Advertisement

The Gene Encoding Disabled-1 (DAB1), the Intracellular Adaptor of the Reelin Pathway, Reveals Unusual Complexity in Human and Mouse*

  • Isabelle Bar
    Correspondence
    Chargée de Recherche of the Fonds National de la Recherche Scientifique. To whom correspondence should be addressed. Tel.: 32-81-724-274; Fax: 32-81-724-280;
    Affiliations
    Neurobiology Unit, University of Namur Medical School, 61, rue de Bruxelles, B5000 Namur, Belgium and the
    Search for articles by this author
  • Fadel Tissir
    Footnotes
    Affiliations
    Developmental Genetics Unit, University of Louvain Medical School, Avenue E. Mounier, B1200 Brussels, Belgium
    Search for articles by this author
  • Catherine Lambert de Rouvroit
    Affiliations
    Neurobiology Unit, University of Namur Medical School, 61, rue de Bruxelles, B5000 Namur, Belgium and the
    Search for articles by this author
  • Olivier De Backer
    Affiliations
    Neurobiology Unit, University of Namur Medical School, 61, rue de Bruxelles, B5000 Namur, Belgium and the
    Search for articles by this author
  • André M. Goffinet
    Affiliations
    Developmental Genetics Unit, University of Louvain Medical School, Avenue E. Mounier, B1200 Brussels, Belgium
    Search for articles by this author
  • Author Footnotes
    * This work was supported in part by Grant 3.4533.95 from the Fonds de la Recherche Scientifique et Médicale, Grants 186 and 248 from the Actions de Recherches Concertées, and by the Fondation Médicale Reine Elisabeth (Belgium).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.The nucleotide sequence(s) reported in this paper has been submitted to the GenBank™/EBI Data Bank with accession number(s) AF525763AY131331AY172736AY174214AY174232
    ‖ Postdoctoral Fellow supported by European Union Program CONCORDE Grant QLG3-CT-2000-00158.
Open AccessPublished:November 21, 2002DOI:https://doi.org/10.1074/jbc.M207178200
      The Disabled-1 (Dab1) gene encodes a key regulator of Reelin signaling. Reelin is a large glycoprotein secreted by neurons of the developing brain, particularly Cajal-Retzius cells. The DAB1 protein docks to the intracellular part of the Reelin very low density lipoprotein receptor and apoE receptor type 2 and becomes tyrosine-phosphorylated following binding of Reelin to cortical neurons. In mice, mutations of Dab1 andReelin generate identical phenotypes. In humans,Reelin mutations are associated with brain malformations and mental retardation; mutations in DAB1 have not been identified. Here, we define the organization of Dab1, which is similar in human and mouse. The Dab1 gene spreads over 1100 kb of genomic DNA and is composed of 14 exons encoding the major protein form, some alternative internal exons, and multiple 5′-exons. Alternative polyadenylation and splicing events generate DAB1 isoforms. Several 5′-untranslated regions (UTRs) correspond to different promoters. Two 5′-UTRs (1A and 1B) are predominantly used in the developing brain. 5′-UTR 1B is composed of 10 small exons spread over 800 kb. With a genomic length of 1.1 Mbp for a coding region of 5.5 kb, Dab1 provides a rare example of genomic complexity, which will impede the identification of human mutations.
      Neuronal migration is a complex process that is affected in a variety of human disorders such as periventricular heterotopias and different types of lissencephalies (
      • Gupta A.
      • Tsai L.H.
      • Wynshaw-Boris A.
      ,
      • Monuki E.S.
      • Walsh C.A.
      ,
      • Lambert de Rouvroit C.
      • Goffinet A.M.
      ). The Disabled-1(Dab1) gene belongs to the Reelin signaling pathway that plays a key role during brain development in mouse and human (
      • Howell B.W.
      • Herrick T.M.
      • Hildebrand J.D.
      • Zhang Y.
      • Cooper J.A.
      ,
      • Rice D.S.
      • Curran T.
      ,
      • Feng Y.
      • Walsh C.A.
      ). Inactivation of Dab1, either by homologous recombination (
      • Howell B.W.
      • Hawkes R.
      • Soriano P.
      • Cooper J.A.
      ) or by spontaneous mutations in scrambleror yotari mutant mice (
      • Sheldon M.
      • Rice D.S.
      • D'Arcangelo G.
      • Yoneshima H.
      • Nakajima K.
      • Mikoshiba K.
      • Howell B.W.
      • Cooper J.A.
      • Goldowitz D.
      • Curran T.
      ,
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ), generates a phenotype similar to that of Reelin-deficient mice. This phenotype is characterized by a poor organization of architectonic patterns at the end of radial neuronal migration (reviewed in Ref.
      • Lambert de Rouvroit C.
      • Goffinet A.M.
      ). The neurons that are the most affected include those of the cortical plate in the cortex and hippocampus, Purkinje cells, and inferior olivary neurons. In human, mutations in Reelin result in a specific lissencephaly with mental retardation and severe abnormalities of the cerebellum, hippocampus, and brain stem (Norman-Roberts type, OMIM257320) (
      • Hong S.E.
      • Shugart Y.Y.
      • Huang D.T.
      • Shahwan S.A.
      • Grant P.E.
      • Hourihane J.O.
      • Martin N.D.
      • Walsh C.A.
      ), a phenotype that shows similarity to its mouse counterpart. Cognitive development is delayed, with little or no language acquisition and no ability to sit or stand unsupported. Thus far, no human disease associated with mutations in DAB1 or other genes in the Reelin pathway has been identified.
      Reelin is an extracellular protein secreted by some neurons such as Cajal-Retzius cells in the marginal zone of the embryonic cerebral cortex and hippocampus, external granule cells in the cerebellum, olfactory mitral cells, and ganglion and amacrine cells in the mouse retina (
      • Lambert de Rouvroit C.
      • Goffinet A.M.
      ,
      • Rice D.S.
      • Curran T.
      ) and in the spinal cord (
      • Carroll P.
      • Gayet O.
      • Feuillet C.
      • Kallenbach S.
      • de Bovis B.
      • Dudley K.
      • Alonso S.
      ,
      • Yip J.W.
      • Yip Y.P.
      • Nakajima K.
      • Capriotti C.
      ). The response of target neurons to Reelin requires the expression of at least one of two surface receptors that belong to the lipoprotein receptor family,viz. the very low density lipoprotein receptor and apoE receptor type 2, as well as the presence of the intracellular adaptor DAB1. The DAB1 protein contains a 180-amino acid N-terminal protein interaction/phosphotyrosine-binding (PTB)
      The abbreviations used are: PTB, phosphotyrosine-binding; RACE, rapid amplification of cDNA ends; P, postnatal day; E, embryonic day; PAC, P1 artificial chromosome; YAC, yeast artificial chromosome; contig, group of overlapping clones; UTR, untranslated region; DMEM, Dulbecco's modified Eagle's medium; RA, all-trans-retinoic acid; RT, reverse transcription; IAP, intracisternal A particle; ORF, open reading frame; uORF, upstream open reading frame; EST, expressed sequence tag
      1The abbreviations used are: PTB, phosphotyrosine-binding; RACE, rapid amplification of cDNA ends; P, postnatal day; E, embryonic day; PAC, P1 artificial chromosome; YAC, yeast artificial chromosome; contig, group of overlapping clones; UTR, untranslated region; DMEM, Dulbecco's modified Eagle's medium; RA, all-trans-retinoic acid; RT, reverse transcription; IAP, intracisternal A particle; ORF, open reading frame; uORF, upstream open reading frame; EST, expressed sequence tag
      domain that docks to the short cytoplasmic tail of the very low density lipoprotein receptor or apoE receptor type 2 at the level of NPXY motifs, with a preference for unphosphorylated motifs (
      • D'Arcangelo G.
      • Homayouni R.
      • Keshvara L.
      • Rice D.S.
      • Sheldon M.
      • Curran T.
      ,
      • Trommsdorff M.
      • Gotthardt M.
      • Hiesberger T.
      • Shelton J.
      • Stockinger W.
      • Nimpf J.
      • Hammer R.E.
      • Richardson J.A.
      • Herz J.
      ,
      • Hiesberger T.
      • Trommsdorff M.
      • Howell B.W.
      • Goffinet A.
      • Mumby M.C.
      • Cooper J.A.
      • Herz J.
      ,
      • Gotthardt M.
      • Trommsdorff M.
      • Nevitt M.F.
      • Shelton J.
      • Richardson J.A.
      • Stockinger W.
      • Nimpf J.
      • Herz J.
      ,
      • Howell B.W.
      • Lanier L.M.
      • Frank R.
      • Gertler F.B.
      • Cooper J.A.
      ). Potential tyrosine phosphorylation sites and a 310-amino acid C-terminal region of unknown function follow the PTB domain. The binding of Reelin to the extracellular part of both receptors induces phosphorylation of tyrosine residues of DAB1, particularly Tyr198 and Tyr220 (
      • Howell B.W.
      • Herrick T.M.
      • Cooper J.A.
      ,
      • Keshvara L.
      • Benhayon D.
      • Magdaleno S.
      • Curran T.
      ). Mice expressing a mutant form of DAB1 in which all the potential tyrosine phosphorylation sites are mutated have a phenotype similar to reeler mice (
      • Howell B.W.
      • Herrick T.M.
      • Hildebrand J.D.
      • Zhang Y.
      • Cooper J.A.
      ), and mice expressing a truncated DAB1 protein missing the C-terminal part have an almost normal phenotype (
      • Herrick T.M.
      • Cooper J.A.
      ). This shows that the PTB domain and tyrosine phosphorylation are both necessary and sufficient to fulfill most of the DAB1 functions. Cytoplasmic tyrosine kinases of the Src family are able to phosphorylate DAB1 in vitro, but the kinase(s) involved in DAB1 phosphorylation in vivo remain to be identified (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ). Similarly, the other downstream effectors of the Reelin signal are not known.
      Various isoforms of the mouse DAB1 protein have been described. The main form contains an open reading frame of 555 amino acids encoding a 80-kDa protein, the predominant form expressed in the brain. Another form, 555*, contains an additional exon inserted in-frame between codons 241 and 242. Form 217 results from alternative polyadenylation, whereas isoform 271 is similar to form 555, except that an additional exon of 270 bp containing a stop codon is inserted between codons 241 and 242. Upon Northern blotting, three transcripts of 5.5, 4.0, and 1.8 kb have been detected with a probe covering the PTB domain, and protein isoforms of 36, 45, 60, 80, and 120 kDa have been observed on Western blots (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ).
      In this work, we defined the genomic organization of the human and mouse Dab1 genes. The structure is highly complex and similar in both species. The gene extends over >1 Mbp of genomic DNA due to the presence of large introns and the wide dispersion of several alternative transcription initiation sites. The presence of several alternative promoters and alternatively spliced forms points to a fine regulation of Dab1 expression and further emphasizes the key position of this gene as a switch in the Reelin signaling pathway. The complexity of the gene may explain why no human disease associated withDAB1 mutations could be identified thus far.

      RESULTS

      The published physical map of the mouse Dab1gene contains inconsistencies (
      • Sheldon M.
      • Rice D.S.
      • D'Arcangelo G.
      • Yoneshima H.
      • Nakajima K.
      • Mikoshiba K.
      • Howell B.W.
      • Cooper J.A.
      • Goldowitz D.
      • Curran T.
      ,
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ). We therefore reconsidered the physical map and genomic organizations of the human and mouseDab1 genes as summarized in Fig.1 and TableII.
      Table IINucleotide sequences of exon-intron boundaries of the human and mouse Dab1 genes
      Figure thumbnail fx2
      The upper part concerns the 5′-UTRs for the mouse (Mm) and human (Hs) genes. The methods of validation were RACE, RT-PCR, and/or GenBank™/EBI Data Bank entries as indicated. The ATG column indicates the number of ATG codons present in each 5′-UTR exon. The lower part indicates the exon-intron organization of the conserved part, with human and mouse in the top and bottom lines, respectively. RLM, RNA ligase-mediated.

      Physical Maps

      The mouse Dab1 gene maps to chromosome 4 at 52.7 centimorgans, whereas human DAB1 maps to chromosome 1p32-p31 (
      • Lambert de Rouvroit C.
      • Goffinet A.M.
      ). In both species, the 5′-end is located on the centromeric side. The gene is flanked by the mouseAK008020 locus (identical to human XM_055482 or LOC115209) on the centromeric side and by the complement factor 8B (C8B) gene on the telomeric side. As shown in Fig. 1 A, Dab1is located between microsatellites D4Mit118 andD4Mit75, but is at least 3 Mbp away from markerD4Mit176, confirming the data published by Ware et al. (
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ). D4Mit331 is located between Dab1exons 1B1 and 1B2 (described below), 565 kb upstream of theDab1 ATG codon and 700 kb centromeric to D4Mit29, in agreement with the reported genetic distance of 0.6 centimorgans between D4Mit29 and D4Mit331 (
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ).D4Mit29 is located in intron 4, 1.5 kb from exon 4 ofDab1 (Fig. 1 B). D4Mit75 maps 440 kb distal to the ATG codon and 300 kb telomeric to D4Mit29.
      In scrambler mutant mice, a portion of an intracisternal A particle (IAP) sequence is inserted in the antisense orientation in theDab1 mRNA by aberrant splicing. The mutation results in production of an enlarged transcript of ∼7 kb, with the introduction of multiple stop codons. The defect results from the use of a cryptic splice acceptor site in intron 4 coupled with a cryptic donor site in the IAP element. In the scrambler mRNA, 28 bases unrelated to Dab1 or to the IAP are inserted betweenDab1 exon 4 and the IAP sequence. BLAST alignment against the mouse genome localized these 28 bases in intron 4, 11 kb distal to exon 4. No IAP sequence is present in this region in the C57BL/6 DNA. Although we cannot exclude that the IAP element was present in the DC/le strain in which the scrambler mutation arose, it appears more likely that an IAP insertion caused the mutation. Inyotari mutant mice, 357 nucleotides corresponding to exons 5–8 are missing from the Dab1 mRNA, and the open reading frame is maintained. At the genomic level, this deletion in the mRNA is due to the insertion of a 962-bp L1 element. This insertion starts at the junction of exon 5-intron 5 and ends in the middle of exon 8 (
      • Kojima T.
      • Nakajima K.
      • Mikoshiba K.
      ).

      Genomic Organization

      The mouse genomic organization was derived from direct sequencing on PACs and sequencing of PCR products derived from YACs and genomic DNA. Data from the NCBI mouse genome sequence (accession number NW_000211.1) confirmed the structure of the gene. The human genomic organization was defined by direct sequencing on PAC clones and confirmed by the NCBI human genome sequence (accession number NT_029223.8). As explained below, several alternative first exons named A–F were identified in both the mouse and human genes and are dispersed over large genomic regions. Genomic YAC clones covering the mouse Dab1 region were described previously (
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ) and are shown in Fig. 1 B. YAC 37G4 (500–800 kb) contains exons 2–15, and YAC 175A2 (1220–1500 kb) extends from the complex 5′-UTR 1B to exon 9. Mouse PACs RPCI21-97L11 and RPCI21-31E11 contain upstream exons 1A–1D and 1B8. The sequence-tagged site content of PACS and YACs was not analyzed in detail. However, the fact that the mouse and human genomic organization is similar suggests that the clones were not rearranged.
      The entire human DAB1 ORF (Fig. 1 B) is contained in four PAC clones covering ∼300 kb of genomic DNA. Clone RPCI6-102O10 (132 kb; accession number AL390243) contains 120 kb of intron 1, exon 2, and 15 kb of intron 2; clone RPCI6-65F20 (107 kb; accession number AL138779) contains exons 3–6; clone RPCI6-239D12 (175 kb; accession number AL161740) contains exons 10–15 and the C8B gene; and clone RPCI6-225E22 (not sequenced) contains exon 5 up to at least exon 15. The human PACs have been characterized by the Sanger Center using fluorescent in situhybridization and sequence determination and are not chimeric.
      The Dab1 coding regions (from exon 2 containing the ATG codon to exon 14 containing the stop codon) are spread over 254 kb of genomic DNA for the mouse gene and over 294 kb for the human gene. The size of the major DAB1 protein is 555 amino acids, which corresponds to a coding capacity of 0.2% compared with a mean genomic coding density of ∼10%. The organization of the mouse and human genes is conserved, and all exon-intron splice junctions conform to the GT/AG rule (Fig.1 C and Table II). With the exception of exons 12 and 15 (549 and >3300 bp, respectively), exons are relatively small, ranging in size from 39 to 140 bp. Introns in the ORF region range in size from 89 bp to 146 kb. The PTB domain is encoded in exons 3–6. Important tyrosine residues are encoded in exons 6 (Tyr185), 7 (Tyr198), 8 (Tyr200 and Tyr220), and 9 (Tyr232) (
      • Howell B.W.
      • Herrick T.M.
      • Hildebrand J.D.
      • Zhang Y.
      • Cooper J.A.
      ,
      • Keshvara L.
      • Benhayon D.
      • Magdaleno S.
      • Curran T.
      ).
      The 3′-end of the mouse Dab1 mRNA was determined using 3′-RACE. Using a primer in exon 15, we found a Dab13′-untranslated segment that extends 1252 bp downstream from the stop codon. This sequence contains several putative polyadenylation signals and aligns with several ESTs. Another set of ESTs align with genomic sequences 1 kb farther downstream (Fig.2 A). RT-PCR with primer 34 defined in this downstream EST set and other primers in exons 14 and 15 showed that the 3′-UTR extends at least until another polyadenylation signal located 3325 bp from the stop codon. To verify that theDab1 3′-UTR is >3 kb long, we performed Northern blot analysis of mouse brain mRNA using a 2-kb probe corresponding to this 3′-sequence (Fig. 2 B). This probe revealed a single band of 5.5 kb, whereas probes corresponding to theDab1 PTB coding region revealed a band of similar size plus two additional bands of lower size (see Fig. 4). In human, four EST clones (accession numbers AA541650, AI799728, R52905, and R67274) contain a polyadenylation signal (followed by a poly(A) tail) localized 3344 bp downstream from the DAB1 stop codon.
      Figure thumbnail gr2
      Figure 2Mouse Dab1 3′-UTR. A, shown is a schematic representation of the mouseDab1 3′-UTR region. Exon 14 contains the stop codon and part of the 3′-UTR. Exon 15 is 3.3 kb in length and contains the rest of the 3′-UTR. The end of the published Dab1 cDNA and fragment obtained by RACE are represented. Potential AATAAA polyadenylation signals are also indicated. A cluster of ESTs was found 1 kb farther downstream. For Northern blot analysis, a 2-kb probe was amplified with primers 34 and 35. B, 2 μg of P0 mouse brain poly(A) RNA was hybridized to randomly primed 32P-labeled probe as illustrated in A. This probe revealed a single band of 5.5 kb, whereas probes corresponding to the Dab1 PTB coding region revealed a band of similar size plus two additional bands of 4 and 1.3 kb (see Fig. ).
      Figure thumbnail gr4
      Figure 4Expression of alternative 5′-UTR 1A.Poly(A)+ RNA (2 μg) from P0 mouse brain was analyzed by Northern blotting using 32P-labeled probes. Transcripts of three different sizes (∼5.5, 4, and 1.3 kb, indicated byarrowheads) were detected with a probe for the PTB domain (left). Three bands of similar sizes plus a band of 1.8 kb were detected with a probe for exon 1A (right).

      Alternative First Exons and Organization of the 5′-Region

      Comparison of the different Dab1 sequences revealed extensive variation in the 5′-regions. In mouse, theDab1 cDNA sequence initially described (accession number Y08379) (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ) contains 263 nucleotides of 5′-UTR encoded by exon 1A and 136 bp of 5′-UTR encoded by exon 2, which contains the ATG codon. The macaque AB05528 and human AK095513 DAB1 cDNAs contain another 5′-UTR (1B) of 497 and 532 bp, respectively. The human AF263547 DAB1 cDNA contains a different 5′-UTR region of 629 bp, and the human XM_010707 cDNA sequence contains yet another different first exon of 508 bp.
      As these data suggest the presence of alternative first exons, we performed 5′-RACE and RNA ligase-mediated RACE on embryonic human and mouse brain RNAs. Using embryonic mouse brain RNA, four different products named 1A–1D were obtained (Fig. 1 C and Table II). Mouse fragment 1A corresponds to Dab1 exon 1 in sequence Y083379 (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ) and is found in 10 mouse and two rat ESTs. RACE product 1B is similar to human DAB1 cDNA sequences AF263547, AB05528, and AK095513 and is present in three mouse ESTs and one human EST. In mouse, RT-PCR and RACE experiments with primers specific for this product 1B revealed that it does not correspond to a single exon, but is composed of combinations of 10 different exons, 1B1–1B10 (Fig. 1 C). Exon 1C does not correspond to any published sequence, but could be amplified by RT-PCR, whereas exon 1D is novel and is present in one mouse and one rat EST sequence; it is conserved in human and mouse and was amplified from human brain RNA by RT-PCR. Attempts to map the transcription initiation sites by primer extension were unsuccessful, possibly because of high GC content and secondary structures of the alternative first exons. Using repeated RNA ligase-mediated RACE reactions on poly(A) RNA from E17 mouse brain, we were unable to extend the Dab1 UTR sequences farther and therefore considered them close to full-length. All the sequences obtained by RNA ligase-mediated RACE were also obtained using classical RACE reactions, and all RACE products were shown to be connected toDab1 by RT-PCR.
      RACE reactions on human brain RNA yielded four different products. One is similar to mouse exon 1A, with no match in human EST data bases. A second RACE product is similar to the highly complex mouse fragment 1B and is present in human cDNA clones AF263547, AB055282, andAK095513. RACE and RT-PCR experiments revealed that 5′-UTR 1B is composed of combinations of a least seven different exons. We were not able to clone the 5′-end of the reconstituted RNA sequence XM_010707 using RACE or to connect it to DAB1 exon 2 using RT-PCR on embryonic human brain RNA. Exon E is not present in EST data bases, but can be connected to DAB1 exon 2 by RT-PCR. Exon F is similar to bases 608–701 of sequence XM_060465, the rest of which is unrelated to DAB1.
      Among the novel 5′-UTRs, three are conserved between human and mouse (Fig. 1 C), viz. exons 1A (90% identity), 1B (mouse 1B1/human 1B1, 65% identity; mouse 1B2/human 1B4, 67% identity; and mouse 1B4/human 1B7, 79% identity), and 1D (55% identity). The sequences are present in EST data bases and have been isolated as cDNA by others, thus confirming their expression.

      Genomic Organization of the 5′-Region

      The Dab15′-UTR spreads over 850 kb in mouse and 961 kb in human (Fig. 1,B and C). Whereas exons 1A, 1C, and 1D in mouse and exons 1A and 1D–1F in human are clustered in a 1.5-kb fragment, the complex 5′-UTR 1B has a highly unusual structure. It is composed of 10 exons (1B1–1B10) in mouse and seven exons (1B1–1B7) in human, with sizes ranging from 63 to >675 nucleotides, separated by introns with sizes between a few hundred nucleotides and >300 kb. The sequence of UTR 1B is dispersed over >800 kb of genomic DNA, which is consistent with physical mapping data. YAC 175A2, which is 1500 kb in length, contains the region between exon 1B8 and coding exon 9, but does not contain exons 1B1 and 1B2 (Fig. 1 C). Exons 1B1–1B10 are flanked with consensus splice sites and obey the GT/AG rule. Both in mouse and man, the alternative exons that compose UTR 1B contain numerous ATG codons and upstream ORFs (uORFs) (Table II). For example, exon 1B1 contains one uORF of 87 codons in human and of 101 codons in mouse, and the first 55 encoded amino acids are highly conserved. uORFs are common in certain genes that are involved in the control of cellular growth and differentiation. This may have implications for the control of DAB1 mRNA translation, as many examples have been described in which ORFs present in the 5′-UTR influence expression levels (
      • van der Velden A.W.
      • Thomas A.A.
      ,
      • Morris D.R.
      • Geballe A.P.
      ).

      Expression of Alternative First Exons

      To assess whether the different 5′-exons have different expression patterns, PCRs were carried out on mouse brain cDNA at different stages from E11 to E18 and at postnatal stages from P0 to adult. We used forward primers in the alternative first exons and reverse primers in exons 2 and 5 (Fig.3 A). As shown in Fig.3 B, exons 1A and 1D were expressed in all stages tested. Amplification of exon 1C was weak, indicating low level expression in the tissues examined (data not shown). The complex UTR 1B was barely detectable at E11 and E12, whereas two main bands were amplified in RNA isolated from brain at E15 and later, including adult (Fig.3 B). We tested the expression of the alternative first exons in P19 cells, which differentiate into neurons in the presence of RA (Fig. 3 C). Exons 1A and 1D were detected in undifferentiated and differentiated P19 cells. In contrast, the expression of UTR 1B was complex. Multiple bands were amplified in undifferentiated cells and up to 4 days after RA induction. When neural induction was complete, only two bands were visible. This developmental regulation was confirmedin vivo. As shown in Fig. 3 (C and D), a pattern of multiple bands amplified from E11 mouse RNA becomes restricted to two main amplicons at P0 and adult. Sequencing of the two main amplicons showed that they are formed of fragments 1B1, 1B2, and 1B4, with alternative inclusion of fragment 1B8. Interestingly, the sequences of fragments 1B1, 1B2, and 1B4 are conserved in mouse, man, and macaque. Sequencing of the other amplicons confirmed that they correspond to different combinations of exons 1B1–1B10, as shown in Fig. 3 E. These fragments contain variable numbers of ATG codons, suggesting that some uORFs are excluded from segment 1B in parallel to neuronal differentiation (Fig. 3 E and Table II). The ability of uORFs to down-regulate translation is documented in mammals, as mentioned above (
      • van der Velden A.W.
      • Thomas A.A.
      ,
      • Morris D.R.
      • Geballe A.P.
      ). Although the brain is the major site of Dab1 expression, low mRNA levels are also detected in the kidney, liver, and uterus (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ). By RT-PCR, exons 1A and 1D were amplified from the brain, testis, kidney, and liver, but not from the spleen, heart, or thymus. By contrast, the mature forms of exon 1B were solely amplified from brain cDNA, confirming its neuron-specific expression in the adult.
      Figure thumbnail gr3
      Figure 3Expression of alternative 5′-UTRs. A, schematic representation of 5′-UTR exons (boxes) and primers used in RT-PCR, with orientation indicated by arrows. All reactions were carried out using 30 cycles of PCR and 25 ng of cDNA. UTRs 1A, 1B, 1C, and 1D are shaded differently. Exons 2–5 are coding exons. Note the complexity of UTR 1B, also shown in C–E. B, RT-PCR analysis of UTR 1A, 1B, and 1D expression during mouse brain development from E11 to adult. The following primer combinations were used: for exon 1A, primers 33 and 22 (amplicon of 666 bp); for UTR 1B, primers 33 and 37 (amplicons of 974 and 1085 bp corresponding to exons 1B1, 1B2, and 1B4 and exons 1B1, 1B2, 1B4, and 1B8, respectively); and for exon 1D, primers 33 and 38 (amplicon of 663 bp). Glyceraldehyde-3-phosphate dehydrogenase (G3PDH) was used as a reference gene. C, RT-PCR analysis of exons 1A, 1B, and 1D during P19 cell differentiation induced by RA from days 0 to 9 and in control P0 brain. The following primer combinations were used: for exon 1A, primers 2 and 22 (375-bp product); for exon 1D, primers 38 and 2 (351 bp); and for UTR 1B, primers 39 and 2 (450 and 560 bp, respectively, and several larger products). Hypoxanthine-guanine phosphoribosyltransferase (HPRT) was used as a reference gene. D, illustration of the complexity of UTR 1B using RT-PCR (primers 39 and 2). Shown is a comparison of undifferentiated (−RA) versus differentiated (+RA) P19 cells and of developing (E11) versus mature (adult (Ad)) mouse brain. Note the simplification of the UTR 1B amplification pattern in parallel to neural maturation. E, organization of UTR 1B. PCR products in C and D were cloned and sequenced (exon terminology as described for Fig. C). RT-PCR on non-neural tissues showed the inclusion of numerous alternative exons of UTR 1B.
      Northern blot analysis was performed using probes corresponding to the different alternative first exons and a probe covering the PTB domain-encoding region as a control. With the PTB domain probe, three bands of ∼5.5, 4, and 1.3 kb were revealed in poly(A) RNAs from E17 and P0 brain and correspond to the pattern described previously (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ,
      • Ware M.L.
      • Fox J.W.
      • Gonzalez J.L.
      • Davis N.M.
      • Lambert de Rouvroit C.
      • Russo C.J.
      • Chua Jr., S.C.
      • Goffinet A.M.
      • Walsh C.A.
      ). Satisfactory results were achieved with the exon 1A probe, which revealed three bands with similar sizes and an additional band of ∼1.8 kb (Fig. 4). No signal could be detected with a probe for exon 1D. A probe for UTR 1B revealed a major 5.5-kb band and several smaller and fainter bands, all of which are absent in RNA extracted from scrambler mouse brain (data not shown). No clear correspondence between the bands revealed on Northern blots and the alternative forms of the Dab1 RNA could be established.
      We analyzed the expression of the different 5′-first exons by in situ hybridization with 33P-labeled riboprobes. The short sequence of the exons and their high GC content (∼75% for exons 1A and 1D) made it difficult to obtain a signal, and satisfactory results could be obtained only with the complex UTR 1B. As shown in Fig. 5 (D–F), the UTR 1B expression pattern at E14 and in the newborn (P0) brain was very similar to that observed with the PTB probe (A–C), with the exception that the ventricular zone appeared to be very weakly labeled at both stages. Both signals were also similar in other parts of the brain. This suggests that 5′-UTR 1B and the form containing exon 1A are the two major Dab1 forms in the brain. This was confirmed using multiplex RT-PCR on adult brain RNA reactions, which yielded approximate proportions of messages containing exons 1A, 1B, and 1D in the brain of 47, 33, and 20%, respectively.

      Mouse Dab1 Promoter Activity in Neurons and Cell Lines

      The presence of multiple 5′-exons suggests that the transcription ofDab1 is regulated by different promoters. Promoter activity was assessed by transient transfection of HEK293 and HepG2 cells, which do not express Dab1, as well as undifferentiated P19 embryonic carcinoma cells and embryonic mouse primary neuronal cultures, which express Dab1 (Fig.6) (
      • Howell B.W.
      • Gertler F.B.
      • Cooper J.A.
      ). A 150-bp fragment upstream of exon 1A (construct A+ in Fig. 6 A) was active in all cells tested (25-fold in neurons, 8-fold in P19, 12-fold in HEK293, and 7-fold in HepG2). A comparable or higher activity was observed when this sequence was cloned in the reverse orientation (construct A in Fig. 6 B), suggesting that this region may function as a bidirectional promoter, at least in vitro. This may be related to its high GC content (75% in mouse with 32 CpG dinucleotides and 80% in human with 38 CpG dinucleotides), with three SP1-binding sites conserved between mouse and human coupled with the absence of TATA and CAAT sequences. Construct C+, containing the region upstream of exon 1C, showed weak promoter activity in all cell lines tested. This segment has a lower GC content of ∼56%. The promoter prediction programs Promoter Inspector, TSSW, and TSSG detected a promoter, a degenerate TATA box, and a transcription initiation site in this region. Construct D+, corresponding to the 500-bp region upstream of exon 1D, was 6-fold more active than the promoterless vector in HEK293 cells and neurons and 3-fold more active in HepG2 and P19 cells. This region had no promoter activity when cloned in the reverse orientation. The programs also predicted a promoter and a degenerate TATA box in this segment. Construct AD+, which contains both regions upstream of exons 1A and 1D, showed promoter activity comparable to that of fragment A+. Construct ACD+, which contains exons 1A, 1C, and 1D, showed promoter activity comparable to that of fragment C+. There was no activity of this segment when cloned in the reverse orientation. Two constructs were used to assay the promoter activity of regions upstream of exon 1B1 (data not shown). A 1.3-kb construct that includes 350 nucleotides of exon 1B1 and three ATG codons was inactive in primary cortical neurons. To avoid possible interference of the ATG codons with translation of the luciferase reporter, another construct was derived by deleting these ATG triplets. However, no promoter activity was detected in primary cortical neurons, indicating that the promoter of form 1B may be located farther upstream in the 260-kb genomic interval between exon 1B1 and the AK008020 gene.
      Figure thumbnail gr6
      Figure 6Promoter activities of sequences upstream of exons 1A, 1C, and 1D in mouse. A, shown is a schematic representation of the genomic region with the constructs used to test promoter activity in the luciferase reporter system. B, reporter activity was tested in P19 cells, primary neuronal cultures, and HepG2 and HEK293 cells. The promoter activity is expressed relative to that of the promoterless control plasmid pGL3-Basic. + and − refer to constructs tested in the forward and reverse orientations. Values correspond to the means ± S.D. of at least three experiments.

      Internal Alternative Splicing Events

      Using PCR on human brain cDNA and alignment of genomic and EST sequences, we were unable to identify exons corresponding to mouse Dab1 forms 217 and 271 in human. Using RT-PCR, the presence of fragment 555* was confirmed in mouse and man. In both species, this sequence corresponds to two exons of 51 and 48 bp separated by an intron of 91 bp in mouse and of 89 bp in human (Fig. 7 A). Both exons were consistently co-amplified. Interestingly, an alternatively spliced product of 57 bp was detected in the corresponding location in theDab1 cDNA in lizard and chick (Fig. 7 B). As shown in Fig. 7 C, the peptide sequences encoded by the two small exons that form fragment 555* in mouse and man and by the single 57-nucleotide exon in lizard and chick are conserved, suggesting a duplication event during evolution. Upon Northern blotting using a probe that includes exons 555* and some adjacent sequences, a major RNA species of ∼5.5 kb, presumably corresponding to the longest form of the Dab1 mRNA, was detected in poly(A) RNA from E17 mouse brain (data not shown). In undifferentiated P19 cells, theDab1 cDNA did include fragment 555*. However, when differentiation of P19 cells was induced with RA, a proportion ofDab1 cDNA without fragment 555* appeared at day 2 and increased progressively to become the major form at day 9 (Fig.7 D). In early embryonic mouse brain (E11 and E12), theDab1 isoform with fragment 555* was predominant, but RNAs from later developmental stages (E12 and later) and from primary neuronal cultures did not contain this fragment (Fig. 7 B). In non-neural tissues such as liver and kidney, the Dab1mRNA contained fragments 555* (data not shown). A similar pattern was found in chick, with inclusion of the small 57-nucleotide exon in RNA from E6 or adult eye, but exclusion of that exon from brain RNA at E20 (Fig. 7 B).
      Figure thumbnail gr7
      Figure 7Alternative exons 555* are excluded from neurons. A, shown is the genomic localization of alternative exons 555* in the mouse Dab1 gene. Form 555* is composed of two exons of 51 (555*1) and 48 (555*2) bp, located between alternative exon 271 and exon 10. Boxes indicate exons, andhorizontal lines indicate introns. Primers are represented by arrows. B, using primers 31 and 36, two products of 746 and 644 bp were obtained, respectively, with and without exons 555*. Alternative exon 271 was never included in the amplified fragments. Exons 555* were expressed during early mouse brain development, but not in P0 brain or primary cortical neuronal cultures. An alternative exon was also included in RT-PCR products amplified from chick (early stage E6 and eye, but not E20) or lizard RNA using the same primers. C, in human (Homo sapiens(Hs)) and mouse (Mus musculus (Mm)), form 555* is composed of two exons, 555*1 and 555*2. In chick and lizard, this alternative form is composed of one small exon. The amino acid sequences coded by these exons are well conserved, and this suggests a possible duplication in mammals. D, shown is the alternative splicing of exons 555* during P19 cell differentiation induced by RA from days 0 to 9. cDNAs prepared from undifferentiated and differentiated P19 cells and control P0 mouse brain RNA were amplified using primers 31 and 36. During neuronal maturation, the larger product of 746 bp containing exon 555* was progressively replaced with a product of 644 bp lacking exon 555*.
      Using in situ hybridization with a cDNA probe covering fragments 555* and adjacent segments (Figs. 5 (G–I) and 7A), a strong signal was detected in ventricular zones of precursor proliferation; the moderate labeling of post-migratory fields could be related to the parts of the probe adjacent to exons 555* (Fig.5, G–I). The predominance of the exon 555*-related signal in ventricular zones was also noted in other parts of the brain. Altogether, these observations suggest that the exclusion of exons 555* parallels neural differentiation.

      DISCUSSION

      Both in man and mouse, the Dab1 gene reveals an unusual complexity that leaves ample room for subtle regulation of its expression and function. Examples of such highly complex genomic organization are few and include the metabotropic glutamate receptor GRM8 gene, which spans >800 kb of genomic DNA for a coding length of 2.3 kb (
      • Scherer S.W.
      • Soder S.
      • Duvoisin R.M.
      • Huizenga J.J.
      • Tsui L.C.
      ), and the human neurotrophin receptor genes NTRK2 and NTRK3 (
      • Valent A.
      • Danglot G.
      • Bernheim A.
      ), which extend over >350 and 380 kb for coding lengths of 3.7 and 2.8 kb, respectively. Intriguingly, the Dab1 paralogous geneDab2 (named DAB2 or DOC2 in human) is much simpler than Dab1, with an ORF extending over <50 kb of genomic sequence compared with 300 kb for Dab1 (
      • Sheng Z.
      • He J.
      • Tuppen J.A.
      • Sun W.
      • Fazili Z.
      • Smith E.R.
      • Dong F.B.
      • Xu X.X.
      ,
      • Sheng Z.
      • He J.
      • Tuppen J.A.
      • Martin W.D.
      • Dong F.B.
      • Xu X.X.
      ). Apparently, this situation is not unusual. For example, theNTRK1 gene, closely related to NTRK2 andNTRK3, spreads over only 20 kb (
      • Valent A.
      • Danglot G.
      • Bernheim A.
      ). A similar feature is found in the two mouse paralogous phospholipase D genes Pld1and Pld2. Whereas Pld1 contains 28 exons and spans ∼147 kb, the whole Pld2 gene is contained in 17 kb of genomic DNA (
      • Redina O.E.
      • Frohman M.A.
      ). From an evolutionary standpoint, it would be interesting to know whether such huge differences in the genomic complexity of paralogous genes result from extension or contraction of the set of introns in one of the genes after duplication. Like mostDrosophila genes, the fly Disabled gene has small introns and extends over 12 kb of genomic DNA (
      • Gertler F.B.
      • Hill K.K.
      • Clark M.J.
      • Hoffmann F.M.
      ), suggesting that the large size of Dab1 might result from intron extension.
      Our results also reveal a remarkable diversity in the 5′-UTR of both the human and mouse Dab1 genes. We have identified six alternative 5′-UTRs in human and four in mouse, three of which are conserved. Fragments with promoter activity were defined for two of them, but we were unable to clone the promoter for 5′-UTR 1B. This 5′-UTR is unusually complex and spreads over 1 Mbp of genomic DNA. It is composed of seven exons in human and 10 exons in mouse, with three exons conserved between both species and always included together in the mRNA. This results in a 5′-UTR of 1 kb or more (with the inclusion of alternative exons), which is much larger than the average size of 210 bp (
      • Pesole G.
      • Liuni S.
      • Grillo G.
      • Licciulli F.
      • Mignone F.
      • Gissi C.
      • Saccone C.
      ,
      • Mignone F.
      • Gissi C.
      • Liuni S.
      • Pesole G.
      ). In situ hybridization using a probe specific for UTR 1B and RT-PCRs clearly showed that it is part of theDab1 mRNA. The long 5′-UTR of Dab1 contains small uORFs and numerous upstream ATG codons that precede the major translation initiation site, some of which are conserved between human and mouse. The ability of uORFs to down-regulate translation of mRNA in mammals is well documented. For example, in mice, this phenomenon is implicated in the 50-fold increase in the concentration of the cyclin-dependent kinase inhibitor p18INK4c during differentiation of skeletal muscle cells. In proliferating myoblasts, this gene is abundantly transcribed, but not detectably translated, because the mRNA carries a 1115-nucleotide-long 5′-UTR with five upstream ATG codons. During differentiation, a downstream promoter produces a second form of mRNA with a much shorter 5′-UTR that efficiently supports translation (
      • Phelps D.E.
      • Hsiao K.M.
      • Li Y.
      • Hu N.
      • Franklin D.S.
      • Westphal E.
      • Lee E.Y.
      • Xiong Y.
      ). It would be interesting to test whether mutations in the long 5′-UTR of Dab1 affect the expression of the major ORF, resulting in altered regulation of gene expression in vivo.
      Large gene size and complexity may be important for the production and processing of the transcripts. Based on a transcription rate of ∼1.4 kb/min (
      • Shermoen A.W.
      • O'Farrell P.H.
      ) and data on the dystrophin gene (
      • Tennyson C.N.
      • Klamut H.J.
      • Worton R.G.
      ), transcription ofDab1 would require at least 13 h. This is close to or larger than the estimated division time of neuronal precursors (
      • Takahashi T.
      • Nowakowski R.S.
      • Caviness Jr., V.S.
      ), suggesting that the promoter associated with form 1B could not be utilized in proliferating cells. In summary, our data show thatDab1 is far more complex than expected and that further work is needed to understand better the control of Dab1expression and the molecular machinery by which it exerts its powerful activity. The detailed genomic structure reported here should facilitate the study of human DAB1 mutations, which are predicted to yield abnormal brain phenotypes similar to those related to Reelin deficiency.

      REFERENCES

        • Gupta A.
        • Tsai L.H.
        • Wynshaw-Boris A.
        Nat. Rev. Genet. 2002; 3: 342-355
        • Monuki E.S.
        • Walsh C.A.
        Nat. Neurosci. 2001; 4 (suppl.): 1199-1206
        • Lambert de Rouvroit C.
        • Goffinet A.M.
        Mech. Dev. 2001; 105: 47-56
        • Howell B.W.
        • Herrick T.M.
        • Hildebrand J.D.
        • Zhang Y.
        • Cooper J.A.
        Curr. Biol. 2000; 10: 877-885
        • Rice D.S.
        • Curran T.
        Annu. Rev. Neurosci. 2001; 24: 1005-1039
        • Feng Y.
        • Walsh C.A.
        Nat. Rev. Neurosci. 2001; 2: 408-416
        • Howell B.W.
        • Hawkes R.
        • Soriano P.
        • Cooper J.A.
        Nature. 1997; 389: 733-737
        • Sheldon M.
        • Rice D.S.
        • D'Arcangelo G.
        • Yoneshima H.
        • Nakajima K.
        • Mikoshiba K.
        • Howell B.W.
        • Cooper J.A.
        • Goldowitz D.
        • Curran T.
        Nature. 1997; 389: 730-733
        • Ware M.L.
        • Fox J.W.
        • Gonzalez J.L.
        • Davis N.M.
        • Lambert de Rouvroit C.
        • Russo C.J.
        • Chua Jr., S.C.
        • Goffinet A.M.
        • Walsh C.A.
        Neuron. 1997; 19: 239-249
        • Lambert de Rouvroit C.
        • Goffinet A.M.
        Adv. Anat. Embryol. Cell Biol. 1998; 150: 1-106
        • Hong S.E.
        • Shugart Y.Y.
        • Huang D.T.
        • Shahwan S.A.
        • Grant P.E.
        • Hourihane J.O.
        • Martin N.D.
        • Walsh C.A.
        Nat. Genet. 2000; 26: 93-96
        • Rice D.S.
        • Curran T.
        J. Comp. Neurol. 2000; 424: 327-338
        • Carroll P.
        • Gayet O.
        • Feuillet C.
        • Kallenbach S.
        • de Bovis B.
        • Dudley K.
        • Alonso S.
        Mol. Cell. Neurosci. 2001; 17: 611-623
        • Yip J.W.
        • Yip Y.P.
        • Nakajima K.
        • Capriotti C.
        Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 8612-8616
        • D'Arcangelo G.
        • Homayouni R.
        • Keshvara L.
        • Rice D.S.
        • Sheldon M.
        • Curran T.
        Neuron. 1999; 24: 471-479
        • Trommsdorff M.
        • Gotthardt M.
        • Hiesberger T.
        • Shelton J.
        • Stockinger W.
        • Nimpf J.
        • Hammer R.E.
        • Richardson J.A.
        • Herz J.
        Cell. 1999; 97: 689-701
        • Hiesberger T.
        • Trommsdorff M.
        • Howell B.W.
        • Goffinet A.
        • Mumby M.C.
        • Cooper J.A.
        • Herz J.
        Neuron. 1999; 24: 481-489
        • Gotthardt M.
        • Trommsdorff M.
        • Nevitt M.F.
        • Shelton J.
        • Richardson J.A.
        • Stockinger W.
        • Nimpf J.
        • Herz J.
        J. Biol. Chem. 2000; 275: 25616-25624
        • Howell B.W.
        • Lanier L.M.
        • Frank R.
        • Gertler F.B.
        • Cooper J.A.
        Mol. Cell. Biol. 1999; 19: 5179-5188
        • Howell B.W.
        • Herrick T.M.
        • Cooper J.A.
        Genes Dev. 1999; 13: 643-648
        • Keshvara L.
        • Benhayon D.
        • Magdaleno S.
        • Curran T.
        J. Biol. Chem. 2001; 276: 16008-16014
        • Herrick T.M.
        • Cooper J.A.
        Development. 2002; 129: 787-796
        • Howell B.W.
        • Gertler F.B.
        • Cooper J.A.
        EMBO J. 1997; 16: 121-132
        • Lambert de Rouvroit C.
        • Goffinet A.M.
        Genomics. 1998; 53: 246-247
        • Kojima T.
        • Nakajima K.
        • Mikoshiba K.
        Brain Res. Mol. Brain Res. 2000; 75: 121-127
        • van der Velden A.W.
        • Thomas A.A.
        Int. J. Biochem. Cell Biol. 1999; 31: 87-106
        • Morris D.R.
        • Geballe A.P.
        Mol. Cell. Biol. 2000; 20: 8635-8642
        • Scherer S.W.
        • Soder S.
        • Duvoisin R.M.
        • Huizenga J.J.
        • Tsui L.C.
        Genomics. 1997; 44: 232-236
        • Valent A.
        • Danglot G.
        • Bernheim A.
        Eur. J. Hum. Genet. 1997; 5: 102-104
        • Sheng Z.
        • He J.
        • Tuppen J.A.
        • Sun W.
        • Fazili Z.
        • Smith E.R.
        • Dong F.B.
        • Xu X.X.
        Genomics. 2000; 70: 381-386
        • Sheng Z.
        • He J.
        • Tuppen J.A.
        • Martin W.D.
        • Dong F.B.
        • Xu X.X.
        Gene. 2001; 268: 31-39
        • Redina O.E.
        • Frohman M.A.
        Gene. 1998; 222: 53-60
        • Gertler F.B.
        • Hill K.K.
        • Clark M.J.
        • Hoffmann F.M.
        Genes Dev. 1993; 7: 441-453
        • Pesole G.
        • Liuni S.
        • Grillo G.
        • Licciulli F.
        • Mignone F.
        • Gissi C.
        • Saccone C.
        Nucleic Acids Res. 2002; 30: 335-340
        • Mignone F.
        • Gissi C.
        • Liuni S.
        • Pesole G.
        Genome Biol. 2002; 3: S0004
        • Phelps D.E.
        • Hsiao K.M.
        • Li Y.
        • Hu N.
        • Franklin D.S.
        • Westphal E.
        • Lee E.Y.
        • Xiong Y.
        Mol. Cell. Biol. 1998; 18: 2334-2343
        • Shermoen A.W.
        • O'Farrell P.H.
        Cell. 1991; 67: 303-310
        • Tennyson C.N.
        • Klamut H.J.
        • Worton R.G.
        Nat. Genet. 1995; 9: 184-190
        • Takahashi T.
        • Nowakowski R.S.
        • Caviness Jr., V.S.
        J. Neurosci. 1995; 15: 6046-6057
        • Bernier B.
        • Bar I.
        • Pieau C.
        • Lambert de Rouvroit C.
        • Goffinet A.M.
        J. Comp. Neurol. 1999; 413: 463-479
        • Simmons D.
        • Arriza J.L.
        • Swanson L.W.
        J. Histotechnol. 1989; 12: 169-181