Oncogenic forms of NOTCH1 lacking either the primary binding site for RBP-Jkappa or nuclear localization sequences retain the ability to associate with RBP-Jkappa and activate transcription.

Truncated forms of the NOTCH1 transmembrane receptor engineered to resemble mutant forms of NOTCH1 found in certain cases of human T cell leukemia/lymphoma (T-ALL) efficiently induce T-ALL when expressed in the bone marrow of mice. Unlike full-sized NOTCH1, two such truncated forms of the protein either lacking a major portion of the extracellular domain (ΔE) or consisting only of the intracellular domain (ICN) were found to activate transcription in cultured cells, presumably through RBP-Jκ response elements within DNA. Both truncated forms also bound to the transcription factor RBP-Jκ in extracts prepared from human and murine T-ALL cell lines. Transcriptional activation required the presence of a weak RBP-Jκ-binding site within the NOTCH1 ankyrin repeat region of the intracellular domain. Unexpectedly, a second, stronger RBP-Jκ-binding site, which lies within the intracellular domain close to the transmembrane region and significantly augments association with RBP-Jκ, was not needed for oncogenesis or for transcriptional activation. While ICN appeared primarily in the nucleus, ΔE localized to cytoplasmic and nuclear membranes, suggesting that intranuclear localization is not essential for oncogenesis or transcriptional activation. In support of this interpretation, mutation of putative nuclear localization sequences decreased nuclear localization and increased transcriptional activation by membrane-bound ΔE. Transcriptional activation by this mutant form of membrane-bound ΔE was approximately equivalent to that produced by intranuclear ICN. These data are most consistent with NOTCH1 oncogenesis and transcriptional activation being independent of association with RBP-Jκ at promoter sites.

NOTCH1 (also referred to as TAN-1), a homolog of Drosophila NOTCH, was first identified at the breakpoint of a recurrent (7;9)(q34;q34.3) chromosomal translocation associated with a subset of human T cell acute lymphoblastic leukemia/ lymphoma (T-ALL) 1 (1)(2)(3). RNA transcribed from the normal NOTCH1 gene is found in most cells but is present at highest levels in developing thymus and brain (2,4). The product of this gene is a 2555 amino acid type I transmembrane receptor protein of about 350 kDa (p350) containing a series of structural motifs originally described in other polypeptides (Fig. 1), including 36 epidermal growth factor-like repeats and three lin-12-like repeats in the extracellular domain and six ankyrinlike repeats in the intracellular domain. All chromosomal breakpoints in human T-ALLs occur within a single intron dividing the coding sequence for epidermal growth factor repeat 34 and result in overexpression of novel mRNAs containing sequences from the 3Ј end of NOTCH1 (5). These transcripts encode a series of truncated NOTCH1 polypeptides of ϳ100 -125 kDa lacking the epidermal growth factor-like and lin-12-like repeats (5).
The oncogenicity of truncated NOTCH1 has been confirmed by inserting cDNAs encoding portions of the NOTCH1 gene into murine bone marrow progenitor cells which were then transplanted into syngeneic recipients (6). In these experiments, NOTCH1 missing the entire extracellular domain except for the leader peptide and 61 amino acids immediately external to the transmembrane region (⌬E), or deleted for all protein sequence up to amino acid 14 of the intracellular domain (ICN) (shown schematically in Fig. 1) induced T-ALL in ϳ50% of the recipient animals. In contrast, normal, full-length NOTCH1 produced no tumors in similar transplant experiments. The tumors induced by ⌬E contained an abundant ϳ120-kDa NOTCH1 that localized to perinuclear and cytoplasmic membranes, while tumors induced by ICN contained an abundant ϳ100 -115 kDa NOTCH1 within the nucleus. Therefore, different amino-terminal deletions created cytoplasmic and nuclear forms of truncated NOTCH1 that appeared to be equally oncogenic.
Removal of extracellular amino acid sequences from NOTCH1 and NOTCH-related proteins has effects on differentiation similar to those observed when cells expressing fulllength versions of these proteins are exposed to ligand (7)(8)(9)(10)(11), findings consistent with the interpretation that truncation leads to constitutive activation of the intracellular domain.
Signaling by NOTCH1 appears to be mediated, at least in part, by the transcription factor RBP-J (12)(13)(14)(15), which is homologous to the proteins Su(H) in the Drosophila and lag-1 in Caenorhabditis elegans. Evidence for this functional link between the proteins derives from genetic studies which have placed Drosophila NOTCH and its nematode counterparts, lin-12 and glp-1 (16), upstream of Su(H) (17)(18)(19) and lag-1 (20, 21), respectively. Furthermore, the murine HES1 promoter activated by NOTCH1 contains an RBP-J-binding site (22), suggesting that RBP-J is also downstream of activated NOTCH1.
When overexpressed, fragments of NOTCH1 representing the intracellular domain can associate with RBP-J bound to DNA (22), and when tethered to a promoter by covalent linkage to a DNA-binding protein, NOTCH1 has transactivating effects (5,23), indicating that NOTCH1 might transactivate through physical interaction with RBP-J bound to promoter elements. If this mechanism of transactivation were involved in NOTCH1 transformation, it might be expected that subcellular localization would influence oncogenicity. However, ⌬E transforms as well as ICN despite its cytoplasmic localization. Moreover, based on studies performed in transiently overexpressing cells, NOTCH1 association with RBP-J has been reported to require a tryptophan residue (Trp 1767 ) lying 10 amino acids internal to the transmembrane domain that is absent from ICN (24). These discrepancies raised the possibility that NOTCH1 oncogenicity might not involve RBP-J or nuclear localization. To investigate these possibilities, we have evaluated the physical and functional interaction of RBP-J and NOTCH1 in a variety of cell types, including NOTCH1-induced T-ALL cell lines.

EXPERIMENTAL PROCEDURES
Antibodies-Production, purification, and characterization of polyclonal rabbit antibodies against two portions of the intracellular domain of NOTCH1 (25), T3 (amino acids 1763-1877) and TC (amino acids 2277-2470), and an amino-terminal portion of RBP-J (26), have been previously described.
Cell Lines-Cell culture reagents were obtained from Life Technologies. All cell lines were cultured at 37°C under 5% CO 2 . SUP-T1, a human T lymphoblastic cell line containing two copies of a balanced (7;9)(q34;q34.3) chromosomal translocation and no normal chromosome 9 (27) was grown in RPMI 1640 supplemented with 10% fetal calf serum. T6E and I22 cells, derived from murine tumors induced by cDNAs encoding ⌬E and ICN (6), respectively, were grown in RPMI supplemented with 20% fetal calf serum and interleukin-2 (4 units/ml). Human embryonic kidney 293 cells and two cell lines derived from 293 cells, 293T and Bosc23 (28), were all grown in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum.
cDNA Expression Constructs-The cDNA constructs NOTCH1 (codons 1-2555), ⌬E (codons 1-22 fused to codons 1673-2555), ⌬E⌬AR (a subclone of ⌬E with a deletion removing codons 1858 -2206), ICN (codons 1770 -2555), and ICN⌬N1⌬N2 (a subclone of ICN with two deletions removing codons 1771-1857 (⌬N1) and codons 2097-2204 (⌬N2)) have been previously described (5,6). Additional subclones with mutations and small deletions in potential nuclear localization signal sequences were made by ligating polymerase chain reaction products with compatible restriction sites to the ICN and ⌬E constructs. The construct ICNW was created by ligating a double stranded DNA linker containing nucleotides Ϫ15 to ϩ6 (relative to the NOTCH1 start codon) to the ICN construct after digestion with the restriction enzymes Bsu36I and BamHI. A cDNA construct encoding RBP-J3 in the eukaryotic expression plasmid pSG5 has been previously described (15). To express RBP-J3 with a myc epitope tag, the cDNA was excised from pSG5 as a BamHI-SfcI restriction fragment and subcloned using a SfcI/NotI adaptor into a derivative of pcDNA3 that encodes three iterated copies of a myc epitope. The resultant fusion cDNA encodes RBP-J3 with myc epitopes replacing the six most carboxyl-terminal amino acids (RBP-J3-myc).
Production of GST Fusion Polypeptides-Expression and purification of GST-NOTCH1 ankyrin repeat (5) (codons 1872-2123) and GST-RBP-J3 (26) fusion proteins have been described. A cDNA spanning the region encoding the NOTCH2 ankyrin repeats (codons 9 -271 in the sequence of Stifani et al. (29)) was amplified by polymerase chain reaction using a human umbilical vein encdothelial cell cDNA library (30) as template. After subcloning into pGEX-2TK (Pharmacia), GST-NOTCH2 ankyrin repeat fusion protein was expressed in Escherichia coli and purified on glutathione-Sepharose beads (Pharmacia) as described for the NOTCH1 ankyrin repeats (5).
In Vitro Binding Assays-NOTCH1 cDNAs in pSP72 (Promega) and RBP-J3 cDNA in pSG5 were transcribed and translated in vitro in the presence of [ 35 S]methionine using rabbit reticulocyte lysates (Promega). [ 35 S]Methionine-labeled polypeptides were stored at Ϫ80°C. Binding of in vitro translated polypeptides to GST fusion proteins was performed under conditions described by Jarriault et al. (22). Briefly, 5 l of labeling mixture containing polypeptides translated in vitro were diluted into 0.5 ml of ice-cold 25 mM Hepes, pH 7.5, containing 0.5% Nonidet P-40, 100 mM KCl, 5 mM MgCl 2 , 5 mM dithiothreitol, 0.2 mM EGTA, 10 g/ml aprotinin, 10 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride (binding buffer). This mixture was precleared by incubation with 50 l of GST-glutathione Sepharose beads at 4°C with mixing for 60 min. After removal of the beads by brief centrifugation in a microcentrifuge, the remaining supernatant was divided into two aliquots of 0.25 ml and incubated with either GST fusion proteinglutathione Sepharose beads or a equivalent volume of GST-glutathione Sepharose beads at 4°C with mixing for 60 min. Beads were washed twice with binding buffer, once with binding buffer containing 200 mM KCl, and once more with binding buffer, and then boiled in SDS-PAGE loading buffer. 35 S-Labeled polypeptides were separated by SDS-PAGE in 10% gels and visualized by fluorography.
Yeast Two-hybrid Analysis-Two-hybrid analysis was performed as described by Fields and Song (31). A 1.1-kilobase NcoI fragment encoding the NOTCH1 ankyrin repeats and immediate flanking sequences (codons 1858 -2206) was subcloned into the NcoI site of the plasmid pAS2, while a NcoI-BamHI fragment containing the complete cDNA for RBP-J was subcloned into the NcoI-BamHI site of the plasmid pAC-TII. Yeast strain Y190 was transformed with pAS2-AR1 and selected on TrpϪ plates, and then retransformed with pACTII-RBP-J3 and selected on Trp-Leu-agar plates in the presence of 25 mM 3-aminotriazole. Resultant colonies were then stained for ␤-galactosidase activity using 5-bromo-4-chloro-3-indoyl ␤-D-galactoside as substrate.
Eukaryotic Expression Studies-For retroviral transduction, NOTCH1 cDNAs were subcloned into pBABEpuro (32) and packaged by transient transfection of Bosc23 cells (6). After addition of Polybrene (4 g/ml), the resultant supernatants containing retrovirus were used to infect NIH 3T3 cells. For transient expression, cDNAs were subcloned into pcDNA3 (Invitrogen) and transfected into cultured cells by calcium phosphate precipitation (28).
Chloramphenicol Acetyltransferase (CAT) Assays-Various NOTCH1 cDNAs cloned into pcDNA3 were used in co-transfections of 293 cells along with pSG5-RBP-J3, a CAT reporter containing a concatamerized RBP-J-binding sequence derived from the DNA of the Epstein-jBarr virus Cp1 promoter (33), and an internal control cDNA encoding ␤-galactosidase expressed from a glucokinase housekeeping promoter. Cell extracts were routinely prepared 48 h post-transfection and analyzed for ␤-galactosidase and CAT activity according to standard protocols (34). CAT activities were normalized with respect to ␤-galactosidase activity and quantitated by comparison to results of control transfections performed with pcDNA3 lacking a cDNA insert and the reporter construct.
Immunoprecipitation-Unless otherwise noted, all incubations described below were performed at 4°C with mixing. To prepare immu- noprecipitates with polyclonal rabbit antibodies, cells were washed twice with ice-cold Hank's buffered saline and then lysed in ice-cold 25 mM Hepes, pH 7.5, containing 1% Nonidet P-40, 70 mM KCl, 20 mM sodium fluoride, 2 mM sodium orthovanadate, 2 mM sodium molybdate, 0.2 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin (lysis buffer). After 30 min on ice, insoluble material was removed by centrifugation at 4°C and 14,000 ϫ g for 15 min. Supernatants were divided into two aliquots which were incubated with 3 l of rabbit serum or 10 l of affinity purified rabbit antibody for 60 min on ice. Ten l of protein A-Sepharose beads (Pharmacia) were then added and the incubation allowed to proceed for an additional 60 min at 4°C with mixing. To prepare immunoprecipitates with anti-myc monoclonal antibody 9E10, 2 l of ascites was pre-bound to 10 l of protein A-Sepharose in 200 l of lysis buffer by mixing for 2 h at 4°C. Cell lysates were precleared by incubation with 10 l of protein A-Sepharose for 30 min and then allowed to mix with 9E10 antibody bound to beads for 2 h. Beads were washed four times with lysis buffer and bound polypeptides eluted by boiling in SDS-PAGE buffer. Immunoprecipitated polypeptides were then separated by SDS-PAGE and analyzed by Western blotting.
Immunolocalization-Cells grown on slides were washed and fixed with 3% paraformaldehyde in phosphate-buffered saline for 15 min at room temperature. For immunoperoxidase staining of NIH 3T3 cells, cells were permeabilized with Ϫ20°C methanol for 1 min and then stained with rabbit anti-NOTCH1 as described previously (5). For immunfluorescent localization, 293T cells were stained with affinity-purified rabbit anti-NOTCH1 (4 g/ml) followed by goat anti-rabbit antibody linked to fluorescein isothiocyanate (1:1000) in the presence of 0.1% Triton X-100, using the conditions described by Kopan et al. (35). Confocal and DIC microscopy was performed on Zeiss epifluorescence microscope equipped with a MRC600 confocal unit (Bio-Rad). For each field imaged, both confocal fluorescence and DIC images were collected using SOM software (Bio-Rad). Image files were translated to TIFF format using NIH Image software and composite figures were subsequently created using Adobe Photoshop 2.5.

NOTCH1 and RBP-J Are Physically Associated in T Cell
Lines-To assess the interaction of RBP-J and oncogenic forms of NOTCH1, immunoprecipitates were prepared from lysates of human SUP-T1 cells, which contain two copies of the t(7;9) and no normal chromosome 9, from lysates of T6E cells, which are derived from a T cell tumor induced by a retrovirus containing the ⌬E cDNA, and from lysates of I22 cells, which are derived from a T cell tumor induced by a retrovirus containing the ICN cDNA.
Human T cells contain two major RBP-J isoforms, J1 and J3, created by alternative splicing of the 5Ј exon. Western blot analysis of immunoprecipitates prepared from SUP-T1 cells ( Fig. 2A) demonstrated that RBP-J3 and trace amounts of RBP-J1 were co-precipitated by affinity-purified antibodies against two different portions of the intracellular domain of NOTCH1, while neither isoform was precipitated by control affinity-purified antibody against GST. Since approximately equal amounts of RBP-J1 and RBP-J3 were precipitated by anti-RBP-J, enrichment for RBP-J3 in the NOTCH1 immunoprecipitates probably reflects a greater association of RBP-J3 with NOTCH1, as has been previously noted with Epstein-Barr virus transcriptional regulators EBNA2 and EBNA3C (36). RBP-J antiserum precipitated only RBP-J3 from T6E and I22 lysates, and RBP-J3 co-precipitated with anti- Equal amounts of lysates from SUP-T1 (SUPT-1), T6E, and I22 cells were incubated with antiserum against RBP-J, affinity-purified anti-GST antibody, or affinity-purified anti-NOTCH antibody directed against the T3 or TC regions, followed by the addition of protein A-Sepharose. Immunoprecipitated proteins were eluted, subjected to SDS-PAGE in 8% gels, and transferred to nitrocellulose. A composite Western blot is shown that was stained with RBP-J antiserum. Input represents 1% of total protein. The position of bands for RBP-J and immunoglobulin heavy chain (HC) are indicated. B, immunoprecipitation of NOTCH1 with antiserum to RBP-J. Lysates were incubated with antiserum to RBP-J or preimmune serum and immunoprecipitates were collected and analyzed as described under A, except that electrophoresis was performed in a 6% gel. A composite Western blot is shown that was stained with affinity-purified NOTCH1 antibody against TC. NOTCH1 antibody but not with anti-GST antibody. The extent of RBP-J association with NOTCH1 forms in SUP-T1, T6E, and I22 cells was substantial, since NOTCH1 antibody precipitated much more than 1% of the RBP-J3 and approximately as much as with the RBP-J antibody.
In converse experiments, RBP-J-specific antibody immunoprecipitated NOTCH1 (Fig. 2B). From SUP-T1 cell lysates, multiple t(7;9)-specific polypeptides of 100 -125 kDa co-precipitated with RBP-J-specific antibody. From T6E lysates, a 120-kDa NOTCH1 corresponding in size to ⌬E co-precipitated with RBP-J antibody, as did smaller amounts of a 350-kDa crossreactive polypeptide. The 350-kDa polypeptide was recognized by multiple NOTCH1 antibodies and is identical in size to human p350 (not shown). Several NOTCH1 polypeptides from I22 lysates were also co-precipitated with RBP-J antibody, with the largest being slightly smaller than ⌬E, as expected for ICN.
The Ankyrin Repeat Region of NOTCH1 Binds RBP-J3-ICN precipitated with RBP-J antibody even though it lacks the residue Trp 1767 previously shown to be important in NOTCH1 association with RBP-J both in vitro and in transiently transfected cells (24). To investigate an alternative ba-sis for the association of ICN with RBP-J, various forms of NOTCH1 (shown schematically in Fig. 3A) were expressed in vitro and tested for binding to RBP-J3-GST fusion protein immobilized on glutathione-Sepharose beads.
As expected, ICNW, a polypeptide consisting of the entire intracellular domain of NOTCH1 and differing from ICN by only nine amino acids at the amino terminus (Fig. 3A), bound RBP-J-GST more strongly than ICN (Fig. 3B). However, when compared with the level of binding observed with control GST beads, ICN also bound specifically to RBP-J-GST (Fig. 3C). Binding was enhanced by deletion of regions flanking the ankyrin repeats (ICN⌬N1⌬N2 in Fig. 3C) and abolished by a deletion that removed the ankyrin repeats (ICN⌬AR in Fig.  3C). The isolated ankyrin repeats of NOTCH1 and of a closely related human homolog, NOTCH2, also bound in vitro translated RBP-J in amounts above that observed with the GST control (Fig. 3D), indicating the presence of a weak RBP-Jbinding site within the ankyrin repeat region that has been conserved during divergence of these two mammalian NOTCH receptors. The ankyrin repeat region of NOTCH1 was also found to interact strongly with RBP-J in the yeast two-hybrid assay (Table I) A, lysates prepared from 293T cells co-transfected 10 g of pcDNA3-RBP-J3myc and 5 g of pcDNA3-NOTCH1 (1), pcDNA3-⌬E (2), pcDNA3-ICNW (3), or pcDNA3-ICN (4) were divided into two aliquots that were incubated with either protein A-Sepharose beads or protein A-Sepharose beads containing pre-bound anti-myc 9E10 antibody. Precipitated polypeptides were separated by electrophoresis in a 8% polyacrylamide gel and then transferred to nitrocellulose. A Western blot is shown that was stained with anti-T3 NOTCH1 antibody (upper panel), then stripped and restained with antiserum to RBP-J3 (lower panel). The input lane contains 1% of the total lysate. In B, three dishes of 293T cells were transfected with 5 g of pcDNA3-NOTCH1, 10 g of pcDNA3-RBP-J3myc, or 5 g of pcDNA3-NOTCH1 and 10 g of pcDNA3-RBP-J3myc. Cells transfected with pcDNA3-NOTCH1 or pcDNA3-RBP-J3myc alone were subsequently mixed into a single cell pellet. Lysates of the mixture of singly transfected cells (1) and of co-transfected cells (2) were prepared in a large volume (5 ml) to minimize association ex vivo. Polypeptides immunoprecipitated from lysates with anti-myc 9E10 antibody were separated by electrophoresis in a 8% polyacrylamide gel and transferred to nitrocellulose. The resultant Western blot was divided into an upper half that was stained with anti-T3 NOTCH1 antibody (upper panel) and a lower half that was stained with anti-myc 9E10 antibody (lower panel). The input lane contains 0.4% of the total lysate.  ␤-galactosidase reporter gene. In contrast, no ␤-galactosidase expression was observed when a control cDNA encoding an unrelated protein, yeast SNF1, was substituted for AR1 or RBP-J3. Therefore, the ankyrin repeat region of NOTCH1 contains a second binding site for RBP-J that could mediate stable association of ICN and RBP-J in I22 cells.

Trp 1767 Is Needed for Stable Association of NOTCH1 and RBP-J in Transiently
Expressing 293 Cells-To further investigate the RBP-J binding sites of NOTCH1, immunoprecipitates were prepared from lysates of 293 cells transiently expressing myc epitope-tagged RBP-J3 and various forms of NOTCH1. NOTCH1, ⌬E, and ICNW all co-precipitated with anti-myc antibody (Fig. 4A, lanes 1-3, respectively), as did ⌬E⌬AR (not shown), while ICN did not co-precipitate in detectable amounts (Fig. 4A, lane 4). These data indicate that the binding site containing Trp 1767 is necessary and sufficient for stable association in transiently expressing cells. In contrast, stable association of RBP-J through the ankyrin repeat region is not detectable in transiently expressing 293 cells.
Co-precipitation of membrane-bound forms of NOTCH1 (fullsized NOTCH1 and ⌬E) in T cell lines and transiently expressing 293 cells suggested that RBP-J can associate with NOTCH1 in the cytoplasm. Alternatively, association might occur ex vivo after cell lysis. To distinguish between these possibilities, immunoprecipitates were prepared from lysates of cells co-expressing NOTCH1 and RBP-J3 (Fig. 4B, lane 2) and mixtures of cells expressing either full-size NOTCH1 or RBP-J3 (Fig. 4B, lane 1). Co-precipitation of NOTCH1 and RBP-Jmyc with myc antibody was only observed in lysates prepared from co-expressing cells, consistent with RBP-J3⅐NOTCH1 complexes being formed in the cytoplasm of intact cells.
Oncogenic Forms of NOTCH1 Activate Transcription from a RBP-J-sensitive Promoter Element-The roles of the two RBP-J-binding regions and of the nuclear localization signal se-quences in transcriptional activation were characterized using the RBP-J binding element from the Epstein-Barr virus Cp1 promoter (33). Transient expression of ⌬E and ICN in 293 cells activated transcription 8 -12-fold, whereas full-length NOTCH1 had no effect (Fig. 5A). Although a plasmid encoding RBP-J3 was routinely included in these transfections, similar results were obtained when this plasmid was excluded, probably because endogenous RBP-J levels in 293 cells are sufficient for full activation. 2 ICN was slightly more active than ⌬E in all experiments (p Ͻ 0.06). ⌬E⌬AR, deleted for the ankyrin repeats, had only 20% of the activity of ⌬E, consistent with a key role for the ankyrin repeat region in transcriptional activation. In contrast, ICNW, ICN, and ICN⌬N1 stimulated transcription to a similar degree (Fig. 5B), despite the absence of Trp 1767 from ICN and ICN⌬N1. Therefore, the strong RBP-Jbinding site around Trp 1767 is not required for transcriptional or oncogenic activity.
Transcriptional Activation Is Not Dependent on Nuclear Localization-Activation of transcription using the RBP-J binding element of Cp1 promoter by ⌬E and ICN, but not by NOTCH1, indicated that deletion of the extracellular domain resulted in activation of transcription through RBP-J. Since ⌬E localizes predominantly to cytoplasmic and nuclear membranes, intranuclear localization may not be required for its effects. Alternatively, a small amount of ⌬E, but not NOTCH1, might be cleaved within the intracellular domain, freeing it from membranes and enabling it to up-regulate transcription through RBP-J bound to DNA. If NOTCH1 must first translocate to the nucleus to be active, mutation of the putative NOTCH1 nuclear localization sequences (NLSs) should inhibit transactivation and ICN should be a more potent transactivator than ⌬E. To test these predictions, the nuclear localization 2 J. C. Aster, unpublished data. signal sequences of NOTCH1 were mapped by mutational analysis in NIH 3T3 cells (Fig. 6, A-E, summarized in Fig. 6F), and NLS mutations were then evaluated for effects on transactivation by ICN and ⌬E.
Previous work has indicated the existence of at least one sequence amino-terminal and one sequence carboxyl-terminal of the ankyrin repeats with NLS activity (5,37). The N1 region amino-terminal to the ankyrin repeats contains two potential NLSs, Lys 1779 -Lys-Lys-Arg-Arg 1783 and Lys 1820 -Lys-Phe-Arg-Phe-Glu 1825 , while the region carboxyl-terminal to the ankyrin repeats contains two closely spaced stretches of basic amino acids, Lys 2156 -Lys-Val-Arg-Lys 2160 and Lys 2177 -Ala-Arg-Arg-Lys-Lys 2182 . When coupled with a deletion termed ⌬N2.1 that removed amino acids 2158 -2206, mutation of the sequence Lys 1820 -Lys-Phe-Arg-Phe-Glu 1825 to LEFRFE (termed MN1) led to a marked increase in cytoplasmic localization (Fig. 6C), with most cells showing approximately equal cytoplasmic and nuclear staining. Amino acid residues 1820 -1825 (designated N1) are conserved among all members of the vertebrate NOTCH receptor family (Table II). In contrast, deletion of the second amino-terminal basic sequence, Lys 1779 -Lys-Lys-Arg-Arg 1783 , had no additional effect on localization, either in concert with ⌬N2.1 alone or together with MN1 and ⌬N2.1 (not shown).
When coupled with MN1, mutation Lys 2156 -Lys-Val-Arg-Lys 2160 to KEFRK (designated MN2a, Fig. 6D) or mutation of Lys 2177 -Ala-Arg-Arg-Lys-Lys 2182 to KARRGT (designated MN2c, Fig. 6E) increased cytoplasmic localization, indicating that these two basic sequences may function as a bipartite NLS. This sequence, designated N2, is also conserved (Table  II). Neither mutation or deletion of both N1 and N2 completely abolished nuclear staining, consistent with previous reports that the ankyrin repeat region has some intrinsic capacity for nuclear localization (37). No attempt was made to further reduce nuclear localization through ankyrin repeat mutations, since involvement of this region in binding and activation of RBP-J would confound interpretation of "loss-of-function" mutations.
To produce a truncated NOTCH1 with limited (if any) capacity for nuclear localization, MN1 and ⌬N2.1 were subcloned into ⌬E to create ⌬EMN1⌬N2.1. Localization of various forms of NOTCH1 in transiently expressing 293 cells was then determined by confocal microscopy (Fig. 7). Inactive, full-sized NOTCH1 localized to perinuclear regions and cytoplasmic vesicles (Fig. 7A). Cells expressing ⌬EMN1⌬N2.1 showed staining of nuclear membrane and variable perinuclear staining compatible with endoplasmic reticulum and/or Golgi, with no detectable nuclear staining (Fig. 7B). In contrast, ICNW produced exclusively nuclear staining (Fig. 7C).
The effect of NLS mutations on transcriptional activation was then determined in transient transfection studies using   (Fig. 8A). ICNMN1⌬N2.1 had similar activity to ICN, whereas ⌬EMN1⌬N2.1 unexpectedly had consistently increased activity relative to ⌬E (p Ͻ 0.05). This increase in activity was probably not due to diminished nuclear localization per se, because a derivative of ⌬EMN1⌬N2.1 with a functional SV40 NLS sequence inserted into the ⌬N2.1 site maintained levels of transactivation similar to ⌬EMN1⌬N2.1 (not shown). To compare the potency of nuclear and membranebound versions of truncated NOTCH1, transactivation by ICNW and ⌬EMN1⌬N2.1 was measured over a range of input plasmid (Fig. 8B). Both constructs had similar dose-response curves, with a trend toward slightly higher stimulation with ICNW at low input levels of plasmid. However, ⌬EMN1⌬N2.1 (Fig. 8C, lane 1) also produced less protein than ICNW at low levels of input plasmid, (Fig. 8C, lane 2), suggesting that minor differences in transactivation could result from differences in protein level rather than differences in potency. These data indicate that both intranuclear and extranuclear forms of truncated NOTCH1 transactivate the Cp1 promoter, suggesting that transactivation may not require association of NOTCH1 and RBP-J at promoter sites.

DISCUSSION
Recurrent, specific chromosomal translocations found in tumors frequently exert their oncogenic effects by increasing some normal activity associated with the proteins encoded by genes located at the translocation breakpoints. It seems likely that this theme will extend to the role of truncated NOTCH1 in T cell neoplasia since normal NOTCH1 is highly expressed in thymocytes; oncogenic forms of NOTCH1 exhibit "gain-of-function" activity in various assays of transcription and differentiation; and these gain-of-function forms are particularly oncogenic in T cell progenitors. However, the precise mechanism by which truncated NOTCH1 contributes to tumor development is not clear and may be complex. Besides interacting physically and functionally with RBP-J, NOTCH1 also interacts with components of the NF-B signaling pathway (5,38) and some NOTCH phenotypes in the fly appear to be independent of Su(H) (18). The data presented here indicate that the oncogenic capacity of truncated NOTCH1 may be mediated at least in part by RBP-J. In this regard, it is relevant that four Epstein-Barr viral proteins required for B cell transformation, EBNA2 (39), EBNA3A, EBNA3B, and EBNA3C (26), bind to and alter the activity of RBP-J (14, 15, 26, 40 -47). Extensive binding of these transforming proteins with RBP-J is also found in Epstein-Barr virus-transformed B lymphocytes (36). Since viral transforming proteins often associate with and dysregulate molecules which seem to have general functions in the control of normal cell proliferation, it may be that RBP-J has a critical role in T cell transformation as well as in B cell transformation induced by Epstein-Barr virus. For similar reasons, dysregulation of RBP-J may be involved in neoplasms beyond the lymphoid system, and abnormal NOTCH1 signaling could be a factor in this process.
Based on our data, the ankyrin repeat region of NOTCH1 is important for transactivation of promoters, whereas neither the strong RBP-J binding site around amino acid residue Trp 1767 (a domain termed RAM23 by Tamura et al. (24)) nor nuclear localization of NOTCH1 is required. A major role for the ankyrin repeats in the effects of NOTCH1 has been previously indicated by studies showing that a point mutation in the fourth ankyrin repeat prevented transactivation of RBP-J (22) and inhibition of myogenesis (48) by NOTCH1, and that a peptide consisting of the ankyrin repeats of glp-1 and immediate flanking sequences was sufficient to produce a gain-offunction phenotype during development in C. elegans (49). Assuming that the basic mechanism of NOTCH signaling has been conserved during vertebrate evolution, this latter finding supports the possibility that association of RBP-J with the ankyrin repeat region is involved in and may be sufficient for some level of transactivation.
Our observation that the region containing Trp 1767 and the RAM23 domain (defined as amino acids 1758 -1857) is not required for transactivation is consistent with experiments showing that murine ICN lacking amino acids 1758 -1818 inhibited myogenic differentiation of NIH 3T3 cells (37), and that human ICN lacking amino acids 1758 -1770 inhibited ganglion cell differentiation in the developing chick retina (11). In the latter of these two studies, the effects of ICN were mimicked by exposure of cells expressing the full-sized NOTCH1 receptor to the NOTCH ligand DELTA, suggesting that Trp 1767 is not required for production of a bona fide NOTCH1 signal. Similarly, we have recently observed that the inhibitory effect of ICN⌬N1 (which lacks amino acids 1758 -1857) on myogenic differentiation of murine C2C12 cells is mimicked by treatment of these cells with the ligands produced by the human JAGGED and SERRATE genes. 3 Since it seems unlikely that binding of RBP-J to the RAM23 sequence of NOTCH1 is without functional significance, alternative functions for this sequence other than a direct role in activation of RBP-J need to be considered. One possible function for RAM23 would be to serve as a docking site for RBP-J with normal, full-sized NOTCH1. Upon binding of ligand to the extracellular domain of NOTCH1, this pool of pre-associated RBP-J could then be activated through a mechanism involving the ankyrin repeat binding site.
The strong oncogenic and transactivating activity of membrane-tethered forms of NOTCH1 (⌬E) and the failure to detect processing of ⌬E to a nuclear polypeptide in either T6E T-ALL cells or transiently expressing 293 cells implies that nuclear localization is not required for NOTCH1 oncogenesis or transactivation of the Cp1 promoter. This interpretation is consistent with the observation that among normal tissues in invertebrates and vertebrates, intranuclear NOTCH1 has been detected in only a small population of post-mitotic retinal cells in the rat (50). Hence, while intranuclear forms of NOTCH (7,8,51,52) and NOTCH1 (6,37,48) produce gain-of-function phenotypes in a variety of assays, intranuclear localization may not be essential for function.
The conclusion that nuclear localization of NOTCH1 is nonessential for at least some functions of NOTCH1 conflicts with one model for NOTCH1 signaling based on observations made using a slightly different form of ⌬E, termed ⌬E-C, lacking two conserved extracellular cysteine residues, Cys 1685 and Cys 1692 . Unlike ⌬E, which contains these two cysteine residues, transient expression of ⌬E-C is accompanied by proteolytic processing to smaller polypeptides that localize to the nucleus and associate with RBP-J, suggesting that polypeptides derived from ⌬E-C transactivate by direct physical interaction with RBP-J bound to DNA (22,35). Failure to detect similar processing of ⌬E could be explained if Cys 1685 and Cys 1692 influence proteolytic processing but not transactivation or oncogenesis per se. In support of this possibility, ⌬E-C-induced T cell tumors contain a series of lower molecular weight polypeptides cross-reactive with NOTCH1 antibody that are absent from ⌬E-induced T cell tumors (6) and also show nuclear staining with anti-NOTCH1 antibodies that is not observed in ⌬E-induced tumors. 3 ⌬E and ⌬E-C are equally potent oncoproteins (6), however, so the observed differences in proteolytic processing and subcellular localization do not appear to affect transforming activity.
In contrast to the nuclear translocation model for NOTCH1 signaling, our data pertaining to NOTCH1 oncogenesis and transactivation are generally consistent with an alternative signaling model based on observations made in Drosophila. This model proposes that transient association of Su(H) and NOTCH in the cytoplasm results in transactivation through Su(H) without nuclear translocation of NOTCH (53). One important additional factor in this activation pathway is deltex, a protein found in Drosophila that binds the ankyrin repeats of NOTCH (54). Deltex can displace Su(H) from the ankyrin repeats and is a positive regulator of NOTCH signaling (55,56), indicating the existence of factors that promote dissociation of NOTCH and Su(H) yet enhance transactivation. Conceivably, cell type-specific variation in expression levels of deltex-like factors could result in stable association of ICN with RBP-J in some cell types (e.g. I22 cells) but not others (e.g. transiently expressing 293 cells). A critical question not explained by this model is whether Su(H) activation stems from post-translational modification or from association with unknown accessory factors. Further elucidation of the mechanism of NOTCH1 transactivation will require identification of mammalian accessory factors and additional investigation of the effects of activated NOTCH1 on RBP-J.