γ-Secretase-regulated Proteolysis of the Notch Receptor by Mitochondrial Intermediate Peptidase

Abstract

Notch is a transmembrane receptor that controls a diverse array of cellular processes including cell proliferation, differentiation, survival, and migration. The cellular outcome of Notch signaling is dependent on extracellular and intracellular signals, but the complexities of its regulation are not well understood. Canonical Notch signaling involves ligand association that triggers sequential and regulated proteolysis of Notch at several sites. Ligand-dependent proteolysis at the S2 site removes the bulk of the extracellular domain of Notch. Subsequent γ-secretase-mediated intramembrane proteolysis of the remaining membrane-tethered Notch fragment at the S3 site produces a nuclear-destined Notch intracellular domain (NICD). Here we show that following γ-secretase cleavage, Notch is proteolyzed at a novel S5 site. We have identified this S5 site to be eight amino acids downstream of the S3 site. Biochemical fractionation and purification resulted in the identification of the S5 site protease as the mitochondrial intermediate peptidase (MIPEP). Expression of the MIPEP-cleaved NICD (ΔNICD) results in a decrease in cell viability and mitochondria membrane potential. The sequential and regulated proteolysis by γ-secretase and MIPEP suggests a new means by which Notch function can be modulated.

Introduction

The Notch signaling pathway is evolutionarily conserved and plays a fundamental role in communication between adjacent cells to influence differentiation, proliferation, survival, and migration (1–3). During its maturation and function as a signal transducer, the Notch receptor undergoes proteolysis by three different enzymes. Notch is processed initially by a furin-like convertase in the trans-Golgi at the S1 site to generate an extracellular domain and a membrane-bound domain, which remain associated during their transport to the plasma membrane. Binding of a ligand to the extracellular domain of Notch stimulates proteolysis of Notch at the S2 site by the metalloprotease tumor necrosis factor α-converting enzyme (TACE) and produces a transmembrane fragment referred to as Notch extracellular truncation (NΔE)³ (4, 5). The membrane-tethered NΔE is recruited to the γ-secretase complex for S3 site cleavage to generate the Notch intracellular domain (NICD) (6–8). The resultant NICD translocates into the nucleus, where it associates with the transcription factor CBF-1 and with coactivators, such as Mastermind, to initiate gene transcription (9–12). Notch has mostly been characterized by its transcriptional role in the nucleus, but several studies suggest that it may have functional roles in the cytoplasm (13, 14) as well as in the mitochondria (15–20) where, in the latter, Notch acts as an interpreter of cellular cues. In this study, we provide evidence that following cleavage by γ-secretase, NICD is further proteolyzed at a novel S5 site. We identify the protease responsible for this cleavage as the mitochondrial intermediate peptidase (MIPEP), which generates ΔNICD (¹⁷⁵²RQHGQ-…). We showed that Notch function can be modulated by MIPEP-mediated cleavage. Our findings raise the possibility that S5 site proteolysis represents a novel regulatory component of Notch signaling.

EXPERIMENTAL PROCEDURES

Materials

Chemicals used included γ-secretase inhibitor N-[N-(3,5-Difluorophenylacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT) (Calbiochem/EMD) and the EDTA-free complete protease inhibitor mixture tablet (Roche). Antibodies used include anti-myc 9E10 (ATCC), anti-cleaved Notch1 (Val-1744) (Cell Signaling Technology, Inc.) and anti-FLAG (Sigma). Western blot analysis was performed as described previously. The constructs NΔE and NICD in the pCS2 vector were obtained from Dr. R. Kopan (Washington University School of Medicine, St. Louis, MO) (7). Mutations in Notch were generated using site-directed mutagenesis. Mammalian Notch constructs were used as PCR templates to clone into pFastBac1 (Invitrogen). PS1^−/− mouse embryonic fibroblast (MEF) cells were obtained from Dr. B. De Strooper (Katholieke Universiteit, Leuven, Belgium).

Proteolytic Activity Assay and Edman Protein Sequencing Analysis

Biochemical analyses, proteolytic assays, and Edman protein sequencing were performed as described (8, 21, 22). Reconstituted recombinant γ-secretase was purified from Sf9 cells as described (8). Mouse Notch1-derived substrates and recombinant MIPEP protein were also purified from baculovirus-infected Sf9 cells. The soluble extracts (S100) from cell lines were used in activity assays with Notch substrate and reconstituted γ-secretase. S100 was obtained from cells lysed in 50 mm PIPES (pH7.0), 150 mm KCl, 10 μg/ml leupeptin, 5 μg/ml trypsin inhibitor, 5 μg/ml Na-Tosyl-Lys-chloromethylketone, 5 μg/ml aprotinin, and 1 mm PMSF using nitrogen cavitation at 800 psi for 10 min at 4 °C. Postnuclear supernatant was collected at 800 × g, and the supernatant was centrifuged at 100,000 × g for 1 h to obtain the S100. Recombinant FLAG/His-tagged N100 was incubated with reconstituted γ-secretase and assayed as described previously (8). For identification of substrate sequences, proteins on the membrane were stained with Amido Black, and the proteolytic fragments were excised and subjected to Edman degradation and sequenced at the Protein Chemistry Technology Center, University of Texas Southwestern Medical Center.

Biochemical Identification of MIPEP

HeLa S100 was subjected to ammonium sulfate precipitation. Precipitated proteins were resuspended in 20 mm PIPES, 150 mm KCl, 10% glycerol, 1 mm DTT plus protease inhibitors and chromatographed over a size exclusion column Sephacryl S300 (Pharmacia). Fractions (1 ml) were collected from the column and assayed for S5 site proteolysis. Fractions that exhibit γ-secretase-dependent cleavage activity were pooled and buffer-exchanged against 20 mm PIPES, 50 mm KCl, 10% glycerol, and 1 mm DTT plus protease inhibitors and passed successively over a Hi-Trap S and Hi-Trap Q (Pharmacia). The flow-through fraction was loaded on ω-aminooctyl-agarose (Sigma) (1 × 5 cm) at 20 mm PIPES, 50 mm KCl, 10% glycerol, and 1 mm DTT plus protease inhibitors. Protein fractions were eluted using a 40-ml gradient from 300 mm to 1 m KCl. The active fractions from ω-aminooctyl-agarose were pooled and chromatographed directly over a 1 × 5 cm hydroxyapatite type I (Bio-Rad) and eluted. Protein fractions were dialyzed against the wash buffer before subjecting the sample for γ-secretase-dependent protease activity assay. The peak of activity was found to elute at 150 mm sodium phosphate (pH 8). The protein fraction containing the activity was subjected to mass spectrometry at the Protein Chemistry Technology Center, University of Texas Southwestern Medical Center. Purification of mitochondria using sucrose gradient fractionation was performed according to standard procedures (23, 24).

MIPEP Knockdown

Cell culture and RNAi experiments were done as reported (22, 25). MIPEP (5′GCCGGGAUCCGGGCCCGAATT3′ and 5′CGUGCAGAGAGGUAUAAUATT3′), presenilin-1 (5′GGUCCACUUCGUAUGCUGGTT3′), or scrambled sequence (5′GAUCACGGAUCUCCAUGGCTT3′) siRNA (Dharmacon) were transfected into mammalian cells using oligofectamine. Scramble oligo duplex, FMR-8, duplex and MIPEP sense oligo were also used as controls for the experiments, but only the data from the scrambled oligo duplex is shown because the data obtained from these controls were similar. More details of the biochemical and cellular methods are described in the figure legends when appropriate. Representative data are presented for experiments performed a minimum of three independent times.

Cell Viability Assay

HeLa cells were maintained in DMEM supplemented with 10% FBS. Equal numbers of cells were transfected with cDNA constructs using Lipofectamine 2000 (Invitrogen). Cells were harvested 24 h post-transfection and distributed in triplicates into 96-well plates. Cell viability was determined at 40 h post-transfection using MTS reagent (Promega). Absorbance was read at 490 nm, and data were normalized to cells expressing empty vector and plotted as % cell viability (mean ± S.D.).

RESULTS

Proteolysis of Notch at the S5 Site by a Novel Peptidase

We had previously established an in vitro assay whereby recombinant γ-secretase purified from insect Sf9 cells coexpressing mammalian presenilin, nicastrin, Aph-1, and Pen-2 faithfully proteolyzes a Notch substrate (N100) at the S3 site, generating a proteolytic fragment beginning with Val-1744 (NICD') (Fig. 1A) (8, 22). In this study, we utilized this assay to search for novel factors that regulate γ-secretase activity or its Notch products in mammalian cell extracts. As expected, soluble protein extracts themselves did not contain γ-secretase activity (data not shown). However, addition of the soluble extracts to the in vitro Notch cleavage assay in the presence of recombinant γ-secretase produced the expected NICD' fragment and another robust and faster-migrating proteolytic fragment referred to as ΔNICD' (Fig. 1B). To identify the nature of the protease responsible for this secondary cleavage, the in vitro Notch cleavage assay was repeated with HeLa soluble extracts and recombinant γ-secretase and in the presence of several types of protease inhibitors in the assay. NICD' production by recombinant γ-secretase was inhibited by γ-secretase inhibitor DAPT but not by dimethyl sulfoxide, the divalent cation chelators EGTA and EDTA, or the Roche EDTA-free protease inhibitors mixture known to inhibit proteases like chymotrypsin, papain, and trypsin (Fig. 1B, top left panel). Addition of HeLa soluble extract, on the other hand, generated the ΔNICD' fragment in samples that produced NICD' except for that treated with EGTA and EDTA (Fig. 1B, top right panel). These observations suggest the existence of a novel γ-secretase- and divalent cation-dependent activity responsible for generating ΔNICD' in vitro.

FIGURE 1.

Notch is proteolyzed by a novel metallopeptidase activity in a γ-secretase-dependent manner. A, summary of the Notch proteolytic fragments analyzed in this study. The ligand-independent, membrane-tethered Notch trimmed of its extracellular domain (NΔE) consists of a transmembrane domain (TM) and an intracellular cytoplasmic domain (NICD). The cytoplasmic region includes a putative mitochondrial targeting signal (MTS), a RBP-Jκ/CBF1-associated module (RAM) domain, four known or putative nuclear localization signals (NLS), and a carboxyl-terminal PEST domain. NΔE is cleaved at the S3 site between Gly-1743 and Val-1744 by γ-secretase to generate NICD. A second γ-secretase-dependent cleavage occurs at the S5 site between Arg-1751 and Arg-1752 to generate ΔNICD. N-terminal Edman protein sequencing confirmed the presence of these three equivalent Notch fragments in our in vitro cleavage assay: N100, a Notch transmembrane substrate harboring Val-1711-Glu-1809 of mouse Notch-1 with a carboxyl-terminal FLAG/His tag. NICD', the γ-secretase proteolytic fragment of N100 with Val-1744 as the first amino acid and ΔNICD', the S5 site proteolyzed fragment of N100 containing Arg-1752 as the first amino acid. LNR, Lin/Notch repeats; NCR, Notch cytokine response domain; TAD, transactivation domain. B, Notch substrate N100 was subjected to an in vitro cleavage assay for 6 h at 37 °C with either no addition (- Extract) or addition (+ Extract) of high-speed soluble HeLa extract. Samples containing dimethyl sulfoxide, 2 μm of the γ-secretase inhibitor DAPT, a Roche protease inhibitor mixture, or a mixture of 5 mm EGTA and 5 mm EDTA was initiated with N100 and recombinant γ-secretase added at the same time (top panel) or with N100 that was preincubated with γ-secretase for 4 h to first generate the γ-secretase-cleaved product, NICD' (+ γ-sec preinc, bottom panel). C, similar to B, extracts from either wild-type (PS1^+/+) or presenilin 1 knockout (PS1^−/−) MEF cells were subjected to the proteolytic assay with N100 only (without γ-sec) or with N100 preincubated with exogenous recombinant γ-secretase (with γ-sec) for 0 h or 20 h. D, N100 substrate mutated at position 1749 (K1749H) and positions 1751 and 1752 (R1751I, R1752L) and wild-type N100 (WT) purified from baculovirus-infected Sf9 cells were subjected to the proteolytic assay with exogenous recombinant γ-secretase only (−Extract) or in the presence of exogenous γ-secretase and the novel enzymatic activity present in the HeLa cell extract (+ Extract) for 0, 6, or 20 h. All samples were monitored by Western blot analysis with anti-FLAG antibody. Note that N100-K1749H is a control substrate for this experiment, as its cleavage is not affected by either γ-secretase or MIPEP.

One possible explanation for our observations is that an unknown factor in the soluble extract modulates γ-secretase activity through reconstitution with the presenilin, nicastrin, Aph-1, and Pen-2 γ-secretase complex to differentially process Notch substrate at separate sites. Alternatively, a novel protease might be responsible for processing Notch after a prerequisite γ-secretase cleavage event. To distinguish between these two possibilities, N100 was preincubated with recombinant γ-secretase to generate NICD' prior to the addition of inhibitors or HeLa extract. In the absence of the extract, the inhibitors had no effect on the preincubated N100 and its proteolytic product NICD' (Fig. 1B, bottom left panel). Addition of HeLa extract to the preincubated substrate generated the ΔNICD' proteolytic product that was sensitive to inhibition by EGTA and EDTA but not by DAPT or general protease inhibitors (Fig. 1B, bottom right panel). Moreover, the production of NICD' and ΔNICD' was markedly reduced in lysates from presenilin-deficient MEF cells (PS1^−/−, Ref. 6) but was restored when recombinant γ-secretase was added back (Fig. 1C). Together, these observations suggest that the ΔNICD'-generating activity is metalloprotease-dependent and regulated by, but distinct from, γ-secretase.

We have previously shown that NICD' produced in our in vitro γ-secretase assay starts from ¹⁷⁴⁴VLLSR… (8), the physiological S3 site cleavage of Notch (27, 28). Similarly, Edman protein sequencing revealed that the first five N-terminal amino acids of ΔNICD' are ¹⁷⁵²RQHGQ… (Fig. 1A), indicating that the novel peptidase activity cleaves NICD between Arg-1751 and Arg-1752, hereafter referred as the S5 site, to remove eight amino acids from the N terminus of NICD. To confirm that we have indeed identified the correct proteolytic site, we replaced both arginine residues that constitute the S5 site with isoleucine and leucine (R1751I, R1752L) and tested the mutated Notch substrate in the in vitro proteolytic assay. We also mutated lysine 1749 to histidine (K1749H) because Lys-1749 is thought to be important for Notch being a proper γ-secretase substrate in cells (27, 29). None of the mutations in Notch interfered with the generation of NICD' after incubation with recombinant γ-secretase (Fig. 1D). Addition of the peptidase activity from soluble HeLa extract produced similar level of ΔNICD' from either the K1749H or the wild-type Notch substrate, suggesting that mutation of the lysine residue did not hinder the ability of the novel peptidase to cleave Notch in the in vitro assay. The peptidase activity, however, was not able to generate ΔNICD' from the R1751I, R1752L mutant substrate, suggesting that mutation of the arginine residues abolished the proteolytic S5 site in Notch. Moreover, the level of NICD' generated from the R1751I, R1752L substrate was much higher than from the other two substrates. A possible explanation for this latter observation is that NICD' could not be further processed to become ΔNICD', supporting a substrate-product relationship between NICD' and ΔNICD' for the S5 site peptidase.

Identification of MIPEP as the Notch S5 Site Peptidase

We next developed a six-step biochemical fractionation procedure to purify the S5 site peptidase activity from HeLa extract (Figs. 2, A–C). The activity was purified 1863-fold (Fig. 2C), and the resultant products were subjected to mass spectrometric analysis. Three tryptic peptides (LNTNVDLYQSLQK, HYQTGQPLPK, GSLEAGIR) matched the MIPEP (GI 14602871, EC3.4.24.59), an oligopeptidase that belongs to the M3 zinc metallopeptidase family. Peptides for programmed cell death 6 interacting protein (DAFDKGSLFGGSVK, SVIEQGGIQTVDQLIKGI; GI 22027538) and RNA-dependent helicase (SSQSSSQQFSGIGR, ELAQQVQQVADDYGK; GI 3122595) were also identified and were not characterized further, as they are not predicted to have peptidase activity. We tested whether recombinant MIPEP can generate the ΔNICD' fragment seen with HeLa soluble extracts. Full-length recombinant MIPEP protein was purified from Sf9 cells using the baculoviral expression system. Recombinant MIPEP, in the presence or absence of reconstituted recombinant γ-secretase, was then included in the in vitro Notch cleavage assay. In the presence of recombinant γ-secretase, MIPEP, but neither BSA, MAP kinase kinase 3 (MEK3), nor a regulator of G-protein signaling molecule (β₅Rgs7), generated a proteolytic fragment that comigrated with the ΔNICD' fragment seen in an assay with HeLa extract (Fig. 2D). As with HeLa extracts (Fig. 1B), MIPEP did not generate ΔNICD' from the N100 in the absence of γ-secretase and is not dependent on γ-secretase when purified NICD' was added as a substrate (Fig. 2E). These data indicate that recombinant MIPEP is responsible for producing ΔNICD' from NICD.

FIGURE 2.

Identification of the S5 site peptidase as MIPEP. A and B, Sephacryl S300 size-exclusion chromatography of a 15–30% ammonium sulfate precipitate from HeLa high-speed S100 supernatant. Proteins (Abs280, dashed line) collected from the column were monitored using the in vitro Notch cleavage assay (B) and replotted in A as activity (solid line). Fractions of high ΔNICD'-generating activity (hatched bar) were pooled and subjected to further purification as detailed in C. D, purified recombinant MIPEP (lanes 4–6), BSA, MEK3, β5Rgs7, or HeLa extract was incubated in the presence (lanes 1–4 and 6-7) or absence (lane 5) of recombinant γ-secretase and subjected to the in vitro Notch cleavage assay for 6 h using the Notch N100 substrate. E, 100 ng (lanes 1, 4, 7, and 10) or 500 ng (lanes 2, 5, 8, and 11) purified recombinant MIPEP or 500 ng β5Rgs7 (lanes 3, 6, 9, and 12) was incubated in the absence (lanes 1–3 and 7-9) or presence (lanes 4–6 and 10-12) of γ-secretase and subjected to the in vitro Notch assay for 0 h or 20 h using either the Notch N100 substrate or a recombinant Notch substrate that begins with Val-1744 (NICD'). F, RT-PCR analysis of cells transfected with no oligo, sense oligo, MIPEP duplex, or presenilin-1 duplex was performed with primers specific for either MIPEP, presenilin-1, or 18S rRNA to monitor for linear amplification of RNA. G, the in vitro Notch cleavage assay using equal proteins from cells transfected with the indicated oligos (lanes 1–4), from parental HeLa cells (lane 5), or with no cell extract (lane 6) was performed with the coaddition of N100 and recombinant γ-secretase for 6 h. Equal protein in the assay was monitored using an anti-β-actin antibody.

To further verify that MIPEP is responsible for generating ΔNICD', we utilized small interfering RNA (siRNA) to examine how reducing endogenous levels of MIPEP affects the S5 site cleavage activity. HeLa cells transfected with MIPEP siRNA duplex but not oligofectamine alone, sense oligo, or presenilin 1 (PS1) siRNA duplex showed specific reduction in the mRNA level of MIPEP (Fig. 2F). Although both the NICD' and ΔNICD' proteolytic fragments were present in assays performed with control cells, knockdown of MIPEP caused the level of ΔNICD' fragment to decline significantly (Fig. 2G).

Recently, Perumalsamy et al. (17) found Notch in the mitochondrial fraction of cellular lysates. Having shown that NICD can be further processed by the mitochondrial protease MIPEP in vitro and in vivo, we next asked whether we, too, could detect NICD and/or the upstream protease γ-secretase in mitochondria. We were able to detect MIPEP, γ-secretase-cleaved NICD (α-Val-1744), NCT, and the mitochondria marker HSP60 in mitochondria but not in the non-mitochondria postnuclear supernatant of HeLa cell lysates (Fig. 3, A and B). Furthermore, these mitochondria preparations (but not the supernatant) were able to generate NICD' and ΔNICD' in our in vitro Notch assay (Fig. 3B).

FIGURE 3.

NICD can be cleaved by MIPEP in mitochondria. A, equal amounts of mitochondrial (Mito) and non-mitochondrial postnuclear supernatant (Sup) from centrifugation were probed with anti-Notch-1(Val-1744) and with antibodies against HSP60, Bim, and MIPEP. B, equal amounts of mitochondria purified from non-mitochondrial postnuclear supernatant were subjected to the in vitro Notch cleavage assay. Anti-HSP60 was used to monitor the integrity and enrichment of mitochondria. C, FACS analysis of S5 site proteolysis in intact cells. HeLa cells treated with siRNA oligos for MIPEP, PS1, or control as in Fig. 2F were transiently transfected with empty vector, NΔE, or NΔE V1744K, each of which contained a C-terminal 6x-myc tag. Cells were fixed, permeabilized, and stained with either the anti-Notch-1(Val-1744) antibody, which specifically recognizes the cleaved Notch intracellular domain starting at Val-1744, or the control anti-c-myc antibody. Cells were then sorted by FACS analysis with an Alexa Fluor 488-coupled secondary antibody to measure the relative fluorescence intensity of NICD. The data are normalized and presented in the representative histogram as % of maximum, where the fluorescence intensity with the greatest number of cells was set at 100%.

Having demonstrated that γ-secretase-generated NICD' is an immediate substrate of MIPEP activity in vitro and that the necessary components for ΔNICD' generation can be found in mitochondria, we next examined whether MIPEP-mediated Notch cleavage occurs in intact cells. This proved to be rather difficult as 1) by the N-end rule, alternate forms of the Notch intracellular domain can be extremely labile ((27, 30)), 2) removal of eight N-terminal residues (∼1 kDa) between the S3 and S5 site cleavage products does not result in a substantial gel shift in the ∼100 kDa Notch intracellular domain, and 3) we were unable to generate specific antibodies that can differentiate the eight residues between S3 and S5 site cleavage products. We therefore tested whether reducing the MIPEP level by siRNA would impair S5 site proteolysis of Notch, thereby resulting in an increased level of the substrate, NICD. Detecting NICD in cells is feasible with the anti-cleaved Notch-1 (α-Val-1744) antibody that specifically recognizes the amino terminus of the S3 site proteolyzed Notch intracellular domain. Taking advantage of this antibody, we cotransfected HeLa cells with various Notch constructs (empty vector, NΔE, or NΔE V1744K, each containing a C-terminal 6x-myc tag) and siRNA (no oligo, MIPEP sense, or MIPEP or PS1 duplex), permeabilized, and stained the cells with anti-cleaved Notch-1 (Val-1744) or control (anti-c-myc) antibodies, and sorted the cells by FACS analysis. In HeLa cells expressing NΔE, knockdown of PS1 resulted in a dramatic loss of α-Val-1744 staining (Fig. 3C, center panel). Moreover, mutation of Val-1744 to lysine resulted in a dramatic loss in α-Val-1744 staining (Fig. 3C, right panel), thus confirming that this analysis specifically detects Notch that has been cleaved by γ-secretase. Knockdown of MIPEP (and, therefore, inhibition of further processing) resulted in an increase in the amount of cleaved Notch, suggesting a substrate-product relationship in cells.

Successive Cleavage of Notch by γ-Secretase and MIPEP Mediates Cell Survival

In the course of our experiments, we noticed considerable differences over time in the survival of the cells that were transfected with different Notch constructs. This observation prompted us to determine whether cell survival can be used as an assay for measuring functional effects of MIPEP-mediated S5 site proteolysis on Notch function. We assessed cell viability using an MTS assay. HeLa cells expressing either NΔE, NICD, or ΔNICD exhibited 20–50% cell survival at 36 h post-transfection as compared with control cells expressing either empty vector or β-galactosidase, with ΔNICD as the most potent form (Fig. 4A). Knockdown of MIPEP (Fig. 4B) rescued the cytotoxicity of NΔE and NICD but not ΔNICD, suggesting that ΔNICD, the product of successive γ-secretase- and MIPEP-mediated proteolysis, leads to the cytotoxicity. As the MTS assay also monitors mitochondrial integrity as a correlate with cell viability, these results further implicate mitochondria in the Notch-based cytotoxicity. Indeed, experiments similar to Fig. 4B but measuring mitochondrial membrane potential (Δψ_m) by the JC-1 reagent showed similar results: ΔNICD leads to cytotoxicity through a decrease in Δψ_m (supplemental Fig. S1). Thus, taken together, these observations suggest that γ-secretase-dependent cell death caused by NΔE and NICD occurs through mitochondria.

FIGURE 4.

Successive cleavage of Notch by γ-secretase and MIPEP yields ΔNICD, a pro-death Notch fragment. A, HeLa cells were transiently transfected with either empty vector (□), β-galactosidase (♢), NΔE (▴), NICD (■), or ΔNICD (●) for the indicated time. The viability of the cells was monitored using the MTS assay. The graph is presented as % cell viability (mean ± S.D.) after normalizing each set of samples performed in triplicate to the empty vector data collected at 24 h. B, HeLa cells were transiently cotransfected with the indicated constructs and siRNA oligos for MIPEP or controls for 40 h. Quantification of % cell viability was performed as in A.

DISCUSSION

In this study, we identify a novel proteolytic event of Notch at the new S5 site: ligand-induced proteolysis of the Notch receptor at the S2 and S3 sites (by TACE and γ-secretase, respectively) yields NICD, which, in turn, is trimmed of its first eight amino acids by MIPEP, resulting in an alternative Notch intracellular fragment, ΔNICD. Previous studies have shown that there exist at least two pools of NICD with different destinations and purposes (13–15, 17). It is conceivable, then, that one pool is marked for MIPEP processing, whereas the other translocates to the nucleus for transcriptional regulation. The molecular details concerning where, when, and how MIPEP processes Notch will be further explored in the future as more powerful tools become available.

The connection between Notch proteolysis and a mitochondrial protease is unexpected and warrants further investigation. There is growing evidence for Notch-mediated cellular events at the mitochondria (15–20), and Perumalsamy et al. (17) recently showed that NICD localizes to the mitochondrial fractions. We note, however, that despite best efforts, cellular studies using a technique such as mitochondrial fractionation can be complicated by contamination of the samples with the closely associated endoplasmic reticulum membranes. Additional studies are therefore needed to further explore the relationship between mitochondria and Notch functions. On the other hand, it appears that the capability of γ-secretase-produced NICD to locate to mitochondria is endowed in the amino acid sequence of Notch. Three computer programs for predicting subcellular localization of proteins (TargetP (31), iPSORT (32), and MitoProt (33)) all indicated that the N-terminal portion of NICD harbors an evolutionarily conserved mitochondrial targeting signal that superimposes two monopartite nuclear localization signal motifs, a putative monoubiquitination site (29), and the RAM domain (Fig. 1A). It is thus possible that NICD is capable of shuttling between nuclei and mitochondria. Alternatively, because a portion of γ-secretase has also been detected in mitochondria (Fig. 3B and Refs. 34, 35), it is possible that the TACE-cleaved Notch fragment could travel to mitochondria, wherein γ-secretase and MIPEP serve roles analogous to those of mitochondrial processing peptidase and MIPEP in transporting the immediate γ-secretase substrate derived from Notch (NΔE), and possibly other membrane proteins, for sequential processing and compartmentalization.

Regulation of Notch signaling and the diverse physiological outcomes are dependent on the context of specific cells and their extracellular and intracellular environments (1, 36). Notch has previously been implicated in context-dependent cell death events (37–40). Depending on cellular context, Notch-mediated transcription can be either pro- or antiapoptotic (41, 42). For example, activated Notch in T-cells promotes tumorigenic properties and protects against T-cell receptor-mediated apoptosis (43, 44). However, Notch enhances apoptosis in malignant B-cells and the HeLa cervical cancer cell line in a dose-dependent manner (45). It is therefore possible that the γ-secretase-regulated processing of Notch at the newly identified S5 site by MIPEP may contribute to the molecular switching mechanism for the context-dependent decisions of different cell fates, including cell death. In this study, we show that ΔNICD negatively affects cell viability and mitochondria membrane potential. It has been suggested that NICD localizes to the mitochondria and acts as a cellular sensor to block staurosporine-induced apoptosis (17). In the context of our study, cellular cues may induce MIPEP cleavage of Notch as a terminal signal. Moreover, several studies have found that Notch and γ-secretase can regulate levels of mitochondrial proteins, including MIPEP, to impact mitochondrial function (18, 46), suggesting the existence of a feedback loop of Notch and MIPEP regulation via γ-secretase. The role of MIPEP in the context of apoptotic cellular cues and the possibility of a feedback loop will need to be addressed in the future.

Our findings raised the possibility that MIPEP is a modulator of Notch signaling. Analysis of MIPEP knockdown in Drosophila supports this hypothesis (supplemental Fig. S2). Mechanistically, MIPEP could modulate Notch signaling by trimming the first eight residues of NICD in mitochondria, resulting in an alternative Notch intracellular fragment. It also removes several key N-terminal residues of the RAM sequence, a domain thought to trigger allosteric changes in the Notch transcriptional activation complex for the derepression of transcription (47–50). It is possible that N-terminal trimming affects the compartmentalization, trafficking, stability, or quantity of the active Notch species and that MIPEP-mediated processing represents a regulatory mechanism to produce the right amount of bioactive Notch intracellular domain at the right time under the right condition. Although provocative, the unexpected implication of a mitochondrial peptidase in a key intrinsic cell signaling pathway will advance our understanding of regulated and sequential proteolysis (51, 52) and expand the roles of mitochondria in cell regulation (53–57).