Murine notch homologs (N1-4) undergo presenilin-dependent proteolysis.

Oncogenic forms of Notch1, Notch2, and Notch4 appear to mimic signaling intermediates of Notch1 and suggest that the role of proteolysis in Notch signaling has been conserved. Here we demonstrate that extracellularly truncated Notch homologs are substrates for a presenilin-dependent gamma-secretase activity. Despite minimal conservation within the transmembrane domain, the requirement for a specific amino acid (P1' valine) and its position at the cleavage site relative to the cytosolic border of the transmembrane domain are preserved. Cleaved, untethered Notch intracellular domains from each receptor translocate to the nucleus and interact with the transcriptional regulatory protein CSL. All four Notch proteins display presenilin-dependent transactivating potential on a minimal promoter reporter. Thus, this study increases the number of biochemically characterized gamma-secretase substrates from two to five. Despite a high degree of structural homology and the presenilin-dependent activity of truncated Notch proteins, the extent that this reflects functional redundancy is unknown.

Notch proteins are type I cell surface receptors of short range signals exchanged between cells. Notch (N1-N4 in mammals) is used repeatedly throughout development (1) and in the adult (2) to determine cell fate selection. In adult mammals, anomalous Notch signaling has been linked to neoplasia (N1, N2, and N4), stroke (N3), and possibly schizophrenia (N4) (3)(4)(5). The biochemical details of the N1 signaling mechanism have been elucidated and are supported by Caenorhabditis elegans and Drosophila genetic experiments (6). It is currently thought that N1 undergoes a conformational change upon DSL 1 ligand binding (Delta, Serrate, lag-2) exposing the extracellular juxtamembrane region to proteolysis at N1 V1711 (site 2 or S2, see Fig. 1A and Ref. 6). The resultant C-terminal product, NEXT (Notch extracellular truncation), then undergoes a presenilin-dependent cleavage at N1 V1744 (S3 cleavage), releasing the N1 intracellular fragment, NICD (Notch intracellular domain). N1 NICD is free to translocate to the nucleus where it acts as a transcriptional regulator of CSL (CBF1, Su(H), Lag-1)-dependent pathways.
It is not clear however, whether all murine NEXT-like Notch proteins are substrates for presenilin. The phenotype of mice deficient in both presenilin genes (7,8) suggests that proteolysis is a key mechanistic component of all mammalian Notch signal transduction. These embryos display more severe phenotypes than any single Notch null (9 -12). Interestingly, oncogenic forms of N1, N2, and N4 loosely mimic N1 NEXT or N1 NICD (13)(14)(15)(16)(17)(18) suggesting that regardless of the final nuclear target(s), Notch proteolysis is a key to neoplastic transformation as well as normal activity of these genes.
In this study, we sought to determine whether N2, N3, or N4 NEXT mimics undergo intramembranous proteolysis in a manner analogous to N1 and whether the cleavage site is conserved. We observe that all undergo presenilin-dependent proteolysis at a conserved transmembrane position to produce NICD fragments that are post-translationally modified. All NICD peptides translocate to the nucleus and interact with CSL proteins. Additionally, all activate a simple reporter containing multiple CSL binding sites in a presenilin-dependent manner.
Reporter Constructs-The 4xCSL-luciferase reporter was constructed from the CBF1/pGL2-GLO TATA CAT plasmid, a gift from Dr. S. Speck. A fragment containing the multimerized high affinity CSL sites (4ϫ CGTGGGAA) was excised by BamHI digest and ligated into a BglII/BamHI-digested RSV-TATA pGL2 vector, a gift from Dr. D. Towler. This modified vector has a TATA box from Rous sarcoma virus inserted into the pGL2-basic vector (Promega) to reduce the basal luciferase activity. Hes1-luciferase and Hes5-luciferase are gifts from Drs. A. Israel and R. Kageyama, respectively. All constructs were sequenced prior to use.
Transfections and Assays-For the presenilin cells, 2.5 ϫ 10 4 cells were plated in 1 ml of the media (DMEM, 10% fetal bovine serum (Sigma), 100 units/ml penicillin/streptomycin (Life Technologies, Inc.)) per well of a 12-well plate. Each well was transfected with 200 ng of 4ϫCSL-luciferase reporter, 100 ng of Notch, 20 ng of pCS2ϩ/␤-gal, and the remaining DNA up to 1 g with pCS2ϩ vector. The DNA mixture was transfected into MEF cells using FuGene (Roche Molecular Biochemicals). Cells were re-fed with medium after 16 h and lysed 48 h after transfection in 200 l of buffer (100 mM KPO 4 buffer, pH 7.8; 0.2% Triton). 20 l of lysate was used to determine ␤-galactosidase concentration to normalize for transfection efficiency. These assays were performed per the Tropix Galacton chemiluminescent substrate protocol. 50 l of lysate incubated with luciferin assay buffer (30 mM Tricine, pH 7.8; 3 mM ATP; 15 mM MgSO 4 ; 10 mM dithiothreitol; 0.2 mM CoA; 1 mM luciferin) was used to determine luciferase activity using a Tropix TR717 luminometer. To transfect 3T3 and 293 cells, cells were plated at ϳ10% density. A total of 10 g of DNA mixture was incubated with 500 l of 0.25 M CaCl 2 and 500 l of 2ϫ BBS (10 min, 23°C). Cells were re-fed with appropriate medium 16 h later.
Immunostaining-3 g of Notch DNA was transfected into 3T3 cells and fixed for 15 min in 4% paraformaldehyde 2 days post-transfection. After three washes with 1ϫ PBS (10 min, 23°C), cells were briefly permeabilized with 0.2% Triton/PBS solution, rinsed with 1ϫ PBS and incubated with 9E10 hybridoma overnight. Cells were rinsed the following morning and visualized with 1:300 ␣-mouse Texas-red secondary antibody.
Peptide Sequencing-30 plates of HEK293 cells were transiently transfected and lysed 2 days post-transfection. Notch proteins were ␣-Myc-immunoprecipitated and separated on 6% SDS-PAGE gels. Proteins were transferred to Problot (polyvinylidene difluoride) membranes and visualized by Coomassie Blue staining. Bands were excised and N-terminally sequenced using an ABI 480A sequencer.
Metabolic Labeling-2 g of Notch and 3 g of carrier DNA were transfected into 60-mm dishes. For pulse-chases, cells were starved for 60 min in methionine-free DMEM and labeled with 500 Ci of [ 35 S]methionine (Amersham Pharmacia Biotech) for 10 min. Proteins were chased in the presence of cyclohexamide for the times noted. Alternatively, cells were 32 PO 4 labeled (Amersham Pharmacia Biotech) in 10% dialyzed fetal bovine serum/phosphate-free DMEM (Life Technologies, Inc.). Labeled proteins were immunoprecipitated and analyzed by immunoblotting. CIP experiments: half of the beads bound with labeled proteins were incubated with or without CIP in 1ϫ CIP buffer (New England Biolabs) at 37 o for 30 min. Beads were then stored in Laemmli buffer and analyzed as described above.

Presenilin-dependent Proteolysis Forms a CSL-associated,
Nuclear NICD for All Notch Homologs-Truncated N1 proteins lacking an extracellular domain (N1 ⌬E ) mimic the N1 NEXT fragment generated as a result of ligand binding (21) are cleaved at Val 1744 (S3) releasing the N1 NICD fragment; these proteins are constitutively active (19). To assess whether N2, N3, or N4 NEXT mimics, analogous to N2 and N4 oncogenic proteins, undergo S3 cleavage and are thus capable of forming NICD in vivo, similarly constructed Notch proteins were compared in parallel biochemically. N2 ⌬E , N3 ⌬E , and N4 ⌬E were constructed to closely resemble N1 ⌬E . Each of the ⌬E constructs include the N1 leader peptide and ϳ20 amino acids of the extracellular domain, the transmembrane domain and extend intracellularly to the region N-terminal to the PEST region, ending with a hexameric Myc tag (Fig. 1A).
To determine whether a stable NICD-like fragment was present in cells expressing Notch ⌬E proteins, Western blot analysis was performed. As shown in Fig. 2A, extracts from 3T3 cells transfected with each of the four Notch ⌬E constructs contain an uncleaved ⌬E polypeptide of the expected size. In addition, a faster migrating band is observed in all lanes. Similar results were obtained in HEK293 and CHO cells (Fig.  4). It has been previously demonstrated that the formation of smaller peptides from N1 ⌬E can result from initiation of translation at an alternative methionine within the transmembrane domain (22). To prevent generation of NICD-sized fragments due to alternative translation from methionines abutting the putative S3 cleavage site in N2 and N3 (Fig. 1B), these amino acids were mutated to leucine (N2 ⌬E⅐M1697L and N3 ⌬E⅐M1663L ). Cells expressing N2 ⌬E⅐M1697L and N3 ⌬E⅐M1663L produced the same pattern of uncleaved ⌬E and NICD-like fragments (Fig.  2B). Therefore, cells expressing any of the four truncated Notch proteins produce, in addition to the full-length ⌬E protein, a stable NICD-like fragment, most likely by proteolysis.
To confirm that a precursor/product relationship exists between all uncleaved ⌬E polypeptides and NICD-like fragments, metabolic pulse-chase labeling experiments were performed, demonstrating that despite the low degree of amino acid conservation within the transmembrane domain, all Notch ⌬E constructs produce a NICD fragment (Fig. 3, A and B). In addition, several higher molecular mass bands were also observed in cells expressing N2 ⌬E and N3 ⌬E (Fig. 3B), suggesting that N2 NICD and N3 NICD fragments may be post-translationally modified more extensively than N1 NICD and N4 NICD . Indeed, a similar shift in mobility was detected with N2 IC and N3 IC proteins after labeling (Fig. 3B). Immunoprecipitation of Notch proteins from orthophosphate ( 32 PO 4 )-labeled cells unequivocally demonstrate that NICD proteins are phosphorylated (Fig. 3, C-F). Incubating immunoprecipitated proteins in the presence of CIP only affects the mobility of N2 peptides (Fig. 3, D and F), suggesting either that additional modifications are present on N3 or that the phosphates on N3 NICD cannot be removed by CIP.
As N1 NICD formation is dependent on presenilin, we sought to determine whether presenilin proteins are required for proteolysis of all murine Notch proteins. Previous experiments have demonstrated that presenilin1 preferentially binds the full-length N1 ⌬E polypeptide (23), suggesting that presenilin1 binds N1 prior to S3 cleavage. Our co-immunoprecipitation analyses demonstrate that uncleaved N2 ⌬E⅐M1697L , N3 ⌬E⅐M1663L and N4 ⌬E proteins preferentially bind to both presenilin1 and presenilin2 (Fig. 4, A and B). The importance of presenilin to the production of NICD in all four proteins was further assessed using three experimental paradigms. First, CHO cells expressing a dominant negative presenilin mutation, known to inhibit both presenilin1 and presenilin 2 activity (24, 25), demonstrate a marked reduction in NICD fragments (Fig. 4C). Second, treatment with an inhibitor specific for ␥-secretase activity (26) decreases the amount of NICD produced from all Notch proteins (Fig. 4D). Lastly, in cells deficient for both presenilin1 and presenilin 2 proteins (27,28), no NICD is detected by Western blot analysis (Fig. 4E).
Collectively, these results demonstrate that presenilin is involved in converting all truncated Notch proteins to NICD. Experiments conducted in Drosophila suggest that transmembrane proteins can undergo presenilin-dependent processing if the extracellular domain has been truncated within a particular size range (29), suggesting a lack of cleavage site specificity. We therefore wished to determine whether Notch proteins are cleaved at an equivalent position to N1 V1744 . As the N4 transmembrane domain is the most divergent from N1 (Fig. 1B), N4 NICD was purified and sequenced. The N-terminal amino

FIG. 2. All murine Notch molecules form a NICD fragment that preferentially interacts with CSL.
In this and all subsequent experiments N1 ⌬E refers to N1 ⌬E⅐M1727V . A, two fragments are detected from Notch ⌬E constructs, the faster migrating one preferentially coimmunoprecipitates with CSL RBP3⅐FLAG . B, N2 and N3 methionine mutants form NICD fragments that interact with CSL RBP3⅐Flag ⅐N2 ⌬E⅐M/L stands for N2 ⌬E⅐M1697L and N3 ⌬E⅐M/L denotes N3 ⌬E⅐M1663L . A and B, closed circles mark the uncleaved ⌬E peptide; open circles denote the cleaved Notch fragment (NICD). Notch ⌬E constructs were co-transfected with CSL RBP3⅐Flag DNA into 3T3 cells and harvested 2 days post-transfection. An aliquot was saved for Western analysis (Lysate), the rest was co-immunoprecipitated (CSL-CoIP) using ␣-FLAG antibody. Equivalent amounts of the lysates were run on SDS-PAGE gels and blotted with an ␣-Myc antibody. For these and subsequent gels, different exposures of the same gel were scanned. C, Notch ⌬E constructs generate tagged proteins that are clearly detected in 3T3 cell nuclei by ␣-Myc antibodies. acid of N4 NICD is Val 1463 placing the N4 cleavage site at the same relative position as the N1 S3 cleavage site (N1 V1744 , Fig.  1B). Importantly, N4 NICD production is severely reduced in constructs in which this valine has been mutated to leucine (N4 V1463L ) in the N4 transmembrane domain (Fig. 4F); an analogous mutation (N1 V1744L ; (19)) similarly effects S3 processing of N1.
Once N1 has undergone cleavage at the cell surface, the non-membrane tethered N1 NICD fragment translocates to the nucleus where it interacts with CSL proteins to regulate transcription of target genes. Immunocytochemical analysis revealed Notch nuclear staining in cells expressing membrane bound N2 ⌬E⅐M1697L , N3 ⌬E⅐M1663L , or N4 ⌬E (Fig. 2C). Additionally, co-immunoprecipitation studies demonstrate that CSL proteins preferentially interact with nuclear N1 NICD fragments rather than the uncleaved N1 ⌬E polypeptide (19,30). In contrast to the enrichment for the NEXT-like fragments by presenilin co-immunoprecipitation (Fig. 4, A and B), CSL RBP3-FLAG co-immunoprecipitations enrich for the NICD fragments produced from cells expressing N2, N3, and or N4 ⌬E constructs (Fig. 2, A and B). The preference of CSL for the low molecular weight NICD may suggest that non-modified NICD (Fig. 3, C-F) accumulates preferentially in the nucleus or that phosphatases are active during co-immunoprecipitation procedure.
These results establish biochemically that presenilindependent ␥-secretase is the predominant S3 proteolytic activity in cultured cells. To determine whether any truncated Notch protein activate a CSL-dependent reporter in the absence of presenilin proteins, MEF cells derived from embryos heterozygous for presenilin1 and presenilin2 or from littermates deficient for both genes were used to compare the CSLdependent activity of truncated Notch proteins (Fig. 5). In the absence of presenilin, the activity is similar to that of the 4xCSL reporter alone. We conclude that no alternative prote-ase exists in these cells capable of releasing sufficient NICD protein to elicit activity.

Regulated Intramembranous Proteolysis May Be a Common
Feature of Notch Signaling-Upon ligand binding, the fulllength N1 receptor is cleaved by a metalloprotease just extracellular to the transmembrane domain (S2 cleavage); the resulting cleavage product, NEXT, is then processed to form NICD. Western blot analysis demonstrates that a NICD-like fragment is produced from all NEXT-like, truncated Notch proteins (Figs. 2 and 3). While such fragments can be the result of alternative translation-initiation (22) or translation from a cryptic internal ribosome entry site (31), we demonstrate unequivocally that proteolysis is responsible for NICD production as mutating methionine residues in the N2 and N3 transmembrane domains do not impact NICD production (Figs. 2B, 3, 4). Pulse-chase analyses confirm that a precursor/product relationship exists between all Notch ⌬E and NICD-like peptides (Fig. 3). Post-translational phosphorylation of all Notch proteins was also observed (Fig. 3), N2 and N3 modifications have been previously noted (32,33). Although the Notch ⌬E constructs used in this paper include the analogous S2 cleavage site (Fig. 1B), we do not detect any S2 cleavage products by N-terminal sequencing. 2 This is consistent with the model suggesting that S2 cleavage is only necessary in molecules containing an inhibitory extracellular domain and that molecules with a short extracellular domain can be directly processed to form NICD (21,29).
Protein sequencing data confirm that the cleaved peptide bond in N4 occurs at a position equivalent to the known N1 S3 FIG. 3. NICD formation from ⌬E constructs. A and B, Notch ⌬E proteins are converted to NICD (highlighted by arrows). IC constructs (NICD mimics) serve as markers. N1 IC , N2 IC and N3 IC appear to be modified after 60 min (asterisk). C, N1 and N4 ⌬E proteins are phosphorylated. Notch proteins were labeled with 32 PO 4 and immunoprecipitated. D, CIP treatment of [ 35 S]methionine-labeled N2 protein indicates that N2 is phosphorylated. E, N2, N3, and N4 IC proteins are phosphorylated. An aliquot of cells expressing 32 PO 4 -labeled N2 IC , N3 IC , or N4 IC proteins were blotted with ␣-Myc (left side). These same samples were also immunoprecipitated and visualized by autoradiography (right side). F, N3 ⌬E is phosphorylated. CIP and 32 PO 4 -labeled proteins were run on the same gel, separated to better visualize the fragments by autoradiography and the images subsequently spliced together. In the above experiments, methionine mutants for N1, N2, and N3 ⌬E constructs were used. Additionally, proteins were immunoprecipitated with an ␣-Myc antibody; immunoprecipitations with an irrelevant antibody were negative (C and D and data not shown).
cleavage site (N4 V1463 , N1 V1744 , Fig. 1B). This conservation is intriguing as N1 and N4 share only 24% identity within the transmembrane domain (Fig. 1B). The fact that intramembranous Notch processing is highly dependent on this conserved valine (Fig. 4F, Refs. 19 and 34) is in stark contrast to the apparent lack of cleavage site specificity demonstrated for amyloid precursor protein (APP), the first known substrate of the presenilin-dependent enzymatic complex, ␥-secretase (35,36). We only detect one proteolytic product from N1 ⌬E and N4 ⌬E by N-terminal sequencing. It is worth noting that while we measure the release of a C-terminal fragment by ␥-secretase, ␥-secretase activity on APP was assayed by measurements of released N-terminal fragments. Recent sequencing of the N terminus of an APP C-terminal fragment identified a presenilindependent cleavage site in APP at Val 50 , a position equivalent to S3 in Notch (37). Three of five ␥-secretase substrates are now known to be cleaved at an equivalent position (N1, N4, APP). It is however possible that the low-level proteolysis observed in Notch S3 valine mutants results from cleavage at position equivalent to APP A␤40. Alternatively, residual proteolysis may occur at the mutant S3 site. Isolation of N ⌬E N-terminal stubs will reveal if such additional proteolytic fragments exist, perhaps resulting from rare cleavage events in the middle of the Notch transmembrane domain as in APP (36). This possibility remains untested.
It has been suggested that many transmembrane proteins can be cleaved by a presenilin-dependent activity (29). Ectopic expression of chimeric, membrane tethered proteins containing a potent transcription activator in Drosophila demonstrated that different transmembrane domains can all be cleaved by a presenilin-dependent activity as long as they lack an extensive extracellular domain and have the appropriate conformation. However, even with this sensitive assay, the authors reported a decline in proteolysis when the conserved valine in Drosophila Notch is mutated. It remains to be determined how many other proteins are natural substrates for presenilins and whether their proteolysis depends on specific sequences at the cleavage site. Finally, the observation that all Notch proteins are dependent on presenilins for proteolysis supports the hypothesis that the more severe phenotype of the murine presenilin 1/presenilin 2 nulls is due to a loss of all Notch activity (7,8). FIG. 4. Notch and presenilins. presenilin1 (A) and presenilin2 (B) form a stable and preferential interaction with the longer, uncleaved forms of Notch in transiently transfected HEK293 cells. A portion of the lysed cell extract was saved for Western analysis (L), the remaining sample was co-immunoprecipitated (C) using a presenilin1-or preseni-lin2-specific antibody. C, functional presenilin is required for S3 processing of all Notch proteins as NICD formation is severely reduced in cells expressing the dominant negative presenilin. ⌬E Notch constructs were transfected into wild type CHO cells (marked Ϫ) or a CHO cell line stably expressing a dominant negative presenilin mutation (D257A, marked ϩ). D, NICD formation is severely reduced in the presence of Compound 11 protease inhibitor (26). Cells expressing Notch were [ 35 S]methionine-labeled with or without Compound 11 prior to immunoprecipitation. E, NICD formation requires the presence of presenilins. Cell lysates from Notch-transfected presenilin-deficient versus presenilin wildtype cells were analyzed. F, N4 NICD production is severely reduced when a valine in the transmembrane domain is mutated to leucine. Cells expressing TM4 and TM4 V1463L were lysed for Western analysis. If the gel is overexposed, small amounts of NICD are seen in the TM4 V1463L lane (data not shown). In the figure, TM4 V/L represents TM4 V1463L . All gels were immunoblotted using an ␣-Myc-specific antibody. Open circles mark the NICD fragment. ⌬E construct mutants for N1, N2, and N3 methionines were used.
FIG. 5. Notch activity in presenilin-deficient versus wild-type cells. MEF cells from mice heterozygous for presenilin1 and presenilin2 (marked ϩ) or from littermates deficient for both genes (marked Ϫ) were used to compare the CSL-dependent activity of truncated Notch proteins. In the absence of presenilin, the activity is similar to that of the 4ϫCSL reporter alone (n ϭ 4).