobes. coun-terstained hematoxylin

The highly conserved WNT genes belong to a growing family of signaling molecules and potential proto-oncogenes that are involved not only in mammary carcinogenesis but also in the regulation of pattern formation during embryogenesis and differentiation of cell lineages (1)(2)(3)(4)(5). In general, WNT genes encode 38 -43-kDa glycoproteins with features typical of secreted growth factors, including a hydrophobic signal peptide, the absence of additional transmembrane domains, highly conserved cysteine residues, and the presence of N-linked glycosylation sites (5). WNT proteins are closely associated with the cell surface (6) and the extracellular matrix (7). Accordingly, the radius of action of the paracrine effect extends across a few cell lengths, consistent with a factor diffusing over short distances in the extracellular matrix (8). Some of the WNT proteins also possess an ability to induce cell proliferation via an autocrine or paracrine route. In this respect, they represent multivalent factors that are capable of governing developmental processes by acting on different cellular systems similarly to the fibroblast growth factor or transforming growth factor-␤ families (9). For example, the neural cell line PC12 can be phenotypically modified upon transfection with the mammalian WNT-1 cDNA (10), and these effects are associated with modulation of plakoglobin and E-cadherin and an accompanying modulation of calcium-dependent cell adhesion (10). In support of this study is the determination that armadillo, a Drosophila homologue of mammalian plakoglobin, is directly involved in the wg-mediated signal transducing pathway (11,12).
The first WNT-5A gene was identified in the mouse where multiple transcripts ranging from 4.6 to Ͼ8 kb 1 encode a cysteine-rich protein of 379 amino acids whose temporal expression correlates with spatial patterning and morphogenesis (13). The Drosophila (14,15), frog (16), axolotl (17), and zebrafish (18) ortholog of the mammalian WNT-5A have been cloned, and their expression is also developmentally regulated. Of note, in embryos of Xenopus laevis, overexpression of WNT-5A leads to complex malformations that are distinct from those achieved by expression of other WNT genes (19). These findings suggest that WNT-5A has a potential activity in modifying the morphogenetic movements of tissues (19). During mouse embryonic development, the expression of WNT-5A gene follows a complex spatial and temporal pattern; it is associated with several distinct sites including the central nervous system, the facial processes, and the posterior region of the fetus; in the adult, expression is restricted to brain, heart, and lungs (13). WNT-5A is expressed at a low level in the mammary gland of virgin animals, but expression is increased during pregnancy, with a peak in early pregnancy followed by a rapid decline to undetectable levels (20,21). Moreover, WNT-5A expression is upregulated during retinoic acid-induced neural differentiation of P19 embryonal carcinoma cells (22). Two recent independent reports showed a lack of WNT-5A transforming activity based upon its inability to induce morphological transformation or altered growth characteristics onto C57MG mouse mammary epithelial cells (23,24). The results of a screening of 10 different WNT members (24) suggest that this family of growth factors should be divided into three distinct groups: (i) a highly transforming group comprising WNT-1, WNT-3A, and WNT-7A, (ii) a moderately transforming group comprising WNT-2, WNT-5B, and WNT-7B, and (iii) a nontransforming group comprising WNT-4, WNT-5A, and WNT-6. Overall, these data demonstrate that WNT genes have distinct effects on cell growth and that they should not be regarded as functionally equivalent (24).
We have recently cloned the human WNT-5A cDNA (25) and discovered an aberrant expression of the gene in several human tumors, including lung and breast carcinomas and melanomas (26). To facilitate the study of human WNT-5A gene and to understand its transcriptional control, we have characterized the entire genomic organization of WNT-5A, established functional promoter activity for the 5Ј-flanking region in transient cell transfection assays, determined a finer chromosomal location for the gene, and investigated its expression in early human embryogenesis. This knowledge should provide the molecular basis for understanding the genetic control and fine tuning of this important gene during developmental processes and should allow the discovery of genetic diseases linked to this signaling molecule.

EXPERIMENTAL PROCEDURES
Materials-All the reagents were of molecular biology grade. Radionucleotides [␣- 32  Genomic Clones and DNA Sequencing-A lambda phage FIX II library (Clontech) prepared from partial MboI digestion of human WI38 lung fibroblast DNA was used for screening. Approximately, 5 ϫ 10 5 plaques were screened using two c-DNA probes, a 1.4-kb fragment derived from the 5Ј end of the gene containing the coding sequence, and a 1.2-kb fragment from the 3Ј-untranslated region (25). Both probes were labeled to high specific activity by the random priming method (27) utilizing a kit from Stratagene. The filters (Duralose from Stratagene) were hybridized separately for each probe overnight at 65°C in a solution containing 5 ϫ SSC (1 ϫ SSC ϭ 0.15 M NaCl, 15 mM sodium citrate), 1 ϫ Denhardt's reagent, 0.1% SDS, and 100 g/ml denatured salmon sperm DNA. After washing the filters at high stringency (0.5 ϫ SSC, 0.1% SDS at 65°C), the filters were exposed to Kodak XAR-5 film at Ϫ70°C in a cassette containing intensifying screens and subsequently eleven positive clones were identified. Two strongly positive phage clones designated W1 and W10 were selected and purified by sequential plaque screening for further characterization. DNA isolated from each phage clone was characterized by restriction enzyme digestion and Southern blotting analysis and hybridizing fragments containing exons were subcloned into pBluescript II KS (Stratagene). Plasmid DNA was sequenced by an automated sequencing system (Applied Biosystems) utilizing T3 or T7 primers in pBluescript or synthetic oligonucleotide primers. Computer analysis of DNA sequences was done using the GCG or PC/GENE computer programs as described previously (28).
RNA Isolation and Analysis-RNA was isolated from confluent monolayer of human embryonic fibroblasts or MG-63 osteosarcoma cells by a modified method of Chomczynski and Sacchi (29) using a commercially available "Tri" reagent (Molecular Research Corp, Cincinnati, OH). RNA was dissolved in Formazol (Molecular Research Corp.) and stored at Ϫ70°C. Approximately 10 -15 g of RNA was separated by electrophoresis in 1% agarose gels containing 0.66 M formaldehyde (30), transferred to nylon membranes, and hybridized to 32 P-labeled DNA probes, as indicated. Hybridization was done overnight at 42°C in 50% formamide, 5 ϫ SSC, 25 mM sodium phosphate (pH 6.8), 0.5% SDS, 10 ϫ Denhardt's reagent, and 250 g/ml denatured salmon sperm DNA. After hybridization, the blots were washed in 2 ϫ SSC, 0.1% SDS at 50°C before exposure to film.
Primer Extension and Reverse Transcriptase Polymerase Chain Reaction-The transcription start site of the human WNT-5A gene was analyzed by primer extension. Total RNA from fetal human fibroblasts (cell line CRL 1262) and human osteosarcoma cells (cell line MG-63) was isolated using "Tri" reagent. Approximately, 8 g of RNA was used as a template for a specific 21-base oligonucleotide (AACAAGTGGAGC-CAGAATTAA) corresponding to the inverse complement of ϩ2 to ϩ22 of the original WNT-5A cDNA sequence (25). The primer (10 pmol) was end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase (Promega) for 10 min at 37°C. The reaction mix was heated to 90°C to inactivate the enzyme and passed over a Sephadex G-25 spin column (5 Prime 3 3 Prime, Inc. Boulder, CO) to remove unincorporated ATP. Annealing of the primer (0.1 pmol) to the RNA was allowed to occur in 10 l of Superscript (RNase-free) reverse transcriptase (Life Technologies, Inc.) buffer (50 mM Tris-HCl, pH 8.3, 40 mM KCl, 6 mM MgCl 2 , and 1 mM dithiothreitol) for 20 min at 58°C followed by slow cooling to room temperature for 10 min. Extension of the primer was done by addition of 10 l of a solution containing 1 ϫ Superscript II reverse transcriptase buffer, 0.5 mM each of dATP, dCTP, dGTP, and dTTP, and 200 units of Superscript reverse transcriptase. Incubation was done at 42°C for 30 min. The reaction mixture was precipitated with 0.3 M sodium acetate (pH 5.2) and 3 volumes of 100% ethanol. After centrifugation the pellet was washed in 70% ethanol, dried at room temperature, and allowed to dissolve in loading buffer (Promega). The samples were heated to 90°C for 10 min prior to loading on a 6% sequencing gel.
For RT-PCR, 5 g of total RNA from human embryonic fibroblasts (CRL-1262) were incubated with an antisense primer 5Ј-TGCAGTTC-CACCGTCGAT-3Ј located in exon 3 in the presence of avian myeloblastosis virus reverse transcriptase according to the manufacturer's recommendations (Invitrogen, CA). For amplifying the cDNAs products, 2 l of RT reaction was incubated in a 50-l volume containing 5 l of 10 ϫ PCR buffer, 1 l of 100 mM of dNTPs, 1 M of a sense primer 5Ј-AGCCCAGGAGTTGCTTTG-3Ј located in exon 2 and the same antisense primer as above, and 4 units of Taq polymerase (Perkin-Elmer). The PCR conditions included a hot start, 5 min at 94°C, followed by 35 cycles for 1 min at 94°C, 2 min at 55°C, and 3 min at 72°C.
Promoter Constructs, Nested and 5Ј Deletions, and Transient Cell Transfection Assays-The largest promoter construct, which contained 3888 bp of WNT-5A sequence upstream of the 307-bp leader exon, was cloned upstream of the CAT gene within the pBS vector, the pBSCAT (31). The pBSCAT contained a promotorless CAT reporter gene within pBluescript and was generated as described before (31). A nested deletion was generated by digesting the first construct of Ϸ4.2 kb with the restriction endonuclease HindIII. This resulted in an internal deletion of a Ϸ750-bp fragment that was subsequently ligated into the corresponding site of the pBSCAT (32). Finally, we generated a 5Ј deletion construct encompassing the 631 bp upstream of the leader exon. Briefly, the largest 3888-bp construct was partially digested with HindIII, producing three DNA fragments; two of them spanning from Ϫ3888 to Ϫ1382 and Ϫ1382 to Ϫ632 bp were discarded, whereas the remaining fragment, comprising a sequence between Ϫ631 and Ϫ1 bp relative to the major transcription start site, was isolated, purified, and subjected to self-ligation. All of these constructs were checked by sequencing both ends with either internal primers or primers based on the T3 or T7 promoter sequence of the pBSCAT vector.
For transient cell transfection, HeLa cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 4 mM L-glutamine, 100 units/ml penicillin at 37°C in a humidified 5% CO 2 atmosphere. Transient transfections of HeLa cells were performed by the calcium phosphate method as described previously (32). Briefly, confluent 75-cm 2 flasks of Hela cells were transfected with 20 g of WNT-5A promoter/pBSCAT plasmid along with 10 g of pRSV-␤-galactosidase plasmid DNA in order to normalize for transfection efficiency. In addition, all the transfection assays were performed with positive (SV40-driven CAT) and negative (promotorless CAT) vectors. The cells were incubated in Dulbecco's modified Eagle's medium for 12 h at 37°C. After an additional 48 h of incubation, the cells were rinsed twice with a phosphate-buffered saline solution (Ca 2ϩ -and Mg 2ϩ -free) and assayed for ␤-galactosidase activity according to standard procedures (33). CAT assays were performed as described before (31). The acetylated prod-ucts were separated in preslotted thin layer chromatography plates and subjected to autoradiography. The resultant autoradiograms were quantified by scanning laser densitometry and computer integration. Additional experimental details are provided in the text and legends to figures.
DNA Isolation, Southern Blotting, and Regional Chromosomal Mapping of the Human WNT-5A Gene-Genomic DNA was isolated as described before (34). Tissue culture cells were centrifuged, and the resulting pellets were mixed with 600 l of a solution containing 50 mM Tris (pH 8.0), 100 mm EDTA, 100 mM NaCl, 1% SDS, and 585 g/ml Proteinase K (Boehringer Mannheim) and incubated at 55°C overnight with occasional mixing. The solution was then extracted once with phenol and again with phenol:chloroform (24:1), and the DNA was precipitated with 2 volumes of 100% ethanol. The DNA precipitate was removed with a capillary pipette and dissolved in 10 mM Tris, 0.1 mM EDTA, pH 8.0 (TE). Approximately 10 g of genomic DNA was digested with HindIII, precipitated, and redissolved in TE, and DNA fragments were separated on a 0.7% Tris borate-EDTA (TBE) agarose gel. The DNA was transferred to nitrocellulose supported by polyester films (30) and hybridized to a WNT-5A cDNA 1.4-kb fragment (ϩ443 to ϩ1904) generated by PCR. Probes labeled with [␣-32 P]dCTP were prepared by random priming using a kit from Stratagene (San Diego, CA). For regional chromosomal mapping, a rodent-human hybrid panel retaining partial chromosome 3 was used as described previously (35). Hybrid DNAs were tested for the presence of specific human WNT-5A restriction fragments detected by radiolabeled WNT-5A cDNA using standard Southern hybridization methods described above.
Human Fetal Material, Synthesis of Riboprobes, and In Situ Hybridization Analysis-Human embryos were obtained from elective therapeutic terminations of pregnancy. Approval was obtained from the Ethical Committee of the Medical Faculty, the University of Hong Kong for work with fresh embryonic materials and archival specimens. The gestational ages of archival specimens was determined by assessing the centers of ossification and the histological characteristics of the stained sections. The crown-rump length was measured for intact fresh specimens. Five embryos, two at 28 days and one each at 33, 38, and 42 days, were studied. Archival samples of human embryos were previously fixed in 10% (v/v) neutral buffered formalin. Fresh embryos were fixed in 4% paraformaldehyde. Both sources of human embryonic tissue were analyzed by in situ hybridization. Sense and antisense riboprobes for in situ hybridization were generated from a recombinant plasmid containing a 498-bp DraII-SacI fragment of human WNT-5A cDNA (951-1448) subcloned into pBluescript II KS (Stratagene). Synthesis of [ 35 S]UTPlabeled riboprobes, in situ hybridization, autoradiography, and histological staining were carried out as described (36,37). Single-stranded sense and antisense riboprobes labeled with [␣ 35 S]UTP were generated using T3 and T7 RNA polymerase, respectively. As controls, hybridizations were also performed using sense riboprobes. Sections were counterstained with Harris hematoxylin and eosin and photographs were taken using Kodak Ektachrome ASA 64 film as described previously (36,37).

Heterogeneity of the 5Ј End within the WNT-5A cDNA Clones and Determination of the Transcription Start
Sites-We previously reported (25) a 483-bp sequence encoding the 5Ј-untranslated region of WNT-5A cDNA. We now have identified the sequence within two adjacent BamHI genomic fragments. A close examination of the sequence data detected possible donor and acceptor splice sites at positions 33 and 444, which would generate a 411-bp intervening sequence. Further analyses of the cDNA clones revealed that the 411-bp region was spliced out in three different clones and retained in the others (Fig.  1A). To determine whether the 411-bp fragment was intronic, we generated a PCR probe of 363 bp within this region and used it as a labeled probe in Northern blotting analyses of the same fibroblast RNA that was used in the cDNA library construction. The probe failed to hybridize to any transcript (not shown) under the same conditions in which we could detect several transcripts by using other portions of the WNT-5A gene (cf. Fig. 4). Therefore, it appears that the 411-bp fragment is a retained intron cloned from an incompletely spliced nuclear RNA present in the original cDNA library. Fig. 1A shows the position of the intron and the "correct" numbering of the cDNA, as well as the location of the three primers used for primer extension and RT-PCR shown below.
Next, we wished to determine the start of transcription, because putative exon 1 was only 33 bp long, as deduced from the cDNA sequence. Primer extension analysis was performed with RNA from skin fibroblasts (CRL-1262) and MG-63 osteosarcoma cells, using a antisense primer located between bp 2 and 22 of the cDNA. Two putative transcription initiation sites were detected, at 155 and 295 bp upstream of the primer (Fig.  1B). This result was confirmed by additional experiments where the primary product generated by reverse transcription was hybridized to a probe containing the 5Ј end of the cDNA. Specifically, first-strand cDNA synthesis was performed using RNA from skin fibroblasts (CRL-1262) and an antisense oligonucleotide between bp 356 and 373. The generation of WNT-5A cDNA was confirmed as control, by specific amplification of a 342-bp fragment between ϩ52 and ϩ373. The products of the RT reaction were hybridized to the 342-bp probe under stringent conditions by Northern blotting analysis. Fig. 1C shows that two bands of Ϸ533 and 740 bp were detected, confirming the presence of two transcription start sites. The sizes of the fragments shown in Fig. 1C corresponded well to those detected by primer extension (Fig. 1B). These data also confirm that the 411-bp segment is an intron inasmuch as the size of the fragments in Fig. 1C would have been 944 and 1151 bp, respectively, if the intron were present in the cDNA. Furthermore, these findings indicate that the WNT-5A messages begin Ϸ160 and 330 bp 5Ј to the site predicted by our cDNA sequence (25).
Cloning of the Human WNT-5A Gene Locus and Intron/Exon Organization-The exon/intron boundaries of the human WNT-5A gene were established by cloning and sequencing various overlapping portions of genomic DNA that hybridized strongly with PCR-generated portions of the human cDNA (25). The complete WNT-5A gene was contained in two overlapping genomic clones (W1 and W10) and spanned about 22 kb of contiguous DNA ( Fig. 2A). The gene comprised five distinct exons ranging in size between 134 and ϳ6800 bp. Both the 5Ј and 3Ј acceptor splice sites in each of the introns followed the GT-AG consensus sequence for eukaryotic genes (Fig. 2B). The 3Ј exon-intron junctions split codons in three different phases, making it unlikely that alternative splicing of the primary transcript would occur because this would alter the reading frame. This is in agreement with the high degree of conservation across species of the primary protein structure in all the members of the WNT gene family (9). Exon 1 (306 bp) encoded most of the 5Ј-untranslated region, whereas exon 2 (134 bp) encoded the remaining 39 bp of the 5Ј-untranslated region plus the first 22 amino acids comprising the signal peptide and an additional 9 amino acid residues. As in the case of most mammalian WNT genes (38), the second intron was at an identical position; that is, the two guanosine residues of the codon for the second of the two conserved Trp residues are split by a phase II intron (Fig. 2B). Exons 3 (251 bp) and 4 (293 bp) coded for the next 83 and 98 amino acids respectively, whereas exon 5 (Ϸ6800 bp) contained the final 153 codons and an extensive 3Ј-untranslated region of Ϸ6.5 kb.
The 5Ј-flanking Region of the Human WNT-5A Gene: A Complex Promoter Active in Transient Cell Transfection Assays-A BamHI genomic fragment of 4.8 kb was found to overlap with the 5Ј end of the human WNT-5A cDNA. This fragment was subcloned in pBluescript and sequenced in total with either specific internal primers or by using the T3 and T7 promoter sequences and overlapping subclones. A 2.5-kb sequence is shown in Fig. 3A that includes the first exon and Ϸ2 kb of 5Ј-flanking region. This DNA sequence contained neither conventional CAAT nor TATA boxes near the transcription start site, a finding that is in agreement with the presence of multiple transcription start sites as shown above. Of note, this region was enriched in GpC dinucleotides with an overall 60% GϩC content and peaks reaching 80% (Fig. 3B). The 5Ј end and 5Ј-flanking region of the human WNT-5A gene were located in a CpG island, a microenvironment in which the number of CpG approximately equals that of GpC (39). Indeed, the CpG/GpC ratio in this region was Ϸ0.9, thus indicating the absence of CpG suppression. This region harbored seven copies of the hexanucleotide GGGCGG, the binding site for the zinc finger transcription factor Sp1, and three GC boxes. In addition, a collection of cognate transcription factor-binding sites, summarized in the legend to Fig. 3, were present in the proximal and distal region of the promoter. In summary, the overall organization of the 5Ј-flanking region strongly suggested that this was the functional promoter of the WNT-5A gene. These GpCrich promoters are usually found in housekeeping genes, genes encoding growth factors, oncoproteins, or extracellular matrix proteins (39).
Next, we wished to determine whether this putative promoter could exhibit functional activity in transient cell transfection assays. To this end, three major CAT constructs harboring the 3.9-kb promoter region, a nested deletion comprised between Ϫ1382 and Ϫ632, and a 5Ј deletion comprised between Ϫ631 and Ϫ1 relative to the major transcription start site were used (Fig. 4). In addition, SV40-driven and promotorless CAT constructs were used, and the values were normalized on ␤-galactosidase activity determined by co-transfecting SV40-driven ␤-galactosidase plasmids (32). All of the three constructs exhibited functional promoter activity, with the highest activity residing in the proximal promoter, the region contained within the 631 bp upstream of the transcription start site (Fig. 4C). Interestingly, both the longest promoter construct and the internal deletion showed an approximate 40% decline in functional promoter activity (Fig. 4C, compare lanes 1 and 2 with  (25). The size of intron 2 was determined by PCR, whereas the sizes of introns 3 and 4 were deduced from the sizes of the genomic fragments hybridizing to each exon. The maximum length of exon 5 (6800 bp) was determined by hybridization studies of 3Ј genomic probes with RNA of human skin fibroblasts (cf. Fig. 5). Because multiple WNT-5A transcripts are present in various tissues and cells, the size of exon 4 can be variable.
lane 3). This suggests that there are negative elements present in the region upstream of the Ϫ631 construct. Nevertheless, the data indicate that the distal promoter region is not required for full transcriptional activity. This is consistent with the fact that the highest concentration of transcription factor-binding elements is within the proximal promoter region (Figs. 3A and 4A). In summary, these data demonstrate for the first time that the 5Ј-flanking region of the WNT-5A gene has features of a functional promoter that can drive the expression of a reporter gene in HeLa cells.
Characterization of the 3Ј End of the WNT-5A Gene and Evidence of Multiple Polyadenylation Sites-Because our Northern blotting studies (26) indicated that the human WNT-5A mRNA should be larger than the cDNA sequence of The major transcription start site, as detected by primer extension (cf. Fig. 1), is indicated by ϩ1. The exon 1 is in italics and underlined, whereas the GT 5Ј splicing junction is boldface. Consensus sequences for the various cognate cis-acting regulatory elements are bold underlined and labeled above the coding strand. The candidate cis-acting factors with minimal mismatching were identified by using the GCG and PCGENE package programs as reviewed before (60)  within the gene (Fig. 5). Hybridization of RNA by Northern blotting with probe A or B revealed the presence of at least four transcripts of approximately 4.4, 5.2, and 6.8 and a poorly resolved doublet of 8.6/9.2 kb, respectively. However, when probe C was used only the 6.8-and 8.6/9.2-kb transcripts were detected. In contrast, probe D did not hybridize with any WNT-5A transcript, indicating that it was 3Ј-flanking region. We therefore conclude that both polyadenylation signals were utilized and that a third may exist upstream from position 7900. In summary, we showed that the large transcript sizes of WNT-5A arose from the utilization of multiple polyadenylation sites within a large 3Ј-untranslated region in the gene. However, we cannot exclude the possibility that yet undiscovered 5Ј exons, encoding mostly untranslated regions of the gene, may contribute to the generation of the largest transcripts.
The WNT-5A Gene Maps in or near 3p14.3 between Two Chromosome Translocation Break Points-Recent studies have demonstrated that chromosomal rearrangements can relocate proto-oncogenes into transcriptionally active DNA loci, and a number of cellular oncogenes have now been mapped at or near the break point involved in tumor-specific translocation. In our previous studies, we mapped the WNT-5A locus by in situ hybridization to a relatively broad region of chromosome 3 encompassing 3p14 to 3p21. To determine more precisely the location of this gene on chromosome 3, we utilized a series of rodent-human hybrid cell lines carrying defined deletions or translocation of 3p. Specifically, human-rodent and parental DNAs were digested with restriction enzyme HindIII, electrophoresed, and blotted to filters. Filters were then hybridized to the radiolabeled full-length WNT-5A cDNA probe, which was generated by PCR amplification. A strongly hybridizing human fragment of Ϸ10 kb was present in hybrids retaining a specific region of chromosome 3 between 3p14.2 and 3p21.1 as illustrated on the left in Fig. 6 and summarized in Fig. 7. The relationship of the map location of the WNT-5A gene to other regions of interest on 3p is shown on the right in Figs. 6 and 7. The results thus independently confirm the previous mapping to chromosome 3 (25) and further refine the assignment of WNT-5A gene to 3p14.2-3p21.1, between two translocation break points.
WNT-5A Expression in Early Human Embryos: Implications for Tissue Differentiation-The expression pattern of WNT-5A mRNAs was studied by in situ hybridization in five human embryos at ages of 28 -42 days. WNT-5A transcripts showed a complex pattern of expression in the developing heart, facial processes, limb, body axis, gonad, and brain (Fig. 8). In all cases, WNT-5A expression in specific sites preceded overt tissue differentiation. In a 28-day human embryo, a low level of WNT-5A transcripts was found in part of the atrium and the ventricle of the heart (Fig. 8A). Expression in the pericardium was transient, positive hybridization being found only in a 33-day embryo (not shown). WNT-5A mRNAs were found to be highly expressed in 28-and 33-day embryos in the frontalnasal mesenchyme (Fig. 8, B and C) and in the branchial arches (Fig. 8G). Particularly high levels of expression were found in cells that were localized toward the tips of the facial prominence (Fig. 8C).
Expression of WNT-5A transcripts was also found in the mesenchyme surrounding the Wolffian and Muellerian ducts of the 42-day embryo (Fig. 8D). In the developing kidney, WNT-5A transcripts were detected in the mesenchyme sur- rounding the nephric tubules (Fig. 8, E and H) in the mesonephros of a 38-day embryo but not in a 42-day one (Fig. 8I).
WNT-5A expression was striking in embryos between 28 and 42 days in the developing central nervous system with a high level of transcripts localized to the tegmentum (floor) of the midbrain (Fig. 8F). Weaker hybridization signals were in the basal region of the caudal hindbrain and rostral spinal cord (not shown). No WNT-5A transcripts were detected in the forebrain and rostral hindbrain, but strong expression was seen in the hypophysis (Rathke's pouch, Fig. 8F).
High levels of WNT-5A expression were found in cells populating the peripheral and lateral but not the core of the branchial arches (Fig. 8G). Particularly high expression was also present in the mesenchyme bordering the pharyngeal clefts and pouches (Fig. 8G). Along parts of the mid-body axis, expression was also detected within the lateral mesoderm (not shown). Expression of WNT-5A was also seen in the developing gonad. In the indifferent gonad of 38-and 42-day embryos, strong WNT-5A expression was seen in the celomic epithelium and in the precursors of supporting cells destined for follicles or seminiferous tubules (Fig. 8, H-J). Primordial germ cells did not express WNT-5A (Fig. 8J).
WNT-5A mRNAs also showed preferential localization in the developing limb. In a 33-day embryo, there was a gradient of WNT-5A expression along the proximal-distal axis of the hind limb bud with highest hybridization signals in the progress zone toward the distal tip of the hind limb bud (Fig. 8K). Expression level of WNT-5A in the developing hind limb increased at 38 and 42 days because most of WNT-5A transcripts were detected in the distal mesenchyme (Fig. 8L) surrounding the condensing prechondrogenic cells and also at the tips of limb digits (Fig. 8M). No expression was seen in differentiating chondrocytes (Fig. 8M) or in the muscle masses. Strong expression was also found in the preaxial and postaxial ectoderm at the rostral and caudal borders of the developing hind limb at the level of joint formation in the 38-day embryo (not shown). DISCUSSION During the past decade there has been a clear realization that proto-oncogenes, the evolutionarily conserved progenitors of mutant genes that contribute to neoplastic growth, are directly implicated in the control of normal development (40). This view has gained support from a number of experimental observations, including the fact that some proto-oncogenes code  Fig. 6. The WNT-5A gene is present in the small region of 3p between the t(3;5)(p21.1;q35) and t(3;8)(p14.2;q24) chromosome translocation breaks as indicated in this sketch. Thus, the WNT-5A gene falls in or near 3p14.3, in agreement with the previous in situ hybridization data (25). Right panel, the position of WNT-5A between the two chromosome breaks is shown relative to some genetic distances (on the far right) in centimorgans (cM) and relative to putative tumor suppressor loci bracketed at 3p21.2-3p21.3 and 3p12-3p13.
FIG. 8. Expression of the WNT-5A gene in early human embryos from 28 to 42 days of gestation. In situ hybridizations were performed using sense and antisense riboprobes specific for WNT-5A (see "Experimental Procedures"). No specific hybridization was found using sense riboprobes. A-G, I, and K-M are dark field pictures taken under dark field illumination where silver grains appear white; the background color is due to the counterstain. H and J are double exposures with a combination of bright field with a blue filter and dark field illumination under a red filter to produce superimposed images where silver grains appear pinkish red. A, weak expression is found in the differentiating trabeculae of the ventricle (vt) and parts of the epimyocardium of the atrium (open arrow) of a 28-day embryo but absent in the pericardium (arrow). B, strong expression in the frontonasal mesenchyme (fn) and weaker expression in the heart (h) are detected in a 33-day embryo. C, hybridization was strong in the frontonasal mesenchyme especially in the region of the tip of the facial prominence of a 38-day embryo (arrows). D, expression is found in the mesenchyme (me) surrounding the Muellerian (mt) and Wolffian (wt) ducts in a 42-day embryo. No hybridization signal was detected in the Muellerian and Wolffian ducts. E, strong expression is found in the mesenchyme (arrow) surrounding the nephric tubules (nt) of the developing kidney of the 38-day embryo. F, restricted expression is localized to the tegmentum of the midbrain (mb) but is absent in the forebrain (fb) and rhombomeres (arrows) of the 28-day embryo. Expression is also seen in the hypophysis (hy). G, transcripts within the branchial arch are concentrated within the peripheral and lateral regions but are absent in the core (co) of a 33-day embryo. There is strong hybridization in the subectodermal mesenchyme in the cleft between the arches (small arrows). The crown-rump axis of the embryo is indicated by the large arrow with the arrowhead pointing toward the crown. H, in the urinogenital complex of a 38-day embryo, expression is found in the mesenchyme (m) but is absent in the tubules (t) and glomerulus (g) of the developing kidney. In the indifferent gonad (gn), strong signal is detected in the celomic epithelium (arrows) and somatic cells. for growth factors, that mutant forms can arrest, distort, or promote differentiation, and that they are temporally and spatially sequestered during embryogenesis (9). In this study, we investigated the genomic organization, promoter activity, and chromosomal localization of the WNT-5A gene, which codes for a cysteine-rich growth factor involved in cell-cell signaling during embryonic development and perhaps tumor growth. The genomic organization of the human WNT-5A gene was notable in several aspects. The gene was found to consist of 5 exons spanning Ϸ25 kb of continuous DNA with the last exon extending for Ϸ7 kb and harboring multiple polyadenylation sites. A genomic 5Ј-flanking region of 4.8 kb was localized that overlapped with the corresponding 5Ј-terminal sequence of the cDNA. We discovered a new leader exon interrupted by a 411-bp intron that was retained in our original cDNA cloning (25). Specifically, the sequence of this first exon was found to be markedly different when compared with the corresponding region of the mouse cDNA (13). This divergence was surprising inasmuch as the remaining coding sequence of the mouse cDNA was Ͼ90% identical to the human cDNA. In addition, our unpublished data on the genomic organization of the murine WNT-5A gene have indicated that the exonic organization of the translated region is also closely conserved. These findings suggest that the 5Ј sequences including the control regions of the murine and human WNT-5A genes have diverged during evolution. However, it is interesting that the initial 39 bp of exon 2, which is also untranslated, were highly conserved. It should be pointed out that because exon 1 comprises most of the untranslated region of the WNT-5A mRNA, the overall characteristics of the WNT-5A protein sequence would not be affected by these differences in DNA sequences.
We have identified a 5Ј-flanking region in the human WNT-5A locus that harbored several features that have been correlated with transcriptional control regions and are typically observed in genes that encode oncoproteins, growth factors, transcription factors, and housekeeping genes (39). This flanking sequence was located in a GpC-rich island and harbored numerous cis-acting elements, including Sp1, AP1, and AP2 binding motifs, and acted as a functional promoter when tested in transient cell transfection assays. Of note, the strongest promoter activity resided in the proximal promoter, the region contained within the 631 bp upstream of the major transcription start site. The fact that both an internal deletion and the long 3.9-kb promoter showed a 40% decline in functional activity suggests that the distal promoter region is not required for full transcriptional activity. This is also supported by the fact that the highest concentration of elements binding to transcription factors is within the proximal promoter region. Collectively, our findings demonstrate for the first time that the 5Ј-flanking region of the WNT-5A gene has features of a functional promoter that can drive the expression of a reporter gene in HeLa cells.
Using a combination of primer extension and RT-PCR, we detected two major transcription start sites for the human WNT-5A gene. Of note, the murine WNT-1 gene is also transcribed from two major transcription start sites (41). The WNT-1 sequence near the more 3Ј start site is reminiscent of a classical promoter with a well defined TATA box. In contrast, the sequence around the more 5Ј start site does not contain a canonical TATA box but is preceded by a GpC-rich stretch of DNA (41) similar to the human WNT-5A. Of note, a WNT-1inducing factor-1 has been recently shown to represent a novel transcription factor that binds a G/C box motif in WNT-1 and regulates the expression of WNT-1 during neuroectodermal differentiation (42). Interestingly, the human WNT-1 gene has an entirely homologous WNT-1-inducing factor-1 binding site at the same relative position (43). Although the human WNT-5A promoter does not possess an identical sequence, it harbors in the proximal promoter several stretches of DNA that are homologous to the WNT-1-inducing factor-1-binding site.
Unusual results were also found at the 3Ј-untranslated region of the WNT-5A gene. The terminal 459 bp coding for 152 amino acids plus the stop codon and the following 6.3 kb were found to comprise a single exon. Two polyadenylation signals were found at Ϸ4.4 and 7.9 kb, accounting for at least some of the observed transcript sizes. However, the appearance of transcripts detected by Northern hybridization that were larger than 8 kb could not be satisfactorily explained by our model of the gene structure and may include alternatively spliced exons in the 5Ј-untranslated region.
The human WNT-5A gene falls between two constitutional chromosome breaks in 3p21.1 and 3p14.2, respectively; thus the WNT-5A gene is perhaps closely telomeric to the 3p14.2 constitutional t(3;8) chromosomal translocation associated with familial renal cell carcinoma (44). Because the murine WNT-5A locus maps to the centromeric region of mouse chromosome 14 (9,45), as do the murine protein tyrosine phosphatase-␥ Ptprg gene (46,47) and the calcium channel ␣1 subunit Cchl1a2 gene (48,49), these loci thus form a small linkage group in mouse and human, whereas most 3p21 markers map to murine chromosome 9. As discussed before (47), abnormalities on murine chromosomal 14 in metastatic murine melanomas (50), suggest that the integrity of the Ptprg locus should be explored in these tumors. Similarly, it would be very interesting to investigate the integrity of the murine WNT-5A locus in such tumors. Interestingly, human chromosome 3 is composed of about 210 million bp, representing about 7% of the human genome (51). Thus, it might be anticipated that the genes responsible for perhaps 7% of human genetic disorders would be located here. To date, however, only a few disease loci have been mapped to the region of chromosome 3 where WNT-5A maps. It is of note that this region of chromosome 3 is often the target of rearrangements and translocation in several malignant neoplasms, most notably in small cell lung carcinomas (52) and, as pointed above, in renal cell carcinomas (44,53).
The expression of WNT-5A in early human embryos was remarkably similar to that reported in the developing mouse (13,54) and chicken (55) with the primary regions of expression of WNT-5A in the midbrain and the developing limb. The murine WNT-5A is the only known member of the WNT gene family that is detected in the developing limb (54) with a pattern of distribution following a proximal-distal gradient along the limb mesenchyme and the ventral half of the limb ectoderm (13). In these examples expression of WNT-5A appears in sites prior to overt tissue differentiation and at times at which important events are occurring in pattern formation. The similarities in expression patterns of the WNT-5A gene in human, mouse, and chicken embryos point to a role for WNT-5A in the patterning and differentiation of the midbrain and limb and reinforces the applicability and importance of using animal models to study the function of this important class of signaling molecules. Of particular interest is the hitherto unreported expression of WNT-5A expression in the developing kidneys and gonads. Commitment of the indifferent gonad to male or female differentiation is likely to be influenced by cells in the supporting cell lineage (56). The SRY gene is involved in initiating that process, but other factors must be involved in the differentiation of germ cells and the gonad. It has also recently been shown that the genital ridge environment is important for providing signals for the initiation of X chromosome reactivation in primordial germ cells (57). Local-ization of WNT-5A transcripts in the somatic cells of the developing gonad suggests that WNT-5A protein may have a role as a signaling molecule in the differentiation of germ cells and/or in X chromosome reactivation in the developing ovary. The diverse but specific pattern of expression of WNT-5A point to pleiotropy in functions. Therefore the possible consequences of mutations in WNT-5A, apart from anomalies in development of the midbrain, craniofacies, and limb, may include abnormalities in sex determination such as sex reversal effects or in germ cell differentiation. Of note is the fact that the WNT-1-inducing factor-1 binding activity is present in nuclear extracts from adult mouse testes, thus suggesting that similar transcription factor(s) may regulate the expression of WNT-5A gene.
We have recently completed a survey of over 100 tissues and 12 cell lines and discovered a widespread distribution of WNT-5A mRNA in human tissues (26), a finding that is in agreement with the in situ hybridization data presented in this study. In addition we found a severalfold increase in the levels of WNT-5A transcripts in tissue extracts of lung and breast carcinomas, as compared with their normal counterpart (26). Our findings, thus, suggest that abnormal expression of this growth factor is involved in the genesis and development of human cancer. This hypothesis is supported by a recent investigation that has shown that both benign and malignant breast tumors exhibit high levels of WNT-5A transcripts (58). In human mammary epithelial cells, the WNT-5A gene is up-regulated by confluence and down-regulated by hepatocyte growth factor (HGF) during the HGF-induced branching in collagen gels (59). These findings suggest that the mammalian WNT-5A is involved in mammary epithelial cell motility and are in agreement with the finding in X. laevis where WNT-5A is involved in embryonal cell migration (19) .
In summary, our studies have characterized the human WNT-5A gene and shown that the WNT-5A gene has a specific and restricted pattern of expression during early human organogenesis with a distribution that is not limited to the central nervous system and limb development. Its presence in the indifferent gonads and kidneys suggests a novel potential role for the WNT-5A in renal and germ cell differentiation. The elucidation of the WNT-5A gene regulatory control regions, together with the mapping data and the expression pattern during development, will make now possible a better understanding of the function and regulation of this gene in normal and pathological states.