Notch2: a second mammalian Notch gene

Notch is a large transmembrane protein that mediates a set of cellular interactions required for correct cell-type differ- entiation during Drosophila development (reviewed in Artavanis-Tsakonas et al., 1991; Campos-Ortega and Jan, 1991; Artavanis-Tsakonas and Simpson, 1991; Greenwald and Rubin, 1992). Although genetic studies first identified Notch as a gene whose activity directly affects the choice between neural and epidermal cell fates in the neurogenic region of the developing Drosophila embryo (Poulson, 1937; Lehman et al., 1983), subsequent studies have also demonstrated roles for Notch in eye (Cagan and Ready, 1989) and sensillum development (Hartenstein and Posakony, 1990), in mesoderm differentiation (Corbin et al., 1991) and in oogenesis (Ruohola et al., 1991). In all of these areas, the loss of Notch leads to the expansion of one cell type at the expense of another cell type. Notch is widely expressed during embryonic and adult development (Hart- ley et al., 1987; Markopoulou and Artavanis-Tsakonas, 1989; Johansen et al., 1989; Kidd et al., 1989; Fehon et al., 1991) and different mutant alleles of the Notch gene affect various tissues during different periods of Drosophila development (Hartley et al., 1987; Kelley et al., 1987; Markapoulou et al., 1989; Xu et al., 1990). Taken together, these results suggest that Notch mediates a large number of different cell-cell interactions during normal fly development.

While there is compelling evidence implicating Notch in processes that determine cell-fate decisions in Drosophila, the mechanism by which this protein exerts its widespread effects remains unclear (Greenspan, 1990; Simpson, 1990; Artavanis-Tsakonas and Simpson, 1991; Greenwald and Rubin, 1992). Results from genetic (Hartley et al., 1987; Kelley et al., 1987; Markopoulou et al., 1989; Xu et al., 1990), biochemical (Fehon et al., 1990; Rebay et al., 1991) and genetic mosaic studies (Hoppe and Greenspan, 1990; Simpson, 1990; Heitzler and Simpson, 1991) are all consistent with the hypothesis that Notch functions as a recep- tor for a signal ligand that regulates the phenotype of the receiving cell. It is thought that Notch might interact with multiple ligands (Rebay et al., 1991), given its wide expression during embryonic and imaginal development, the presence of multiple repeats related to epidermal growth factor (EGF) in its ligand-binding domain and the pleiotropic nature of Notch mutations. Two strong candidates for such ligands are the proteins encoded by the Delta (Vässin et al., 1987; Kopcyznski et al., 1988) and Serrate genes (Fleming et al., 1990; Thomas et al., 1991). Delta has been implicated as a potential Notch ligand during neuroblast segregation, since loss-of-function mutations in Delta produce the same neurogenic phenotype seen in Drosophila embryos carrying loss-of-function mutations in Notch (Lehman et al., 1983; Vässin and Campos-Ortega, 1987). Furthermore, imaginal epithelium cells mutant for Delta behave non-autonomously in genetic mosaics, sug- gesting that Delta could be the signal for the cell sorting that occurs there (Heitzler and Simpson, 1991). The mostconvincing data supporting a direct interaction between Notch and these putative ligands comes from cell aggrega- tion studies (Fehon et al., 1990; Rebay et al., 1991). Cul- tured Drosophila S2 cells expressing Notch form aggre- gates, in a calcium-dependent manner, specifically with cells expressing either Delta or Serrate. Consistent with the direct protein-protein interaction seen in cell aggregation studies, Notch and Delta proteins colocalise in multi-vesic- ular structures in Notch-expressing cells (Artavanis- Tsakonas and Simpson, 1991) and are found complexed together in immunoprecipitates from Drosophila embryo extracts (Fehon et al., 1990).

Notch cDNA clones have been isolated from a number of vertebrate species and the proteins that these cDNAs encode are closely related to the Drosophila protein (Whar- ton et al., 1985a; Kidd et al., 1986; Coffman et al., 1990; Ellisen et al., 1991; Weinmaster et al., 1991). Furthermore, the temporal and spatial distributions of Notch transcripts in murine embryos suggest that the function of mammalian Notch may be analogous to its Drosophila counterpart (Weinmaster et al., 1991). The vertebrate and invertebrate Notch genes also share sequence homology with the C. ele - gans glp-1 and lin-12 genes, which also direct cell-fate deci- sions (Yochem et al., 1988; Yochem and Greenwald, 1989; Austin and Kimble, 1989; Greenwald and Rubin, 1992) and with the recently isolated mouse int-3 oncogene (Robbins et al., 1992). This relatedness is evidenced both in the extra- cellular and cytoplasmic domains of the predicted proteins (Greenwald and Rubin, 1992). The relationship between vertebrate members of this gene family has been established only through sequence identity, since a functional role has yet to be demonstrated for any of these genes. However, it is notable that both TAN-1, the human Notch homolog (Ellisen et al., 1991), and int-3 (Robbins et al., 1992; Jhap- pan et al., 1992) were identified as mutated sequences that interfered with cell differentiation and resulted in neoplas- tic transformation.

It has been suggested that interactions between specific numbers or combinations of the highly conserved 36 EGF- like repeats in Drosophila Notch with different ligands could account for the great diversity of cell-cell interactions regulated by Notch (Rebay et al., 1991). If the Notch sig- naling system regulates mutiple cell-cell interactions in ver- tebrates, then such diversity might also be generated through the activation of different Notch-like receptors. In this report, we present direct evidence that supports this second hypothesis. Here we describe the cloning, sequenc- ing and expression of a second rat Notch gene, designated Notch2. Although Notch2 has retained all the structural motifs that are hallmarks of a Notch protein, it is expressed in a distinct set of cells and at different times in develop- ment than the previously isolated rat Notch gene.

A rat Schwann cell cDNA library was hybridised to a ³²P-radio- labelled EcoRI fragment derived from the EGF-like repeat region of Xenopus Notch cDNA clone X-2 (Coffman et al., 1990) as pre- viously described (Weinmaster et al., 1990). Briefly, Hybond-N membrane filters (Amersham) were hybridized in a solution con- taining 1 M NaCl, 50 mM Tris-HCl (pH 7.5), 1.0% SDS, 5× Den- hart’s solution and 0.2 mg/ml salmon sperm DNA at 50°C for 24 hour. Filters were then washed in 5× SSC with 0.5% SDS at 21°C followed by washes at 50°C. Clone SN12, identified in this screen, was used to isolate cDNA clones encoding additional Notch- related sequence (Fig. 1). All of these clones were isolated under high-stringency conditions (0.2× SSC containing 1.0% SDS at 65°C). Initially a 1.5 kbp PCR fragment, sythesized from SN12 sequence that encoded 13 EGF-like repeats, was used to screen a random-primed rat forebrain cDNA library (J. Boulter, Salk Insti- tute). This screen yielded 21 clones that provided novel sequence both 5′ and 3′ of the original SN12 sequence. The first clone, H11S-1, (1466 bp in size) contained 449 bp that extended 5′ of SN12 and the remaining sequences overlapped SN12 by 1017 bp. The 449 bp in H11S-1 that extended 5′ of SN12 contained 86 bp of untranslated sequence, a potential ATG translation start site, a putative signal peptide sequence and additional sequence that encoded approximately two EGF-like repeats. The second clone, H10, contained 3340 bp of sequence that overlapped H11s-1 by 572 bp and SN12 by 1231 bp, while the remaining 2109 bp encoded novel sequences related to the 35th EGF-like repeat of rat Notch (Weinmaster et al., 1991). A 287 bp PCR fragment from the 3′ end of H10 was used to screen the same library to obtain clones containing sequence 3′ of H10. From this screen, a number of clones were identified, of which H10-6 was the largest (2462 bp) and overlapped clone H10 by 670 bp. The 3′ end of H10-6 was found to contain sequences related to CDC10 repeats. A 458 bp PCR fragment containing this 3′ sequence was used to probe the originally screened Schwann cell library. This screen yielded three clones that all overlapped clone H10-6. The largest overlap was found with RSC-3, which shared 1161 bp in common with H10-6 and extended 3′ of H10-6 for more than 4.8 kbp. RSC-3 and was found to have a stop codon 1473 bp downstream of the 3′ end of H10-6. All three of the cDNA clones isolated from the Schwann cell library had at least 3 kbp of untranslated sequence at their 3′ ends (data not shown) of which only 361 bp, follow- ing the stop codon, are reported for the Notch2 sequence. The clones presented in Fig. 1 were sequenced using the dideoxy chain termination method (Sequenase, US Biochemicals).

Poly (A)⁺RNA was isolated and northern blot hybridisation analy- sis performed as described previously (Weinmaster and Lemke, 1990; Lai and Lemke, 1991). Methylene blue staining was used to verify that equal amounts of RNA were transferred to nylon membranes (MSI). Random-primed ³²P-radiolabelled probes were prepared from cDNA inserts isolated from SN6 (Weinmaster et al., 1991) to identify rat Notch1 transcripts and from pooled cDNA inserts isolated from H10, H10-6 and RSC-3 to detect rat Notch2 mRNA.

Preparation of rat embryo sections and in situ hybridisation analy- sis were carried out essentially as described previously (Roberts et al., 1991). Pregnant Sprague-Dawley female rats were anes- thetised with 35% chloral hydrate and perfused transcardially with 4.0% paraformaldehyde in 0.1 M sodium tertraborate buffer, pH 9.5. The rat embryos (E16), were removed and submersion-fixed for 3-6 months at 4°C in the same fixative. Frozen sections, 20 μm thick, were cut on a cryostat, mounted on polylysine-coated slides and desiccated overnight. Preparations of adult rat brain tiss- sue was essentially the same except that brains from perfused rats were post-fixed for 2-4 weeks and sectioned at 30 μm thick. Iden- tical conditions were used to detect transcripts for Notch1 and Notch2 in adult rat brain and whole rat embryo sections. The specificity of the antisense ³⁵S-radiolabelled riboprobe, synthe- sised from cDNA insert encoding amino acid residues 1755-1903 in the intracellular portions of rat Notch1, subcloned into Bluscript SK⁺ (SN6-7), has been described previously (Weinmaster et al., 1991). For detection of rat Notch2 transcripts, an antisense ³⁵S- riboprobe was synthesised from the linearised SN12-79 plasmid that contained sequences encoding amino acids 351-535 of rat Notch2 (see Fig. 2). The details of the prehybridisation, hybridi- sation and autoradiographic localisation techniques have been published previously (Simmons et al., 1989; Roberts et al., 1991). After film exposure, slides were defatted and dipped in nuclear track emulsion (NTB-2, Kodak) and exposed for appropriate times. Following developing (Kodak D19) and fixation (Kodak, Ektaflo), brain sections were stained with bisbenzimide (Sigma Chemical Co., St Louis, Mo.), which provides a fluorescent blue ribosomal counter stain. Embryo sections were counter-stained with hematoxylin and eosin-Y. Sections were photographed using Kodak Gold 35 mm Kodacolor 100 ASA film and printed on Fuji paper.

A rat Schwann cell cDNA library was screened with a Xenopus Notch cDNA probe (Coffman et al., 1990) under low-stringency hybridisation conditions (see Material and methods). From this screen a number of cDNA clones were isolated, some of which were shown to encode the rat Notch gene (Weinmaster et al., 1991), while others contained related Notch-like sequences. One of the isolated cDNA clones, designated SN12, was found to encode 13 EGF-like repeats. These EGF-like repeats were aligned with EGF- like repeats 3 through 16 of rat Notch with 56% identity at the amino acid level (Fig. 1). This sequence analysis suggested that clone SN12 encoded a novel, yet related, Notch gene.

Rescreening of the rat Schwann cell library failed to yield any additional SN12-related clones. Since in situ hybridisation data indicated that SN12 was expressed in the adult rat brain (see below), we screened a random-primed rat forebrain cDNA library to obtain cDNA clones that over- lapped and extended either 5′ or 3′ of the SN12 clone (for details see Materials and methods). All of the overlapping clones, shown in Fig. 1, were sequenced to obtain 7860 bp of cDNA sequence. This sequence carries a large open read- ing frame of 2472 amino acids, starting with an initiator methionine at nucleotide 86 and terminating with a stop codon at nucleotide 7413. The nucleotide and deduced amino acid sequence of rat Notch2 have been deposited with GenBank under the accession number M93661.

Analysis of the aligned amino acid sequences reported for Notch cDNA clones isolated from Drosophila (Wharton et al., 1985a; Kidd et al., 1986), Xenopus (Coffman et al., 1990), human (Ellisen et al., 1991) and rat (Weinmaster et al., 1991) reveals a number of characteristic and highly con- served motifs. Inspection of the amino acid sequence, deduced from the nucleotide sequence of the cDNA clones described above, established the presence of all the previ- ously identified Notch structural domains (Figs 1, 2). We have therefore designated the gene that encodes this sequence rat Notch2 and refer to the previously published rat Notch gene as Notch1. Fig. 2 shows the alignment of the deduced amino acid sequence for rat Notch1 and Notch2. The two proteins exhibit overall amino acid sequence identity of 56%.

The extracelluar domain of Drosophila Notch contains two cysteine-rich regions (Wharton et al., 1985; Kidd et al., 1986; Yochem et al., 1988). The most N-terminal of these regions contains 36 tandemly arranged repeats that share homology with EGF. Both Notch1 and Notch2 have retained each of these EGF-like repeats in the exact arrange- ment described for Drosophila Notch. In the EGF-like repeat domain, Notch1 and Notch2 exhibit 58% amino acid sequence identity. Of these 36 EGF-like repeats, only repeats 11 and 12 are necessary and sufficient for Drosophila Notch to bind Delta or Serrate proteins, when measured in an in vitro aggregation assay (Rebay et al., 1992). The residues between the first and second cysteine of EGF-like repeat 11 appear crucial for this interaction, since a leucine-to-proline change in this region of Xenopus Notch obliterates Delta and Serrate binding (see Fig. 3). At this site, Notch2 has one and two additional amino acids,respectively, relative to vertebrate Notch and Drosophila Notch sequences (see Fig. 3). The second cysteine-rich region consists of three contiguous lin12/Notch repeats (LNR), which are also present in the C. elegans lin-12 and glp-1 proteins (Yochem et al., 1988; Yochem and Green- wald, 1989; Austin and Kimble, 1989). Notch1 and Notch2 also have these repeats, which are related to each other with 58% amino acid sequence identity.

Hydropathy analysis of the Notch2 amino acid sequence identified two hydrophobic domains. One region, 27 residues in length, located at the extreme N terminus, prob- ably corresponds to a cleaved signal peptide sequence. The second hydrophobic region, located between amino acids 1677-1699 (see Fig. 2), is a putative transmembrane domain. Between this putative transmembrane domain and the LNR are 2 cysteines that are conserved in all previously described Notch proteins as well as C. elegans lin-12 and glp-1. In C. elegans, these highly conserved cysteines are thought to promote dimerization of lin-12 following ligand binding (Greenwald and Seydoux, 1990). They are found in the Notch2 amino acid sequence at positions 1632 and 1639 (Fig. 2).

The intracellular region of Drosophila Notch contains CDC10 repeat elements that were first described in yeast cell cycle control proteins (Breeden and Nasmyth, 1987; Andrews and Herskowitz, 1989) and have subsequently been identified in a diverse set of proteins including ankyrin (Lux et al., 1990) and a number of transcription factors (Thompson et al., 1991; LaMarco et al., 1991; Hoffman, 1991). Notch2 was found to have six of these CDC10 repeats, a number that is characteristic of the Notch family of proteins. Notch1 and Notch2 show 76% amino acid sequence identity in this region. Immediately carboxy-ter- minal to the CDC10 repeats Notch1 and Notch2 have sev- eral short areas of amino acid homology. However, the homology between Notch1 and Notch2 begins to break down near residues 2210 and 2171, respectively, which is also an area where Drosophila Notch has a large insert of approximately 100 residues relative to the vertebrate Notch proteins (Weinmaster et al., 1991). The amino acid sequence of Notch2 from residues 2171 to 2385 shows no signifigant homology with either Drosophila Notch or rat Notch1 protein sequences from the equivalent region.

The OPA region of Drosophila Notch contains a stretch of 30 glutamine residues (Wharton et al., 1985b). These glutamines are only partially conserved in Notch1 (10 glu- tamines) and in Notch2 only 3 glutamines were detected in the general OPA area (Fig. 2). It is possible that the OPA repeats are similar to the unstable CAG repeats associated with mutations in the androgen receptor gene, which are thought to underly to X-linked spinal and bulbar muscular atrophy (Spada et al., 1991).

The amino acid sequences near the C termini of the Drosophila and vertebrate Notch proteins are highly con- served and contain a number of putative phosphorylation sites (Pearson and Kemp, 1991) nested in a PEST consen- sus sequence (Rogers et al., 1986). The Notch2 protein sequence also contains a PEST sequence near its C termi- nus, as well as a number of putative phosphorylation sites. In this region, the Notch1 and Notch2 proteins exhibit 79% amino acid sequence identity. Finally, 12 out of the last 14 amino acids in the human, Xenopus and rat Notch1 pro- teins are identical, while the terminal 20 residues in Drosophila Notch appear unrelated to these sequences. Interestingly, the Notch2 protein sequence ends with 28 amino acids that have no sequence homology with any of the previously described Notch proteins.

A phylogenetic tree for the Notch gene family was obtained as described by Feng and Doolittle (1987) (Fig. 4). This representation of the Notch gene family suggests that rat Notch1, human Notch and Xenopus Notch are homologs and that they are more closely related to each other than to either rat Notch2 or Drosophila Notch. More- over, since the amino acid sequences of rat Notch1 and Notch2 are equally related to Drosophila Notch, it is diffi- cult from sequence alone to discern which of the two rat genes is the true homolog of Drosophila Notch (see Dis- cussion). The C. elegans proteins, lin-12 and glp-1, and mouse int-3 are more distantly related to the vertebrate and Drosophila Notch genes and may represent distinct sub- families.

Fig. 5 illustrates the amount of Notch1 and Notch2 mRNA present in embryonic (E14 and E17), postnatal (P10 and 2 mos) and adult rat brain. As previously reported, Notch1 expression was high in embryonic brain and decreased thereafter (Weinmaster et al., 1991). In contrast, the level of Notch2 expression in the rat brain remained nearly constant throughout this same period (Fig. 5). To determine if the two rat Notch genes were expressed outside the ner- vous system, we examined RNA isolated from various tissues obtained from a 27-day-old rat. The results presented in Fig. 5 indicate that both Notch1 and Notch2 are expressed in a number of postnatal tissues; however, the level of RNA expressed by the two Notch genes in the var- ious tissues examined was markedly different. Most notable is the signal obtained for Notch2 in the spleen, which was approximately 50-fold higher than that observed for Notch1. A similar ratio of mRNA expression for Notch2 versus Notch1 was also found with IG-positive mouse splenocytes (G. W. and J. Lesley, unpublished data).

The rat Notch1 gene is transcribed in both neural and non- neural cells (Weinmaster et al., 1991). To determine if the related Notch2 gene had a similar pattern of expression, we used in situ hybridisation. This analysis revealed both over- lapping as well as differential expression for the two Notch genes. Table 1 summarises the sites of Notch1 and Notch2 gene expression identified in adjacent sections of E16 rat embryos.

Both genes were expressed in the developing central ner- vous system (CNS) and peripheral nervous system (PNS), but in different cell types. A strong signal for Notch1 was clearly visible around the ventricles in the developing rat brain (Fig. 6A,B), with the signal restricted to the periven- tricular zone (Fig. 7B). In contrast, no signal above back-ground was detected for Notch2 in the ventricular linings (Fig. 6C,D). Conversely, a strong Notch2 signal was detected in the epithelial cells of the choroid plexus (Fig. 7D), although no Notch1 signal was detected in this tissue (Fig. 7B).

Notch1 and Notch2 also had complementary expression patterns in distinct regions of the eye and dorsal root gan- glia. Notch1 was expressed in proliferating cells of the retina (Fig. 7F), while Notch2 mRNA was detected in mes- enchyme external to the pigmented epithelium (presump- tive choroid) and not in the retina (Fig. 7H). Examination at higher power identified silver grains over nuclei of pig- mented epithelial cells for both Notch1 and Notch2 (data not shown). The entire dorsal root ganglion (DRG) appeared positive for Notch1 expression (Fig. 8F), while the Notch2 probe labelled only the perineurium surround- ing the DRG (Fig. 8H). The branches of the trigeminal gan- glion were positive with both Notch probes (Fig. 8B,D); however, the Notch1 signal appeared more intense than that detected for Notch2. Finally, while both genes were expressed in the glomeruli of the kidney (Fig. 9B,D), only Notch1 was expressed in the bronchiole tubes in the lung (Fig. 9F,H) and matrix cells of the developing vibrissae (Fig. 9J,L). In both of these areas, the Notch1 signal was restricted to epithelial cells.

The northern blot hybridisation analyses described above indicated that both Notch1 and Notch2 were expressed in the adult brain, but at dissimilar levels. The results pre- sented in Fig. 10 show that the most prominent site of dif- ferential expression between Notch1 and Notch2 was the choroid plexus. A dense autoradiographic signal was found over the choroid plexus in the lateral (Fig. 10B), third and fourth ventricles (data not shown) with the Notch2 probe. In contrast to the robust Notch2 signal detected in the choroid plexus, Notch1 was not expressed at this site (Fig. 10A). Coronal sections through the hippocampal formation in the forebrain identified a second site of differential expression. A strong Notch2 signal localised over the gran- ule cells of the dentate gyrus (Fig. 10D), yet other areas of the hippocampus appeared negative. In contrast, Notch1 expression was below the level of detection in all areas of the hippocampus examined (Fig. 10C). The leptomeninges that surround the brain were strongly positive for Notch2 expression and negative for Notch1 (Fig. 10E,F). The ependymal cells lining the third (Fig. 10G,H), lateral (Fig. 10A) and fourth ventricles (not shown) in the brain were the only cell types identified that were positive for Notch1 and negative for Notch2 expression. Finally, hybridisation to coronal sections of adult rat cerebellum identified shared expression for the two Notch genes in the Purkinje cell layer (Fig. 10E,F). Examination at higher power localised the Notch2 signal to small cells surrounding the Purkinje cells and adjacent to the granule cells, indicative of Bergmann glia (data not shown).

We have identified a cDNA clone that exhibits substantial nucleotide and deduced amino acid sequence identity with both Drosophila Notch and the vertebrate Notch proteins. Based on conservation of the number, arrangement and type of amino acid sequence motifs present in the encoded pro- tein, this gene has been designated Notch2. Although amino acid sequence analysis indicates that Notch1 and Notch2 are equally related to Drosophila Notch, the following observations suggest that Notch1 may be closer in function to the Drosophila protein. First, the strong expression and pronounced developmental regulation of Notch1 during neurogenesis is similar to that reported for Drosophila Notch. In addition, the Notch1 sequence in the region of EGF-like repeats 11 and 12 is more closely related to the sequence of the 11/12 repeats in Drosophila Notch than it is to the equivalent Notch2 sequence. As noted above, these repeats in Drosophila Notch are required for the binding of Delta and Serrate. Finally, the OPA repeats described for Drosophila Notch are more highly conserved in Notch1 than in Notch2.

Northern blot and in situ hybridisation analyses have indicated that Notch1 and Notch2 have partially overlap- ping, yet distinct, patterns of expression in the developing and adult rat. Both of these genes are expressed in devel- oping tissues in which cell-cell interactions are critical. For example, the nervous and immune systems require precise cell-cell interactions for their development, and both Notch1 and Notch2 were expressed in dissimilar cell types, at different times during their development. Most notable was the strong Notch1 expression in the ventricular zone of the developing brain, which contains precursor cells that are rapidly proliferating and beginning to differentiate (Fujita, 1967; Rakic, 1988; McConnell, 1988; Price and Thurlow, 1988; Gao et al., 1991). All the cell types required to form the brain originate from these precursor cells (Jacobson, 1990) and it is these cells that display strong Notch1 expression. However, in the adult rat brain the Notch1 signal was greatly reduced and restricted to the ependymal cells lining the ventricles, indicating that Notch1 transcripts are down regulated in many cell types during brain development. In contrast, Notch2 was not expressed by the precursor cells in the ventricular zone of the embry- onic brain, yet the granule cell neurons in the dentate gyrus of the adult brain did contain transcripts from this gene. Therefore, Notch2 appears to be up-regulated in certain cells during brain development, unlike the down-regulation observed for Notch1 in similar cell types.

High Notch2 expression was detected both in the newly forming choroid plexus and the adult choroid plexus. This finding raises the possibility that Notch2 not only plays a role in establishing the choroid plexus, but that it may also function in its maintenance. Notch2 expression was con- fined to the choroid plexus epithelium, which produces the cerebrospinal fluid (Barr and Kiernan, 1988). Notch2 was also expressed by the leptomeninges surrounding the brain and by the perineurium surrounding the DRGs. It is inter- esting to note that each of these tissues contains special polarized cells that form an occluding epithelium and that function as diffusion barriers in the nervous system (Barr and Kiernan, 1988; Dyck et al., 1984).

Specific cell-cell interactions in the fetal thymus direct the correct differentiation of thymocytes (vonBoehmer, 1988; Shortman, 1990; Scollay, 1991). Notch1, but not Notch2, was expressed strongly in the thymus during this critical period of T-cell development. Consistent with the idea that Notch1 may function during T-cell differentiation, the human Notch gene, TAN-1, was identified through its association with a T-cell-specific acute lymphoblastic leukemia (Ellisen et al., 1991). In the adult spleen, this expression pattern was reversed; high levels of Notch2 tran- scripts were detected in IG-postive splenocytes that had weak Notch1 expression. It will be interesting to determine whether the expression of Notch1 or Notch2 correlates with different lymphocyte subtypes and/or their differentiated state.

During development of the kidney (Saxen et al., 1986), lung and vibrissae (Kopan and Fuchs, 1989; Hirai et al., 1992), specific interactions between mesenchymal and epithelial cells take place to direct their correct morpho- genesis. In each of these cases, Notch1 is expressed by epithelial cells at a time when they are undergoing mor- phogenesis through receiving signals from the adjacent mesenchyme. Moreover, neither Notch1 nor Notch2 was expressed by the mesenchymal cells adjacent to the devel- oping Notch-expressing epithelial cells. Although specula- tive, this spatial and temporal expression pattern is not inconsistent with a role for the Notch genes in pattern for- mation and inductive tissue interactions (Cunhan et al., 1985).

Notch mediates a diverse set of cell-cell interactions that direct cell-fate decisions in Drosophila. The Notch1 and Notch2 expression data described above support the hypothesis that these proteins play similar developmental roles in mammals. However, we have made the unantici- pated finding that the rat differs from Drosophila in having more than one Notch gene. Furthermore, the two rat Notch genes are differentially expressed, suggesting that these genes are not redundant and that the two different recep- tors they encode may respond to different ligands. The amino acid differences between Notch1 and Notch2 in sequences critical for Delta and Serrate binding lend cre- dence to the idea that Notch1 and Notch2 interact with dif- ferent ligands. The existence of two Notch genes is remi- niscent of the two Notch-related C. elegans genes, lin-12 and glp-1. The lin-12 and glp-1 proteins function in the receiving cell to direct cell fate during nematode develop- ment. Each gene controls a distinct set of cell fates, yet when expressed in the appropriate cells they have been shown to be functionally redundant during embryogenesis (Mango et al., 1991; Lambie and Kimble, 1991). It has been suggested that the glp-1 and lin-12 proteins are inter- changeable, but that differential gene regulation accounts for their divergent roles in development (Lambie and Kimble, 1991). We do not know whether Notch1 and Notch2 are interchangeable in the rat, or if the differences in expression reported for these two closely related genes means that they could regulate different cell-fate decisions during mammalian development. Functional analysis in transgenic animals, identification of ligands and the devel- opment of cell-culture systems designed to dissect the Notch signal transduction pathway, will help to define the role that Notch genes play during mammalian development.

We thank Jim Boulter for providing the random-primed rat forebrain cDNA library, Cary Lai for RNA samples, Evan Deneris for rat brain sections, Oliver Bögler for help with graphics and Anne Marie Quinn for help with computer analyses. We also thank our colleagues Oliver Bögler, Jim Boulter, Chris Kintner, Peter Maycox and John Thomas for helpful discussions and comments. Technical assistance from Sara Barth, Chris McGaugh and Danny Ortuño was greatly appreciated. Supported by grants from the NIH and the Rita Allen Foundation (G. L.) and the National Neurofi- bromatosis Foundation, Inc. (G. W.)