IE84505B1 - Human notch and delta, binding domains in toporythmic proteins, and methods based thereon - Google Patents

Description

HUMAN NQEEE AND DELTA, BINDING DOMAINS IN TOPORYTHMIC PROTEINS, AND METHODS BASED THEREON.

This invention was made in part with government support under Grant numbers GM 19093-and NS 26084 awarded by the Department of Health and Human Services. ‘The government has certain rights in the invention.

. INTRODUCTION The present invention relates to the human Delta genes and their encoded products. The also relates to sequences (termed herein sequences") within the proteins encoded by Notch and invention "adhesive toporythmic genes which mediate honotypic or heterotypic binding to sequences within proteins encoded by toporythmic genes. Such genes include but are not limited to Notch, Delta, and serrate.

. BACKGROUND OF THE INVENTION Genetic analyses in Drosophila have been extremely useful in dissecting the complexity of developmental pathways and identifying interacting loci. However, understanding the precise nature of the processes that underlie genetic interactions requires a knowledge of the biochemical properties of the protein products of the genes in question.

Null mutations in any one of the zygotic neurogenic loci --_Notcg (3), Qglgg (Q1), gggtggyiné epidermal structures. This effect is due to the misrouting of epidermal precursor cells into a neuronal pathway, and implies that neurogenic gene function‘is necessary to divert cells within the neurogenic region from a neuronal fate to an vi) 84505 epithelial fate. Studies that assessed the effects of laser ablation of specific embryonic neuroblasts in grasshoppers (Doe and Goodman 1985, Dev. Biol. 111, 206-219) have shown that cellular interactions between neuroblasts and the surrounding accessory cells serve to inhibit these accessory cells from adopting a neuroblast fate. Together, these genetic and developmental observations have led to the hypothesis that the protein products of the neurogenic loci function as components of a cellular interaction mechanism necessary for proper epidermal development (Artavanis-Tsakonas, 1988, Trends Genet. 4, 95-100). sequence analyses (Wharton et al., 1985, Cell 43, 567-581; Kidd et a1., 1936, Mol. Cell. Biol. 6, 3094-3108; Vassin et al., 1987, EMBO J. 6, 3431- 3440; Kopczynski et al., 1988, Genes Dev. 2, 1723- 1735) have shown that two of the neurogenic loci, ggtgh and Qglta, appear to encode transmembrane proteins that span the membrane a single time. The figtgn gene encodes a ~300 kd protein (we use "Notch" to denote this protein) with a large N~termina1 extracellular domain that includes 36 epidermal growth factor (EGF)-like tandem repeats followed by three other cysteine-rich repeats, designated ugtgh/lip-12 repeats (Wharton et al., 1985, cell 43, 567-581; Kidd et al., 1936, M01. Cell Biol. 6, 3094-3108; Yochem et a1., 1988, Nature 335, 547-550). pelt; encodes a ~100 kd protein (we use "Delta" to denote DLZH, the protein product of the predominant zygotic and maternal transcripts; Kopczynski et a1., 1988, Genes Dev. 2, 1723-1735) that has nine EGF—like repeats within its extracellular domain (Vassin et a1., 1987, EMBO J. 6, 3431-3440; Kopczynski et a1., 1988, Genes Dev. 2, 1723-1735). Although little is known about the functional significance of these repeats, the EGF-like motif has been found in a variety of proteins, including those involved in the blood clotting cascade (Furie and Furie, 1988, Cell 53, 505-518). In particular, this motif has been found in extracellular proteins such as the blood clotting factors IX and X (Rees et a1., 1988, EH80 J. 7, 2053-2061; Furie and Furie, 1988, Cell 53, 505-518), in other Qrgsgphila genes (Knust et a1., 1987, EMBO J. 761-766; Rothberg et a1., 1988, Cell 55, 1047-1059), and in some cell- surface receptor proteins, such as thrombomodulin (Suzuki et al., 1987, EMBO J. 6, 1891-1897) and LDL receptor (Sudhof et al., 1985, Science 228, 815-822).

A protein binding site has been mapped to the EGF repeat domain in thrombomodulin and urokinase (Kurosawa et a1., 1988, J. Biol. Chem 263, 5993-5996; Appella et al., 1937, J. Biol. chem. 252, 4437-4440).

An intriguing array of interactions between 39:93 and Qgltg mutations has been described (Vassin, et al., 1985, J. Neurogenet. 2, 291-308; Shepard et al., 1989, Genetics 122, 429-438; xu et al., 1990, Genes Dev., 4, 464-475). A number of genetic studies (summarized in Alton et al., 1989, Dev. Genet. 10, 261-272) has indicated that the gene dosages of flgtgh and Qgltg in relation to one another are crucial for normal development. A 50% reduction in the dose of Qgltg in a wild-type flgtgh background causes a broadening of the wing veins creating a "delta" at the base (Lindsley and Grell, 1968, Publication Number 627, Washington, D.C., Carnegie Institute of Washington). A similar phenotype is caused by a 50% increase in the dose of ggtgg in a wild-type Qglgg background (a "confluens" phenotype; welshons, 1965, Science 150, 1122-1129). This Delta phenotype is partially suppressed by a reduction in the notch dosage. Recent work in our laboratories has shown that lethal interactions between alleles that correlate with alterations in the EGF—like repeats in Notch can be rescued by reducing the dose of Qgltg (Xu et al., 1990, Genes Dev. 4, 464-475). Xu et al. (1990, Genes Dev. 4, 464-475) found that null mutations at either Qglgg or mam suppress lethal interactions between heterozygous combinations of certain flgtgh alleles, known as the Abrupgeg (fig) mutations. A3 alleles are associated with missense mutations within the BGP-like repeats of the Notch extracellular domain (Kelley et al., 1987, Cell 51, 539~S48; Bartley et al., 1987, EMBO J. 6, 3407-3417).

Notch is expressed on axonal processes during the outgrowth of embryonic neurons (Johansen et al., 1989, J. Cell Biol. 109, 2427-2440; Kidd et al., 1989, Genes Dev. 3, 1113-1129).

A study has shown that certain 55 alleles of flgtgh can severely alter axon pathfinding during sensory neural outgrowth in the imaginal discs, although it is not yet known whether aberrant flgtgh expression in the axon itself or the epithelium along which it grows is responsible for this defect (Palka et al., 1990, Development 109, 167-175).

,The present specification describes n"¢13°t1d° sequences of the human figtgh and Qgltg genes, and amino acid sequences of their encoded proteins, as well as fragments thereof containing an antigenic determinant or which are functionally active. The ‘specification also describes fragments (termed herein "adhesive fragments"), and the sequences thereof, of the proteins ("toporythmic proteins") encoded by toporythmic genes which mediate homotypic or heterotypic binding to toporythmic proteins.

Toporythmic genes, as used herein, refers to the genes ggtgh, Qglta, and serrate, as well as other members of the Qglta/serrate family which may be identified, e.g., by the methods described in Section 5.3, infra.

Analogs and derivatives of the adhesive fragments which retain binding activity are also described Antibodies to human Notch and to adhesive fragments are additionally described. - .

In aspecific embodiment of the invention, the adhesive fragment of Notch is that fragment —comprising the Notch sequence most homologous to _Drosophila Notch EGF-like repeats 11 and 12; in SPeCifi¢ eiamplesz the adhesive fragment of Delta mediating heterotypic binding is that fragment comprising the sequence most homologous to Qgosophjla Delta amino acids 1-230; the adhesive fragment of Delta mediating homotypic binding is that fragment comprising the sequence most homologous to Drosophila Delta amino acids 32-230; and the adhesive fragment of serrate is that fragment comprising the sequence most homologous to prgsgghila serrate amino acids 85-283 or 79-282. .1. QEELEITIONS As used herein, the following terms shall have the meanings indicated: AA = V amino acid EGF = epidermal growth factor ELR = EGF-like (homologous) repeat IC = intracellular PCR = polymerase chain reaction As used herein, underscoring the name of a gene shall indicate the gene, in contrast to its encoded protein product which is indicated by the name_ of the gene in the absence of any underscoring. For example, "Notch" shall mean the flotch gene, whereas "Notch" shall indicate the protein product of the flotgn gene.

. R ION 0 IG 8 Figure 1. Expression Constructs and Experimental Design for Examining Notch-Delta Interactions. 82 cells at log phase growth were transiently transfected with one of the three constructs shown. Notch encoded by the MG11a minigene (a CDNA/genomic chimeric construct: CDNA-derived sequences are represented by stippling, genomically derived sequences by diagonal-hatching (Ramos et al., 1989, Genetics 123, 337-348)) was expressed following insertion into the metallothionein promoter vector pRmHa-3 (Bunch et al., 1988, Nucl. Acids Res. 16, 1043-1061). Delta encoded by the D11 cDNA (Kopczynski et al., 1988, Genes Dev. 2, 1723-1735) was expressed after insertion into the same vector. The extracellular Notch (ECN1) variant was derived from a genomic cosmid containing the complete figggh locus (Ramos et a1., 1989, Genetics 123, 337-348) by deleting the coding sequence for amino acids 1790~2625 from the intracellular domain (denoted by 6; Wharton et al., 1985, Cell 43, 567-581), leaving 25 membrane- proximal residues from the wi1d—type sequence fused to a novel 59 amino acid tail (see Experimental Procedures, Section 6.1, ingra). This construct was expressed under control of the notch promoter region.

For constructs involving the metallothionein vector, expression was induced with CuSo,fo11owing transfection. Cells were then mixed, incubated under aggregation conditions, and scored for their ability to aggregate using specific antisera and immunofluorescence microscopy to visualize expressing cells. MT, metallothionein promoter; ATG, translation start site; TM, transmembrane domain; 3' N, Ngtgh gene polyadenylation signal; 3' Adh, polyadenylation signal from Adh gene; 5' N, Notch gene promoter region.

Figure 2. Expression of Notch and Delta in Cultured Cells. (A) Lysates of nontransfected (S2) and figtgh-transfected (N) cells induced with 0.7 mM CuSQ.for 12-16 hr were prepared for sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), run on 3%-15% gradient gels, and blotted to nitrocellulose. Notch was visualized using a monoclonal antibody (MAb C17.9C6) against the intracellular domain of Notch. Multiple bands below the major band at 300 kd may represent degradation products of Notch. (B) Lysates of nontransfected (S2) and Qgltg-transfected (D1) cells visualized with a monoclonal antibody (MAb 201) against Delta. A single band of ‘I05 Rd is detected. In both cases, there is no detectable endogenous Notch or Delta in the S2 cell line nor are there cross—reactive species. In each lane, 10 ul of sample (prepared as described in Experimental Procedures) was loaded.

Figure 3. S2 Cells That Express Notch and Delta Form Aggregates. In all panels, Notch is shown in green and Delta in red.

(A) A single Notch‘ cell. Note the prominent intracellular stain, including vesicular structures as well as an obviously unstained nucleus.

Bright-field micrograph of same field, showing specificity of antibody staining.

A single Delta‘ cell. primarily at the cell surface.

Bright-field micrograph of same field. staining is (C) Aggregate of Delta’ cells from a 24 hr aggregation experiment. Note against that staining is primarily at the cell surface.

(D)-(F) An aggregate of Notch‘ and Delta* cells formed from a 1:1 mixture of singly transfected cell populations that was allowed to aggregate overnight at room temperature. (D) shows Notch‘ cells in this aggregate; (B) shows Delta’ cells; and (F) is a double exposure showing both cell types. Bands of Notch and Delta are prominent at points of contact In (F), these bands appear yellow because of the coincidence of green and red at these points. The apparently doubly stained single cell C) is actually two cells (one on top of the other), one expressing Notch and the other.De1ta.

(G) and (H) Pseudocolor confocal micrographs Note that in (G) extensions (arrows) formed by at least two Delta‘ cells completely encircle the Notch‘ cell in the (H) shows an aggregate formed between Notch* and Delta’ cells (arrows). of Notch‘-Delta‘ cell aggregates. center of the aggregate. from a 2 hr aggregation experiment performed at 4°C.

.Intense bands of Notch are apparent within regions of contact with Delta‘ cells.

(I) An aggregate composed of Delta* cells and cells that express only the extracellular domain of Notch (ECN1 construct). scale bar = 10 pm.

Figure 4. Notch and Delta are Associated in Cotransfected Cells. Staining for Notch is shown in the left column (A, C, and E) and that for Delta is shown in the right column (B, D, and F).

(A) and (B) 82 cell cotransfected with both Ngtgh and Qglta constructs. In general, there was a good correlation between Notch and Delta localization at the cell surface (arrows).

(C) and (D) Cotransfected cells were exposed to polyclonal anti-Notch antiserum (a 1:250 dilution of each anti-extracellular domain antiserum) for 1 hr at room temperature before fixation and staining with specific antisera. Note punctate staining of Notch and Delta and the correlation of their respective staining (arrows).

'(E) and (F) Cells cotransfected with the Aextracellular Notch (ECN1) and Qgltg constructs, induced, and then patched using anti-Notch polyclonal antisera. There was a close correlation between ECN1 and Delta staining at the surface as observed for full-length Notch.

Figure 5.

Delta and Notch are Associated in Lysates from Transfected S2 and Drosophila Embryonic Cells. experiments, Delta was precipitated from NP- 40/deoxycholate lysates using a polyclonal anti-Delta rat antiserum precipitated with fixed staph A cells, and proteins in the precipitated fraction were visualized on western blots (for details, see Experimental Procedures). Lanes 1, 2, 3, and 5: Notch visualized with Mhb C17.9C6; Lanes 4 and 6: Delta visualized using Mhb 201.

In (A), lanes 1 and 2 are controls for these Scale bar = 10 um.

Coimmunoprecipitation Shows that In all experiments. Lane 1 shows a polyclonal anti-Delta immunoprecipitation from cells that express Notch alone visualized for Notch. No Notch was detectable in this sample, indicating that the polyclonal anti- Delta does not cross-react with Notch. Lane 2 shows with Staph A without initial treatment with anti-Delta antiserum and visualized for.Notch, demonstrating that Notch is not precipitated nonspecifically by the Staph A or secondary antibody. Lane 3 shows protein precipitated with anti-Delta antiserum visualized for Delta (D1), and lane 4 shows the same sample visualized for Notch (N). Lane 4 shows that Notch coprecipitates with immunoprecipitated Delta. Note that Notch appears as a doublet as is typical for Notch in immunoprecipitates.

(B) shows the same experiment using embryonic lysates rather than transfected cell lysates. Lane 5 shows protein precipitated with anti- Delta antiserum visualized for Delta (D1), and lane 6 shows the same sample visualized for Notch (N). These lanes demonstrate that Notch and Delta are stably lysates. Bands (in all lanes) are from Staph A (SA) and the heavy (H) and light (L) chains.

Notch Expression Constructs and of the Delta/serrate Binding log phase growth were transiently series of expression constructs associated in embryo below the Delta band anti-Delta antiserum Figure 6. the Deletion Mapping Domain. S2 cells in transfected with the shown; the drawings represent the predicted protein products of the various Ngtgh deletion mutants created. All expression constructs were derived from construct #1 pMtNMg. Transiently transfected cells were mixed with Delta expressing cells from the stably transformed line L497 or with transiently transfected serrate expressing cells, induced with cusou incubated under aggregation conditions and then scored for their ability to aggregate using specific antisera and immunofluorescence microscopy.

Aggregates were defined as clusters of four or more cells containing both Notch and Delta/serrate expressing cells. The values.given for % Aggregation refer to the percentage of all Notch expressing cells found in such clusters either with Delta (Dl) (left column) or with serrate (Ser) (right column). The various Notch deletion constructs are represented diagrammatically with splice lines indicating the ligation junctions. Each EGF repeat is denoted as a stippled rectangular box and numbers of the EGF repeats on either side of a ligation junction are noted. At the ligation junctions, partial EGF repeats produced by the various deletions are denoted by open boxes and closed brackets (for example see #23 AC1a+EGF(10-12)). Constructs #3-13 represent the claI deletion series. As diagrammed, four of the claI sites, in repeats 7, 9, 17 and 26, break the repeat in the middle, immediately after the third cysteine (denoted by open box repeats; see Figure 7 for further clarification), while the fifth and most 3' site breaks neatly between EGP repeats 30 and 31 (denoted by closed box repeat 31; again see Figure 7). In construct #15 split, EGF repeat 14 which carries the split point mutation, is drawn as a striped box. In construct #33 Ac1a+XEGF(10-13), the xgngpgg Notch derived EGF repeats are distinguished from Qggsgphilg repeats by a different pattern of shading. SP, signal peptide; EGF, epidermal growth factor repeat; N, flgtgh/lip-12 repeat; TH, transmembrane domain; cdclo, ggglo/ankyrin repeats; PA, putative nucleotide binding consensus sequence; opa, polyglutamine stretch termed opa; D1, Delta; Ser, serrate.

Figure 7. Detailed Structure of Notch Deletion Constructs #19-24: Both EGF Repeats 11 and 12 are Required for Notch-Delta Aggregation. EGF repeats -13 are diagrammed at the top showing the regular spacing of the six cysteine residues (C). PCR products generated for these constructs (names and numbers as given in Figure 6) are represented by the heavy black lines and the exact endpoints are noted relative to the various EGF repeats. Ability to aggregate with Delta is recorded as (+) or (-) for each construct. The PCR fragments either break the EGF repeats in the middle, just after the third cysteine in the same place as four out of the five Clal sites, or exactly in between two repeats in the same place as the most C-terminal C1aI site.

Figure 8. Comparison of Amino Acid Sequence of EGF Repeats 11 and 12 from Qrgsgphilg and xggopus Notch. The amino acid sequence of EGP repeats 11 and 12 of Drosophila Notch (Wharton et al., 1935, Cell 43:567-581; Kidd et a1., 1986, Mol. Cell Biol. 6:3094- 3108) is aligned with that of the same two EGF repeats from xenogus Notch (coffman et al., 1990, Science 249:1438-1441). Identical amino acids are boxed. The six conserved cysteine residues of each EGF repeat and the Ca‘* binding consensus residues (Rees et al., , EH80 J. 7:20$3-2061) are marked with an asterisk (*). The leucine to proline change found in the xegopgs PCR clone that failed to aggregate is noted underneath.

Figure 9. study. Schematic diagrams of the Delta variants defined in Table IV are shown. Extracellular, amino- proximal terminus is to the left in each case. 5, signal peptide; "EGF", EGF-like motifs; M, membrane- Constructs Employed in this spanning helix; H, stop-transfer sequence; solid lines, other Delta sequences; hatched lines, Arrowheads indicate sites of Sea, Scal; Nae, Nael; neuroglian sequences. translatable linker insertions.

Bam, Bamul; Bgl, BglII; ELR, EGF-like repeat; Bst, BstEII; Dde, Ddel; Stu, Stul; NG1~NG5, Delta- neuroglian chimeras.

Figure 9A. Dependence of Aggregation on Input DNA Amounts. ‘A, Heterotypic aggregation observed using 82 cell populations transiently - 13 _ transfected, respectively, with varied amounts of pHTDl1 DNA (2, 4, 10 or 20 ug/plate) that were subsequently incubated under aggregation conditions with 82 cell populations transiently transfected with a constant amount of pMtNHg DNA (20 ug/plate). Data presented are mean fraction (%) of Delta cells in aggregates of four or more cells 1 standard error for each input DNA amount (N = 3 replicates, except 2 pg and 10 pg inputs for which N = 2).

Delta-expressing cells were counted for each replicate. B, Homotypic aggregation observed using 32 cell populations transiently transfected, respectively, with varied amounts of pMTD11 DNA (2, 4, or 20 pg/plate) that were subsequently incubated A minimum of 100 under aggregation conditions. Data presented are mean fraction (t) of Delta cells in aggregates of four or more cells : standard error for each input DNA amount (N = 3 replicates). A minimum of 500 Delta-expressing cells were counted for each replicate.

Figure 10. Delta-serrate Amino-Terminal Sequence Alignment. Residues are numbered on the basis of conceptual translation of Delta (D1, upper sequence (SEQ ID No:3); beginning at amino acid 24, ending at amino acid 226) and figgrate (Ser, lower sequence (SEQ ID NO:4); beginning at amino acid 85, ending at amino acid 283) coding sequences. Vertical lines between the two sequences indicates residues that are identical within the Delta and serrate sequences, as aligned. Dots represent gaps in the alignment. Boxes enclose cysteine residues within the aligned regions. N1, amino-proximal domain 1; N2, amino-proximal domain 2; N3, amino-proximal domain 3.

Translatable insertions associated with STU B [replacement of Delta amino acid 132 (A) with GKIFP] and NAB B [insertion of RKIF between Delta amino acid and amino acid 198] constructs, respectively, are depicted above the wild type Delta sequence.

Potential Geometries of Delta- Notch Interactions. A, Potential register of Delta (left) and Notch (right) molecules interacting between B, Potential register of Figure 11. opposing plasma membranes.

Delta (left) and Notch (right) molecules interacting within the same plasma membranes. ELR, EGF-like repeat; open boxes, EGF-like repeats; dotted boxes, LNR repeats; solid boxes, membrane-spanning helices.

Delta amino~termina1 domain and Delta and Notch intracellular domains represented by ovals.

Figure 12. Potential Geometries of Delta- Delta Interactions. A and B, Potential register of Delta molecules interacting between opposing plasma 8, Potential register of Delta molecules Open membranes. interacting within the same plasma membranes. boxes, EGF-like repeats; solid boxes, membrane- spanning helices. Delta amino-terminal extracellular and intracellular domains represented by ovals.

Figure 13. Primary Nucleotide Sequence of the Qglta cDNA D11 (SEQ ID NO:5) and Delta amino acid sequence (SEQ ID N026) The DNA sequence of the 5'-3' strand of the D11 CDNA is shown, which contains a number of corrections in comparison to that presented (1988, 1723-1735).

Figure 14. Primary Nucleotide Sequence of the Neuroglian CDNA 1B7A-250 (SEQ ID NO:7). This is the DNA sequence of a portion of the 5'-3' strand of the 1B7A-250 cDNA (A.J. Bieber, pers. comm.; Hortsch et al., 1990, Neuron 4, 697-709). Nucleotide 2890 corresponds to the first nucleotide of an isoleucine codon that encodes amino acid 952 of the conceptually translated neuroglian-long form protein. in Kopczynksi et a1. Genes Dev. 2, -15.- Figure 15. Nucleic Acid Sequence Homologies Between gggrggg and Delta. A portion of the nrgsgnhila figrrgte nucleotide sequence (SEQ ID No:8), with the encoded serrate protein sequence (SEQ ID NO:9) written below, (Fleming et al., 1990, Genes & Dev. 4, 2188-2201 at 2193-94) is shown. The four regions showing high sequence homology with the Drosgphila Delta sequence are numbered above the line and indicated by brackets. The total region of homology spans nucleotide numbers 627 through 1290 of the serrate nucleotide sequence (numbering as in Figure 4 of Fleming et a1., 1990, Genes 8 Dev. 4, 2188-2201). _ Figure 16. Primers used for PCR in the cloning of Human flgtgh. The sequence of three primers used for PCR to amplify DNA in a human fetal brain CDNA library are shown. The three primers, cdcl (SEQ ID N0:10), cdc2 (SEQ ID NO:l1), and cdc3 (SEQ ID N0:12), were designed to amplify either a 200 bp or a 400 bp fragment as primer pairs cdcl/cdcz or cdcl/cdc3, respectively. I: inosine.

Figure 17. schematic Diagram of Human ugggn Clones. A schematic diagram of human flgggh is shown.

Heavy bold-face lines below the diagram show that portion of the ugggh sequence contained in each of the four cDNA clones. The location of the primers used in PCR, and their orientation, are indicated by arrows.

Figure 18. Human ugtgh Sequences Aligned with Qrosopnjlg E9393 sequence. Numbered vertical lines correspond to Drosgpgila ygggh coordinates.

Horizontal lines below each map show where clones lie relative to stretches of sequence (thick horizontal lines).V , Figure 19. Nucleotide Sequences of Human ugggh Contained in Plasmid cDNA Clone hN2k. Figure - 15 _ Ai The DNA sequence (SEQ ID NO:13) of a portion of the human Ngtgn insert is shown, starting at the EcoRI site at the 3' end, and proceeding in the 3' to 5' direction. Figure 193: The DNA sequence (SEQ ID N0:14) of a portion of the human Ngtgh insert is shown, starting at the EcoRI site at the 5' end, and proceeding in the 5' to 3' direction. Figure 19C: The DNA sequence (SEQ ID N0:15) of a portion of the human Ngggh insert is shown, starting 3' of the sequence shown in Figure 193, and proceeding in the S’ to 3' direction. subject to confirmation by determination of The sequences shown are tentative, overlapping sequences.

Figure 20. Nucleotide Sequences of Human Notch Contained in Plasmid CDNA clone hN3k. 20A: The DNA sequence (SEQ ID NO:16) of a portion of the human Notch insert is shown, starting at the EcoRI site at the 3' end, and proceeding in the 3' to 5' direction. Figure 203: The DNA sequence (SEQ ID N0:17) of a portion of the human Ngggh insert is shown, starting at the EcoRI site at the 5' end, and proceeding in the 5' to 3' direction. Figure.20c: The DNA sequence (SEQ ID NO:18) of a portion of the human flgtgh insert is shown, starting 3' of the and proceeding in the 5‘ Figure sequence shown in Figure 205, to 3' direction. Figure 20D: The DNA sequence (SEQ ID NO:19) of a portion of the human Ngtgh insert is shown, starting 5' of the sequence shown in Figure ZQA, and proceeding in the 3' to 5' direction. The sequences shown are tentative, subject to confirmation by determination of overlapping sequences.

Figure 21. Nucleotide Sequences of Human Egtgh Contained in Plasmid cDNA clone hN4k. Figure 21A: The DNA sequence (SEQ ID NO:20) of a portion of the human Notch insert is shown, starting at the EcoRI ~ 17 — site at the 5‘ end, and proceeding in the 5‘ to 3' direction. Figure 21B: The DNA sequence (SEQ ID N0:21) of a portion of the human ugtgh insert is shown, starting near the 3' end, and proceeding in the 3' to 5' direction. The sequences shown are tentative, subject to confirmation by determination of overlapping sequences.

Figure 22. 39593 Contained in Plasmid CDNA Clone hN5k. 22A: The DNA sequence (SEQ ID N0:22) of a portion of the human Notch insert is shown, starting at the EcoRI site at the 5' end, and proceeding in the 5' to 3' direction. Figure 22B: The DNA sequence (SEQ ID N0:23) of a portion of the human Notch insert is shown, starting near the 3' end, and proceeding in the 3° to 5' direction. Figure 22C: The DNA sequence (SEQ ID N0:24) of a portion of the human Notch insert is shown, starting 3' of the sequence shown in Figure 22A, and proceeding in the 5' to 3‘ direction. Figure 22D: The DNA sequence (SEQ ID N0:25) of a portion of the human Notch insert is shown, starting 5' of the sequence shown in Figure 228, and proceeding in the 3' to 5' direction. The sequences shown are tentative, subject to confirmation by determination of overlapping sequences.

Figure 23. DNA (sso ID uo:31) and Amino Acid (SEQ ID NO:34) Sequences of Human figggh Contained in Plasmid CDNA Clone hN3k.

Figure 24. DNA (SEQ ID N0:33) and Amino Acid (SEQ ID NO:34) Sequences of Human figtgh Contained in Plasmid cDNh Clone hNSk.

' Comparison of hN5k with other Schematic representation Nucleotide Sequences of Human Figure Figure 25.

Notch Homologs. Figure 25A. of Drosopniig Notch. Indicated are the signal sequence (signal), the 36 EGF-like repeats, the three flgggh/L13-12 repeats, the transmembrane domain (TM), the six CDC1o repeats, the OPA repeat, and the PEST (proline, glutamic acid, serine, threonine)-rich region. Figure 258. Alignment of the deduced amino acid sequence of hN5k with sequences of other Notch homologs. Amino acids are numbered on the left side.

The cdclo and PEST-rich regions are both boxed, and individual cdclo repeats are marked. Amino acids which are identical in three or more sequences are highlighted. The primers used to clone hNSk are indicated below the sequences from which they were designed. The nuclear localization sequence (NLS), casein kinase II (CKII), and cdcz kinase (cdcz) sites of the putative CcN motif of the vertebrate Notch honologs are boxed. The possible bipartite nuclear targeting sequence (BNTS) and proximal phosphorylation sites of Qrosophila Notch are also boxed.

S. QEEALLED DESCRIPTION OF Qflfi IEEENTION The present invention provides nucleotide- sequences of the human flgggh genes, and amino acid sequences of their encoded proteins.

The nucleic acid and amino acid sequences and antibodies thereto of the invention can be used for the detection and quantitation of mRNA for human Notch molecules, to study d expression thereof, to produce human Notch sequences, in the study and manipulation ‘of differentiation processes.

For clarity of disclosure, and not by way of limitation, the detailed description below. will be divided into the following sub-sections: (i) »Identification of and the sequences of toporythmic protein domains that mediate binding to toporythmic protein domains; The cloning and sequencing of human E2222 and Qsltaz A Identification of additional members of the Delta/Serrate family; i The expression of toporythmic genes; Identification and purification of the expressed gene product; and A Generation of antibodies to toporythmic proteins and adhesive sequences (iv) (V) (Vi) thereof.

IDENTIFICATION OF AND THE SEQUENCES OF TOPORYTHIC PROTEIN DOMAINS THAT MEDIATE BIEQING E9 TOPORY2flfl1C EROTEIN DOMAINS .1.

‘Toporythmic protein fragments; and analogs or derivatives thereof, which mediate homotypic or heterotypic binding (and thus are termed herein "adhesive"), and nucleic acid sequences relating to the foregoing are described below.

In a specific example, the adhesive fragment Of NOCCQ which is part of the present invention, is that comprising the portion of Notch most homologous to ELR 11 and 12. 1-e.. amino_acid numbers 447 through 527 (SEQ 19 N031) Of the Qggsoghila Notch sequence (see Figure 8).

In 3"°ther SPBCifi0 °X8mP1e,the adhesive fragment of Delta mediating homotypic binding is that comprising the portion of Delta most homologous to about amino acid numbers 32-230 of the Qrgsgghilg Delta_sequence (SEQ ID NO:6). In yet another specific examplemthe adhesive fragment of Delta mediating binding to notch is that comprising the portion of Delta most homologous to about amino acid numbers 1-230 of the Dggggghilg Delta sequence (SEQ ID N026). In a speci£ic_examp1e relating to an adhesive fragment of serrate, such fragment is that comprising the portion of serrate most homologous to about amino acid numbers 85-283 or 79-282 of the Drosognila serrate sequence (see Figure 10 (SEQ ID NO:4), and Figure 15 (see ID uo:9)).

The nucleic acid sequences encoding toporythmic adhesive domains can be isolated from porcine, bovine, feline, avian, equine, or canine, as well as primate sources and any other species in which homologs of known toporythmic genes [including but not limited to the following genes (with the publication of sequences in parentheses): ugggg (Wharton et al., 1935, Cell 43, 567-581), ggigg (Vassin et a1., 1937, EKBO J. 6, 3431-3440; Kopczynski et al;, 1988, Genes Dev. 2, 1723-1735; note corrections to the Kopczynski et al. sequence found in Figure 13 hereof (SEQ ID N0:5 and SEQ ID NO:6)) and ggrgate (Fleming et a1., 1990, Genes & Dev. 4, 2188-2201)] can be identified. Such sequences can be altered by substitutions, additions or deletions that provide for functionally equivalent ,(adhesive) molecules. Due to the degeneracy of nucleotide coding sequences, other DNA sequences which encode substantially the same amino acid sequence as the adhesive sequences may be used in the practice of the present invention. These include but are not limited to nucleotide sequences comprising all or portions of the notch, Qelta, or serrate genes which -21.. are altered by the substitution of different codons that encode a functionally equivalent amino acid residue within the sequence, thus producing a silent change. Likewise, the adhesive protein fragments or derivatives thereof, include, but are A not limited to, those containing, as a primary amino acid sequence, all or part of the amino acid sequence of the adhesive domains including altered sequences in which functionally equivalent amino acid residues are substituted for residues within the sequence resulting in a silent change. For example, one or more amino acid residues within the sequence can be substituted by another amino acid of a similar polarity which acts as a functional equivalent, resulting in a silent alteration. substitutes for an amino acid within the sequence may be selected from other members of the class to which the amino acid belongs. For example, the nonpolar (hydrophobic) amino acids include alanine, leucine, isoleucine, valine, proline, phenylalanine, tryptophan and methionine. The polar neutral amino acids include glycine, serine, threonine, cysteine, tyrosine, asparagine, and glutamine. The positively charged (basic) amino acids include arginine, lysine and histidine. The negatively charged (acidic) amino acids include aspartic acid and glutanic acid.

Adhesive fragments of toporythmic proteins and potential derivatives, analogs or peptides related to adhesive toporythmic protein sequences, can be tested for the desired binding activity e.g., by the in vitro aggregation assays described in the examples herein. Adhesive derivatives or adhesive analogs of adhesive fragments of toporythmic proteins include but are not limited to those peptides which are substantially homologous to the adhesive fragments, ..‘22 _ whose encoding nucleic acid is capable of hybridizing to the nucleic acid sequence encoding the adhesive fragments, and which peptides and peptide analogs have positive binding activity e.g., as tested in yitrg by an aggregation assay such as described in the examples sections igfiga. Such derivatives and analogs are envisioned and within the scope of the present invention; The adhesive-protein related derivatives, analogs, and peptides can be produced by various methods known in the art. The manipulations which result in their production can occur at the gene or protein level. For example, the cloned adhesive protein-encoding gene sequence can be modified by any of numerous strategies known in the art (Naniatis, T., 1990, Molecular Cloning, A Laboratory Manual, 2d ed., Cold spring Harbor Laboratory, Cold Spring Harbor, New York). The sequence can be cleaved at appropriate sites with restriction endonuclease(s), followed by further enzymatic modification if desired, isolated, and ligated in yigrg. In the production of the gene encoding a derivative, analog, or peptide related to an adhesive domain, care should be taken to ensure that the modified gene remains within the same translational reading frame as the adhesive protein, uninterrupted by translational stop signals, in the gene region where the desired adhesive activity is encoded.

Additionally, the adhesive-encoding nucleic acid sequence can be mutated in vitro or in giyg, to create and/or destroy translation, initiation, and/or ‘termination sequences, or to create variations in coding regions and/or form new restriction endonuclease sites or destroy preexisting ones, to facilitate further in gitrg modification. Any technique for mutagenesis known in the art can be used, including but not limited to, in vitro site- directed mutagenesis (Hutchinson, C., et al., 1978, J.

Biol. Chem 253, 6551), use of new linkers (Pharmacia), etc.

Hanipulations of the adhesive sequence may Any of numerous chymotrypsin, papain, V8 protease, NaBHg acetylation, formylation, oxidation, reduction; metabolic synthesis in the presence of tunicamycin; etc.

In addition, analogs and peptides related to adhesive fragments can be chemically synthesized. For example, a peptide corresponding to a portion of a toporythmic protein which mediates the desired aggregation activity in xitrg can be synthesized by use of a peptide synthesizer. A Another specific example described herein relates to fragments or derivatives of a Delta protein which have the ability to bind to a second Delta protein or fragment or derivative thereof, but do not bind to Notch. Such binding or lack thereof can be assayed in yitrg as described in Section 8. example, but not limitation, such a Delta derivative is that containing an insertion of the tetrapeptide By way of Arg-Lys-Ile-Phe between Delta residues 198 and 199 of the Dgosophila protein.

THE CLONING AND SEQUENCING OF HUMAN NOTCH AND DELTA The invention further relates to the amino acid sequences of human Notch and .2. fragments and derivatives thereof which comprise an antigenic determinant (i.e., can be recognized by an antibody) or which are functionally active, as well as nucleic acid sequences encoding the foregoing.

Also described herein are the amino acid and nucleic acid sequences of Delta and corresponding fragments and derivates thereof.‘ "Functionally active" material as used herein refers to that material displaying one or more known functional activities associated with the full-length (wild-type) protein product, e.g., in the case of Notch, binding to Delta, binding to serrate, antigenicity (binding to an anti-Notch antibody), etc.

In specific embodiments, the invention provides fragments of a human Notch protein consisting of at least 40 amino acids, or of at least 77 amino In other embodiments, the proteins of the invention comprise or consist essentially of the acids. intracellular domain, transmembrane region, extracellular domain, cdclo region, flgtgh/li -12 repeats, or the EGF-homologous repeats, or any combination of the foregoing, of a human Notch protein. Fragments, or proteins comprising fragments, lacking some or all of the EGF-homologous repeats of human Notch are also provided.

In other specific embodiments, the invention is further directed to the nucleotide sequences and subsequences of human flgtgh Consisting of at least 25 nucleotides, at least 50 nucleotides, or at least 121 nucleotides. Nucleic acids encoding the proteins and protein fragments described above are also provided, as well as nucleic acids complementary to and capable of hybridizing to such nucleic acids.

In one embodiment, such a complementary sequence may be complementary to a human Notch CDNA sequence of at least 25 nucleotides, or of at least 121 nucleotides.

In a preferred aspect, the invention relates to cDNA sequences encoding human Notch or a portion thereof.

In a specific embodiment, the invention relates to the nucleotide sequence of the human flgtgh gene or CDNA, in particular, comprising those sequences depicted in Figures 19, 20, 21 and/or 22 (SEQ ID No:13 through No:25), or contained in plasmids hN3k, hN4k, or hN5k (see Section 9, infra), and the encoded flgtgh protein sequences. As is readily apparent, as used herein, a "nucleic acid encoding a fragment or portion of a Notch protein" shall be construed as referring to a nucleic acid encoding only the recited fragment or portion of the Notch protein and not other portions of the Notch protein.

A human expression library is constructed by methods known in the art. For example, human mRNA is isolated, CDNA is made and ligated into an expression vector (e;g., a bacteriophage derivative) such that it is capable of being expressed by the host cell into which it is then introduced. Various screening assays can then be used to select for the expressed human Delta product. In one embodiment, selection can be carried out on the basis of positive binding to the adhesive domain of human Notch, (i.e., that portion of -26.. human Notch most homologous to Drosophila ELR 11 and 12 (SEQ ID NO:1)). In an alternative embodiment, anti-Delta antibodies can be used for selection.

In another preferred aspect, PCR is used to amplify the desired sequence in the library, prior to selection. For example, oligonucleotide primers representing part of the adhesive domains encoded by a homologue of the desired gene can be used as primers in PCR.

The above-methods are not meant to limit the following general description of methods by which clones of human Notch and Delta may be obtained.

Any human cell can potentially serve as the nucleic acid source for the molecular cloning of the Notch and Delta gene. The DNA may be obtained by standard procedures known in the art from cloned DNA (gygy, a DNA "library"), by chemical synthesis, by cDNA cloning, or by the cloning of genomic DNA, or fragments thereof, purified from the desired human cell. (See, for example Maniatis et al., 1982, Molecular Cloning, A Laboratory Manual, Cold Spring Harbor Laboratory, Cold spring Harbor, New York; Glover, D.M. (ed.), 1985, DNA Cloning: A Practical Approach, MRL Press, Ltd., Oxford, U.K. Vol. I, II.) Clones derived from genomic DNA may contain regulatory and intron DNA regions in addition to coding regions; clones derived from cDNA will contain only exon sequences. Whatever the source, the gene should be molecularly cloned into a suitable vector for propagation of the gene.

In the molecular cloning of the gene from genomic DNA, DNA fragments are generated, some of which will encode the desired gene. The DNA may be cleaved at specific sites using various restriction enzymes. Alternatively, one may use DNAse in the presence of manganese to fragment the DNA, or the DNA can be physically sheared, as for example, by sonication. The linear DNA fragments can then be separated according to size by standard techniques, including but not limited to, agarose and polyacrylamide gel electrophoresis and column chromatography. once the DNA fragments are generated, identification of the specific DNA fragment containing the desired gene may be accomplished in a number of ways. For example, if an amount of a portion of a flgtgh or Qglga (of any species) gene or its specific RNA, or a fragment thereof e.g., the adhesive domain, is available and can be purified and labeled, the generated DNA fragments may be screened by nucleic acid hybridization to the labeled probe (Benton, W. and Davis, R., 1977, Science 196, 180; Grunstein, M.

And Hogness, D., 1975, Proc. Natl. Acad. Sci. U.S.A. 72, 3961). Those DNA fragments with substantial homology to the probe will hybridize. It is also possible to identify the appropriate fragment by restriction enzyme digestion(s) and comparison of fragment sizes with those expected according to a known restriction map if such is available. Further selection can be carried out on the basis of the properties of the gene. Alternatively, the presence of the gene may be detected by assays based on the physical, chemical, or immunological properties of its expressed product. For example, CDNA clones, or DNA clones which hybrid-select the proper mRNAs, can be selected which produce a protein that, egg‘, has similar or identical electrophoretic migration, isolectric focusing behavior, proteolytic digestion maps, in yitgg aggregation activity ("adhesiveness") or antigenic properties as known for Notch or Delta.

If an antibody to Notch or Delta is available, the Notch or Delta protein may be identified by binding of labeled antibody to the putatively Notch or Delta synthesizing clones, in an ELISA (enzyme-linked immunosorbent assay)-type procedure. .

The neeeh or Delee gene can also be identified by mRNA selection by nucleic acid hybridization followed by in giere translation. In this procedure, fragments are used to isolate complementary mRNAs by hybridization. Such DNA fragments may represent available, purified Neeeh or Delee DNA of another species (e.g., Dresophila).

Immunoprecipitation analysis or functional assays (eege, aggregation ability ig yigrg; see examples igfre) of the in yiegg translation products of the isolated products of the isolated mRNAs identifies the mRNA and, therefore, the complementary DNA fragments that contain the desired sequences. In addition, specific mRNAs may be selected by adsorption of polysomes isolated from cells to immobilized antibodies specifically directed against Notch or Delta protein. A radiolabelled ueteh or Delte cDNA can be synthesized using the selected mRNA (from the adsorbed polysomes) as a template. The radiolabelled mRNA or cDNA may then be used as a probe to identify the neeeh or Delee DNA fragments from among other genomic DNA fragments.

Alternatives to isolating the Neeen or genomic DNA include, but are not limited to, chemically synthesizing the gene sequence itself from a known sequence or making cDNA to the mRNA which encodes the fleteh or Delte gene. For example, RNA for CDNA cloning of the ueeeh or Delge gene can be isolated from cells which express Notch or Delta.

Delta other methods are possible and within the scope of the invention.

The identified and isolated gene can then be inserted into an appropriate cloning vector. A large number of vector-host systems known in the art may be used. Possible vectors include, but are not limited to, plasmids or modified viruses, but the vector system must be compatible with the host cell used.

Such vectors include, but are not limited to, bacteriophages such as lambda derivatives, or plasmids such as PBR322 or pUC plasmid derivatives. The insertion into a cloning vector can, for example, be accomplished by ligating the DNA fragment into a cloning vector which has complementary cohesive termini. However, if the complementary restriction sites used to fragment the DNA are not present in the cloning vector, the ends of the DNA molecules may be enzymatically modified. Alternatively, any site desired may be produced by ligating nucleotide sequences (linkers) onto the DNA termini; these ligated linkers may comprise specific chemically synthesized oligonucleotides encoding restriction endonuclease recognition sequences. In an alternative method, the cleaved vector and flgtgh or Qglta gene may be modified by homopolymeric tailing. Recombinant molecules can be introduced into host cells via transformation, transfection, infection, electroporation, etc., so that many copies of the gene sequence are generated.

In an alternative method, the desired gene may be identified and isolated after insertion into a suitable cloning vector in a "shot gun" approach.

Enrichment for the desired gene, for example, by size fractionization, can be done before insertion into the cloning vector. " specific examples transformation of host cells with recombinant DNA molecules that incorporate the isolated ugtgh or Delta gene, cDNA, or synthesized DNA sequence enables generation of multiple copies of the gene. Thus, the gene may be obtained in large quantities by growing transformants, isolating the recombinant DNA molecules from the transformants and, when necessary, retrieving the inserted gene from the isolated recombinant DNA.

The human flotch sequences provided by the instant invention include those nucleotide sequences encoding substantially the same amino acid sequences as found in human otch, and those encoded amino acid sequences with functionally equivalent amino acids, all as described gupra in Section 5.1 for adhesive portions of toporythmic proteins.

IDENTIFICATION OF ADDITIONAL MEMBERS 0 T LT S A rational search for additional members of the Qelta/ggrzgtg gene family may be carried out using SCSC ‘ an approach that takes advantage of the existence of the conserved segments of strong homology between serrate and Qglta (see Figure 10, SEQ ID NO:3 and N0:4). For example, additional members of this gene family may be identified by selecting, from among a diversity of nucleic acid sequences, those sequences that are homologous to both Sgrratg and Qgltg (see Figure 13 (SEQ ID N0:5), and Figure 15 (SEQ ID N0:8)), and further identifying, from among the selected sequences, those that also contain nucleic acid sequences which are non-homologous to serrate and Qgltg. The term "non-homologous" may be construed to mean a region which contains at least about 6 contiguous nucleotides in which at least about two nucleotides differ from gerrate and Qelta sequence.

For example, a preferred method is as follows.

Corresponding to two conserved segments between Delta and gerratg, Delta AA 63-73 and Delta AA 195-206 (see Figure 13, SEQ ID No:6), sets of degenerate . oligonucleotide probes of about 10-20 nucleotides may be synthesized, representing all of the possible coding sequences for the amino acids found in either Delta and serrate for about three to seven contiguous codons. In another embodiment, oligonucleotides may be obtained corresponding to parts of the four highly conserved regions between Delta and serrate shown in Figure 15 (SEQ ID No:8 and NO:9). i.e., that represented by serrate AA 124-134, 149+158, 214-219, and 250-259. The synthetic oligonucleotides may be utilized as primers to amplify by PCR sequences from a source (RNA or DNA) of potential interest. (PCR can be carried out, e.g., by use of a Perkin-Elmer Cetus Taq polymerase (Gene Amp")). This A DNA from any a polypeptide By carrying out thermal cycler and might include mRNA eukaryotic species or cDNA or genomic that could express closely related to serrate and Delta. the PCR reactions, it may be possible or gene product sharing the above—noted segments of If one to detect a gene conserved sequence between serrate and Delta. chooses to synthesize several different degenerate primers, it may still be possible to carry out a complete search with a reasonably small number of PCR reactions. It is also possible to vary the stringency of hybridization conditions used in priming the PCR reactions, to allow for greater or lesser degrees of nucleotide sequence similarity between the unknown gene and serrate or Qgltg. If a segment of a -32.. previously unknown member of the serrate/Qgltg gene family is amplified successfully, that segment may be molecularly cloned and sequenced, and utilized as a probe to isolate a complete cDNA or genomic clone.

This, in turn, will permit the determination of the unknown gene's complete nucleotide sequence, the analysis of its expression, and the production of its protein product for functional analysis. In this fashion, additional genes encoding "adhesive" proteins may be identified. serrate and biological function. comprising portions of any one or more members of the toporythmic gene family (e.g., notch) may be constructed. .4. mag EXEBESSION OF TOPORYTHMIC GENES The nucleotide sequence coding for an adhesive fragment of a toporythmic protein Delta), or an adhesive (preferably, notch, gerrate, or analog or derivative thereof, or human Notch or Delta or bacteria transformed with bacteriophage, DNA, The expression elements plasmid DNA, or cosmid DNA. of vectors vary in their strengths and specificities.

Depending on the host-vector system utilized, any one of a number of suitable transcription and translation elements may be used. In a specific °XamP1e: the adhesive portion of the ugtgh gene, e.g., that encoding EGF-like repeats 11 and 12, is expressed. In another example, the adhesive portion of the Qelta gene: e.g., that encoding amino acids l~230, is expressed. In other specific examples,the human 59393 °" human D9122 gtne is expressed, or a sequence °"°°d1"9 3 functionally active portion of human Notch or Delta._ In yet another example. the adhesive.

Any of the methods previously described for the insertion of DNA fragments into a vector may be used to construct expression vectors containing a chimeric gene consisting of appropriate -34.- transcriptional/translational control signals and the protein coding sequences. These methods may include in gitrg recombinant DNA and synthetic techniques and in 2129 recombinants (genetic recombination).

Expression of nucleic acid sequence encoding a toporythmic protein or peptide fragment may be regulated by a second nucleic acid sequence so that the toporythmic protein or peptide is expressed in a host transformed with the recombinant DNA molecule.

For example, expression of a toporythmic protein may be controlled by any promoter/enhancer element known in the art. Promoters which may be used to control toporythmic gene expression include, but are not limited to, the sv4o early promoter region (Bernoist and Chambon, 1981, Nature 290, 304-310), the promoter contained in the 3' long terminal repeat of Rous sarcoma virus (Yamamoto, et al., 1980, cell 22, 787- 797), the herpes thymidine kinase promoter (Wagner et al., 1981, Proc. Natl. Acad. Sci. U.S.A. 78, 1441- 1445), the regulatory sequences of the metallothionein gene (Brinster et a1., 1982, Nature 296, 39-42); prokaryotic expression vectors such as the B-lactamase promoter (Villa-Kamaroff, et al., 1978, Proc. Natl.

Acad. Sci. U.S.A. 75, 3727-3731), or the tag promoter (DeBoer, et a1., 1983, Proc. Natl. Acad. Sci. U.s.A. 80, 21-25); see also "Useful proteins from recombinant bacteria" in Scientific American, 1980, 242, 74-94; plant expression vectors comprising the nopaline synthetase promoter region (Herrera-Estrella et al., Nature 303, 209-213) or the cauliflower mosaic virus $ RNA promoter (Gardner, et a1., 1981, Nucl. Acids Res. 9, 2871), and the promoter of the photosynthetic enzyme ribulose.biphosphate carboxylase (Herrera- Estrella et al., 1984, Nature 310, 115-120); promoter elements from yeast or other fungi such as the Gal 4 promoter, the ADC (alcohol dehydrogenase) promoter, PGK (phosphoglycerol kinase) promoter, alkaline phosphatase promoter, and the following animal transcriptional control regions, which exhibit tissue specificity and have been utilized in transgenic animals: elastase I gene control region which is active in pancreatic acinar cells (Swift et al., 1984, Cell 38, 639-646; Ornitz et al., 1986, Cold Spring Harbor Symp. Quant. Biol. 50, 399-409; MacDonald, 1987, Hepatology 7, 425-515); insulin gene control region which is active in pancreatic beta cells (Hanahan, 1985, Nature 315, 115-122), immunoglobulin gene control region which is active in lymphoid cells (Grosschedl et al., 1984, Cell 38, 647-658; Adames et al., 1985, Nature 318, 533-538; Alexander et al., 1987, M01. Cell. Biol. 7, 1436-1444), mouse mammary tumor virus control region which is active in testicular, breast, lymphoid and mast cells (Leder et al., 1986, Cell 45, 485-495), albumin gene control region which is active in liver (Pinkert et al., 1987, Genes and Devel. 1, 268-276), alpha-fetoprotein gene control region which is active in liver (Krumlauf et al., 1985, Mol. Cell. Biol. 5, 1639-1648; Hammer et al., 1987, science 235, 53-58; alpha 1-antitrypsin gene control region which is active in the liver (Kelsey et al., 1987, Genes and Devel. 1, 161-171), beta-globin gene control region which is active in myeloid cells (Mogram at al., 1935, Nature 315, 333- 340; Kollias et al., 1986, cell 46, 89-94; myelin basic protein gene control region which is active in oligodendrocyte cells in the brain (Readhead et al., 1987, Cell 48, 703-712); myosin light chain-2 gene control region which is active in skeletal muscle (Sani, 1985, Nature 314, 283-286), and gonadctropic releasing hormone gene control region which is active - 36 _ in the hypothalamus (Mason et al., 1986, Science 234, 1372-1373).

Expression vectors containing toporythmic gene inserts can be identified by three general approaches: (a) nucleic acid hybridization, (b) presence or absence of "marker" gene functions, and (c) expression of inserted sequences. In the first approach, the presence of a foreign gene inserted in an expression vector can be detected by nucleic acid hybridization using probes comprising sequences that are homologous to an inserted toporythmic gene. In the second approach, the recombinant vector/host system can be identified and selected based upon the presence or absence of certain "marker" gene functions (gégg, thymidine kinase activity, resistance to antibiotics, transformation phenotype, occlusion body formation in baculovirus, etc.) caused by the insertion of foreign genes in the vector. For example, if the toporythmic gene is inserted within the marker gene sequence of the vector, recombinants containing the toporythmic insert can be identified by the absence of the marker gene function. In the third approach, recombinant expression vectors can be identified by assaying the foreign gene product expressed by the recombinant. Such assays can be based, for example, on the physical or functional properties of the toporythmic gene product in yitgg assay systems, e.g., aggregation (adhesive) ability (see Sections 6-8, infra).

Once a particular recombinant DNA molecule is identified and isolated, several methods known in the art may be used to propagate it. once a suitable host system and growth conditions are established, recombinant expression vectors can be propagated and prepared in quantity. As previously explained, the -37.. expression vectors which can be used include, but are not limited to, the following vectors or their derivatives: human or animal viruses such as vaccinia virus or adenovirus; insect viruses such as baculovirus; yeast vectors} bacteriophage vectors (g;g&, lambda), and plasmid and cosmid DNA vectors, to name but a few.

In other specific examples,the adhesive toporythmic protein, fragment, analog, or derivative may be expressed as a fusion, or chimeric protein product (comprising the protein, fragment, analog, or derivative joined to a heterologous protein sequence).

Such a chimeric product can be made by ligating the appropriate nucleic acid sequences encoding the desired amino acid sequences to each other by methods known in the art, in the proper coding frame, and expressing the chimeric product by methods commonly known in the art. Alternatively, such a chimeric product may be made by protein synthetic techniques, e.g., by use of a peptide synthesizer.

Both CDNA and genomic sequences can be cloned and expressed.

In other examples, a human Notch CDNA sequence may be chromosomally integrated and expressed. Homologous recombination procedures known in the art may be used.

IDENTIFICATION AND PURIFICATION Of THE EXPRESSED GENE PRODUCT Once a recombinant which expresses the .4.1. toporythmic gene sequence is identified, the gene product may be analyzed. This can be achieved by assays based on the physical or functional properties of the product, including radioactive labelling of the product followed by analysis by gel electrophoresis. once the toporythmic protein is identified, it may be isolated and purified by standard methods including chromatography (eygy, ion exchange, affinity, centrifugation, differential solubility, or by any other standard technique for the purification of and sizing column chromatography), proteins. The functional properties may be evaluated using any suitable assay, including, but not limited to, aggregation assays (see Sections 6-8). .5. GENERATION OF ANTIBODIES TO TOPORYTHMIC PROTEINS AND ADHESIVE SEQUENCES THEREOF Toporythmic protein fragments or analogs or derivatives thereof -39.. which mediate homotypic or heterotypic binding, or human Notch or human Delta proteins or fragments thereof, may be used as an immunogen to generate anti- toporythmic protein antibodies. Such antibodies can be polyclonal or monoclonal. In a specific embodiment, antibodies specific to EGF-like repeats 11 and 12 of Notch may be prepared. In other embodiments, antibodies reactive with the "adhesive portion" of Delta can be generated. one example of such antibodies may prevent aggregation in an in 11:19 assay. In another embodiment, antibodies specific to human Notch are produced.

Various procedures known in the art may be used for the production of polyclonal antibodies to a toporythmic protein or peptide. In a particular embodiment, rabbit polyclonal antibodies to an epitope of the human Notch protein encoded by a sequence depicted in Figure 19, 20, 21 or 22 (SEQ ID N0:l3 through N0:25), or a subsequence can be obtained. For the production of host animals can be immunized by thereof, antibody, various injection with the native toporythmic protein; or a synthetic version, or fragment thereof, including but not limited to rabbits, mice, rats, etc. various adjuvants may be used to increase the immunological response, depending on the host species, and including but not limited to Freund's (complete and incomplete), mineral gels such as aluminum hydroxide, surface active substances such as lysolecithin, pluronic polyols, polyanions, peptides, oil emulsions, keyhold limpet hemocyanins, dinitrophenol, and potentially useful human adjuvants such as BCG (bacille Calmette-Guerin) and corynebacterium parvum.

For preparation of monoclonal antibodies directed toward a toporythmic protein sequence, any technique which provides for the production of antibody molecules by continuous cell lines in culture may be used. For example, the hybridoma technique originally developed by Kohler and Milstein (1975, Nature 256, 495-497), as well as the trioma technique, the human B—cell hybridoma technique (Kozbor et al., 1983, Immunology Today 4, 72), and the BBV—hybridoma technique to produce human monoclonal antibodies (Cole et al., 1985, in Monoclonal Antibodies and cancer Therapy, Alan R. Liss, Inc., pp. 77-96).

Antibody fragments which contain the idiotype of the molecule can be generated by known techniques. For example, such fragments include but are not limited to: the F(ab'), fragment which can be produced by pepsin digestion of the antibody molecule; the Fab‘ fragments which can be generated by reducing the disulfide bridges of the F(ab'), fragment, and the Fab fragments which can be generated by treating the antibody molecule with papain and a reducing agent.

In the production of antibodies, screening for the desired antibody can be accomplished by techniques known in the art, e.g. ELISA (enzyme-linked immunosorbent assay). For example, to select antibodies which recognize the adhesive domain of a toporythmic protein, one may assay generated hybridomas for a product which binds to'a protein fragment containing such domain. For selection of an antibody specific to human Notch, one can select on the basis of positive binding to human Notch and a lack of binding to Drosophila Notch.

The foregoing antibodies can be used in methods known in the art relating to the localization and activity of the protein sequences of the invention. For example, various immunoassays known in the art can be used, including but not limited to competitive and non-competitive assay systems using techniques such as radioimmunoassays, ELISA (enzyme linked immunosorbent assay), "sandwich" immunoassays, precipitin reactions, gel diffusion precipitin reactions, immunodiffusion assays, agglutination assays, fluorescent immunoassays, protein A immunoassays, and immunoelectrophoresis assays, to name but a few. .6. V O GENTS I ND C -E SS N C L The invention also provides methods for delivery of agents into Notch-expressing cells. As discussed in Section 8 i fra, upon binding to a Notch protein on the surface of a Notch-expressing cell, Delta protein appears to be taken up into the Notch- expressing cell. The invention thus provides for delivery of agents into a Notch-expressing cell by agent to a Delta protein or an or derivative thereof capable of and exposing a Notch-expressing cell to the conjugate, such that the conjugate is taken up by the cell. The conjugated agent can be, but is not limited to, a label or a biologically active agent.

The biologically active agent can be a therapeutic agent, a toxin, a chemotherapeutic, a growth factor, an enzyme, a hormone, a drug, a nucleic acid, (e.g., antisense DNA or RNA), etc. In one embodiment, the label can be an imaging agent, including but not limited to heavy metal contrast agents for X-ray imaging, magnetic resonance imaging agents, and radioactive nuclides (i.e., isotopes) for radio- imaging. In a preferred aspect, the agent is conjugated to a site in the amino terminal half of the conjugation of an adhesive fragment binding to Notch, Delta molecule.

The Delta~agent conjugate can be delivered to the Notch-expressing cell by exposing the Notch- expressing cell to cells expressing the Delta-agent conjugate or exposing the Notch-expressing cell to the Delta-agent conjugate in a solution, suspension, or Where delivery is in 3139, the Delta- agent conjugate can be formulated in a pharmaceutically acceptable carrier or excipient, to comprise a pharmaceutical composition. The pharmaceutically acceptable carrier can comprise saline, phosphate buffered saline, etc. The Delta- agent conjugate can be formulated as a liquid, tablet, pill, powder, in a slow-release form, in a liposome, etc., and can be administered orally, intravenously, intramuscularly, subcutaneously, intraperitoneally, to other carrier. name but a few routes, with the preferred choice readily made based on the knowledge of one skilled in the art.

. MOLECULAR INTERACTIONS BETWEEN THE PROTEIN PRODUCTS OF THE NEUROGENIC LOCI NOTCH AND D O EGF-HO O OUS E E IN DROSOP To examine the possibility of intermolecular association between the products of the flgtgh and Qelta genes, we studied the effects of their expression on aggregation in Drosophila Schneider's 2 (S2) cells (Fehon et al., 1990, Cell 61, 523-534). we present herein direct evidence of intermolecular interactions between Notch and Delta, and describe an assay system that will be used in dissecting the components of this interaction. We show that normally nonadhesive Drosophila S2 cultured cells that express Notch bind specifically to cells that express Delta, and that this aggregation is calcium dependent.

Furthermore, while cells that express Notch do not bind to one another, cells that express Delta do bind to one another, suggesting that Notch and Delta can compete for binding to Delta at the cell surface. We also present evidence indicating that Notch and Delta form detergent-soluble complexes both in cultured cells and embryonic cells, suggesting that Notch and Delta interact directly at the molecular level in vitro and in viva. Our analyses suggest that Notch and Delta proteins interact at the cell surface via their extracellular domains.

For the Notch expression construct, the 6 Kb Hpal fragment from the 5' end of the Notct coding A sequence in MgIIa (Ramos et a1., 1989, Genetics 123, 337-348) was blunt-end ligated into the metallothionein promoter vector pRmHa—3 (Bunch, et a1., 1988, Nucl. Acids Res. 16, 1043-1061) after the vector had been cut with EcoRI and the ends were filled with the Klenow fragment of DNA polymerase I (Maniatis et al., 1982, Molecular Cloning: A Laboratory Manual (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory)). A single transformant, incorrectly oriented, was isolated. DNA from this transformant was then digested with SacI, and a resulting 3 kb fragment was isolated that contained the 5' end of the Notch coding sequence fused to the polylinker from pRmHa-3. This fragmgnt was then ligated into the SacI site of pRmHa-3 in the correct orientation. DNA from this construct was digested with KpnI and XbaI to remove must of the Notch sequence and all of the Adh polyadenylation signal in pRmHa-3 and ligated to an 11 kb KpnI—XbaI fragment from HgIIa containing the rest of the Notch coding sequence and 3' sequences necessary for A V CDNA (Kopczynski et al., -44... polyadenylation. In the resulting construct, designated pMtNMg, the metallothionein promoter in pRmHa-3 is fused to Notch sequences starting 20 nucleotides upstream of the translation start site.

For the extracellular Egtgn construct (ECN1), the CosP479BE flgtgn cosmid (Ramos et al., 1989, Genetics 123, 337-348), which contains all figtgh genomic sequences necessary for normal flgtgh function in 2139, was partially digested with AatII. Fragment ends were made blunt using the exonuclease activity of T4 DNA polymerase (Maniatis et al., 1982, Molecular Cloning: A Laboratory Manual (Cold Spring Harbor, New York: Cold spring Harbor LaboratorY)). and the fragments were then redigested completely with Stul.

The resulting fragments were separated in a low melting temperature agarose gel (SeaPlaque, FMC BioProducts), and the largest fragment was excised.

This fragment was then blunt-end ligated to itself.

This resulted in an internal deletion of the Hgtgn coding sequences from amino acid 1790 to 2625 inclusive (Wharton et al., 1985, Cell 43, 567-581), and a predicted frameshift that produces a novel 59 amino acid carboxyl terminus. (The ligated junction of this construct has not been checked by sequencing.) For the Qgltg expression construct, the D11 , Genes Dev. 2, 1723- 1735), which includes the complete coding capacity for Delta, was inserted into the EcoRI site of pRmHa-3.

This construct was called pMTDl1. .1.2. AQIIBODY B3EEfiBATlOfl Hybridoma cell line C17.9C6 was obtained from a mouse immunized with a fusion protein based on a 2.1 kb SalI-HindIII fragment that includes coding sequences for most of the intracellular domain of -.45- Notch (amino acids 1791-2504; Wharton et al., 1985, Cell 43, 567-581). The fragment was subcloned into pUR289 (Ruther and Muller-Hill, 1983, EMBO J. 2, 1791- 1794), and then transferred into the pATH 1 expression vector (Dieckmann and Tzagoloff, 1985, J. Biol. chem. 260, 1513-1520) as a Bg1II-HindIII fragment. Soluble fusion protein was expressed, precipitated by 25% (NHJZSOH resuspended in 6 M urea, and purified by preparative isoelectric focusing using a Rotofor (Bio- Rad) (for details, see Fehon, 1989, Rotofor Review No. 7, Bulletin 1518, Richmond, California: Bio-Rad Laboratories).

Mouse polyclonal antisera were raised against the extracellular domain of Notch using four BstYl fragments of 0.8 kb (amino acids 237-501: "Wharton et al., 1985, Cell 43, 567-581), 1.1 kb (amino acids 501-868), 0.99 kb (amino acids 868-1200), and 1.4 kb (amino acids 1465-1935) length, which spanned from the fifth EGF-like repeat across the transnembrane domain, singly inserted in-frame into the appropriate pGEX expression vector (Smith and Johnson, 1988, Gene 67, 31-40). Fusion proteins were purified on glutathione-agarose beads (SIGMA). Mouse and rat antisera were precipitated with 50% (NHQ,sO4 and resuspended in PBS (150 mu Nacl, 14 mM Nafiuxu, 6 mid Naa,1>o,) with o.o2% Nam. flybridoma cell line 201 was obtained from a mouse immunized with a fusion protein based on a 0.54 kb ClaI fragment that includes coding sequences from the extracellular domain of Delta (Kopczynski et a1., 1988, Genes Dev. 2, 1723-1735) subcloned into the ClaI site within the Lag; gene of pUR 288 (Ruther and Muller-Hill, 1983, EMBO J. 2, 1791-1794). This fragment includes sequences extending from the fourth through the ninth EGF-like repeats in Delta (amino -46.. acids 350-529). Fusion protein was prepared by isolation of inclusion bodies (Gilmer et a1., 1982, Proc. Natl. Acad. Sci. USA 79, 2152-2156); inclusion bodies were solubilized in urea (Carroll and Laughon, 1987, in DNA Cloning, Volume III, D.M. Glover, ed.

(Oxford: IRL Press), pp. 89-111) before use in immunization.

Rat polyclonal antisera were obtained following immunization with antigen derived from the same fusion protein construct. In this case, fusion protein was prepared by lysis of IPTG-induced cells in SDS-Laemmli buffer (Carroll and Laughon, 1987, in DNA Cloning, Volume III, D.H. Glover, ed. (Oxford: IRL Press), pp. 89-111), separation of proteins by SD8- PAGE, excision of the appropriate band from the gel, and electroelution of antigen from the gel slice for use in immunization (Harlow and Lane, 1988, Antibodies: A Laboratory Manual (Cold Spring Harbor, New York: Cold spring Harbor Laboratory)).

Embryol. Exp. Morph. 27, 353-365) was grown in H3 . medium (prepared by Hazleton Co.) supplemented with .5 mg/ml Bacto-Peptone (Difco), 1 mg/ml TC Yeastolate (Difco), 11% heat-inactivated fetal calf serum (FCS) (Hyclone), and 100 U/ml penicillin-100 ug/ml streptomycin-0.25 pg/ml fungizone (Hazleton). growing in log phase at ~2 x 10‘ cells/ml were transfected with 20 pg of DNA-calcium phosphate coprecipitate in 1 ml per 5 ml of culture as previously described (Wigler et al., 1979, Proc. Natl.

Acad. Sci. USA 78, 1373-1376), with the exception that BES buffer (SIGMA) was used in place of HEPES buffer (Chen and Okayama, 1987, Mol. Cell. Biol. 7, 2745- Cells ). conical centrifuge tubes, pelleted in a clinical After 16-18 hr, cells were transferred to centrifuge at full speed for 30 seconds, rinsed once with 1/4 volume of fresh complete medium, resuspended in their original volume of complete medium, and returned to the original flask. were then allowed to recover for 24 hr before Transfected cells induction. .1.4. AGGRBGATIOfl 5SSAYS Expression of the flgtgg and Qgltg metallothionein constructs was induced by the addition of CuSQ,to 0.7 mu. Cells transfected with the ECN1 construct were treated similarly. Two types of aggregation assays were used. In the first assay, a total of 3 ml of cells (5-10 x II? cells/ml) was placed in a 25 ml Erlenmeyer flask and rotated at 40-50 rpm on a rotary shaker for 24-48 hr at room temperature.

For these experiments, cells were mixed 1-4 hr after induction began and induction was continued throughout the aggregation period. In the second assay, ~0.6 ml of cells were placed in a 0.6 ml Eppendorf tube (leaving a small bubble) after an overnight induction (12-16 hr) at room temperature and rocked gently for 1-2 hr at 4°C. The antibody inhibition and Ca’* dependence experiments were performed using the latter assay. For Ca"'dependence experiments, cells were first collected and rinsed in balanced saline solution (BS5) with 11% FCS (BS8-PCS; FCS was dialyzed against 0.9% Nacl, 5mM Tris [pH 7.5)) or in Ca"'free BSS—FCS containing 10 mm EGTA (Snow et al., 1989, Cell 59, 313-323) and then resuspended in the same medium at the original volume. For the antibody inhibition experiments, flgtgn-transfected cells were collected and rinsed in M3 medium and then treated before _ 48 - aggregation in M3 medium for 1 hr at 4°C with a 1:250 dilution of immune or preimmune sera from each of the four mice immunized with fusion proteins containing segments from the extracellular domain of Notch (see Antibody Preparation above).

Cells were collected by centrifugation (3000 rpm for 20 seconds in an Eppendorf microcentrifuge) and fixed in 0.6 ml Eppendorf tubes with 0.5 ml of freshly made 2% paraformaldehyde in PBS for 10 min at room temperature. After fixing, cells were collected by centrifugation, rinsed twice in PBS, and stained for 1 hr in primary antibody in PBS with 0.1% saponin (SIGMA) and 1% normal goat serum (Pocono Rabbit Farm, Canadensis, PA). Monoclonal antibody supernatants were diluted 1:10 and mouse or rat sera were diluted 1:1000 for this step. Cells were then rinsed once in PBS and stained for 1 hr in specific secondary antibodies (doublevlabeling grade goat anti-mouse and goat anti-rat, Jackson Immunoresearch) in PBS—saponin- normal goat serum. After this incubation, cells were rinsed twice in PBS and mounted on slides in 90% glycerol, 10% 1 H Tris (pH 8.0), and 0.5% n-propyl gallate. Cells were viewed under epifluorescence on a Leitz orthoplan 2 microscope.

Confocal micrographs were taken using the Bio-Rad NRC 500 system connected to a zeiss Axiovert compound microscope. Images were collected using the BHS and GHS filter sets, aligned using the ALIGN program, and merged using MERGE. Fluorescent bleed- through from the green into the red channel was reduced using the BLEED program (all software provided by Bio-Rad). Photographs were obtained directly from the computer monitor using Kodak Ektar 125 film. .1.6. CELL LYSATES, IMMUNOPRECIPITATIONS, AND WESTERN BLOTS Nondenaturing detergent lysates of tissue culture and wild-type Canton-S embryos were prepared on ice in ~10 cell vol of lysis buffer (300 mM Nacl, so mu Tris [pH 3.0], 0.5% NP-40, 0.5% deoxycholate, 1 mM CaCl,, 1 mu MgCl,) with 1 mn phenylmethysulfonyl fluoride (PMSF) and diisopropyl fluorophosphate diluted 1:2500 as protease inhibitors. sequentially triturated using 186, 216, and 256 needles attached to 1 cc tuberculin syringes and then centrifuged at full speed in a microfuge 10 min at 4°C Immunoprecipitation was Lysates were to remove insoluble material. performed by adding -1 pg of antibody (1-2 pl of polyclonal antiserum) to 250-500 pl of cell lysate and incubating for 1 hr at 4°C with agitation. To this mixture, 15 ug of goat anti-mouse antibodies (Jackson Immunoresearch; these antibodies recognize both mouse and rat IgG) were added and allowed to incubate for 1 hr at 4°C with agitation. This was followed by the addition of 100 ul of fixed staphylococcus aureus (staph A) bacteria (Zysorbin, Zymed; resuspended according to manufacturer's instructions), which had been collected, washed five times in lysis buffer, and incubated for another hour. Staph A-antibody complexes were then pelleted by centrifugation and washed three times in lysis buffer followed by two 15 min washes in lysis buffer. After being transferred to a new tube, precipitated material was suspended in 50 ml of SDS-PAGE sample buffer, boiled immediately for 10 min, run on 38-15% gradient gels, blotted to nitrocellulose, and detected using monoclonal antibodies and HRP-conjugated goat antiémouse secondary antibodies as previously described (Johansen et al.. 1989, J. Cell Biol. 109, 2427-2440). For total cellular protein samples used on Western blots -50..

(Figure 2), cells were collected by centrifugation, lysed in 10 cell vol of sample buffer that contained 1 mM PMSF, and boiled immediately.

Delta, we examined the behavior of cells expressing these proteins on their surfaces using an aggregation assay. We chose the 32 cell line (Schneider, 1972, J.

Embryol. Exp. Morph. 27, 353-365) for these studies First, these cells are .2.1. for several reasons. relatively nonadhesive, grow in suspension, and have been used previously in a similar assay to study fasciclin III function (snow at al., 1989, Cell 59, 313-323). second, they are readily transfectable, and an inducible metallothionein promoter vector that has been designed for expression of exogenous genes in Drosophila cultured cells is available (Bunch et al., 1988, Nucl. Acids Res. 16, 1043-1061). Third, S2 cells express an aberrant ugtgn message and no detectable Notch_due to a rearrangement of the 5' end of the flgtgh coding sequence (see below). These cells also express no detectable Delta (see below).

Schematic drawings of the constructs used are shown in Figure 1 (see Experimental Procedures, Section 6.1, for details). To express Notch in cultured cells, the notch minigene MG1la, described in Ramos et al. (1989, Genetics 123, 337-348) was inserted into the metallothionein promoter vector pRmHa-3 (Bunch et al., 1988, Nucl. Acids Res. 16, 1043-1061). The Delta expression construct was made by inserting D11 cDNA, which contains the entire coding sequence for Delta from the major embryonic Delta transcript (5.42; Kopczynski et a1., 1988, Genes -1735), into the same vector. A third construct, designated ECN1 for "extracellular Notch 1", contains the 5' flgtgh promoter region and 3' ugggg polyadenylation signal together with coding capacity for the extracellular and transmembrane regions of the Dev. 2, Ngtch gene from genomic sequences, but lacks coding sequences for 835 amino acids of the ~1000 amino acid intracellular domain. In addition, due to a predicted 'frameshift, the remaining 78 carboxy-terminal amino acid residues are replaced by a novel 59 amino acid ‘carboxyterminal tail (see Experimental Procedures).

For all of the experiments described in this paper, expression constructs were transfected into 52 cells and expressed transiently rather than in stable transformants. Expressing cells typically composed 1%-5% of the total cell population, as judged by immunofluorescent staining (data not shown). A Western blot of proteins expressed after transfection' is shown in Figure 2. Nontransfected cells do not‘ express detectable levels of Notch or Delta. However, after transfection, proteins of the predicted apparent molecular weights are readily detectable using monoclonal antibodies specific for each of these proteins, respectively. In the case of Notch, multiple bands were apparent in transfected cells below the -300 kd full-length product. We do not yet know whether these bands represent degradation of Notch during sample preparation or perhaps synthesis or processing intermediates of Notch that are present within cells, but we consistently detect them in samples from transfected cells and from embryos. addition, we performed immunofluorescent staining of live transfected cells with antibodies specific for the extracellular domains of each protein to test for cell surface expression of these proteins. In each case we found surface staining as expected for a surface antigen. Taken together, these results clearly show that the ugtgh and Qglta constructs support expression of proteins of the expected sizes and subcellular localization.

To test the prediction that Notch and Delta interact, we designed a simple aggregation assay to detect these interactions between proteins expressed on the surface of S2 cells. We reasoned that if Notch and Delta are able to form stable heterotypic complexes at the cell surface, then cells that express these proteins might bind to one another and form aggregates under appropriate conditions. A similar assay system has recently been described for the fasciclin III protein (snow et al., 1989, Cell 59, 313-323). cells in log phase growth were separately transfected with either the Notch or Delta metallothionein promoter construct. After induction with CuSO,, transfected cells were mixed in equal numbers and allowed to aggregate overnight at room temperature (for details, see Experimental Procedures, Section 6.1). Alternatively, in some experiments intended to reduce metabolic activity, cells were mixed gently at 4°C for 1-2 hr. To determine whether aggregates had formed, cells were processed for immunofluorescence microscopy using antibodies specific for each gene product and differently labeled fluorescent secondary antibodies. As previously mentioned, expressing cells usually constituted less than 5% of the total cell population because we used transient rather than stable transformants. The remaining cells either did not express a given protein .—, .w.~--~..,s—_....¢swv.- «-53. or expressed at levels too low for detection by immunofluorescence microscopy. As controls, we performed aggregations with only a single type of transfected cell.

Figure 3 shows representative photomicrographs from aggregation experiments, and Table I presents the results in numerical form. As is apparent from Figure 3C and Table I, while Notch- expressing (Notch*) cells alone do not form aggregates in our assa , Delta-expressing (De1ta*) cells do.

Y .mucmBaummxm cowuawouwwa H: vw aouu mum wumu uonuo aaa .mucmEHummxm cowuaooumuo us wv Eouu «van .v ucmawuwmxm mo cowuuwumcauu wfiow Eouu ucoawuomxm wasp Baum uammo .mmumvuumUm +muawoI+:ouoz cw waamo +ouH0o you name cowudmmumvm .mwuwmmumma +muHmau+couoz cu maaoo +aua0o van +nouoz anon uou sumo coavnmwumwm vocwneoo .>mmmm cowvoawuomm 0:» cm EOMUNQSOCH umuuu maaoo aouucou .>ammn cowuwomuwmm wnu ca COMHMQSOEM uvonuws wxmmau cowuomumsmuu EOHH xauowuwu cwxuu waamo aouucoo .u:..nmmo.umx0 wuaoo Cuuawo "mcwmwonmxvtnouoz .+nouoz .Uouoom 0H¢3 Amumumsao aamo no muawo wawcmmv mafia: Have wcwmmouaxo ooa umamn uflvs-M5 _ M an .mm:am> Haw mom .uaaoo mcwmmwumxo wuoa uo udou no mumumzao ma nucnuwc mwummmuum< . 2 8 3 S 3 o o _.w ucofiummxm Nm 3 3 8. .. n s ...m ucoawuamxm am 2. 3 u ow m m w ucmauummxm 2. 2. .2. .. mm a o n ucwawummxu on S 3 S. u H u N ucofiumaxm 3 on mm 3 3 o c H ucwawuomxm .mH~oo .a~aoo .n~aoo .ua~oo .u.3¢U an 3300 2 ~:nu0>o Uounovumot aouucoo vouaoouood aouucoo 3%i‘| dwfluo huuoz hmezummooa ZH WAAMU Eaqun oz: .=u.aoz mo muofizmommm Hlflflfifl The tendency for Delta‘ cells to aggregate was apparent even in nonaggregated control samples (Table I), where cell clusters of 4-8 cells that probably arose from adherence between mitotic sister cells commonly occurred. However, clusters were more common after incubation under aggregation conditions (e.g., 19% of Delta‘ cells in aggregates before incubation vs. 37% of Delta‘ cells in aggregates after .incubation; Experiment 1 in Table I), indicating that Delta‘ cells are able to form stable contacts with one another in this assay. It is important to note that while nonstaining cells constituted over 90% of the cells in our transient transfections, we never found them within aggregates. on rare occasions, nonstaining cells were found at the edge of an aggregate. Due to the common occurrence of weakly staining cells at the edges of aggregates, it is likely that these apparently nonexpressing cells were transfected but expressed levels of Delta insufficient to be detected by immunofluorescence.

In remarkable contrast to control experiments with Notch‘ cells alone, aggregation of mixtures of Notch‘ and Delta‘ cells resulted in the formation of clusters of up to 20 or more cells (Figures 3D-3H, Table I). As Table I shows, the fraction of expressing cells found in clusters of four or more stained cells after 24 hr of aggregation ranged from 32%-54% in mixtures of Notch‘ and Delta‘ cells. This range was similar to that seen for Delta‘ cells alone (37%-40%) but very different from that for Notch‘ cells alone (only 0%—5%). Although a few clusters that consisted only of Delta‘ cells were found, Notch‘ cells were never found in clusters of greater than four to five cells unless Delta‘ cells were also present. Again, all cells within these -56.- clusters expressed either Notch or Delta, even though transfected cells composed only a small fraction of the total cell population. At 48 hr (Table I, experiments 5 and 6), the degree of aggregation appeared higher (63%-71%), suggesting that aggregation had not yet reached a maximum after 24 hr under these conditions. Also, cells cotransfected with figtgh and pelt; constructs (so that all transfected cells express both proteins) aggregated in a similar fashion under the same experimental conditions.

These results indicate that the aggregation observed in these experiments requires the expression of Notch and Delta and is not due to the fortuitous expression of another interacting protein in nontransfected S2 cells. We further tested the specificity of this interaction by diluting Notch‘ and Delta* cells 10-fold with nontransfected S2 cells and allowing them to aggregate for 24 hr at room temperature. In this experiment, 39% of the expressing cells were found in aggregates with other expressing cells, although they composed less than 0.1% of the total cell population. Not surprisingly, however, these aggregates were smaller on average than those found in standard aggregation experiments. In addition, to control for the possibility that Notch‘ cells are nonspecifically recruited into the Delta‘ aggregates because they overexpress a single type of protein on the cell surface, we mixed Delta‘ cells with cells that expressed neuroglian, a transmembrane cell-surface protein (Bieber et al., 1989, Cell 59, 447-460), under the control of the metallothionein promoter (this metallothionein-neuroglian construct was kindly provided by A. Bieber and C. Goodman). We observed no tendency for neuroglian* cells to adhere to Delta‘ aggregates, indicating that Notch-Delta aggregation is not merely the result of high levels of protein expression on the cell surface.

We also tested directly for Notch involvement in the aggregation process by examining the effect of a mixture of polyclonal antisera directed against fusion proteins that spanned almost the entire extracellular domain of Notch on aggregation (see Experimental Procedures, section 611). To minimize artifacts that might arise due to a metabolic response to patching of surface antigens, antibody treatment and the aggregation assay were performed at 4°C in these experiments. Notch’ cells were incubated with either preimmune or immune mouse sera for 1 hr, Delta* cells were added, and aggregation was performed for 1-2 hr. While Notch‘ cells pretreated with preimmune sera aggregated with Delta’ cells (in one of three experiments, 23% of the Notch‘ cells were in Notch*—Delta* cell aggregates), those treated with immune sera did not (only 2% of Notch* cells were in aggregates). This result suggests that the extracellular domain of Notch is required for Notch‘-Delta‘ cell aggregation, although we cannot rule out the possibility that the reduced aggregation was due to inhibitory steric or membrane structure effects resulting from exposure of Notch* cells to the antiserum.

Three other observations worth noting are apparent in Figure 3. First, while Delta was almost always apparent only at the cell surface (Figures 33 and 3c), Notch staining was always apparent both at the cell surface and intracellularly, frequently associated with vesicular structures (Figure 3A). second, we consistently noted a morphological difference between Delta’ and Notch’ cells in mixed aggregates that were incubated overnight. Delta* -58.- cells often had long extensions that completely surrounded adjacent Notch‘ cells, while Notch‘ cells were almost always rounded in appearance without noticeable cytoplasmic extensions (Figure 3G). Third, Notch and Delta often appeared to gather within regions of contact between Notch* and Delta‘ cells, producing a sharp band of immunofluorescent staining (Figures 3D-3P). These bands were readily visible in optical sections viewed on the confocal microscope (Figure 3H), indicating that they were not merely due to a whole-mount artifact. we also observed that these bands formed rapidly (within 2 hr of mixing cells) and at 4°C, indicating that their formation probably did not depend upon cellular metabolism.

These observations would be expected if, within regions of cell contact, Notch and Delta bind to one another and therefore become immobilized. This pattern of expression is also consistent with that observed for other proteins that mediate cell aggregation (Takeichi, 1988, Development 102, Snow et al., 1989, Cell 59, 313-323). -655; like repeats that contain a particular consensus .2.3. sequence may serve as calcium (Ca") binding domains (Morita et al., 1984, J. Biol. Chem. 259, 5698-5704; sugo et al., 1984, J. Biol. Chem. 259, 5705~5710; Rees et al., 1988, EMBO J. 7, 2053-2061; Handford at al., 1990, EH80 J. 9, 475-480). For at least two of these proteins, C and Cl, Ca"'binding has further been shown to be a necessary component of their interactions with other proteins (Villiers et al., 1980, FEBS Lett. 117, 289-294; Esmon et al., 1983, J. Biol. Chem. 258, 5548- ; Johnson, et al., 1983, J. Biol. Chem. 258, S554- ).

Notch and most of those within Delta contain the necessary consensus sequence for Ca" binding (Rees et Many of the EGF-homologous repeats within ’ al., 1933, EMBO J. 7, 205352051; Stenflo et al., 1987, Proc. Natl. Acad. Sci. USA 84, 368-372; Kopczynski et al., 1988, Genes Dev. 2, 1723-1735; Handford et al., 1990, EMBO J. 9, 475-480), although it has not yet been determined whether or not these proteins do bind calcium. We therefore tested the ability of expressing cells to aggregate in the.presence or absence of ca?‘ ions to determine whether there is a Ca"'ion requirement for Notch-Delta aggregation. To minimize possible nonspecific effects due to metabolic responses to the removal of Ca") these experiments were performed at 4°C. Control mixtures of Notch‘ and Delta‘ cells incubated under aggregation conditions in Ca" -containing medium at 4°C readily formed aggregates (an average of 34% 1 13%, mean 1 SD, n = 3; Table II). In contrast, cells mixed in medium that lacked Ca" ions and contained EGTA formed few aggregates (5% 1 5%). These results clearly demonstrate a dependence of Notch-Delta-mediated aggregation on exogenous Ca"’and are in marked contrast to those recently published for the Drosophila fasciclin III and fasciclin I proteins in S2 cells (snow et al., 1989, Cell 59, 313-323; Elkins et al., 1990, J. Cell Biol; 110, 1825-1832), which detected no effect of Ca"’ion removal on aggregation mediated by either protein. .m0uammumvc +ouawoa+nouoz aw maaoo +muamo now u Had aowuamwumma u .w¢uovmumma +muHwoI+nouoz ca maaou +£ouoz you duau aomuawouwva . vwumwm vwnanaou » .AH wnnafi cw may wovomoummm ca venom maaoo mcwmmmumxw wo wmwunwuuom ma Uwucwmwum MUMD . . 2 mm mm o o o on mm mm 3 o 3 3 mm . mm m u e n ucoawummxm N ucwfiauwmxw A uawaaummxm vWHH0U um.H._m0U addd. vWHH0U »W.HH0U nH.HM HQ Z lH0>O HO Z |H0>0 O ...N U. . W + U. . .zoHH¢ummww¢ +4.5mo-+mu.uoz zo zoo wsozmwoxm mo aummmm jg .2.4. OTC D INT W NG C We asked whether Notch and Delta are associated within the membrane of one cell that expresses both proteins by examining the distributions As shown in Figures 4A and 48, these two proteins often show of Notch and Delta in cotransfected cells. very similar distributions at the surface of cotransfected cells. To test whether the observed colocalization was coincidental or represented a stable interaction between Notch and Delta, we treated live cells with an excess of polyclonal anti-Notch . This treatment resulted in "patching" of Notch on the surface of expressing cells into discrete patches as detected by immunofluorescence. a distinct correlation between the distributions of Notch and Delta on the surfaces of these cells after this treatment (Figures 4C and 40), indicating that these proteins are associated within the membrane. is important to note that these experiments do not address the question of whether this association is direct or mediated by other components, such as the cytoskeleton. To control for the possibility that Delta is nonspecifically patched in this experiment, we cotransfected cells with Notch and with the previously mentioned neuroglian construct (A. Bieber and C. Goodman, unpublished data) and patched with anti-Notch antisera. In this case there was no apparent correlation between Notch and neuroglian. antiserum.

There was DO NOT REQUIRE O OTCH In addition to a large extracellular domain that contains BGF-like repeats, Notch has a sizeable intracellular (IC) domain of ~940 amino acids. The IC domain includes a phosphorylation site (Kidd et al., .2.5. INTERACTIONS WITH DELTA D ~52- , Genes Dev. 3, 1113-1129), a putative nucleotide binding domain, a polyglutamine stretch (Wharton et a1., 1985, Cell 43, 567-581; Kidd, et a1., 1986, Mol.

Cell. Biol. 6, 3094-3108), and sequences homologous to the yeast ggggg gene, which is involved in cell cycle control in yeast (Breeden and Nasmyth, 1987, Nature 329, 651-654). Given the size and structural complexity of this domain, we wondered whether it is required for Notch-Delta interactions. we therefore used a variant Notch construct from which coding sequences for ‘B35 amino acids of the IC domain, including all of the structural features noted above, had been deleted (leaving 25 membrane-proximal amino acids and a novel 59 amino acid carboxyl terminus; see Experimental Procedures and Figure 1 for details).

This construct, designated ECNI, was expressed constitutively under control of the normal flgtgh promoter in transfected cells at a level lower than that observed for the metallothionein promoter constructs, but still readily detectable by immunofluorescence.

In aggregation assays, cells that expressed the ECNI construct consistently formed aggregates with Delta‘ cells (31% of ECNl-expressing cells were in aggregates in one of three experiments; see also Figure 31), but not with themselves (only 4% in aggregates), just as we observed for cells that expressed intact Notch. We also observed sharp bands of ECN1 staining within regions of contact with Delta* cells, again indicating a localization of ECN1 within regions of contact between cells. To test for interactions within the membrane, we repeated the surface antigen co-patching experiments using cells cotransfected with the ECN1 and Qglta constructs. As observed for intact Notch, we found that when ECN1 was patched using polyclonal antisera against the extracellular domain of Notch, ECN1 and Delta colocalized at the cell surface (Figures 4E and 4F).

These results demonstrate that the observed interactions between Notch and Delta within the membrane do not require the deleted portion of the Ic domain of Notch and are therefore probably mediated by the extracellular domain. However, it is possible that the remaining transmembrane or Ic domain sequences in ECN1 are sufficient to mediate interactions within a single cell. .2.6. NOTCH AND DELTA FORM DETERGENT-SOLUBLE INTERMOLECULAR COMBLEXES Together, we take the preceding results to indicate molecular interactions between Notch and Delta present within the same membrane and between these proteins expressed on different cells. As a further test for such interactions, we asked whether these proteins would coprecipitate from nondenaturing detergent extracts of cells that express Notch and Delta. If Notch and Delta form a stable intermolecular complex either between or within cells, then it should be possible to precipitate both proteins from cell extracts using specific antisera directed against one of these proteins. We performed this analysis by immunoprecipitating Delta with polyclonal antisera from NP-40/deoxycholate lysates (see Experimental Procedures) of cells cotransfected with the Notch and Delta constructs that had been allowed to aggregate overnight or of 0-24 hr wild-type embryos. We were unable to perform the converse immunoprecipitates because it was not possible to discern unambiguously a faint Delta band among background Staph A bands. It is important to note that we tested this polyclonal anti-Delta antiserum for cross-reactivity against Notch in cell lysates (Figure 5A, lane 1) and by immunofluorescence (e.g., After repeated washing to remove nonspecifically adhering compare Figures 3D and 3E) and found none. proteins, we assayed for coprecipitation of Notch using a monoclonal antibody (MAb c17.9c6) against Notch on western blots.

As Figure 5 shows, we did detect coprecipitation of Notch in Delta immunoprecipitates from cotransfected cells and embryos. However, coprecipitating Notch appeared to be present in much smaller quantities than Delta and was therefore difficult to detect. This disparity is most likely due to the disruption of Notch-Delta complexes during the lysis and washing steps of the procedure.

However, it is also possible that this disparity reflects a nonequimolar interaction between Notch and Delta or greatly different affinities of the antisera used to detect these proteins. The fact that immunoprecipitation of Delta results in the coprecipitation of Notch constitutes direct evidence that these two proteins form stable intermolecular complexes in transfected S2 cells and in embryonic cells. protein products of two of the neurogenic loci, flgtgh and Delta, in order to understand their cellular functions better. Using an in vitro aggregation assay that employs normally nonadhesive S2 cells, we showed that cells that express Notch and Delta adhere specifically to one another. The specificity of this interaction is apparent from the observation that Notch‘-Delta* cell aggregates rarely contained nonexpressing cells, even though nonexpressing cells composed the vast majority of the total cell population in these experiments. We propose that this aggregation is mediated by heterotypic binding between the extracellular domains of Notch and Delta present on the surfaces of expressing cells. consistent with this proposal, we find that antisera directed against the extracellular domain of Notch inhibit Notch-De1ta- mediated aggregation, and that the ECNl Notch variant, which lacks almost all of the Notch intracellular domain, can mediate aggregation with cells that express Delta. we also found that cells that express only Delta aggregate with one another, while those that express only Notch do not. These findings suggest that Delta can participate in a homotypic interaction when present on apposed cell surfaces but that Notch cannot under our assay conditions.

The proposal that Notch and Delta interact at the cell surface is further supported by three lines of evidence. First, we find an intense localization of both proteins within regions of contact which Notch* and Delta* cells, implying that Notch and Delta interact directly, even when expressed in different cells. Second, Notch and Delta colocalize on the surface of cells that express both proteins, suggesting that these proteins can interact within the cell membrane. Third, Notch and Delta can be coprecipitated from nondenaturing detergent extracts of cultured cells that express both proteins as well as from extracts of embryonic cells.

Together, these results strongly support the hypothesis that Notch and Delta can interact heterotypically when expressed on the surfaces of either the same or different cells.

The underlying basis for the observed genetic interactions between Ngtgh and Delta and between figtgh and.mgm (Xu et al., 1990, Genes Dev. 4, 464-475) may be a dose-sensitive interaction between the proteins encoded by these genes.

Two lines of evidence suggest that the Notch and Delta proteins function similarly in gitgg and in 2119. First, the genetic analyses have indicated that the stoichiometry of Notch and Delta is crucial for their function in development. our observations that both Notch-Delta and Delta-Delta associations may occur in vitro imply that Notch and Delta may compete for binding to Delta. Thus, dose—sensitive genetic interactions between Notch and Delta may be the result of competitive binding interactions between their‘ protein products. Second, we were able to detect Notch-Delta association in lysates of cultured cells and in lysates of Drosophila embryos using immunoprecipitation. Taken together, these genetic and biochemical analyses suggest that Notch and Delta do associate in vivo in a manner similar to that which we propose on the basis of our aggregation assays.

Genetic and molecular analyses of Ngtgg have also raised the possibility that there may be interactions between individual Notch proteins (Portin, 1975, Genetics 81, 121-133; Kelley et al., 1987, Cell 51, 539-548; Artavanis—Tsakonas, 1988, Trends Genet. 4, 95-100). Indeed, Kidd et al. (1989, Genes Dev. 3, 1113-1129) have proposed that this protein forms disulfide cross-linked dimers, although this point has not yet been rigorously proven. with or without the formation of covalent cross-links, such interactions could presumably occur either within a our find that single cell or between cells. However, Notch‘ cells do not aggregate homotypically suggests -67.- that Notch-Notch associations are likely to occur within a single cell and not between cells.

Alternatively, it is possible that homotypic Notch interactions require gene products that are not expressed in S2 cells. ' The Notch-Delta interactions indicated by our analysis are probably mediated by the extracellular domains of these proteins. Aggregation experiments using the ECNl construct, from which almost the entire intracellular domain of Notch has been removed or altered by in vitro mutagenesis, confirmed this conclusion. Further experiments that demonstrate ECNl—Delta associations within the membrane on the basis of their ability to co-patch indicated that these interactions are also likely to be mediated by the extracellular domains of Notch and Delta, although in this case we cannot exclude possible involvement of the transmembrane domain or the remaining portion of the Notch intracellular domain. These results are especially interesting in light of the fact that both Notch and Delta have EGF- like repeats within their extracellular domains (Wharton et al., 1985, Cell 43, 567-581; Kidd et a1., 1986, Mol. Cell Biol. 6, 3094-3108; Vassin et al., 1987, EMBO J. 6, 3431-3440; Kopczynski et al., 1988, Genes Dev. 2, 1723-1735).

A second issue of interest regarding BGP domains is the proposal that they can serve as Caz‘ binding domains when they contain a consensus sequence consisting of Asp, Asp/Asn, Asp/Asn, and Tyr/Phe residues at conserved positions within EGF~1ike repeats (Rees et al., 1988, EMBO J. 7, 2053-2061; Handford et al., 1990, EMBO J. 9, 475-430).

Comparisons with a proposed consensus sequence for Ca" binding have revealed that similar sequences are found -.68- within many of the EGF-like repeats of Notch (Rees et al., 1988, EH80 J. 7, 2053-2061) and within most of the EGF-like repeats of Delta (Kopczynski et al., 1988, Genes Dev. 2, 1723-1735). Furthermore, analyses of ugtgh mutations have shown that certain 55 alleles are associated with changes in amino acids within this putative Ca" binding domain (Kelley et al., 1987, Cell 51, 539-548; Hartley et al., 1987, EHBO J. 6, 3407-3417; Rees et al., 1988, EMBO J. 7, 2053-2061). For example, the Ax‘? mutation, which correlates with a His to Tyr change in the 29th EGF- like repeat, appears to change this repeat toward the Conversely, the Ax"’ sequence consensus for Caz‘ binding. mutation appears to change the 24th EGF-like repeat away from this consensus as a result of an Asp to Val change. Thus, the genetic interactions between fig alleles and Qgltg mutations (Xu et al., 1990, Genes Dev., 4, 464-475) raise the possibility that Ca"'ions play a role in Notch-Delta interactions. our finding that exogenous Ca" is necessary for Notch-Delta- mediated aggregation of transfected S2 cells supports this contention.

As we have argued (Johansen et al., 1939, J.

Cell Biol. 109, 2427-2440; Alton et al., 1989, Dev.

Genet. 10, 261-272), on the basis of previous molecular and genetic analyses one could not predict with any certainty the cellular function of either Notch or Delta beyond their involvement in cell-cell interactions. However, given the results presented here, it now seems reasonable to suggest that Notch and Delta may function in yixg to mediate adhesive interactions between cells. At the same time, it is quite possible that the observed Notch-Delta interactions may not reflect a solely adhesive function and may in addition reflect receptor-ligand binding interactions that occur in yigg. Indeed, the presence of a structurally complex 1000 amino acid intracellular domain within Notch may be more consistent with a role in signal transduction than with purely adhesive interactions. Given that Notch may have an adhesive function in concert with Delta, axonal expression of Notch may play some role in axon guidance.

. EGF REPEATS 11 AND 12 OF NOTCH ARE REQUIRED AND SUFFICIENT FOR NOTCH-DELTA-MEDIATED AGGREGATION In this study, we use the same aggregation assay as described in Section 6, together with deletion mutants of Notch to identify regions within the extracellular domain of Notch necessary for interactions with Delta. We present evidence that the EGF repeats of Notch are directly involved in this interaction and that only two of the 36 EGF repeats appear necessary. We demonstrate that these two EGF repeats are sufficient for binding to Delta and that the calcium dependence of Notch—Delta mediated aggregation also associates with these two repeats.

Finally, the two corresponding EGF repeats from the xenopus homolog of Notch also mediate aggregation with Delta, implying that not only has the structure of Notch been evolutionarily conserved, but also its function. These results suggest that the extracellular domain of Notch is surprisingly modular, and could potentially bind a variety of proteins in addition to Delta. .1.1.

The constructs described are all derivatives of the full length Notch expression construct fl pMtNMg (see Section 6, supra). All ligations were - performed using DNA fragments cut from low melting temperature agarose gels (sea Plaque, FMC BioProducts). The 6 kb EcoRI-Xhol fragment from pMtNHg containing the entire extracellular domain of ‘Notch was ligated into the EcoRI—XhoI sites of the Bluescript vector (Stratagene), and named RI/XBS. All subsequent deletions and insertions of EGF repeats were performed in this subclone. The Qgtgh sequence containing the EcoRI-XhoI fragment of these RI/XBS derivatives was then mixed with the 5.5 kb xhol-Xbal fragment from pMtNMg containing the intracellular domain and 3' sequences needed for polyadenylation, and then inserted into the EcoRI€XbaI site of pRMHa-3 (Bunch et al., 1988, Nucl. Acids Res. 16, i043-1061) in a three piece ligation. All subsequent numbers refer to nucleotide coordinates of the Egtgh sequence according to Wharton et al. (1985, Cell 43, 567-581).

For construct f2 Dsph, RI/XBS was digested to completion with Sphl and then recircularized, resulting in a 3.5 kb in~frame deletion from SphI(996) to SphI(4S4S).

For construct #3 Acla, RI/XBS was digested to completion with C1aI and then religated, producing a 2.7 kb in-frame deletion from C1aI(1668) to C1aI(4407). The ligation junction was checked by double strand sequencing (as described by xu et al., 1990, Genes Dev. 4, 464-475) using the Sequenase Kit (U.S. Biochemical Corp., Cleveland). we found that although the ClaI site at position 4566 exists according to the sequence, it was not recognized under our conditions by the ClaI restriction enzyme.

For constructs #4-12, RI/XBS was partially digested with ClaI and then religated to produce all possible combinations of in-frame deletions: construct #4 AEGF7-17 removed the sequence between -71.- ClaI(1668) and C1aI(2820); Construct #5 AEGF9-26 removed the sequence between ClaI(l905) and ClaI(38S5); construct I6 AEGF17-31 removed the sequence between ClaI(2820) and ClaI(44o7); construct #7 AEGF7—9 removed the sequence between C1aI(1668) and C1aI(1905); construct is AEGF9—17 removed the sequence between ClaI(19oS) and ClaI(2820); construct #9 AEGF17-26 removed the sequence between claI(2820) and ClaI(3855); construct I10 AEGF 26~3O removed the sequence between ClaI(3855) and C1aI(44o7); construct #11 AEGF9-30 removed the sequence between claI(1905) and C1aI(4407); construct #12 AEGF 7-26 removed the sequence between ClaI(l668) and C1aI(3855).

For constructs #13 ACla+EGF9-17 and #14 AC1a+EGF17-26, the ~0.9 kb fragment between ClaI(1905) and C1aI(2820), and the -1.0 kb fragment between C1aI(2820) and ClaI(38S5), respectively, were inserted into the unique ClaI site of construct #3 Acla.

For construct #16 split, the l1—kb KpnI/xbal fragment of pMtNMg was replaced with the corresponding KpnI/XbaI fragment from a Notch minigene construct containing the split mutation in EGF repeat 14.

For constructs #17-25, synthetic primers for polymerase chain reaction (PCR) were designed to amplify stretches of EGF repeats while breaking the EGF repeats at the ends of the amplified piece in the same place as the common C1aI sites just after the third cysteine of the repeat (see Figure 7). The PCR products were gel purified as usual and ligated into the C1aI site of construct #3 Acla which was made blunt by filling with the Klenow fragment of DNA Polymerase I (Maniatis et al., 1990, Molecular Cloning, A Laboratory Manual, Cold spring Harbor Laboratory, Cold spring Harbor, New York). The correct orientation of the inserts was determined by PCR using a sense strand primer within the insert together with an antisense strand primer in EGF repeat . number of the nucleotide at their 5' end, according to the nucleotide coordinates of the figtgh sequence in Wharton et al. (1985, cell 43, 567-581), and 8 refers to a sense strand primer while A refers to an antisense strand primer. Construct #16 AC1a+EGF(9-13) used primers 81917 and A2367. Construct #17 Ac1a+EGF(l1-15) used primers 82141 and A2591.

Construct #18 Ac1a+EGF(13-17) used primers 82375 and A2819. Construct #19 AC1a+EGF(l0-13) used primers 32013 and A2367. Construct #20 AC1a+EGF(11-13) used primers S2141 and A2367. Construct #21 AC1a+EGF(10- 12) used primers $2018 and A2015. Construct #22 ACla+EGF(l0-11) used primers 82018 and A2322.

Construct #23 ACla+EGF(10-12) used primers S2018 and A2322. Construct #24 ACla+EGF(11-12) used primers and A2322.

For construct #25 AEGF, construct R1/XBS was digested to completion with sphI(996) and partially digested with BamHI(5135). The resulting incompatible ends were joined using a synthetic linker designed to create a unique ClaI site. This produced an in frame deletion which removed all 36 EGF repeats with the exception of the first half of repeat 1. For constructs #26-29, the EGF fragments were inserted into this ClaI site as previously described for the All primers were 20-mers, and were named with the corresponding constructs #13, 16, 19, and 23.

For construct #30 AECN, construct R1/XBS was digested to completion with BglI, EcoRI and xhoI. The ~0.2 kb EcoRI-Bgll fragment (722-948) and the -0.7 kb Bgll-Xhol (5873-6627) fragments were ligated with EcoRI-XhoI cut Bluescript vector and a synthetic linker designed to create a unique ClaI site, resulting in an in-frame deletion from BglI(941) to BglI(5B73) that removed all 36 EGF repeats except for ~the first third of repeat 1 as well as the 3 flgtgh/lin-12 repeats.

EGF fragments were inserted into the unique ClaI site as previously described for constructs I19 and 23.

For constructs #33 and 34, PCR primers 81508 and A1859 based on the xggopus Notch sequence (Coffman et al., 1990, Science 249, 1438-1441; numbers refer to nucleotide coordinates used in this paper), were used to amplify EGF repeats 11 and 12 out of a xenopus stage 17 CDNA library (library was made by D. Melton and kindly provided by M. Danilchek). The fragment was ligated into construct I3 Dcla and sequenced.

‘For constructs #31 and 32, the The Qrosphila S2 cell line was grown and transfected as described in Section 6, supra. The Delta-expressing stably transformed 82 cell line L 6-7 (kindly established by L. Cherbas) was grown in as medium (prepared by Hazleton Co.) supplemented with 11% heat inactivated fetal calf serum (FCS) (Hyclone), 100 U/ml penicillin-100 pg/ml streptomycin-0.25 ug/ml fungizone (flazleton), 2 x 10" M methotrexate, 0.1 mm hypoxanthine, and 0.016 mM thynidine. .1.3. AGQBEQAIION ASSAYS AHQ IHMQHOELUOBESCENCE Aggregation assays and Ca** dependence experiments were as described ggggg, Section 6. Cells were stained with the anti-Notch monoclonal antibody 9c6.C17 and anti—De1ta rat polyclonal antisera (details described in Section 6, ggggg). Surface expression of Notch constructs in unpermeabilized cells was assayed using rat polyclonal antisera raised against the 0.8 kb (amino acids 237-501; wharton et al., 1985, Cell 43, 567-581) BstYI fragment from the extracellular domain of Notch. Cells were viewed under epifluorescence on a Leitz Orthoplan 2 microscope. 7.2. ggsugms 7.2.1. EGF REPEATS 11 AND 12 OF NOTCH ARE REQUIRED EQB flQICfl-DELTA MEQIAIEQ AGQBEQQTION We have undertaken an extensive deletion analysis of the extracellular domain of the Notch protein, which we have shown (supra, section 6) to be involved in Notch-Delta interactions, to identify the precise domain of Notch mediating these interactions. we tested the ability of cells transfected with the various deletion constructs to interact with Delta using the aggregation assay described in Section 6.

Briefly, notch deletion constructs were transiently transfected into Qrgsgphilg S2 cells, induced with cuso" and then aggregated overnight at room temperature with a small amount of cells from the stably transformed Delta expressing cell line L49—6- 7(Cherbas), yielding a population typically composed of ~1% Notch expressing cells and ~5% Delta expressing cells, with the remaining cells expressing neither protein. To assay the degree of aggregation, cells were stained with antisera specific to each gene product and examined with immunofluorescent microscopy (see experimental procedures for details). Aggregates were defined as clusters of four or more cells containing both Notch and Delta expressing cells, and the values shown in Figure 6 represent the percentage of all Notch expressing cells found in such clusters.

All numbers reflect the average result from at least two separate transfection experiments in which at least 100 Notch expressing cell units (either single cells or clusters) were scored.

Schematic drawings of the constructs tested and results of the aggregation experiments are shown in Figure 6 (see Experimental Procedures for details).

All expression constructs were derivatives of the full length Notch expression construct fl pMtNMg (described in Section 6, sgprg).

The initial constructs (#2 Dsph and #3 Acla) deleted large portions of the EGF repeats. Their inability to promote Notch-Delta aggregation suggested that the EGF repeats of Notch were involved in the interaction with Delta. We took advantage of a series of six in-frame C1aI restriction sites to further dissect the region between EGF repeats 7 and 30. Due to sequence homology between repeats, five of the C1aI sites occur in the same relative place within the EGF repeat, just after the third cysteine, while the sixth site occurs just before the first cysteine of EGF repeat 31 (Figure 7). Thus, by performing a partial C1aI digestion and then religating, we obtained deletions that not only preserved the open reading frame of the Notch protein but in addition frequently maintained the structural integrity and conserved spacing, at least theoretically, of the three disulfide bonds in the chimeric EGF repeats produced by the religation (Figure 6, constructs #4-14).

Unfortunately, the most 3' C1aI site was resistant to digestion while the next most 3' ClaI site broke between EGF repeats 30 and 31. Therefore, when various Clal digestion fragments were reinserted into the framework of the complete Clal digest (construct #3 Acla), the overall structure of the EGF repeats was apparently interrupted at the 3‘ junction.

Several points about this series of constructs are worth noting. First, removal of the ClaI restriction fragment breaking in BGF repeats 9 ~76.- and 17 (construct is AEGF9-17) abolished aggregation with Delta, while reinsertion of this piece into construct #3 Acla, which lacks EGF repeats 7-30, restored aggregation to roughly wild type levels (construct #13 ACla+EGF9-17), suggesting that EGF repeats 9 through 17 contain sequences important for binding to Delta. Second, all constructs in this series (#4-14) were consistent with the binding site mapping to EGF repeats 9 through 17. Expression constructs containing these repeats (#6, 7, 9, 10, 13) promoted Notch-Delta interactions while constructs lacking these repeats (#4, S, 8, 11, 12, 14) did not.

To confirm that inability to aggregate with Delta ‘ cells was not simply due to failure of the mutagenized Notch protein to reach the cell surface, but actually reflected the deletion of the necessary binding site, we tested for cell surface expression of all constructs by immunofluorescently staining live transfected cells with antibodies specific to the extracellular domain of Notch. All constructs failing to mediate Notch—Delta interactions produced a protein that appeared to be expressed normally at the cell surface. Third, although the aggregation assay is not quantitative, two constructs which contained EGF repeats 9-17, #9 AEGFI7-26 or most noticeably #10 AEGF26-30, aggregated at a seemingly lower level.

Cells transfected with constructs #9 AEGF17-26 and 10 AEGF26-30 showed considerably less surface staining than normal, although fixed and permeabilized cells reacted with the same antibody stained normally, indicating we had not simply deleted the epitopes recognized by the antisera. By comparing the percentage of transfected cells in either permeabilized or live cell populations, we found that roughly 50% of transfected cells for construct #9 - 77 _ AEGF17-26 and 10% for construct #10 AEGF26-30 produced detectable protein at the cell surface. Thus these two constructs produced proteins which often failed to reach the cell surface, perhaps because of misfolding, thereby reducing, but not abolishing, the ability of transfected cells to aggregate with Delta-expressing cells.

Thus, to further map the Delta binding domain within EGF repeats 9~17, we used specific oligonucleotide primers and the PCR technique to generate several subfragments of this region. To be consistent with constructs #4-14 which produced proteins that were able to interact with Delta, we designed the primers to splice the EGF repeats just after the third cysteine, in the same place as the common ClaI site (Figure 7). The resulting PCR products were ligated into the ClaI site of construct #3 Acla. were produced, only one of which, #16 Acla+EGF9-13, when transfected into 82 cells, allowed aggregation with Delta cells. Construct #19 ACla+EGF(l0-13), which lacks EGF repeat 9, further defined EGF repeats -13 as the region necessary for Notch-Delta Three overlapping constructs, I16, 17 and 18 interactions, T Constructs #20-24 represented attempts to break this domain down even further using the same PCR strategy (see Figure 7). We asked first whether both EGF repeats 11 and 12 were necessary, and second, whether the flanking sequences from EGF repeats 10 and 13 were directly involved in binding to Delta.

Constructs #20 ACla+EGF(11-13), in which EGF repeat 12 is the only entire repeat added, and #21 ACla+EGF(10- 12), in which EGF repeat 11 is the only entire repeat added, failed to mediate aggregation, suggesting that the presence of either EGF repeat 11 or 12 alone was not sufficient for Notch—Delta interactions. However, since the 3' ligation juncture of these constructs interrupted the overall structure of the EGF repeats, it was possible that a short "buffer" zone was needed to allow the crucial repeat to function normally.

Thus for example in construct #19 AC1a+EGF(l0-13), EGF repeat 12 might not be directly involved in binding to Delta but instead might contribute the minimum amount of buffer sequence needed to protect the structure of EGF repeat 11, thereby allowing interactions with Delta. Constructs #22-24 addressed this issue. we designed PCR primers that broke at the end of the EGF repeat and therefore were less likely to disrupt the EGF disulfide formation at the 3' ligation juncture.

Constructs #22 ACla+EGF(10-11), which did not mediate aggregation, and #23 ACla+EGF(l0-12), which did, again suggested that both repeats 11 and 12 are required while the flanking sequence from repeat 13 clearly is not. Finally, construct #24 ACla+EGF(1l-12), although now potentially structurally disrupted at the 5' junction, convincingly demonstrated that the sequences from EGP repeat 10 are not crucial. Thus based on entirely consistent data from 24 constructs, we propose that EGF repeats 11 and 12 of Notch together define the smallest functional unit obtainable from this analysis that contains the necessary sites for binding to Delta in transfected S2 cells. .2.2. EGF REPEATS 11 AND 12 OF NOTCH ARE SUFFICIENT The large ClaI deletion into which pen ..8o... repeats were also absent, again successfully mediated Notch-Delta aggregation. Thus EGF repeats 11 and 12 appear to function as independent modular units which are sufficient to mediate Notch-Delta interactions in S2 cells, even in the absence of most of the extracellular domain of Notch.

EGF REPEATS 11 AND 12 OF NOTCH MAINTAIN THE CALCIUM DEPENDENCE OF NOTCH-DELTA HEDIATED AGGREGATION V2.3I As described in Section 6, sgpzg (Fehon et a1., 1990, Cell 61,’ 523-534), weshowed that Notch- Delta-mediated 32 cell aggregation is calcium dependent. We therefore examined the ability of cells expressing certain deletion constructs to aggregate with Delta expressing cells in the presence or absence of ca** ions. We tested constructs #1 pMtNMg as a control, and #13, 16, 19, 23, 24, 26, 27 and 28, and found that cells mixed in Ca** containing medium at 4°C readily formed aggregates while cells mixed in Ca** free medium containing EGTA failed to aggregate (Table III).

Tgagg III EEEEQI OE EXOGEEOUS Ca** ON NOTCH - DELTA AGGREGATION' Without Ca** Ions with Ca** 1on5 1. pMtNMg O 37 13. ACla+EGF(9-17) 0 31 . ACla+EGI-‘(9-13) o 33 19. ACla+EGF(10-l3) 0 42 23. ACla+EGF(l0-12) o 48 29. AEGF+EGF(10-12) 0 44 32. AECN+EGF(10-12) 0 39 33. AC1a+XEGF(10-13 o 34 ‘Data presented as percentage of Notch-expressing cells found in aggregates (as in Figure 6).

Clearly, the calcium dependence of the interaction has been preserved in even the smallest construct, ’ consistent with the notion that the minimal constructs containing EGF repeats 11 and 12 bind to Delta in a manner similar to that of full length Notch. This result is also interesting in light of recent studies suggesting EGF-like repeats with a particular consensus sequence may act as Ca** binding domains (Morita et al., 1984, J. Biol. Chem. 259, 5698-5704; Sugo et al., 1984, J. Biol. Chem. 259, 5705-5710; Rees et al., 1988, EMBO J. 7, 2053-2061; Handford et a1., 1990, EMBO J. 9, 475-480). Over half of the EGF repeats in Notch, including repeats 11 and 12, conform to this consensus, further strengthening the argument that EGF repeats 11 and 12 are responsible for promoting Notch-Delta interactions.

THE DELTA BINDING FUNCTION OF EGF REPEATS 11 AND 12 OF NOTCH IS CONSERVED IN THE XENOPUS HOMOLDG OF NOTCH ~ Having mapped the Delta binding site to EGF repeats 11 and 12 of Notch, we were interested in asking whether this function was conserved in the Notch homolog that has been identified in xeggpgg (Coffman et al., 1990, Science 249, 1433-1441). This protein shows a striking similarity to Drosgphila Notch in overall structure and organization. For example, within the EGP repeat region both the number .2.4. and linear organization of the repeats has been preserved, suggesting a possible functional conservation as well. To test this, we made PCR primers based on the ggngpgg Notch sequence (Coffman et a1., 1990, Science 249, 1433-1441) and used these to obtain an ~3S0 bp fragment from a xgggpgg Stage 17 cDNA library that includes EGF repeats 11 and 12 flanked by half of repeats 10 and 13 on either side.

This fragment was cloned into construct #3 Acla, and three independent clones were tested for ability to interact with Delta in the cell culture aggregation assay. Two of the clones, f33a&bACla+xEGF(10—13), when transfected into 82 cells were able to mediate Notch—Delta interactions at a level roughly equivalent to the analogous Qrgsgphila Notch construct fl9AC1a+EGF(10-13), and again in a calcium dependent manner (Table III). However, the third clone #33cACla+XEGF(10-13) failed to mediate Notch-Delta interactions although the protein was expressed normally at the cell surface as judged by staining live unpermeabilized cells. sequence comparison of the xenopus PCR product in constructs 133a and 33c revealed a missense mutation resulting in a leucine to -83.. ‘. proline change (amino acid #453, Coffman, et a1., 1990, Science 249, 1438-1441) in EGP repeat 11 of construct #33c. Although this residue is not conserved between Qrgggghila and xgngpug Notch (Figure 8), the introduction of a proline residue might easily disrupt the structure of the EGF repeat, and thus prevent it from interacting properly with Delta.

Comparison of the amino acid sequence of EGF repeats 11 and 12 of Qgosoghilg and xgggggg Notch reveals a high degree of amino acid identity, including the calcium binding consensus sequence (Figure 8, SEQ ID Nozl and N0:2). However the level of homology is not strikingly different from that shared between most of the other EGF repeats, which overall exhibit about 50% identity at the amino acid level. This one to one correspondence between individual EGF repeats suggests that perhaps they too -Delta interactions, again in a calcium ion-dependent manner. may comprise conserved functional units. .3. QIQQQSSIOE binding domains. -.84- Recent studies have demonstrated that EGF domains containing a specific consensus sequence can bind Ca** ions (Morita et al., 1984, J. Biol. Chem. 259, 5698-5704; Sugo et al., 1934, J. Biol. Chem. 259, 5705-5710; Rees et al., 1988, EMBO J. 7, 2053-2061; Handford et al., 1990, EMBO J. 9, 475-480). In fact, about one half of the EGF repeats in Notch, including repeats 11 and 12, conform to this consensus. We have shown that exogenous Ca‘* was necessary for Notch- Delta mediated aggregation of transfected S2 cells (see Section 6; Fehon et al., 1990, Cell 61, 523-534).

We tested a subset of our deletion constructs and found that EGF repeats 11 and 12 alone (#32AECN+EGF(11-12)) were sufficient to maintain the Ca" dependence of Notch-Delta interactions.

A number of studies have suggested that the genetic interactions between flgtgh and Qglta may reflect a dose sensitive interaction between their protein products. Genetic studies have indicated that the relative gene dosages of flgtgh and Delta are crucial for normal development. For example, xu et al. (1990, Genes Dev. 4, 464-475) found that null mutations at Delta could suppress lethal interactions between heterozygous combinations of Abrugteg (53) alleles, a class of flgtgh mutations that correlate with missense mutations within the EGF repeats (Hartley et al., 1937, EMBO J. 6, 3407-3417; Kelley et al., 1987, Mol. Cell Biol. 6, 3094-3108). The in yitgg interactions we have described in which we observe both Notch-Delta and Delta-Delta associations (see section 6) imply that a competitive interaction between Notch and Delta for binding to Delta may reflect the underlying basis for the observed genetic interactions. Furthermore, we were able to coimmunoprecipitate Notch and Delta from both tissue culture and embryonic cell extracts (see Section 6), indicating a possible in vivo association of the two proteins.’ and Delta expression patterns in the embryo suggest that expression of the two is overlapping but not identical (Kopczynski and Muskavitch, 1989, Development 107, 623-636; Hartley et al., 1987, EMBO J. 6, 3407-3417). Detailed antibody analysis of Notch protein expression during development have recently revealed Notch expression to be more restricted at the _tissue and subcellular levels than previous studies had indicated (Johansen et al., 1989, J. Cell Biol. 109, 2427-2440; Kidd et al., 1989, Genes Dev. 3, 1113- 1129). ‘ our finding that the same two EGF repeats from the xenopus Notch homolog are also able to mediate interactions with Delta in tissue culture cells argues strongly that a similar function will have been conserved in vivo. Although these two EGF repeats are sufficient in gitrg, it is of course possible that in yixg more of the Notch molecule may be necessary to facilitate Notch-Delta interactions.

In fact, we were somewhat surprised for two reasons to find that the Delta binding site did not map to EGF repeats where several of the 53 mutations have been shown to fall, first, because of the genetic screen (Xu et al., 1990, Genes Dev. 4, 464-475) demonstrating interactions between Ax alleles and Delta mutations, and second, because of sequence analyses that have shown certain 53 alleles are associated with single amino acid changes within the putative ca** binding consensus of the EGP repeats. For example, the AX" mutation changes EGF repeat 29 toward the Ca** binding consensus sequence while the AX" mutation moves EGF repeat 24 away from the consensus. It is possible In addition, mRNA in situ analyses of Notch .perhaps modulating Notch-Delta interactions in that in yiyg these regions of the Notch protein may be involved in interactions, either with Delta and/or other proteins, that may not be accurately modelled by our cell culture assay. . our in yitrg mapping of the Delta binding domain to EGF repeats 11 and 12 of Notch represents the first assignment of function to a structural domain of Notch. In fact, the various deletion constructs suggest that these two EGF repeats function as a modular unit, independent of the immediate context into which they are placed. Thus, neither the remaining 34 EGF repeats nor the three flgtgh/li -12 repeats appear necessary to establish a structural framework required for EGF repeats 11 and 12 to function. Interestingly, almost the opposite effect was observed: although our aggregation assay does not measure the strength of the interaction, as we narrowed down the binding site to smaller and smaller fragments, we observed an increase in the ability of the transfected cells to aggregate with Delta expressing cells, suggesting that the normal flanking EGF sequences actually impede association between the proteins. For two separate series of constructs, either in the background of construct #3 Acla (compare #9, 16, 19, 23) or in the background of construct #25 AEGF (compare #26, 27, 28), we observed an increase in ability to aggregate such that the smallest constructs (#19, 23, Q8, 29) consistently aggregated above wild type levels (#1 pMtNMg). These results imply that the surrounding EGF repeats may serve to limit the ability of EGF repeats 11 and 12 to access Delta, thereby vivo.

Notch encodes a structurally complex transmembrane protein that has been proposed to play a pleotropic role throughout Drosophila development.

The fact that EGF repeats 11 and 12 appear to function as an independent modular unit that is sufficient, at least in cell culture, for interactions with Delta, immediately presents the question of the role of the hypothesis is that these may also form modular binding domains for other proteins interacting with Notch at various times during development. iIn addition to xegopgs Ngtgh, 11_-12 and glp~1, two genes thought to function in cell-cell interactions involved in the specification of certain cell fates during 9. elggans development, encode EGF homologous transmembrane proteins which are structurally quite similar to Qgosopnila and xegopyg Notch. All four proteins contain BGF homologous repeats followed by three other cysteine rich repeats (Notch/;ig~12 repeats) in the extracellular domain, a single transmembrane domain, and six cdclo/ankyrin repeats in the intracellular region. Unlike zenopgs Notch, which, based on both sequence comparison as well as the results of our Delta binding assay, seems likely to encode the direct functional counterpart of Qrgggghila notch, 115-12 and glp—1 probably encode distinct members of the same gene family. Comparison of the predicted protein products of lip-12 and glp-1 with Notch reveal specific differences despite an overall similar organization of structural motifs.

The most obvious difference is that lip-12 and glp-1 proteins contain only 13 and 10 EGF repeats, respectively, as compared to the 36 for both zgngpus and Qgggggnilg Notch. In addition, in the nematode genes the array of EGF repeats is interrupted after the first EGP repeat by a distinct stretch of sequence absent from Notch. Furthermore, with respect to the Delta binding domain we have defined as EGF repeats 11 and 12 of Notch, there are no two contiguous EGF , repeats in the lig-12 or gig-1 proteins exhibiting the Ca‘* binding consensus sequence, nor any two contiguous repeats exhibiting striking similarity to EGF repeats 11 and 12 of Notch, again suggesting that the li_-12 and glp-1 gene products are probably functionally distinct from Notch.

Our finding that EGF repeats 11 and 12 of Notch form a discrete Delta binding unit represents the first concrete evidence supporting the idea that each EGF repeat or small subset of repeats may play a unique role during development, possibly through The homologies seen between the adhesive domain of Delta and serrate (see Section 8.3.4, igjrg) suggest that the homologous portion of serrate is "adhesive" in that it mediates binding to other toporythmic In addition, the gene so brous, which direct interactions with other proteins. proteins. encodes a secreted protein with similarity to fibrinogen, may interact with Notch.

A In addition to the EGF repeat, multiple copies of other structural motifs commonly occur in a variety of proteins. one relevant example is the cdclo/ankyrin motif, six copies of which are found in the intracellular domain of Notch. Ankyrin contains 22 of these repeats. Perhaps repeated arrays of structural motifs may in general represent a linear assembly of a series of modular protein binding units.

Given these results together with the known structural, genetic and developmental complexity of notch, Notch may interact with a number of different ligands in a precisely regulated temporal and spacial pattern throughout development. Such context specific interactions with extracellular proteins could be mediated by the EGF and flgtgg/lip-12 repeats, while interactions with cytoskeletal and cytoplasmic V proteins could be mediated by the intracellular cdclo/ankyrin motifs.

. THE AMINO-TERMINUS OF DELTA IS AN EGF-BINDING HA RACT W T NO C LT Aggregation of cultured cells programmed to express wild type and variant Delta proteins has been employed to delineate Delta sequences required for heterotypic interaction with Notch and homotypio Delta interaction. We have found that the amino terminus of the Delta extracellular domain is necessary and sufficient for the participation of Delta in heterotypic (Delta-Notch) and homotypic (Delta-Delta) interactions. We infer that the amino terminus of Delta is an EGF motif-binding domain (BED), given that Notch EGF-like sequences are sufficient to mediate heterotypic interaction with Delta. The Delta EBD apparently possesses two activities: the ability to bind EGF-related sequences and the ability to self- associate. We also find that Delta is taken up by cultured cells that express Notch, which may be a reflection of a mechanism by which these proteins interact in vivo.

, J. Embryol. Exp. Morph. 27, 353-365)) used in these experiments was grown as described in Section 6. .1.2. U CAL Immunohistochemistry was performed as described in section 6, supra, or sometimes with minor modifications of this procedure. Antisera and antibodies employed included mouse polyclonal anti- Delta sera raised against a Delta ELR array segment that extends from the fourth through ninth ELRs (see Section 6); rat polyclonal anti-Delta sera raised against the same Delta segment (see Section 6); rat polyclonal anti-Notch sera raised against a Notch ELR array segment that extends from the fifth through thirteenth ELRs; mouse monoclonal antibody C17.9C6 (see section 6), which recognizes the Notch intracellular domain; and mouse monoclonal antibody BP-104 (Hortsch et a1., 1990, Neuron 4, 697-709), which recognizes the long form of Drosophila neuroglian. wild type Delta (pMTDl1) and wild type Notch (pHTNMg) are described in Section 6, ggpgg. Constructs that direct expression of variant Delta proteins were generated using pMTDl1, the D11 cDNA cloned into Bluescript+ (pBsDll; Kopczynski et al., 1988, Genes Dev. 2, 1723-1735), and pRmHa3-104 (A.J. Bieber, pers. comm.), which consists of the insertion of the lB7A- 250 CDNA into the metallothionein promoter vector pRmHa-3 (Bunch et al., 1988, Nucl. Acids Res. 16, 1043-1061) and supports inducible expression of the long form of Drosophila neuroglian (Hortsch et al., 1990, Neuron 4, 697-709).

Briefly, constructs were made as follows: Del(sca-Nae) - Cut pBSDl1 with SalI (complete digest) and ScaI (partial), isolate vector- containing fragment. Cut pBSDll with Nael (partial) and sall (complete), isolate Delta carboxyl—terminal Ligate fragments, transform, and Transfer EcoRI insert into pRmHa-3. coding fragment. isolate clones. ‘ molar excess of 5‘-TTAAGTTAACTTAA—3' Del(Bam+Bgl) - Cut pBSDl1 with BglII (complete) and BamHI (partial), fill ends with Klenow DNA polymerase, ligate, transform, and isolate clones.

Transfer EcoRI insert into pRmHa—3.

Del(ELR1-ELR3) - PCR-amplify basepairs 236- 830 of the D11 CDNA using 5-ACTTCAGCAACGATCACGGG-3' (SEQ ID NO:26) and 5'-TTGGGTATGTGACAGTAATCG~3'(SEQ ID NO:27), treat with T4 DNA polymerase, ligate into pBSDl1 cut with ScaI (partial) and BglII (complete) and end-filled with Klenow DNA polymerase, transform, and isolate clones. Transfer BamHI-SalI Delta carboxyl-terminal coding fragment into pRmHa-3.

Del(ELR4-ELR5) - pBSDl1 was digested to completion with Bglll and partially with PstI. The .6 kb vector-containing fragment was isolated, circularized using T4 DNA ligase in the presence of a 1oox molar excess of the oligonucleotide 5'—GATCTGCA- 3', and transformed and clones were isolated. The resulting EcoRI insert was then transferred into pRmHa-3.

Ter(Dde) - Cut pBSDl1 with DdeI (partial), end—fill with Klenow DNA polymerase, ligate with 100x (SEQ ID NO:28), transform, and isolate clones. Transfer EcoRI insert into pRmHa-3.

Ins(Nae)A - Cut pMTDl1 with NaeI (partial), isolate vector-containing fragment, ligate with 100x molar excess of 5'-GGAAGATCTTCC-3' (SEQ ID NO:29), transform, and isolate clones.

NAB B - pMTDl1 was digested partially with NaeI, and the population of tentatively linearized circles approximately 5.8 kb in length was isolated.

The fragments were recircularized using T4 DNA ligase in the presence of a 1oox molar excess of the oligonucleotide 5'-GGAAGATCTTCC-3' (SEQ ID NO:29) and transformed, and a clone (NAB A) that contained multiple inserts of the linker was isolated. NAB A was digested to completion with BglII, and the resulting 0.4 kb and 5.4 kb fragments were isolated, ligated and transformed, and clones were isolated.

Ins(Stu) e Cut pMTDl1 with StuI (complete), isolate vector—containing fragment, ligate with 100X molar excess of 5'-GGAAGATCTTCC-3'(SEQ ID No:29), transform and isolate clones.

STU B - pMTDl1 was digested completely with Stul, and the resulting 5.8 kb fragment was isolated.

The fragment was recircularized using T4 DNA ligase in the presence of a 1oox molar excess of the oligonucleotide S’-GGAAGATCTTCC-3' (SEQ ID NO:29) and transformed, and a clone (STU A) that contained multiple inserts of the linker was isolated. STU B was digested to completion with BglII, and the resulting 0.6 kb and 5.2 kb fragments were isolated, ligated and transformed, and clones were isolated.

NGl - Cut pRmHa3-104 with BglII (complete) and EcoRI (complete), isolate vector—containing fragment. Cut Ins(Nae)A with EcoRI (complete) and BglII (complete), isolate Delta amino-terminal coding fragment. Ligate fragments, transform and isolate clones.

NG2 - Cut pRmHa3-104 with BglII (complete) and ficoRI (complete), isolate vector-containing fragment. Cut Del(ELR1-ELR3) with EcoRI (complete) and BglII (complete), isolate Delta amino-terminal coding fragment. Ligate fragments, transform and isolate clones.

NG3 - Cut pRmHa3-104 with BglII (complete) and EcoRI (complete), isolate vector-containing fragment. Cut pMTDl1 with EcoRI (complete) and BglII (complete), isolate Delta amino-terminal coding fragment. Ligate fragments, transform and isolate clones. .

NG4 - Cut pRmHa3-104 with BglII (complete) and EcoRI (complete), isolate vector containing fragment. Cut Del(Sca-Nae) with EcoRI (complete) and BglII (complete), isolate Delta amino-terminal coding fragment. Ligate fragments, transform and isolate clones.

NG5 - Generate De1(Sca-Stu) as follows: cut pMTD11 with ScaI (complete) and StuI (complete), isolate ScaI—scaI amino-terminal coding fragment and StuI-ScaI carboxy1—terminal coding fragment, ligate, transform and isolate clones. Cut Del(Sca-Stu) with EcoRI (complete) and BglII (complete), isolate Delta amino terminal coding fragment. Cut pRmHa3~104 with BglII (complete) and EcoRI (complete), isolate vector- containing fragment. Ligate fragments, transform and isolate clones.

The sequence contents of the various Delta variants are shown in Table IV. Schematic diagrams of the Delta variants defined in Table IV are shown in Figure 9.

B IV SEQUENCE CONTENTS OF DELTA VARIANTS Wild type De1(Sca-Nae) De1(Bam-B91) gggpgxgp [E TflIS STQQI flgcleogides 1-2892* -235/734-2892 -713/1134-2992 -31/W/199191/332-833 De1(ELRl-ELR3) 1-830/1134-2892 1-230/332-333 De1(ELR4-ELR5) 1-1137/14o5~2892 1-332/422-833 Ter(Dde)~ 1—2o21/TTAAcTTAAcTTAAE/ 1-626/H 2227-2392 Ins(Nae)A 1-733/(ccAAcATcTTcc)f/ 1-197/(RKIF), 734-2392' 198-833 NAB B 1-733/GGAAGATCTTCCF/ 1-197/RKIF -2892 198-833 Ins(Stu) 1-s3s/(ccnAcATcTTcc)f/ 1-131/ -2892‘ G(KIFR),, KIFP/133-833 STU B 1-s3s/GcAAGATcTTcc‘/ 1-131/GKIFP 536-2892 133-333 NG1 1-733/GGAA/2889-3955(NG)° 1-198/K/9S2- " N62 1-830/2889-39S5(NG) 1-23o/9s2- 1302 NG3 1-1133/2339-39ss(uc) 1-331/952- 1302 NG4 1-235/734-1133/ 1-31/199-331/ 2339-3955(uc) 952-13oz NG5 1-23s/s35-1133/ 1-31/s/133- zaaa-39ss(na) 952-1302 A Coordinates for Delta sequences correspond to the sequence of the D11 CDNA (Figure 12).

B The exact number of linkers inserted has not been determined for this construct.

C Coordinates for neuroglian (Bieber et al., 1989, Cell 59, 447-460; Hortsch et al., 1990, Neuron 4, 697-709) nucleotide sequences present in Delta- neuroglian chimeras correspond to the sequence of the 187A-250 CDNA (Figure 13, SEQ ID NO:5) and are indicated in bold face type.

D Neuroglian amino acid sequences are derived from conceptual translation of the 1B7A-250 cDNA bnucleotide sequence (Figure 13, SEQ ID N0:5) and are indicated in bold face type.

SEQ ID N0:28 SEQ ID N0:29 ‘'81 .1.4. AGGREGATLOE P3o1ocoLs _Cell transfection and aggregation were performed as described in section 6, su ra. or with minor modifications thereof. .2. BESULQS AMINO-TERMINAL SEQUENCES WITHIN THE DELTA EXTRACELLULAR DOMAIN ARE NECESSARY AND SUFFICIENT FOR THE HETEROTYPIC INTERACTIQN WITH NOTCH Because we anticipated that some Delta variants might not be efficiently localized on the cell surface, we investigated the relationship between the level of expression of wild type Delta and the extent of aggregation with Notch-expressing cells by varying the input amount of Delta expression construct in different transfections. We found that the heterotypic Delta-Notch interaction exhibits only slight dependence on the Delta input level over a 10- fold range in this assay (Figure 9A). Given the robustness of the heterotypic interaction over the range tested and our observations that each of the Delta variants we employed exhibited substantial .2.1. surface accumulation in transfected cells, we infer that the inability of a given Delta variant to support ‘.96- heterotypic aggregation most probably reflects a functional deficit exhibited by that variant, as opposed to the impact of reduced levels of surface expression on heterotypic aggregation. _ The results of the heterotypic aggregation experiments mediated by Delta variants and wild-type Notch are shown in Table V. mm mma mm «ea «Ha on we awa do me mad «Ha um mm nma mag 02 oz cm can vs Asa oz oz mu ma~ an ma~ Nw mwfl an pm on ma am mea mm maw «Ha aw Manama unmmqa Uovmvmumwwca aw on w «am an em a man ma on N «on Na mm a woa an an v b¢ ma mu m »- ad a N om pa ad a man oz oz n. .oo~ o o N oom A n a mwm oz aoz v com O o M nan o o N wmm o o A now am cu m me an ma v «nu nu an m mom aw um m new ad mu H aha mumu4«a.+nmqma uammuu nunnuunaduaa vwuovouvwd vmuauoummocb unuauuauuxauaqumuazuuumqamamnaaama um oma¢Homz zouamomaooc unmuaommemm mwuquaa Am.Hm Amvnm .m.wm Axvmn Ammqmuqmamvaao Amvuwa am.- .m.o~ Amvmm Ammqmnamqmvamo Am.o Amvo Am.» nHomaaam.Hwo .=.o .=Vo Amvo .:vo Awmzanomvaoo Amvmna Amvmm Amvwn Amvwm oAmVnn Umuoowuuot Amuavuwa mam» wads .u0.DHMWfldmw .voUuoomu uo: oumz Umunowuwwmas can mwuomoumoa cw A+nouoz can +auawo .mw.«V mwmxu aaao ao:uw>qU:w uou sumo 0 .nmvonuvz new mnawumuwx mmmv ouuaa nmuuuouowe aaozvma a cw uouawmuoma mum: maamo van» moumoaocw Amy .xmaau ummwacmaum He mm m cw oouaumumuo mum: maaoo van» moumomoca Amv U . Amuonuuz Una mamwumuwz mmmv cwwavousmc no CHUEOO umadaawowuucw on» mmuwcvoooh pan» aoonwucm amcoaoocoe a wcwms cwuuuuwo wuwz "cozy uuoauumcoo ummanocnaaoouaw: mmmumxu van» maawo m .ma~wu ouoa.uo udou awavcou uunu m¢uoamuwoa ca maamo mcwmmoumxw no uwnasc amuoa ¢ boa mm o o a com .mvo own mm o o m maw Amvo «ca oma o o a «ma Amvo «oz do «w an ac H mua Anya» noz mna an em mg n. oma .m.mn «ma cm ma mm ~ mam Amvom wm em 5 5 H cod .m.ea woz mod or o o n waw .mVo on an o o N «am .mvo mm ova o o a mom Hmvo aoz on «ma o o m HON Aavo mm «ea o o w qmw .m.o vm «Ha o o H mow .m.o 4.moz.mcH ow oaa mm on n oma Amvmm Delta amino acids (AA) 1-230 is the current minimum sequence interval defined as being sufficient for interaction with Notch. This is based on the success of NG2-Notch aggregation. Within this interval, Delta AA198-230 are critical because their deletion in the NG1 construct inactivated the Notch-binding activity observed for the N62 construct. Also within this interval, Delta AA32-198 are critical because their deletion in the NG4 construct also inactivated the Notch-binding activity observed for the N63 construct.

The importance of Delta AA192-230 is also supported by the observation that the Del(ELRl-ELR3) variant, which contains all Delta amino acids except AA231-331, possessed Notch-binding activity, while the Del(Bam- Bgl) variant, which contains all Delta amino acids except AA192—331, was apparently inactivated for Notch-binding activity.

Conformation and/or primary sequence in the vicinity of Delta AA197/198 is apparently critical because a multimeric insertion of the tetrapeptide - Arg-Lys-Ile-Phe [in one letter code (see e.g.

Lehninger et a1., 1975, Biochemistry, 2d ed., p. 72), RKIF] (SEQ ID N0:30) - between these two residues, as in the Ins(Nae)A construct, inactivated the Notch- binding activity observed with wild type Delta.

In addition, the observation that the De1(ELRl-ELR3) construct supported aggregation implies that ELR1-ELR3 are not required for Delta—Notch interaction; the observation that the De1(ELR4-ELR5) construct supported aggregation implies that ELR4 and ELR5 are not required for Delta-Notch interaction, and the observation that the Ter(Dde) construct supported aggregation implies that the Delta intracellular domain is not required for Delta-Notch interaction. - 1oo - AMINO-TERMINAL SEQUENCES WITHIN THE DELTA EXTRACELLULAR DOMAIN ARE NECESSARY AND ' F0 0 PIC TERA The results of the homotypic aggregation experiments mediated by Delta variants is shown in Table VI. .2.2.

;; V; cc E on o v R N s Conggrggt Aggzegated Ugaggrggated Eggt. 1 wild type 38(H)‘ . 175 1 (H) 171 2 13(H) . 95 3 33(8) 173 4 134(3) 72 5 De1(Sca—Nae) 0(H) 200 1 O(H) 200 2 0(H) 200 3 De1(Bam-B91) 0(H) zoo 1 0(5) zoo 2 0(H) 200 3 De1(ELR1-ELR3) 160(8) 62 1 55(3) so 2 0(3) 200 3 4(3) 203 4 41(5) 234 5 4(3) 366 6’ 23(3) 325 (1:20) 0(3) 400 7" (5) 347 (1:5) (3) 223 (1:20) 0(B) 400 3" (3) 345 (1:5) (3) 268 (1:20) " 101 ' (3) 500 96 18(8) 500 (1:5) 12(8) 271 (1:20) 7(3) 128 (1:50) 0(5) 500’ 10° 0(3). 500 (1:5) 0(3) 500 (1:20) 21(3) 246 (1:50) 0(3) 500 11° (3) 500 (1:5) 3(3) 177 (1:20) 4(3) 69 (1:50), De1(ELR4-ELR5) 21(H) 175 ‘ 1 29(3) 243 2 (3) 179 3 Ter(Dde) 53(H) 164 ’ 1 33(3) 178. 2 36(H) 203 3 Ins(Nae)A 0(B) 200 1 0(3) 200 2 0(3) 200 3 (H) indicates that cells were aggregated in a 25 ml Erlenmeyer flask; (B) indicates that cells were aggregated in a 12-well microtiter plate (see Materials and Methods).

Transfected cells were incubated under aggregation conditions overnight, then diluted into the appropriate volume of log-phase S2 cells in the presence of inducer and incubated under aggregation conditions for an additional four to six hours.

Transfected cells to which inducer had been added were diluted into the appropriate volume of log- phase S2 cells to which inducer had been added, and the cell mixture was incubated under aggregation conditions overnight.

Deletion of Delta AA32-198 [Del(Sca-Nae)] or Delta AA192-331 [De1(Bam-Bgl)] from the full-length Delta_ protein eliminated the Delta-Delta interaction.

Deletion of Delta AA23l-331 [Del(ELR1éELR3)] did not eliminate the Delta-Delta interaction. Therefore, sequences within the Delta AA32-230 are required for the Delta-Delta interaction.

Conformation and/or primary sequence in the vicinity of Delta AA197/198 is apparently critical for the Delta-Delta interaction because a multimeric insertion of the tetrapeptide -Arg-Lys-Ile-Phe- (SEQ ID NO:30) between these two residues, as in the Ins(Nae)A construct, inactivated Delta-Delta interaction.

In addition, the observation that the De1(ELR1-ELR3) construct could support aggregation implies that ELR1~ELR3 are not required for Delta- Delta interaction; the observation that the De1(ELR- ELR5) construct supported aggregation implies that ELR4 and ELR5 are not required for Delta-Delta interaction, and the observation that the Ter(Dde) construct supported aggregation implies that the Delta intracellular domain is not required for Delta-Delta interaction.

A summary of the results of assays for heterotypic and homotypic aggregation with various constructs are shown in Table VI A. " 103 - -r B v AGGREGATION MEDIAN-.‘..D av WILD TYPE AND VARIANT DELTA PROTEINS mrrsnorvpxc HOMOTYPIC AGGREGA'1‘ION' AGGREGATION‘ cousraucr DELTA NOTCH DELTA wild Type 33 i 12* 26 1 11° 27 3 10* ‘Del(Sca-Nae) 0 0 0 Del(Bam-Bgl) 0.4 + 0.4 0.5 1 0.6 o Del(ELR1—ELR3) . 2s ; 11‘ 15; 3" 324315‘ De1(ELR4-ELR5) 17 i 2 13 J; 2 13 1- 2 Ter(Dde) 22 ; 1 18 it 2 13; 3 was 3 25 _~g 5 o 27 3 7 sum B o o 0 N61 0 o o NG2 13 :1 23:6 43:1‘ NG3 16 :1 13:1 27;17 N64 0 ¥ o 045 i 0.3 a: Mean fraction (%) of Delta or Notch cells in aggregates of four or more cells (3 standard error). N=3 replicates, unless otherwise noted. bx Mean fraction (%) of Delta cells in aggregates of four or more cells (1 standard error). N= 3 V replicates, unless otherwise noted. c: N = 5 replicates. d: N = 4 replicates. .2.3. DELTA SEQUENCES INVOLVED IN HETEROTYPIC AND HOMOTYPIC INTERACTIONS ARE QUALITATIVELY DISTINCT The respective characteristics of Delta sequences repaired for heterotypic and homotypic interaction were further defined using Delta variants in which short, in-frame, translatable linker insertions were introduced into the Delta amino terminus (i.e., NAE B and STU B; Figure 9, Table VI A). Replacement of Delta residue 132 (A) with the - 104 ~ pentapeptide GKIFP (STU B variant) leads to the inactivation of heterotypic and homotypic interaction activities of the Delta amino terminus. This suggests that some Delta sequences required for these two distinct interactions are coincident and reside in ' proximity to residue 132. on the other hand, insertion of the tetrapeptide RKIF between Delta residues 198 and 199 (NAB B variant) eliminates the ability of the Delta amino terminus to mediate heterotypic interaction with Notch, but has no apparent effect on the ability of the altered amino terminus to mediate homotypic interaction. The finding that the NAB B insertion affects only one of the two activities of the Delta amino terminus implies that the Delta sequences that mediate heterotypic and homotypic interactions, while coincident, are qualitatively distinct. aggregation experiments, we have noted that Delta protein can sometimes be found within cells that have been programmed to express Notch, but not Delta. We conduct heterotypic aggregation assays by mixing initially separate populations of S2 cells that have been independently transfected with expression constructs that program expression of either Delta or Notch. Yet, we often detect punctate staining of Delta within Notch-expressing cells found in heterotypic aggregates using De1ta~specific antisera. our observations are consistent with Delta binding directly to Notch at the cell surface and subsequent clearance of this Delta-Notch complex from the cell surface via endocytosis.

AMINO-TERMINAL SEQUENCES UNRELATED T0 EGF ARE INVOLVED IN THE INTERACTION BEIEEEN DELTA AND fl0ICfl We have employed cell aggregation assays to define a region within the amino-proximal region of the Delta extracellular domain that is necessary and .3.1. . sufficient to mediate the Delta-Notch interaction.

Functional analyses of a combination of deletion and sufficiency constructs revealed that this region extends, maximally, from Ahl through AA23o. It is striking that this region does not include any of the EGF-like sequences that reside within the Delta extracellular domain. It is probable that the particular Delta sequences within the sufficient interval required for interaction with Notch include AA198-230 because deletion of these residues eliminates Notch-binding activity. The fact that deletion of AA32-198 also inactivates Notch~binding activity suggests that sequences amino-proximal to AA198 are also required, although the deleterious impact of this deletion could result from the removal of additional amino acids in the immediate vicinity of AAl98.

Sequences within Delta sufficient for interaction with Notch can be grouped into three subdomains - N1, N2, and N3 — that differ in their respective contents of cysteine residues (Figure 10, SEQ ID NO:3). The N1 and N3 domains each contain six cysteine residues, while the N2 domain contains none.

The even number of cysteines present in N1 and N3, respectively, allows for the possibility that the respective structures of these subdomains are dictated, in part, by the formation of particular disulfide bonds- The broad organizational pattern of the Delta amino-terminus is also generally analogous ' 106 ' to that of the extracellular domain of the vertebrate EGF receptor (Lax et al., 1988, Mol. Cell. Biol. 3, 1970-1978), in which sequences believed to interact with EGF are bounded by two cysteine-rich subdomains. .3.2. DELTA SEQUENCES REQUIRED FOR HOMOTYPIC AND FOR HOMOTYPIC HETEROTYPIC ION B 0 N B Our results also indicate that sequences essential for homotypic Delta interaction reside within the interval AA32-230. or insertion of additional amino acids within this Deletion of sequences amino-proximal domain eliminate the ability of such Delta variants to singly promote cell aggregation.

Thus, sequences required for Delta-Delta interaction map within the same domain of the protein as those required for Delta-Notch interaction.

THE DELTA AMINO TERMINUS CONSTITUTES G -B ING T The work described in examples ggggg has revealed that Notch sequences required for Delta-Notch interaction in the cell aggregation assay map within the EGF-like repeat array of the Notch extracellular domain. This finding implies that Delta and Notch interact by virtue of the binding of the Delta amino- terminus to EGF-like sequences within Notch and, therefore, that the.amino-terminus of the Delta extracellular domain constitutes an EGF-binding domain (Figure 11).

These results also raise the possibility that homotypic Delta interaction involves the binding of the Delta amino-terminus to EGF~like sequences within the Delta extracellular domain (Figure 12).

However, none of the EGF-like repeats within the Delta extracellular domain are identical to any of the EGF- .3.3. ' 107 - like repeats within the Notch extracellular domain (Figure 13, SEQ ID NO:6; Wharton et al., 1985, Cell 43, 567-581). Given this fact, if Delta homotypic interactions are indeed mediated by interaction between the Delta amino-terminus and Delta BGF-like. repeats, then the Delta EGF-binding domain has the capacity to interact with at least two distinct EGF- like sequences.

DELTA SEQUENCES INVOLVED IN THE DELTA-NOTCH INTERACTION ARE CONSERVED IN THE SERRATE PROTEIN Alignment of amino acid sequences from the amino termini of the Delta (Figure 13, sea ID rims, and Figure 15, SEQ ID NO:9) and serrate (Fleming et al., 1990, Genes 8 Dev. 4, 2188-2201; Thomas et a1., 1991, Devel. 111, 749-761) reveals a striking conservation of structural character and sequence composition. The general N1-N2-N3 subdomain structure of the Delta amino terminus is also observed within the serrate amino terminus, as is the specific occurrence of six cysteine residues within the Delta N1- and Delta N3-homologous domains of the serrate protein. Two notable blocks of conservation correspond to Delta AA63-73 (8/11 residues identical) and Delta AA195-206 (10/11 residues identical). The latter block is of particular interest because insertion of additional amino acids in this interval can eliminate the ability of Delta to bind to Notch or Delta. .3.4. .3.5. CIS AND TRANS INTERACTIONS BETWEEN DELTA AND NOTCH MAY INVOLVE DIFFERENT SEQUENCES WITHIN NOICH Inspection of the overall structures of Delta and Notch suggests that Delta-Notch interaction could involve contacts between the Delta EGF-binding ' 108 - domain with either of two regions within Notch, depending on whether the interaction were between molecules that reside on opposing membranes or within the same membrane (Figure 11). The cell aggregation ~ assays, which presumably detect the interaction of molecules in opposing membranes, imply that the Delta EGF-binding domain interacts with Notch EGF-like repeats 11 and 12 (see examples guprg). If tandem arrays of EGF-like motifs do form rod-like structures (Engel, 1989, FEBS Lett. 251, 1-7) within the Delta and Notch proteins, then the estimated displacement of the Delta EGF—binding domain from the cell surface would presumably be sufficient to accommodate the rigid array of Notch EGF-like repeats 1°10. It is also intriguing to note that the displacement of the Delta EGF-binding domain from the cell surface could place this domain in the vicinity of the Notch EGF- like repeats (25-29) that are affected by Abruptex mutations (Hartley et al., 1987, EMBO J. 6, 3407-3417; Kelley et al., 1987, Mol. Cell. Biol. 6, 3094-3108) and could allow for interaction of Delta and Notch proteins present within the same membrane. .3.6. INTERACTIONS ANALOGOUS TO THE Given the interaction between Delta and Notch in Drosophila, it is quite probable that a Delta homologue (flelta?) exists in vertebrates and that the qualitative and molecular aspects of the Delta-Notch and Delta-Delta interactions that we have defined in Drosophila will be highly conserved in vertebrates, including humans. Such homologs can be cloned and sequenced as described supra, section 5.2.

. S C S T -S I B IO S ' 109 - we report a novel molecular interaction between Notch and serrate, and show that the two EGF repeats of Notch which mediate interactions with Delta, namely EGF repeats 11 and 12, also constitute a serrate binding domain.

To test whether Notch and serrate directly interact, S2 cells were transfected with a serrate expression construct and mixed with Notch expressing cells in our aggregation assay. For the serrate expression construct, a synthetic primer containing an artificial_BamHI site immediately 5' to the initiator AUG at position 442 (all sequence numbers are according to Fleming at al., 1990, Genes 8 Dev. 4:2188-2201) and homologous through position 464, was used in conjunction with a second primer from position 681-698 to generate a DNA fragment of ~26o base pairs.

This fragment was cut with BamHI and KpnI (position 571) and ligated into Bluescript KS+ (Stratagene).

This construct, BTser5'PCR, was checked by sequencing, then cut with KpnI, The serrate Kpnl fragment (571 - 2981) was inserted and the proper orientation selected, to generate BTSer5'PCR-Kpn. The 5' sacII fragment of BTSer5'PCR-Kpn (SacII sites in Bluescript polylinker and in serrate (1199)) was isolated and used to replace the 5' SacII fragment of cDNA Cl (Fleming et al., 1990, Genes & Dev. 4:2188-2201), thus _regenerating the full length gerrate cDNA minus the 5' untranslated regions. This insert was isolated by a SalI and partial BamHI digestion and shuttled into the Bamfll and SalI sites of pRmHa-3 to generate the final expression construct, Ser-mtn.

We found that serrate expressing cells adhere to Notch expressing cells in a calcium dependent manner (Figure 6 and Table VII). However, unlike Delta, under the experimental conditions - 110 ~ tested, serrate does not appear to interact homotypically. In addition, we detect no interactions between serrate and Delta.

IABLE VII Effect of Exogenous ca** on Notch - serrate Aggregation‘ Notch-serrate fiitnout ca** with cg** 1. pMtNMq 0 15 32 . AECN+EGF(l0-12) o 13 33 . ACla+XEGF(10-13) o 15 ‘ Data presented as percentage of Notch expressing All numbers are from single transfection experiments (rather than an average of values from several separate experiments as in Figure 6). cells found in aggregates (as in Figure 6).

We have tested a subset of our Notch deletion constructs to map the serrate-binding domain and have found that EGF repeats 11 and 12, in addition to binding to Delta, also mediate interactions with serrate (Figure 6; Constructs #1, 7-10, 13, 16, 17, 19, 28, and 32). In addition, the serrate-binding function of these repeats also appears to have been conserved in the corresponding two EGF repeats of Xenopus Notch (f33ACla+XEGF(10—13)). unambiguously show that Notch interacts with both Delta and serrate, and that the same two EGP repeats of Notch mediate both interactions. We were also able to define the serrate region which is essential for These results ' 111 ' the Notch/serrate aggregation. Deleting nucleotides 676-1287 (i.e. amino acids 79-282) (See Figure 15) eliminates the ability of the serrate protein to aggregate with Notch.

Notch and serrate appear to aggregate less efficiently than Notch and Delta, perhaps because the Notch-serrate interaction is weaker. For example, when scoring Notch-Delta aggregates, we detect ~40% of all Notch expressing cells in clusters with Delta expressing cells (Figure 6, #1 pMtNMg) and -40% of all Delta expressing cells in contact with Notch expressing cells. For Notch-serrate, we find only 920% of all Notch expressing cells (Figure 6; pMtNMg) and ~15? of all serrate expressing cells in aggregates. tested, we consistently detect a reduction in the amount of aggregation between Notch and serrate as compared to the corresponding Notch-Delta levels (Figure 6), with the possible exception of constructs #9 and 10 which exhibit severely reduced levels of aggregation even with Delta. one trivial explanation for this reduced amount of aggregation could be that our serrate construct simply does not express as much protein at the cell surface as the Delta construct, thereby diminishing the strength of the interaction.

Alternatively, the difference in strength of interaction may indicate a fundamental functional difference between Notch-Delta and Notch-serrate interactions that may be significant in 3139. .1. S F ‘N NO Clones for the human Notch sequence were originally obtained using the polymerase chain For the various Notch deletion constructs ' 112 ° reaction (PCR) to amplify DNA from a 17-18 week human fetal brain cDNA library in the Lambda Zap II Vector (Stratagene). Degenerate primers to be used in this reaction were designed by comparing the amino acid sequences of the ggggpgg homolog of ugtcn with Qrgsgphila flggch. Three primers (cdcl (SEQ ID N0:10), cdcz (SEQ ID Nozll), and cdc3 (SEQ ID NO:12); Figure 16) were designed to amplify either a 200 bp or a 400 bp fragment as primer pairs cdcl/cdc2 or cdcl/cdc3, respectively.

The 400 bp fragment obtained in this manner was then used as a probe with which to screen the same library for human ugtgh clones. The original screen yielded three unique clones, hN3k, hN2K, and hNSk, all of which were shown by subsequent sequence analysis to fall in the 3' end of human flgggh (Figure 17). A second screen using the 5' end of hN3k as probe was undertaken to search for clones encompassing the 5' end of human figtgh. one unique clone, hN4k, was obtained from this screen, and preliminary sequencing data indicate that it contains most of the 5' end of the gene (Figure 17). Together, clones hN4k, hN3k and hNSk encompass about 10 kb of the human figggh homolog, beginning early in the EGP-repeats and extending into the 3' untranslated region of the gene. All three clones are CDNA inserts in the EcoRI site of paluescript SK'(Stratagene). The host 3. 99;; strain is XLl-Blue (see Maniatis, T., 1990, Molecular Cloning, A Laboratory Manual, 2d ed., Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, p.

A12).

The sequence of various portions of flgtgn contained in the CDNA clones was determined (by use of Sequenase0, U.S. Biochemical Corp.) and is shown in Figures 19-22 (SEQ ID No:13 through N0:25).

The complete nucleotide sequences of the human Ngtgn cDNA contained in hN3k and hN5k was determined by the dideoxy chain termination method using the Sequenasew kit (U.S. Biochemical Corp.).

Those nucleotide sequences encoding human Notch, in the appropriate reading frame, were readily identified since translation in only one out of the three possible reading frames yields a sequence which, upon comparison with the published Drosophila Notch deduced amino acid sequence, yields a sequence with a substantial degree of homology to the Qrosophila Notch sequence. since there are no introns, translation of all three possible reading frames and comparison with Qzgsgpnilg Notch was easily accomplished, leading to the ready identification of the coding region. The DNA and deduced protein sequences of the human Notch cDNA in hN3k and hN5k are presented in Figures 23 and 24, respectively. Clone hN3k encodes a portion of a Notch polypeptide starting at approximately the third notch/113-12 repeat to several amino acids short of ‘ the carboxy-terminal amino acid. Clone hN5k encodes a portion of a Notch polypeptide starting approximately before the cdclo region through the end of the polypeptide, and also contains a 3' untranslated region.

Comparing the DNA and protein sequences presented in Figure 23 (SEQ ID NO:31 and N0:32) with those in Figure 24 (SEQ ID N0:33 and N0:34) reveals significant differences between the sequences, suggesting that hN3k and hN5k represent part of two distinct flgtgh-homologous genes. Our data thus suggest that the human genome harbors more than one Ngtgn-homologous gene. This is unlike Qggsgphila, where Notch appears to be a single-copy gene. comparison of the DNA and amino acid sequences of the human figtgh homologs contained in hN3k and hN5k with the corresponding Qrgsophila ngtgh sequences (as published in Wharton et a1., 1985, Cell 43:567~581) and with the corresponding xeggpus flgtgh sequences (as published in Coffman et al., 1990, science 249:1438~1441 or available from Genbank@q (accession number M33874)) also revealed differences.

. The amino acid sequence shown in Figure 23 (hN3k) was compared with the predicted sequence of the TAN-1 polypeptide shown in Figure 2 of Ellisen et al., August 1991, Cell 66:649-661. some differences were found between the deduced amino acid sequences; however, overall the hN3k Notch polypeptide sequence ~ is 99% identical to the corresponding TAN-1 region (TAN-1 amino acids 1455 to 2506). Four differences were noted: in the region between the third flgtgh/113-12 repeat and the first cdclo motif, there is an arginine (hN3k) instead of an X (TAN-1 amino acid 1763); (2) there is a proline (hN3k) instead of an X (TAN-1, amino acid 1787); (3) there is a conservative change of an aspartic acid residue (hN3k) instead of a glutamic acid residue (TAN-1, amino acid 2495); and (4) the carboxyl-terminal region differs substantially between TAN-1 amino acids 250? and 2535.

The amino acid sequence shown in Figure 24 (hN5k) was compared with the predicted sequence of the TAN—1 polypeptide shown in Figure 2 of Ellisen et al., August 1991, cell 66:649-661. Differences were found between the deduced amino acid sequences. The deduced Notch polypeptide of husk is 79% identical to the TAN- l polypeptide (64% identical to Qggggphilg Notch) in the cdclo region that encompasses both the cclo motif (TAN-1 amino acids 1860 to 2217) and the well- conserved flanking regions (Fig. 25). The cdclo region covers amino acids 1860 through 2217 of the TAN—1 sequence. In addition, the hN5k-encoded polypeptide is 65% identical to the TAN-1 polypeptide ,(44% identical to Drgsophila Notch) at the carboxy- terminal end of the molecule containing a PEST (proline, glutamic acid, serine, threonine)-rich region (TAN—l amino acids 2432 to 2551) (Fig. 25B).

The stretch of 215 amino acids lying between the aforementioned regions is not well conserved among any of the notch-homologous clones represented by hN3k, hN5k, and TAN-1. Neither the hN5k polypeptide nor Qrosgphila Notch shows significant levels of amino acid identity to the other proteins in this region (e.g., hN5k/TAN-1 = 24% identity; hN5k/Qrgggphilg Notch = 11% identity; TAN-1/Qrgggphila Notch = 17% identity). In contrast, Xenopgs Notch (Xotch) (SEQ ID NO:35), rat Notch (SEQ ID NO:36), and TAN-1 (SEQ ID No:37) continue to share significant levels of sequence identity with one another (e.g., TAN-1/rat Notch = 75% identity, TAN-1/genopus Notch = 45% identity, rat Notch/Xenopus Notch = 50% identity).

Finally, examination of the sequence of the intracellular domains of the vertebrate Notch homologs shown in Figure 25B revealed an unexpected finding: all of these proteins, including hN5k, contain a putative CcN motif, associated with nuclear targeting function, in the conserved region following the last of the six cdclo repeats (Fig. 25B). Although Drosophila Notch lacks such a defined motif, closer inspection of its sequence revealed the presence of a possible bipartite nuclear localization sequence (Robbins et al., 1991, Cell 64:6l5-623), as well as of possible CK II and cdcz phosphorylation sites, all in relative proximity to one another, thus possibly defining an alternative type of CcN motif (Fig. 25B). .2. EXPRESSION OF UMAN 0 C Expression constructs were made using the human Notch cDNA clones discussed in Section 10.1 In the cases of hN3k and hN2k, the entire clone was excised from its vector as an EcoRI restriction fragment and subcloned into the EcoRI restriction site of each of the three pGEx vectors (Glutathione S-Transferase expression vectors; Smith and Johnson, 1988, Gene 7, 31-40). This allows for the expression of the Notch protein product from the above. subclone in the correct reading frame. In the case of hNSk, the clone contains two internal EcoRI restriction sites, producing 2.6, 1.5 and 0.6 kb fragments. Both the 2.6 and the 1.5 kb fragments have also been subcloned into each of the pGEX vectors.

The pGEX vector system was used to obtain expression of human Notch fusion (chimeric) proteins The cloned Ngtgh DNA in each case was inserted, in phase, into the from the constructs described below. appropriate pGEX vector. Each construct was then electroporated into bacteria (5. ggli), and was expressed as a fusion protein containing the Notch protein sequences fused to the carboxyl terminus of glutathione S~transferase protein. Expression of the fusion proteins was confirmed by analysis of bacterial protein extracts by polyacrylamide gel electrophoresis, comparing protein extracts obtained from bacteria containing the pGEX plasmids with and without the inserted Notch DNA. The fusion proteins were soluble in aqueous solution, and were purified from bacterial lysates by affinity chromatography using glutathione—coated agarose (since the carboxyl terminus of glutathione s—transferase binds to The expressed fusion proteins were glutathionine). bound by an antibody to Drosophila Notch, as assayed by Western blotting.

The constructs used to make human Notch- glutathione S-transferase fusion proteins were as follows: hN£P£g ~ PCR was used to obtain a fragment starting just before the cdclo repeats at , nucleotide 192 of the hN5k insert to just before the PEST-rich region at nucleotide 1694. The DNA was then digested with BamHI and Smal and the resulting fragment was ligated into pGEx-3.

After expression, the fusion protein was purified by binding to glutathione agarose. The purified polypeptide was quantitated on a 4-15% gradient polyacrylamide gel. The resulting fusion protein had an approximate molecular weight of 83 kD. hN3Fg£l - The entire hN3k DNA insert (nucleotide 1 to 3235) was excised from the Bluescript (SK) vector by digesting with EcoRI.

The DNA was ligated into pGEX-3. hN3Eg£3 — A 3' segment of hN3k DNA (nucleotide 1847 to 3235) plus some of the polylinker was cut out of the Bluescript (SK) vector by digesting with Xmal. The fragment was ligated into pGEX—1.

Following purification, these fusion proteins are used to make either polyclonal and/or monoclonal antibodies to human Notch.

. QEPOSIT OF HICRO0RGANISflS The following recombinant bacteria, each carrying a plasmid encoding a portion of human Notch, were deposited on May 2, 1991 with the American Type Culture Collection, 1201 Parklawn Drive, Rockville, Maryland 20852, under the provisions of the Budapest Treaty on the International Recognition of the Deposit of Microorganisms for the Purposes of Patent Procedures.

Bacteria carrying Elgsmid AIQC hccessiog 39.

E. coli XL1-Blue hN4k 68610 E. coli XL1—B1ue hN3k 68609 E. 09 1 XL1-Blue hN5k 68611 The present invention is not to be limited in scope by the microorganisms deposited or the specific embodiments described herein. Indeed, various modifications of the invention in addition to those described herein will become apparent to those skilled in the art from the foregoing description and accompanying.figures. such modifications are intended to fall within the scope of the appended claims.

Various publications are cited herein, the disclosures of which are incorporated by reference in their entireties.

Claims (2)

Claims

1. A purified Notch protein encoded by a first nucleic acid that is obtainable from human DNA or cDNA and that is hybridisable to a second nucleic acid consisting of the human Notch sequence, or its complement, contained in (a) plasmid hN3k as deposited with the ATCC and assigned accession number 68609, or (b) plasmid hN5k as deposited with the ATCC and assigned the accession number 68611; or a purified fragment of said human Notch protein; or which protein, or fragment is capable of binding to a Delta protein or is capable of binding to an antibody, specific to human Notch protein.

2. A purified protein comprising the amino acid sequence depicted in