Ectodomain shedding and intramembrane cleavage of mammalian Notch proteins is not regulated through oligomerization

biotinylated with biotin were lysed Co-IP into equal aliquots and and biotinylated after Myc


Introduction
Notch signaling is a highly conserved signaling pathway that mediates cell fate decisions during multiple stages of development in multicellular eukaryotes; in the adult, Notch signaling is involved in several diseases including cancer, stroke and muscular dystrophy (for a review see (1,2)). Notch receptors are single pass type I transmembrane proteins characterized by a large extracellular ligand binding domain containing 11-36 EGF repeats, three Ca 2+ binding Lin12-Notch repeats (LNR) that maintain the receptor in an inactive state, and a single predicted TMD (3)(4)(5). Mammalian cells have four different Notch receptors and up to six different ligands, five of which are membrane-tethered type I proteins that belong to the DSL family (Delta/Serrate(Jagged)/Lag-2). Notch ligands are also type I proteins containing extracellular EGF repeats and a conserved DSL domain that mediates receptor binding.
Binding of DSL ligands to EGF repeats 11-12 leads to Notch activation (6) via a poorly understood mechanism requiring ubiquitination-induced ligand endocytosis (reviewed in (7)).
Ligand binding to Notch receptors triggers juxtamembrane cleavage by ADAM metalloproteases in the extracellular domain of Notch to produce NEXT (Notch Extra Cellular Truncation (8). Shedding of the Notch extracellular domain is followed by intramembranous cleavage of Notch by γ-secretase, a novel multi-protein complex containing four transmembrane proteins: the aspartyl protease Presenilin (PS; containing the catalytic residues), Pen 2, Aph-1 and Nicastrin (9). Consequently, the Notch Intracellular Domain (NICD) is released from the membrane and translocates to the nucleus where its association with the transcription factor RBPjk leads to activation of target genes (reviewed in (10,11)). Other γ-secretase substrates include the receptor tyrosine kinase ErbB4, and the cell adhesion molecules CD44, E-Cadherin and Nectin (reviewed in (12,13)). γ-secretase is best known for its involvement in the intramembranous cleavage of the Amyloid Precursor Protein (APP) that generates the neurotoxic amyloid β (Aβ) peptides that form aggregates associated with senile plaques in Alzheimer's Disease. Given that mutations in PS lead to increased production of by guest on March 17, 2020 http://www.jbc.org/ Downloaded from amyloidgenic peptides and cause early onset Familial Alzheimer's Disease, γ-secretase inhibitors could be useful therapeutic tools in the treatment of this disease (reviewed in (14)). Therefore, further insight into the regulation of Notch proteolysis and γ-secretase activity is of great importance in understanding both development and disease.
Current models suggest that for iCLiP-mediated proteolysis to occur, a preceding Juxtamembrane Proteolytic step (JP) is necessary. This is thought to permit dissociation of substrate TMD (reviewed (16,18)). The prototype I-CLiP, S2P (site-2 protease) is a metalloprotease involved in the release of sterol regulatory element binding protein (SREBP) from the Golgi as part of a cholesterol/fatty acid sensing system in the cell.
The substrate, SREBP, has two TMDs connected by a short loop; after SREBP is transported by the sterol sensing protein SCAP to the Golgi, the loop is cleaved by S1P (site-1 protease) and S2P cleavage of a single SREBP TMD follows. Another I-CLiP, related to the aspartic peptidase PS/γ-secretase is signal peptide peptidase (SPP), which catalyzes cleavage of signal peptides from type II membrane proteins. SPP substrates have two lipid-embedded domains that are separated after cleavage of the precursor protein in the secretory pathway by signal peptidase (SP) (19). The single TMDs of γ-secretase substrates were proposed to exist as dimers prior to ectodomain shedding (see below; (20)) which has been demonstrated for most presenilin substrates Caenorhabditis elegans raised the possibility that ligand drives Notch dimerization (24).
The Notch intracellular domain contains seven ankyrin repeats, of which repeat 2-7 assume an ankyrin fold (11) and was shown to dimerize in yeast two-hybrid assays (25).
However, structural and biochemical analyses using the purified ANK domain have shown unequivocally that the Notch intracellular domain is monomeric (11,26).
Why juxtamembrane cleavage is a prerequisite for intramembrane cleavage is unclear, as is the mechanism by which these proteases recognize and cleave scissile bonds within the lipid bilayer. Activating mutations in Notch (Drosophila), Glp-1 and Lin-12 (C. elegans) indicate that the extracellular juxtamembrane region of Notch functions to inhibit receptor activation (4,(27)(28)(29)(30). It has been proposed that juxtamembrane cleavage of Notch, and by inference of other I-CLiP substrates, is needed because in the absence of ligand the extracellular domain imposes a conformational state protecting scissile bonds from hydrolysis (3,20,29,31). This model of I-CLiP regulation emerged from elegant in vivo reporter assays in Drosophila utilizing a transcription factor (TMD-gal4VP16) that depended on γ-secretase mediated cleavage of its TMD to activate an UAS-LacZ reporter (20). These studies demonstrated that LacZ activation was attenuated in transgenic flies expressing gal4VP16 proteins containing the Glycophorin-A (GpA) TMD; this domain forms a stable dimer within the lipid bilayer (32).
Furthermore, engineered dimerization of the Notch TMD in TMD-gal4VP16 using a heterologous, extracellular leucine zipper also impaired reporter gene activation (20). In contrast, engineering mutations in the GpA TMD or in the extracellular leucine zipper forcing them to adopt a monomeric conformation resulted in LacZ activation that was comparable to that seen with TMD-gal4VP16 containing a wild type Notch TMD. These findings supported a model in which in the absence of ligand, oligomerization, mediated by the extracellular domain, prevents ectodomain shedding and maintains the TMD domain in a dimeric 'inaccessible' conformation. This model predicts that constitutively active Notch proteins are monomers whereas intact or inactive proteins are dimers (20); it also predicts that TMD lacking the extracellular domain will not self associate. While the behavior of chimeric Notch proteins supported this interpretation (29), no direct measurement of oligomerization state or its effect on proteolysis have been performed.
To test the hypothesis that oligomerization status functionally influences Notch processing, we investigated the oligomerization state of the Notch proteins that are direct substrates for γ−secretase and compared them to others that require ligand binding in order to become substrates. We find that in contrast to the predictions of the current model, all Notch1 proteins exist predominantly as monomers regardless of their ability to act as efficient γ−secretase substrates. We identified the Notch 1 TMD as the minimal substrate sufficient for PS dependent cleavage, indicating that the intracellular domain does not contribute to the interaction with γ-secretase in the cell. In order to evaluate the association tendencies of the Notch TMD and to determine if cleavageimpairing TMD mutations altered the oligomerization state of Notch TMD, we used the established bacterial TOX-CAT system (33). We discovered that the Notch and APP TMDs are dimers irrespective of their ability to act as an effective substrate in mammalian cells. Together, our results indicate that controlled dimerization of TMD by the Notch extracellular domain and its reversal by ligand binding is unlikely to underlie the regulatory mechanism of ectodomain shedding and hence intramembranous cleavage. Instead, we favor induced conformational change as the regulatory mechanism.

Plasmids and vectors
All Notch plasmids were initially cloned into pCS2+6MT as described (34). Notch1∆E contained a TMD substitution (M1726V) facilitating biochemical analysis (34). Sitedirected mutagenesis was performed using the Quickchange kit according to manufacturer's instructions (Stratagene). A construct containing 4xCSL synthetic binding sites in tandem was used for Notch transcription assays (35).
Notch Renilla fusions were made by PCR-directed cloning and removal of the 6MT (6Myc tag) for replacement with the Renilla Luciformis gene from pCMV-RL (Promega).

Cell culture and transfection
HEK293 and PS1/2dKO cells (39) were grown and maintained in DMEM with 10 % FBS.

Western Blot and Immunoprecipitation analysis
Whole cell lysates were prepared by directly adding Laemmli SDS sample buffer (+10 mM DTT) to PBS washed cells. For bacterial lysates freshly inoculated o/n cultures were normalized for OD (OD595), grown for 2 1/2 hrs at 37 0 C, centrifuged at 9000 rpm and resuspended in SDS sample buffer and boiled. Western blotting was performed as previously described (29)

Active and Inactive Notch proteins are monomeric.
We reasoned that if dimerization was an important component in regulation of Notch signaling, active and inactive Notch molecules should differ in their dimerization/oligomerization state. We previously reported that engineered oligomerization of the Notch ectodomain inhibited the metalloprotease dependent extracellular cleavage (S2) and monomeric forms of such proteins allowed S2 cleavage to occur (29). Likewise, Fibroblast growth factor receptor (FGFR)-Notch fusion proteins are constitutively cleaved (and thus active); addition of the ligand FGF promotes receptor dimerization and reduces their activity by about 30% (40). Similarly, placement of the an extracellular dimerization domain or a dimer-forming TMD inhibited processing of a TMD-galVP16 fusion protein in Drosophila (20). None of these studies, however, established the intrinsic ability of Notch proteins to autodimerize. To rigorously test whether inactive Notch proteins interacted with each other, we adapted a quantitative co-immunoprecipitation method developed by Taipale et al., (38). This method employs fusion proteins with the Renilla Luciferase (RL) gene; the degree of association between the tagged protein and an interacting partner was directly measured by RL activity recovered after immunoprecipitation (IP) with the partner ( Figure 1C) corrected for remaining RL activity in the supernatant ( Figure 1B). This specific association was presented as fold over non-specific interactions measured with a non-interacting control, the 12 pass transmembrane receptor mPatched1-RL (mPtc; Figure 1D).
First, we asked whether Notch proteins with the RL fused to their carboxy termini ( Figure 1A) interact with the Notch partner, RBPjk. When co-expressed in HEK293 cells, N 1 ICV-RL proteins efficiently immunoprecipitated with Flag-tagged RBPjk ( Figure 1D) as did N 1 ICV-RL produced from the membrane-tethered, extracellular domain deleted Notch (N 1 ∆E-RL). As expected, this latter interaction was blocked by adding γ-secretase inhibitor (DAPT) to the growth medium or when using a N 1 ∆E mutant (V1744G) that is cleaved less efficiently by γ−secretase. Intact Notch proteins are not γ-secretase substrates in the absence of ligand, and they too interacted poorly with RBPjk. Finally, we assayed the ability of constitutively active N 1 ∆E-RL proteins to activate a Notch reporter construct composed of four tandem CSL-binding sites (4xCSL-Luc; (41); Figures 2B, 6C). These control experiments confirmed that Notch RL fusion proteins behave as their Myc-tagged counterparts and that RL permits quantitative analysis of Notch-partner interactions.
Intact Notch and N 1 LNR molecules lacking the extracellular EGF repeats are inactive in signaling ( Figure 1A; (29)); the current models would thus predict they are oligomeric. Mutation of a cysteine pair (CC1682SS) can activate Notch signaling in flies (28) and also activate N 1 LNR (29). Additional activating mutations in Notch proteins which map to the juxtamembrane region include S1597N (42) and A1695T (24). The current models predict that these molecules are monomers and that these mutations affect the ability of these molecules to dimerize. To determine if these mutations affected the oligomerization state of Notch proteins they were introduced into the mouse N 1 LNR proteins and found that although far inferior to N 1 ∆E, this resulted in a 2-3 fold greater reporter activity compared to the parental N 1 LNR-RL protein ( Figure 3A). Further biochemical analyses indicated that all mutants are expressed to a similar extent, undergo furin (S1) cleavage with equal efficiency ( Figure 2B  To test if oligomerization state of N 1 LNR variants varied in correlation with their activity, extracts from cells transfected with LNR-RL and LNR-6MT wild type proteins or harboring the activating CC>SS or S1597N mutation were subjected to Myc-IP. Recovered luciferase activity from LNR-RL and mPtc1-RL in anti-Myc immunoprecipitates was comparable ( Figure 3). This could indicate that LNR proteins failed to interact with each other. However, IP experiments are normally conducted in buffers lacking calcium (Ca 2+ ) which is required for folding of the LNR domain in-vitro (43,44). Repeating the experiments in the presence of physiological amounts of Ca 2+ confirmed that active and inactive LNR proteins fail to specifically associate with each other (see below and data not shown,).
One limitation of these experiments and those performed in Drosophila (20) Figure S1 and not shown), indicating the associations were not mediated through the Myc epitope. Next we examined whether the interaction between N 1 EGF repeats was confined to the highly conserved EGF repeats of Notch proteins and their ligands or whether it extended to unrelated EGF repeat-containing proteins as well. We co-expressed N 1 FL-RL with an HA tagged Low Density Lipoprotein Receptor (hLDLR) (47) that has similar Ca ++ -binding-EGF repeats (45). We found that hLDLR preferentially associated with mPtc1-RL (Supplementary Figure S2). Finally, to exclude the possibility that truncated Notch proteins could not interact with full length Notch proteins because they were targeted to different membrane subdomains, we also co-expressed RL equivalent forms of the truncated proteins (e.g N 1 ∆E-RL with N 1 ∆E-6MT, N 1 LNR-RL with N 1 LNR-6MT; Figure   1A). The RL-tagged proteins failed to form homotypic interactions in the absence of the EGF repeats (i.e., they interacted with each other as efficiently as they interacted with mPtc-1; Figure. 3 and not shown).
If all surface full length Notch proteins are dimeric, our results would be consistent with the notion that intact Notch molecules are kept inactive by dimerization.
To determine the fraction of dimers at the surface, we biotinylated surface proteins and measured RL activity recovered by Myc immunoprecipitation before or after depletion of surface protein from the same lysate. If all the dimers we observed were at the cell surface, depletion of biotinylated surface proteins from the lysate (with streptavidinagarose) prior to IP should greatly reduce the recovery of Notch dimers compared to non-biotinylated extracts (experimental scheme in Figure 4E). Instead, we found that the majority (81-88 %) of Notch dimers survive depletion and are thus intracellular. Only a fraction (0.94 to 1.90 %) is present at the cell surface. In contrast, N 1 FL-RL monomers were 3.6 to 7.0 fold more abundant at the cell surface than dimers, yet they were not cleaved and did not interact with RBPjk ( Figure 1). The observation that most surface Notch proteins are inactive in a monomeric state is in agreement with the observation that changes in oligomerization state cannot discriminate between active (∆E and LNR mutants) and inactive (LNR, FL) molecules.  Figure 5B, C). The characterized GpA TMD and its monomeric mutant GpA (G83I) TMD served as controls (33). G83I TMD was only able to weakly activate the expression of CAT compared to the GpA TMD ( Figure 5B). As an additional control, constructs expressing the FGFR1 TMD were also tested; they too activated CAT expression weakly. Surprisingly, we found that the Notch1 TMD elicited significant expression of CAT in comparison to the reference monomer G83I TMD ( Figure 5B; (33)).
If dimerization contributed to or detracted from TMD recognition or cleavage by γsecretase, mutations that abolish PS-dependent cleavage might elicit changes in the dimerization state of substrate TMDs. To test this we analyzed the effect of most TMD mutations shown in Figure 6A on dimerization of the Notch 1 TMD. We found no obvious difference between mutations that completely abolished PS cleavage or transcriptional activity (i.e. GCGV>LLFF, V1744G and V1744L) and those that had no effect on activity (i.e. V1744A, shown in Figure 5 and V1740L) compared to wild type N1 (N1) ( Figure   6B). To assess the oligomerization state of TMDs from additional PS substrates, we subjected the TMDs of the mammalian N2, N3 and N4 receptors and APP to the TOX-CAT test. All showed robust CAT expression when compared to the monomeric TMDs of GpA or FGFR1 ( Figure 5C). We observed slight differences in the propensity to dimerize among the Notch TMDs, with the Notch2 TMD consistently being the strongest dimer. A slight reduction in CAT expression was observed between N1∆QLH (N1TMD lacking the QLH residues) relative to N1, however the significance of this remains unclear.
Moreover, others using the same assay have demonstrated dimerization of the ErbB4 TMD (48). Thus, contrary to the notion that γ−secretase substrate TMDs are monomeric we find that all can self-associate in this assay.

Changes in self-association motifs do not affect γ-secretase cleavage
It has been shown that PS-dependent cleavage of Notch in Drosophila does not critically depend on the primary amino-acid composition within the TMD (Struhl and Adachi 2000). Furthermore, no significant homology or motif has emerged as more bona fide γ−secretase substrates are being identified (table1 in (13)), yet Val1744 mutations impact Notch proteolysis in mammalian cells. To identify additional sequences within the Notch1 TMD that could impact its utilization as a substrate, we first substituted residues at or in close proximity to the S3 cleavage site ( Figure 6A) in the context of a signaling molecule (N 1 ∆ERL). We compared the ability of all N 1 ∆E TMD mutants to activate 4xCSL-Luc. The previously characterized G, K and L substitution for Val1744 reduced cleavage of N 1 ∆E in a range of 30-80% of wild type, V1744G being the weakest and V1744L being the strongest, at the Notch DNA concentration used here ( Figure 6C).
Alanine at V1744A was tolerated, only modestly affecting cleavage ( Figure 6B) and transcriptional activity. Replacing the entire GCGV1744 sequence to LLFF completely abolished activity as reported previously (37,49).
While helix breaking residues such as glycine are critical in the TMDs of substrates cleaved by other intramembranous cleaving enzymes (50,51) the V1744G substitution in the Notch1 TMD attenuates cleavage. We introduced several additional mutations to further test the role of helix-breaking residues in Notch cleavage while leaving Val1744 intact (underlined below and in Figure 6A). Replacement of both Gly in the sequence GCGV with Ala, a residue known to promote helix formation (ACAV), did not markedly affect reporter activation, suggesting it was still a substrate for γ-secretase ( Figure 6C). Substitution of these glycines (an important component within the dimerization motif in GpA) with alanines did not affect dimerization of N1 in the TOX-CAT assay ( Figure 5B and not shown).
Next, we probed the role of putative self-association motifs within the TMD.
Leucine zippers are well-known protein-protein interaction domains and have been implicated in dimerization of TMD sequences (52). We tested the effect of substituting the di-Leu motifs in LLFFVGCGVLLS with Alanine (AAFFVGCGVAAS), to maintain the helical structure of the TMD. Notch1 AAS mutants as well as the quadruple mutant behaved as wild type in 4xCSL activation assay with LLFF>AAFF mutant consistently showing a small increase in activity relative to the wild type. Finally, the V1744L substitution (VGCGV-VGCGL) may have enhanced the helical structure by generating a new interaction between V1740 and L1744. If this was the case, the reciprocal substitution, replacing V1740 with Leucine, should have an identical negative impact on proteolysis (i.e. VGCGV-LGCGV). However, we found that V1740L had no impact on the ability to activate reporter transcription. The activity of all these mutants was completely abolished in PS1/2 dKO cells (not shown) indicating that the luciferase activity was PS dependent. Collectively, these experiments could not support a role for self-association in either positive or negative regulation of S3 cleavage nor could they establish a role for helix breaking or helix enhancing residues in regulation of Notch proteolysis by γ-secretase.
Phenylalanine scanning of the APP TMD generated shifts in scissile bond preferences by γ-secretase (53). We therefore asked whether the scissile bond cleaved in any of the mutants described above was indeed G1743-V1744, as it is in vivo (54)(55)(56)(57).
To identify the scissile bond we used an antibody recognizing the amino-terminal sequence VLLS exposed after γ−secretase cleavage. We found that in the mutants  Figure 6E). Moreover, the same scissile bond generated in vivo is also generated in the minimal N 1 ∆ICE substrate ( Figure 6E, D). These results indicate that the transmembrane sequences of Notch 1 (with a possible contribution from the juxtamembrane region) was sufficient to act as a PS substrate and that sequences in the intracellular domain do not contribute to the characteristic cleavage by γ−secretase. However, we could not determine with certainty the self-association properties of N 1 TMDs in mammalian cells transfected with N1∆ICE molecules because of their non-specific association with control proteins (e.g. mPtc1-RL), most likely due to the hydrophobic nature of the TMD.

Discussion
Intramembrane proteolysis is emerging as a widespread and evolutionary conserved regulatory mechanism for a variety of signaling pathways. With the exception of Rhomboid substrates (59), all other I-CLiP require an ectodomain shedding event prior to intramembrane proteolysis of the substrate (18). In the regulation of Notch proteolysis, shedding of the extracellular domain is induced by ligand binding and carried out by a disintegrin metalloprotease (29,49,60). Similarly, ectodomain processing of APP by γ−secretase requires cleavage of APP by BACE, a transmembrane aspartyl protease (61), or by the disintegrin metalloprotease TACE or ADAM10 (62,63). It has been proposed that the role of this shedding event is to sever the extracellular domain thus removing an auto-inhibitory domain (e.g., the LNR), inhibiting I-CLiP cleavage. In the case of SPP and SREBP substrates this inhibitory domain is thought to be another helical TMD. Progressive shortening of the extracellular domain of chimeric Notch constructs in transgenic flies leads to an increase in γ-secretase cleavage (20), whereas forced dimerization of TMD substrates reduced proteolysis. Here, we examined two predictions made by prevailing models explaining the regulation of I-CliP activity: (i) TMDs of γ-secretase substrates are protected from proteolysis by virtue of their forcedassociation and (ii) following ectodomain shedding the TMDs will dissociate.

Ectodomain shedding of Notch is not regulated through oligomerization
We established that the highly conserved EGF repeats in the Notch extracellular domain are necessary and sufficient to interact with itself as well as with other Notch family members and the Notch ligand Jagged1 (Figure 2, 4, Supplementary Figure S1 and (46)). Notch can act as a sensor for extracellular Ca 2+ concentration in vivo (64) and coordination of Ca 2+ ions is important in the folding of the EGF and LNR domain in vitro (43,45). Depletion of Ca 2+ by EDTA from the culture medium from Notch expressing cells leads to their activation by dissociating the heterodimeric bond holding the Notch extracellular domain; this region may fold in a manner that obscures access to the S2 site (3). However, we found no role for Ca 2+ ions in regulating homotypic interactions between the EGF repeats of Notch ( Figure S1) suggesting perhaps that oligomerization involves EGF repeats that do not bind Ca 2+ . Intracellular association between Notch molecules and their ligands has been reported previously (46), but the association with Notch2 prompted us to test for specificity. We find no association between Notch and hLDLR, a protein containing multiple tandem EGF repeats ( Figure S2). The role for homodimerization between different Notch molecules (Notch1 and Notch2) remains unclear.
The model predicts that inactive Notch proteins exist mostly in a dimeric state to keep the TMD from drifting apart and becoming subject to proteolysis. To our surprise, we found that Notch and APP all have a propensity to dimerize in the TOX-CAT assay. This is also the case for the TMDs of other γ−secretase substrates (ErbB4 (48) Alternatively, it has been proposed that the length of the extracellular domain could be the major determinant in the regulation of Notch extracellular cleavage (20,54).
This too fails too explain the differences in activity since active and inactive N 1 LNR molecules are of the exact same length. Therefore, regulation of the S2 extracellular cleavage is likely to involve structural changes that may be mimicked by mutations (24,42). The nature of these structural changes remains uncertain.

The helical content of the Notch TMD
It is accepted that helical substrate TMDs are resistant to cleavage by I-CliPs.
Whereas Rhomboid (51), SPP (50) and SREBP (70) all require helix-breaking residues in the TMDs of their substrates to elicit intramembrane cleavage; we found no such requirement for γ-secretase. We have described several mutations at the S3 cleavage site that affect the activity of Notch proteins from mildly (V>G, V>K) to severely (GCGV>LLFF); of these the V>G substitution was predicted to enhance cleavage as it inserted a helix-breaking residue (Gly). Introduction of strong helix-promoting residues (V>A) do not seem to affect the activity of the Notch protein in mammalian cells. The V>L substitution impacts proteolysis but appears more severe due to rapid degradation of NICD due to an N-terminal Leucine (71). To investigate the possibility that we suppressed proteolysis by creating a helix-stabilizing pair (V1744L with V1740), we tested the reciprocal substitution (V1740L with V1744) but no impact on proteolysis was observed. Helix-breaking residues are found in the vicinity of the conserved Valine in vertebrate and Drosophila Notch and the C.elegans Notch receptor, Lin-12. Substitution of Glycine (GCGV) with the helix-promoting residue Alanine (ACAV) was predicted to reduce cleavage; the cleavage site may have shifted to another position (which remains to be determined) but no marked reduction in signaling capacity was observed. Thus, it appears that γ-secretase substrates differ from other I-CliP substrates in their tolerance to helix-forming residues and Notch1 differs from other γ-secretase substrates in its intolerance of helix-breaking residues at the P1′ site.
In conclusion, we are still left with the conundrum of monomeric substrate proteins (N1FL, N1LNR) that are bound to γ-secretase but are not proteolytic substrates (72,73) until after S2 cleavage (29,60). We present evidence that Notch signaling is not simply regulated by oligomeric changes in the extracellular and/or transmembrane domain, nor is the length of the extracellular domain regulating cleavage. While we cannot rule out that ligand binding may alter the oligomerization state of Notch, we propose that Notch cleavage (and perhaps all γ-secretase substrates) is more likely to be regulated by structural changes induced upon ligand binding to Notch receptors. We propose that intact proteins cannot be transferred to the active site of γ-secretase where the helical TMD is relaxed. Activating mutations in the conserved region carboxyterminal to the LNR repeats could act by enforcing a conformation that is competent to enter the active site. Conversely, helix-breaking mutations in the Notch TMD do not enhance cleavage because they may lock the transmembrane domain in the wrong conformation, preventing hydrolysis. Given that γ-secretase is the only known multiprotein I-CLiP that cleaves substrates multiple times within their TMDs, the mechanism controlling cleavage of its substrates may be unique among I-CLiPs. Similar experiments will have to be conducted with other I-CLiP substrates.         immunoprecipitates. For unknown reasons HA immunoprecipitation consistently retrieved fewer N-N dimers than with anti-Myc but did not affect the association with JAG1. Ptc1 seems to have a higher affinity for LDLR than LDLR has for Notch1.