γ-Secretase is a protease involved in important signaling cascades like the Notch, CD44 and cadherin pathways (1). It is critically involved in Alzheimer's disease by mediating the release of the Aβ peptide (2) and in many tumors where the Notch pathway is upregulated (3). γ-Secretase is composed of four different subunits: presenilin 1 or 2 (PS), nicastrin (Nct), Aph1 and Pen2 (for review, see Ref. 2). It assembles in the endoplasmic reticulum (ER) or the early secretory pathway by subsequent recruitment of PS to an initial assembly intermediate composed of Nct and Aph1, followed by the final recruitment of Pen2 (for review, see Refs. 4–6). ER-retention/retrieval signals in the subunits ensure that only fully assembled γ-secretase leaves the ER. Such signals have been identified in PS1 (7), Pen2 (8) and Nct (9). Interestingly, these motifs are located in transmembrane domains (TMDs) and therefore do not belong to one of the well-characterized ER-retention/retrieval signals like RXR found in cytoplasmic domains of many ion channels (10). Little is known about the machinery mediating ER retention/retrieval of TMD-based signals. We recently identified Rer1 as a protein interacting with TMD1 of unassembled Pen2 and localizing it to the ER (8). How TMD-based ER-retention/retrieval signals are masked in the fully assembled complex to bypass quality control mediated by Rer1 and other proteins is unknown. We here describe dual functions for both TMD1 of Pen2 and TMD4 of PS1 in ER retention/retrieval and in γ-secretase complex assembly. In the fully assembled γ-secretase complex, the ER-retention/retrieval signals in both TMDs mask each other, permitting ER export.
γ-Secretase is critically involved in the Notch pathway and in Alzheimer's disease. The four subunits of γ-secretase assemble in the endoplasmic reticulum (ER) and unassembled subunits are retained/retrieved to the ER by specific signals. We here describe a novel ER-retention/retrieval signal in the transmembrane domain (TMD) 4 of presenilin 1, a subunit of γ-secretase. TMD4 also is essential for complex formation, conferring a dual role for this domain. Likewise, TMD1 of Pen2 is bifunctional as well. It carries an ER-retention/retrieval signal and is important for complex assembly by binding to TMD4. The two TMDs directly interact with each other and mask their respective ER-retention/retrieval signals, allowing surface transport of reporter proteins. Our data suggest a model how assembly of Pen2 into the nascent γ-secretase complex could mask TMD-based ER-retention/retrieval signals to allow plasma membrane transport of fully assembled γ-secretase.
The TMD4 of PS1 contains an ER-retention/retrieval signal
A WNF sequence in TMD4PS1 was shown to interact with Pen2 and mutation of WNF abolished γ-secretase complex formation (11,12). Using a simple cell-based γ-secretase reconstitution assay (7), we confirmed these findings. PS1-EGFP, but not PS1mut-EGFP, where WNF was mutated to three alanines, reconstituted γ-secretase as indicated by Notch cleavage in PS1/PS2 double knockdown (KO) mouse embryonic fibroblasts (MEFs) (Figure S1A).
The WNF sequence in TMD4PS1 contains an asparagine, a polar amino acid. We previously showed that an asparagine in TMD1Pen2 was important for ER localization of unassembled Pen2 (8) (for comparison of the two TMDs, see Figure S1B). To test if TMD4PS1 is important for ER localization, the TMD in CD4 was replaced with TMD4PS1 and its localization determined by immunofluorescence microscopy (Figure 1). CD4 localized to the plasma membrane (PM) while CD4-RXR, carrying a cytosolic ER-retention/retrieval signal (13), localized to the ER. Likewise, CD4-TMD4PS1 localized to the ER in most cells, but in some additionally also to the PM. We then determined if the WNF sequence is involved in ER retention/retrieval and mutated it to three alanines (TMD4PS1mut). The WNF/AAA mutation led to a mixed distribution of the reporter in most cells, with prominent PM localization and some ER localization remaining (indicated by the nuclear envelope staining). This suggests that the three amino acids WNF are important for ER retention/retrieval (Figure 1). The TMD4PS1 mediated ER localization independent of the cell line used (Figure S2A shows HEK 293 cells stably expressing CD4 constructs) and independent of the reporter [Figure S2B shows identical results using tac (interleukin 2 receptor α-subunit) as a reporter]. As an alternative method to determine the localization of the reporter constructs, we employed the selective sensitivity of endoglycosidase H (endoH) to cleave only high-mannose sugars and thus proteins localized in early secretory compartments. As shown in Figure 2A, CD4 was partially endoH resistant, resulting in faster mobility compared with untreated CD4 but clearly slower mobility than the completely deglycosylated CD4. This endoH resistance is indicative of a localization of CD4 beyond early Golgi compartments. In contrast, CD4-RXR was found to be completely endoH sensitive, indicating ER localization (Figure 2A). Likewise, CD4-TMD4PS1 was endoH sensitive, indicating that TMD4 confers ER localization. In contrast, the partial endoH resistance/sensitivity of TMD4PS1mut indicated that a substantial fraction of the construct escaped the ER and was transported beyond the Golgi (Figure 2A). This mixed distribution was in line with the immunofluorescence data (Figure 1). To further substantiate our findings, the cells from Figure 2A were used to determine the localization of the CD4 variants by fluorescence-activated cell sorting (FACS) (Figure 2B). Cells expressing CD4 were used as control for PM localization, cells expressing CD4-TMD1Pen2(8) or CD4-RXR as control for ER localization. As shown in Figure 2B, CD4-TMD4PS1 was retained efficiently, comparable to CD4-RXR. In contrast, twofold increase of CD4-TMD4PS1mut reached the PM, showing a highly significant increase. Taken together, these data show that TMD4PS1 contains an ER-retention/retrieval signal based on the amino acids WNF that is sufficient to localize a reporter protein to the ER.
Dual functions of TMD1Pen2 and TMD4PS1 in ER localization and complex formation
The data presented so far suggest that TMD4PS1 is bifunctional, it carries an ER-retention/retrieval signal and an important PS1–Pen2 interaction domain. Having shown this dual role for TMD4PS1, we also tested if TMD1Pen2 would have such a dual role. TMD1Pen2 contains an asparagine-based ER localization signal (8), and we next tested whether it would be important for complex formation. To this end, mutations were generated in TMD1 of green fluorescent protein (GFP)-Pen2. The asparagine important for ER retention/retrieval (GFP-Pen2N/L) as well as the asparagine plus two adjacent amino acids (GFP-Pen23L) were mutated to leucine, and the 9 amino acids around the ER-retention/retrieval signal exchanged with CD4 TMD sequences (GFP-H2Pen2 , all constructs are schematized in Figure S1D). The constructs were stably expressed in HEK293 cells stably overexpressing Amyloid Precursor Protein (APP) with the Swedish mutation (Swe) and a short hairpin RNA (shRNA) against Pen2 (14). Cells were analyzed for complex assembly and γ-secretase activity (Figure 3A). As shown earlier (8), GFP-Pen2 fully reconstituted γ-secretase assembly and activity, as indicated by increased Nct maturation, PS1 endoproteolysis and reduction of APP C-terminal fragments (APP CTF) accumulation. GFP-Pen2N/L and GFP-Pen23L reconstituted Nct maturation, PS1 endoproteolysis and APPCTF accumulation but not to the same extent as GFP-Pen2. In addition, GFP-Pen2N/L and GFP-Pen23L were expressed in lower amounts, suggesting they are not as well stabilized by incorporation into γ-secretase as GFP-Pen2. In contrast, GFP-H2Pen2 was severely compromised in γ-secretase assembly, indicated by reduced Nct maturation, accumulation of PS1 holoprotein and of APPCTF (Figure 3A). To further show that GFP-H2Pen2 was not able to reconstitute a functional γ-secretase complex, we exploited the fact that unassembled Pen2 has a much shorter half-life than γ-secretase complex-associated Pen2 (14). As expected, GFP-Pen2 is very stable after cycloheximide treatment, confirming that it is assembled into γ-secretase complexes (Figure 3B). In contrast, already after 3 h of a cycloheximide chase GFP-H2Pen2 is not detectable anymore, indicating it is not assembled into γ-secretase and therefore rapidly degraded (Figure 3B).
To test if the two TMDs directly interact with each other, we measured direct interactions of isolated TMD peptides in vitro by fluorescence spectroscopy and fluorescence resonance energy transfer (FRET). The fluorophore-labeled TMD peptides displayed the same helicity when measured with CD spectroscopy (data not shown). The fluorescence emission of the donor-labeled TMD4PS1 peptide (peptide sequence in Figure S1E) was significantly reduced when the acceptor-labeled TMD1 was present (Figure 3C) because of an interaction of the two peptides and energy transfer from the donor to the acceptor fluorophore. The donor fluorescence was less quenched when the acceptor-labeled TMD4PS1mut was used instead of the TMD4PS1 peptide, suggesting that the interaction is specific and dependent on the WNF sequence. Quantification of the FRET measurements indicated a reduction in the energy transfer efficiency from 48 to 33%, corresponding to a reduction of 31% (Figure 3D). The dependence of energy transfer between the donor and acceptor dye on the mole fraction of the acceptor can be used to determine the number of peptides associated within an oligomer (15). In Figure 3E, the determined FRET efficiencies are shown as a function of acceptor mole fraction. The FRET efficiency depends linearly on the acceptor mole fraction, which clearly shows that the formed oligomer is a dimer. Based on these results, it can therefore be concluded that the TMD1Pen2 and TMD4PS1 peptides form a dimer.
Although we cannot completely rule out the role of spatial separation because of the destruction of ER-retention/ retrieval signals in the assembly of γ-secretase, our data suggest that the domain in TMD1Pen2 carrying the ER-retention /retrieval signal is also critical for γ-secretase complex assembly by direct interaction with TMD4PS1. Thus, like TMD4PS1, TMD1Pen2 has a dual function in ER retention/retrieval and γ-secretase assembly.
Masking of ER-retention/retrieval signals by direct interaction of TMD1Pen2 and TMD4PS1
An important question in understanding the assembly and ER export of γ-secretase is how the ER-retention/retrieval signals are masked in the fully assembled complex, because otherwise it could never reach the PM. An elegant way for the cell to achieve this would be if domains in subunits would serve two purposes, on one hand carrying ER-retention/retrieval signals and on the other hand carrying subunit–subunit interaction domains. Complex assembly and therefore interaction of such domains might mask the respective ER-retention/retrieval signals and allow export of the now fully assembled complex. We tested if the direct interaction of TMD1Pen2 with TMD4PS1 would have consequences on the accessibility of the ER-retention/retrieval signals. Such experiments are not possible in the context of γ-secretase with its 19 not yet fully characterized TMDs and the various assembly intermediates, which complicate the analysis. Therefore, the TMDs were analyzed in reporter proteins. To allow distinction between the constructs, CD4 and tac (interleukin 2 receptor α-subunit) were used for TMD1Pen2 and TMD4PS1, respectively. HeLa cells stably expressing ER-localized CD4-TMD1Pen2-EGFP were transiently transfected with tac constructs. As expected, transfection of tac-RXR did not change the ER localization of CD4-TMD1Pen2-EGFP, suggesting the two proteins did not interact (Figure 4A, middle). In contrast, when tac-TMD4PS1 was transfected, in a number of cells coexpressing both proteins, a clear PM localization of CD4-TMD1Pen2-EGFP and tac-TMD4PS1 was observed, suggesting TMD1Pen2 and TMD4PS1 interact with each other, promoting ER export (Figure 4A, bottom). Similar results were obtained when CD4-RXR or CD4-TMD4PS1 was transfected. Only when TMD1Pen2 and TMD4PS1 were coexpressed, a PM localization of the reporters could be observed (Figure S2C). To assess this interaction in a more quantitative manner, COS cells were transiently transfected separately with CD4-TMD1Pen2-EGFP and CD4-TMD4PS1-myc or with both plasmids together and the ratio of surface/total CD4 immunoreactivity was determined by FACS. When expressed alone, CD4-TMD4PS1-myc and CD4-TMD1Pen2-EGFP are localized to only 6.9 and 8.6% at the PM, demonstrating efficient ER retention/retrieval as shown in Figure 1 and before (8). However, when CD4-TMD1Pen2-EGFP and CD4-TMD4PS1-myc were coexpressed together, 14% of CD4 reporter was localized to the PM, a statistically significant twofold increase. This increase is not because of a saturation of the ER-retention/retrieval machinery by overexpression, as shown by western blotting using CD4 antibodies (Figure 4C). In addition, increasing amounts of expression plasmid ranging from 2 to 8 µg did not change the ratio of surface/total amounts of CD4-TMD1Pen2 (data not shown). Moreover, CD4-TMD1Pen2 and CD4-TMD4PS1 do not seem to compete for the same ER-retention/retrieval machinery, making it unlikely that the observed effect is due to competition with endogenous factors. Despite the similar occurrence of a polar asparagine in both ER-retention/retrieval signals, only CD4-TMD1Pen2 is localized via the sorting receptor Rer1 to the ER, as shown using Rer1 siRNA in stably expressing HEK293 cells (Figure 4D). As shown earlier (8), upon Rer1 downregulation almost threefold more CD4-TMD1Pen2 localized to the PM, whereas CD4-TMD4PS1 localized to the ER irrespective of Rer1 levels. As an additional control, CD4-RXR did not change ER localization, irrespective of Rer1 levels (Figure 4D). This suggests that there must exist a yet to be discovered additional Rer1-independent ER-retention/ retrieval mechanism sensing TMD-based signals.
Further evidence for the relevance of this interaction for γ-secretase complex assembly and transport came from HEK293 cells stably overexpressing CD4-TMD4PS1. In these cells, endogenous Nct maturation was abolished completely and immature, ER-localized, not fully assembled Nct, accumulated (Figure 5A). The maturation defect was similar to HEK293 cells stably expressing an anti-Pen2 shRNA (14). The defect in Nct maturation/transport resulted in a strong reduction in γ-secretase activity, as indicated by the strong reduction of Notch intracellular domain (NICD) production, the product of γ-secretase-dependent Notch cleavage (Figure 5A,B). CD4-TMD4PS1 directly binds endogenous Pen2 (11), suggesting that it acts in a dominant-negative manner by titrating away endogenous Pen2.
Taken together, these data indicated that the two TMDs directly interact with each other and masked their respective ER-retention/retrieval signals, allowing to bypass the quality control that would otherwise sense such signals and localize its bearers to the ER.
For proper cell function, it is essential that only correctly assembled, fully functional oligomeric complexes like ion channels or cell surface receptors are transported beyond early secretory compartments. Also in the case of γ-secretase, such quality control mechanisms prevent the export of unassembled subunits to the distal secretory pathway, where they might have deleterious effects. ER retention/retrieval is mediated by specific signals in the unassembled subunits. In the fully assembled complex, these signals must be masked or modified, otherwise an assembled complex could never pass the quality control and get exported. In the case of cytoplasmic ER-retention/retrieval signals, such masking was characterized in some instances down to the molecular detail (16–18), whereas in the case of TMD-based signals little is known how they could be masked or modified to overcome ER retention/retrieval.
We here show that PS1, in addition to the ER-retention/ retrieval signal in the C-terminal hydrophobic domain described before (7), carries an ER-retention/retrieval signal in TMD4 based on the amino acids WNF. The WNF sequence was previously shown to be a binding site for Pen2 essential for γ-secretase assembly (11,12). We confirmed this finding by showing that a PS1 mutated in the WNF sequence cannot reconstitute active γ-secretase and by showing that overexpressing the TMD4 acts in a dominant-negative way on γ-secretase assembly, probably by titrating away Pen2. The WNF motif has no resemblance to other TMD-based ER-retention/retrieval signals like the ones described in acetylcholine receptor (AChR) subunits (19) or Pen2 (8), apart from the fact that a polar amino acid, asparagine, seems to be important. As shown here it is not Rer1 dependent, suggesting that there must exist a yet to be discovered additional ER-retention/retrieval mechanism sensing TMD-based signals. The analysis of the various TMD-based signals is not possible in the context of γ-secretase with its 19 not yet fully characterized TMDs and the various assembly intermediates, necessitating the use of reporter proteins like CD4. We therefore at present cannot analyze the relative relevance of the ER-retention/retrieval signals in PS1, nor can rule out the existence of even a third or fourth signal.
The TMD1 in Pen2 also has a dual function: it contains an ER-retention/retrieval signal and it is also an interaction domain essential for γ-secretase assembly. We here show that the C-terminal part of the TMD carrying the ER-retention/retrieval signal is important for assembly, whereas Kim et al. showed that the N-terminal half is essential (20). At present we do not have an explanation for this discrepancy, maybe mutations in either half of the TMD1 have an impact on the overall structure of the TMD, thus compromising interaction with TMD4 of PS1.
Our data suggest an elegant way by which masking of ER-retention/retrieval signals could be achieved by direct binding of the two domains to each other. In Figure 6, a model is proposed about the last steps in γ-secretase complex assembly. The trimeric Nct-Aph1-PS1 complex intermediate could be localized to the ER by a combination of the ‘open’ WNF-containing signal on TMD4PS1 and the unfolded ectodomain of Nct, which binds to ER chaperones like calnexin (21) (Figure 6A). Upon binding of Pen2, known to be the last component to be added (reviewed in 4,5), this signal and the one in Pen2 are masked and thus no longer recognizable for the retention machinery (Figure 6B). In addition, binding of Pen2 could induce the terminal folding of the Nct ectodomain, which might bury exposed hydrophobic domains releasing it from the ER chaperones like calnexin. This is speculative at present, but the ectodomain is the likely candidate for being involved in ER localization of unassembled or not fully folded Nct, because we could not identify an ER-retention/retrieval signal in the TMD or cytoplasmic tail of Nct (8,22). In addition, upon Pen2 recruitment, the Nct ectodomain undergoes a conformational change (22,23), which might serve as a switch from an ER export restrictive to a permissive state. Given the high number of 19 TMDs in γ-secretase, future work will likely identify more ER-retention/retrieval signals, interactions between TMDs and their role in assembly and transport of γ-secretase. It will be interesting to identify ER-retention/retrieval signals in Aph1, where the recently characterized histidines in TMD5 and 6 might be of interest (24).
Materials and Methods
The following antibodies were used: anti-myc monoclonal 9E10 (Santa Cruz) and polyclonal A-14 (both Santa Cruz) were used for immunofluorescence and FACS analysis, respectively. To detect CD4, the following antibodies were used: monoclonal CD4-APC M-T466 (Miltenyi Biotec) for FACS analysis, monoclonal EDU-2 (Diatec Monoclonals) for immunoprecipitation (IP) and immunofluorescence and polyclonal H-370 (Santa Cruz) for western blot. Anti-GRP78 polyclonal C-20 (Santa Cruz) was used as an ER marker in immunofluorescence. Anti-β-actin polyclonal ab8227 (Abcam) was used for western blot. Polyclonal rabbit anticleaved Notch 1 (Val1744) (Cell signaling) was used for western blot. Polyclonal anti-Nct N1660 was obtained from Sigma. Antibodies against PS1, APP C-terminus and affinity-purified antibody against the N-terminus of PEN-2 (1638) have been described before (25) and references therein. Rat anti-mTac antibody (PC61.53) was obtained from J. Stirnweiss (FSU Jena). Secondary horseradish-peroxidase (HRP)-conjugated antibodies for western blot were obtained from Promega. Secondary antibodies conjugated to fluorophores [Alexa 488, Alexa 555, Alexa 660, phycoerythrin (PE)] for immunofluorescence and FACS analysis were obtained from Invitrogen.
HEK293 cells stably expressing swedish mutant APP (Swe) (26) and Swe cells stably expressing shRNA against Pen2 (14) were described earlier. MEF from PS1−/−/PS2−/− mice are described in (27) and were kindly provided by B. De Strooper. Hela Kyoto cells were kindly provided by Rainer Pepperkok (EMBL). All cells were grown in standard culture conditions with DMEM and 10% fetal calf serum (FCS).
FACS analysis was carried out using a BD FACSCantoII and standard protocols. The staining for the extra- and intracellular CD4 antigen was performed as described earlier (8). For the FACS-based masking assay, we gated for GFP + and myc + cells and determined the amounts of CD4 antigen. The percentage of surface CD4 antigen was calculated from the geometric mean of the signal intensities in permeabilized (total CD4 antigen) or unpermeabilized (surface CD4 antigen) cells.
cDNA constructs and transfections
CD4-TMD1Pen2 has been described earlier (8). CD4-RXR was obtained from Lily E. Jan [Howard Hughes Medical Institute; (13)]. Constructs containing TMD4PS1 were generated using standard polymerase chain reaction (PCR)-based mutagenesis. The amino acid sequences of TMDs swapped into CD4 and mTac and adjacent domains are shown in Figure S1B/C. To clone C-terminally tagged versions of the CD4 reporters, the stop codon was removed and a NotI site was introduced by PCR. The CD4 fragment was subsequently cloned into pcDNA4/myc-His A. The enhanced green fluorescent protein (EGFP) tag was amplified from pmaxGFP (AMAXA) generating a NotI site at the 5′ end and CD4 and EGFP were joined in pcDNA3.1/Hygro. mTac was amplified from spleen and lymph node cDNA of a complete Freund's adjuvant (CFA)-immunized C57BL/6 mouse using appropriate primers and subsequently cloned into pcDNA3.1/Hygro (Invitrogen). To generate mTac-RXR, the stop codon was removed and a NotI site was introduced by PCR. This fragment was swapped with the CD4 part in CD4-RXR. PS1-EGFP (PE) and GFP-Pen2 have been described earlier (8,28). NotchΔE with six C-terminal myc-tags was obtained from Raphael Kopan [Washington University; (29)]. To clone NotchΔE-EGFP, NotchΔE was amplified excluding a linker and the six C-terminal myc-tags and introducing an XbaI site at the 3′ end. EGFP was amplified from pEGFP-N1 (Clontech) introducing an XbaI site at the 5′ end. Both fragments were joined in pcDNA3.1/Hygro. Mutations were introduced using the QuikChange site-directed Mutagenesis Kit (Stratagene). Cells were transfected using Lipofectamine 2000 (Invitrogen).
Deglycosylation and immunoblotting
Deglycosylation experiments were carried out as described previously (28). Cells were lysed in STEN lysis buffer [50 mm Tris pH 7.6, 150 mm NaCl, 2 mm ethylenediaminetetraacetic acid (EDTA), 1% Nonidet P-40 (NP-40) and protease inhibitor mix]. Proteins were separated on SDS–PAGE and transferred to polyvinyl difluoride (PVDF) membranes. Membranes were cut at appropriate positions and blotted with antibodies as indicated.
Carboxyfluorescein- (FRET donor: TMD4PS1;TMD4PS1mut) and Tetramethyl-6-Carboxyrhodamine-labeled (FRET acceptor: TMD1Pen2) peptides were kindly provided by Jens Rothemund (Core Unit Peptide Technologies IZKF). For the FRET analysis, fluorescence emission was measured in the range of 480–650 nm after excitation of the fluorescence donor at 450 nm in a 5-mm path length cuvette for an equimolar mixture of donor and acceptor peptides. The concentration of each peptide was 230 nm. The spectra were collected with a SHIMADZU RF-5301PC spectrofluorometer in a 10 mm phosphate-buffered solution (pH 7.4) containing 5 mmn-dodecyl-β-d-maltoside. Energy transfer, E, was calculated from the donor fluorescence intensity at 521 nm in the absence and presence of an acceptor according to E = 1 − (FDA/FD), where FD and FDA are the donor intensities of samples containing only donor-labeled peptides and samples with both donor- and acceptor-labeled peptides, respectively.
Immunofluorescence was performed using standard protocols (30). Fixed cells were analyzed on a Zeiss Axiovert 200 microscope (Zeiss) equipped with a ×63/1.25 objective and standard fluorescein isothiocyanate (FITC), tetramethylrhodamine isothiocyanate (TRITC) and Alexa 660 fluorescence filter sets, using an Axiocam Mrm Camera and AxioVision software. For some images, a Zeiss Apotome was used (Carl Zeiss). Images were assembled and processed using Adobe Photoshop.
Means of numerical data were compared using Student's t-test. A difference in means was considered statistically significant (*) with p < 0.05 and highly significant (**) with p < 0.01. All error bars depict the standard error (SE) with the number of independent replicates indicated in the figure legend.
This work was supported by the Hans-and-Ilse-Breuer Stiftung (M. F.), the Ministry of Science, Research and Arts of Baden-Württemberg (D. S. and A. V.) and the Deutsche Forschungsgemeinschaft (SFB604 to C. K., SCHN 690/2-3 to D. S.). We thank Sven Lammich (DZNE), Bart de Strooper (K.U. Leuven) and Rainer Pepperkok (EMBL Heidelberg) for providing cell lines, Raphael Kopan (Washington University) and Lily E. Jan (Howard Hughes Medical Institute) for NotchΔE and CD4-RXR cDNA, respectively. Simone Tänzer and Daniela Reichenbach are acknowledged for excellent technical assistance, Matthias Görlach for help with peptides and Kerstin Hünniger for NotchΔE-EGFP cDNA (all Fritz Lipmann Institut). We are also grateful to Jörg Stirnweiss (FSU Jena) for providing anti-mTac antibody and Jens Rothemund (IZKF Leipzig) for peptide synthesis.