Presenilin-dependent “γ-Secretase” Processing of Deleted in Colorectal Cancer (DCC)

The presenilin-γ-secretase complex plays a critical role in mediating intramembranous proteolysis of several type I membrane proteins, including β-amyloid precursor protein (APP) and Notch. We now show that deleted in colorectal cancer (DCC) is subject to proteolysis within the ectodomain segment both in cultured cells and in vivo and that the residual membrane-tethered DCC “stub” is subsequently processed by γ-secretase to generate a derivative termed DCC-intracellular domain (ICD). The production of DCC-ICD is inhibited by selective γ-secretase inhibitors, and by the expression of the dominant negative PS1 D385A variant. Moreover, the membrane-tethered DCC “stubs” accumulate to high levels in PS1-deficient embryos. We also demonstrate that expression of a DCC-Gal4 chimera is capable of activating transcription in a luciferase-based reporter assay and this activity is dependent on γ-secretase activity. Our findings offer the proposal that DCC performs dual roles both as a cell surface receptor that modulates intracellular signaling pathways and as a transcriptional coactivator that relies on γ-secretase-dependent production and nuclear translocation of the cytoplasmic domain.

The presenilin-␥-secretase complex plays a critical role in mediating intramembranous proteolysis of several type I membrane proteins, including ␤-amyloid precursor protein (APP) and Notch. We now show that deleted in colorectal cancer (DCC) is subject to proteolysis within the ectodomain segment both in cultured cells and in vivo and that the residual membrane-tethered DCC "stub" is subsequently processed by ␥-secretase to generate a derivative termed DCC-intracellular domain (ICD). The production of DCC-ICD is inhibited by selective ␥-secretase inhibitors, and by the expression of the dominant negative PS1 D385A variant. Moreover, the membrane-tethered DCC "stubs" accumulate to high levels in PS1-deficient embryos. We also demonstrate that expression of a DCC-Gal4 chimera is capable of activating transcription in a luciferase-based reporter assay and this activity is dependent on ␥-secretase activity. Our findings offer the proposal that DCC performs dual roles both as a cell surface receptor that modulates intracellular signaling pathways and as a transcriptional coactivator that relies on ␥-secretase-dependent production and nuclear translocation of the cytoplasmic domain.
Presenilin 1 and 2 (PS1 and PS2) 1 are polytopic membrane proteins that are mutated in the vast majority of pedigrees with early-onset familial Alzheimer's disease (1). PS plays an essential role in intramembranous, "␥-secretase" processing of several type I membrane proteins, including the ␤-amyloid precursor protein (APP) (2,3), Notch1 (4,5), ErbB-4 (6), N-and E-cadherins (7), low density lipoprotein receptor-related protein (8), CD44 (9), and nectin-1␣ (10). It is now well established that ␥-secretase processing of transmembrane proteins is preceded by proteolysis near the interface of the ectodomain and transmembrane segments that generates a soluble ectodomain that is "shed" and one, or more, membrane-tethered derivatives. In the case of APP, a set of membrane-tethered APP derivatives, termed APP-CTFs, are the substrates for ␥-secretase, and intramembranous processing leads to the production of A␤ peptides. Similarly, ␥-secretase is responsible for processing of a membrane-tethered Notch1 derivative, termed S2/ NEXT, resulting in the generation of a Notch derivative, termed S3/NICD that translocates to the nucleus and activates transcription of target genes (4,11).
Intrigued by the observation that deleted in colorectal cancer (DCC) is a substrate for cleavage by a metalloprotease that leads to "shedding" of the ectodomain segment (12), we asked whether the resulting membrane-tethered "stub" might be processed by ␥-secretase. In the present report, we show that in cultured mammalian cells, the truncated DCC stub is subject to a PS-dependent processing event that is inhibited by a highly potent ␥-secretase inhibitor. Moreover, the DCC stub accumulates to high levels in spinal cords of PS1-deficient mice, leading us to postulate that this species is also a substrate for ␥-secretase processing in vivo. Finally, we demonstrate that the ␥-secretase-generated intracellular domain of DCC, termed DCC-ICD, can translocate to the nucleus and that this fragment has an intrinsic transcriptional activation domain.

EXPERIMENTAL PROCEDURES
Construct-pSecTagA/DCC encodes c-myc-epitope tagged rat fulllength DCC expression plasmid (Fig. 1B). To generate pCMVneo/DCC-GIC, pCMVneo/DCC (13) was partially digested with BstXI and the fragment encoding Gal4 DNA-binding domain was inserted into the BstXI site located between the transmembrane region and the conserved P1 domain of DCC intracellular domain. The V1117A and V1119A mutant DCC constructs were generated by overlapping PCRbased mutagenesis (14) using wild type pSecTaqA/DCC as a template. The sequences of the constructs were verified by DNA sequencing.
Cell Culture and Transfection-HEK293 or N2a cells were transfected using LipofectAMINE Plus (Invitrogen). 24 h after transfection, the cells were treated with vehicle, 50 nM compound E, 10 M lactacystin or both compound E and lactacystin for additional 24 h. Detergent lysates were prepared as described previously (14) and analyzed by immunoblotting with the 9E10 antibody. For luciferase assays, cells were transfected with 0.5 g of expression plasmid, 167 ng of pG5E1Bluc (15), and 17 ng of pRL-SV40 Renilla luciferase reporter gene (Promega). Transcriptional activities were determined 24 h after transfection using dual luciferase reporter assay system (Promega).

PS-dependent
Cleavage of DCC Protein-DCC, first isolated as a gene deleted in colorectal cancer (13), encodes a type I integral membrane protein of 1,445 amino acids that is a receptor for the axonal chemoattractant, termed netrin-1 (16). DCC is subject to a metalloprotease-dependent ectodomain shedding event (12). Like DCC, Notch1 is also subject to ectodomain shedding (17,18), and in this case, the residual membrane-tethered stub, termed S2/NEXT, is subject to an intramembranous, presenilin-dependent, ␥-secretase processing reaction that leads to the production of cytosolic derivatives termed S3/NICD (11). A valine residue at the P1Ј site is essential for ␥-secretase-dependent processing of Notch (11), and a valine residue occurs at a similar position within the trans-membrane domains of other ␥-secretase substrates (Fig. 1A). Intrigued by the finding that DCC is subject to proteolysis in the ectodomain, coupled with the presence of several valine residues within the transmembrane domain proximal to the cytoplasmic segment, we asked whether DCC might also serve as a ␥-secretase substrate.
We transiently expressed a myc epitope-tagged full-length rat DCC (pSecTag-DCC; Fig. 1B) into HEK293 cells, then treated cells with 0.5% Me 2 SO (vehicle), or compound E (19), a highly potent inhibitor of ␥-secretase. Treatment of the cells with 50 nM compound E resulted in the accumulation of ϳ56 -58 kDa, myc antibody-reactive fragments that we refer to as "␣" fragments ( Fig. 2A, lane 3). As the DCC cytoplasmic tail is degraded by the proteasome (20), we treated cells with the proteasome inhibitor, lactacystin. Lactacystin treatment of DCC-expressing cells resulted in the accumulation of closely spaced fragments with an apparent molecular weight of between 48 -52 kDa ( Fig. 2A, lane 4). By virtue of the fact that these fragments accumulate to high levels with lactacystin treatment and are not present in cells treated with the ␥-secretase inhibitor, we refer to these ␥-secretase-generated, DCC derivatives as DCC-ICD fragments. Treatment with both compound E and lactacystin resulted in the accumulation of ␣ but not DCC-ICD fragments, as expected ( Fig. 2A, lane 5). Lactacystin treatment also resulted in the accumulation of an ϳ20-kDa myc antibody-reactive fragment ( Fig. 2A, lanes 4 and 5), the production of which was fully abolished upon treatment of cells with the broad spectrum caspase inhibitor, z-Val-Ala-Aspfluoromethylketone (data not shown).
To determine whether ␥-secretase cleavage of DCC occurs in an in vivo setting, we performed Western blot analysis of detergent extracts prepared from the spinal cords of E14.5 wild type and PS1-deficient (22) mouse embryos. We observed a prominent species of ϳ180 kDa, representing full-length DCC in the spinal cords from wild type embryos (Fig. 2B, lanes 1 and  2). In contrast, we observed ϳ54 -56-kDa fragments that accumulated in extracts from PS1-deficient mice (Fig. 2B, lanes 3  and 4) that are reminiscent of the ␣ fragments that accumulate in cultured cells treated with the ␥-secretase inhibitor ( Fig. 2A,  lanes 3 and 5). Although we failed to detect a band correspond-ing to DCC-ICD in extracts from wild type animals, we surmise that this derivative is generated, but rapidly degraded. In any event, we interpret these collective results to suggest that in the developing spinal cord, DCC is subject to ␥-secretase cleavage.
To further confirm that the generation of the ␥-secretasegenerated DCC-ICD fragments is PS-dependent, we transiently expressed DCC in mouse neuroblastoma N2a cells stably expressing either human wild type PS1 (WT.7) or the dominant negative D385A mutant PS1 (D385A. 16;Ref. 14). Similar to the results obtained in HEK293 cells, treatment of N2a WT.7 cells with compound E (Fig. 2C, lane 3) and lactacystin (Fig. 2C, lane 4) led to the accumulation of ␣ and DCC-ICD fragments, respectively. In contrast, in lysates of N2a D385A.16 cells that transiently express DCC, we observed robust accumulation of ␣ fragments (Fig. 2C, lane 7), and neither lactacystin, nor compound E, had an appreciable effect on stabilization of this fragment (Fig. 2C, lanes 8 -10). Taken together, these results indicate that following ectodomain shedding, the membrane-tethered DCC ␣ derivative is subject to PS-dependent ␥-secretase cleavage.
In the case of Notch, a mutation of the P1Ј valine residue results in the marked reduction of transcriptional activation by Notch (11,21). To determine whether the conserved valine residues in the transmembrane domain of DCC are required for ␥-secretase processing, we transiently expressed DCC mutants with amino acid substitutions either at valine 1117 or valine 1119 (marked with asterisk in Fig. 1A) in HEK293 cells and examined the accumulation of DCC derivatives in the presence of lactacystin and compound E. Remarkably, both mutants appeared to be processed in a ␥-secretase-dependent manner similar to wild type DCC (Fig. 2D, compare lanes 3, 7, and 11  with lanes 4, 8, and 12). These results reveal that valine residues within the DCC transmembrane domain that are immediately proximal to the cytoplasmic segment are dispensable for ␥-secretase processing of the DCC ␣ fragment. The sequence, or structural, determinants required for ␥-secretase cleavage of DCC remain to be determined.
The DCC Intracellular Domain Has Intrinsic Transcriptional Activity-It is well established that ␥-secretase cleavage of the membrane-tethered Notch S2/NEXT fragment results in the generation of a soluble Notch S3/NICD fragment that translocates to the nucleus where it is a transcriptional coactivator with RBP-J (11). Similarly, the APP AICD derivative forms a transcriptionally competent ternary complex that includes Fe65 and Tip60 (15,23). To examine whether DCC-ICD has intrinsic transcriptional activity, we transiently expressed a chimera in which the Gal4 DNA-binding domain was inserted between the DCC transmembrane and cytoplasmic domains (pCMVneoDCC-GIC; Fig. 1B) along with a luciferase-based reporter gene into HEK293 cells. We obtained considerable transcriptional activation (Fig. 3A) that was largely blocked by treatment of cells with compound E or another unrelated, but potent, ␥-secretase inhibitor (compound 1) (24). The DCC-GIC "ICD" fragments are readily visualized either in the absence, or presence, of lactacystin (Fig. 3B, lanes 1 and 2), and treatment with compound E results in accumulation of the DCC-GIC "␣ fragment" (Fig. 3B, lane 3), results similar to those obtained with full-length DCC. Thus, the ␥-secretase-generated DCC cytoplasmic domain contains elements sufficient to activate transcription of heterologous promoters.
In an attempt to determine whether DCC-ICD could be visualized in the nucleus, we expressed a DCC-EGFP chimera in HEK293 cells. We only observed the accumulation of this chimera on the plasma membrane, but not in the nucleus (data not shown). These results suggest that like Notch S3/NICD (11), the transcriptional stimulatory activity of DCC is likely mediated by limiting levels of DCC-ICD in the nucleus. DISCUSSION Presenilins play an essential role in the intramembranous "␥-secretase" processing of membrane-tethered stubs of several type I membrane proteins, including the APP/APLP (2, 3), Notch receptors (4, 5), ErbB-4 (6), E-cadherin (7), low density lipoprotein receptor-related protein (8), CD44 (9), and nectin-1␣ (10). The finding that the netrin 1 receptor, DCC, is subject to a metalloprotease-mediated ectodomain shedding event (12), and the presence of several valine residues in the DCC transmembrane domain, led us to hypothesize that this molecule might also serve as substrate for ␥-secretase. In the present report, we confirm this prediction and offer several insights relevant to the molecular apparatus responsible for intramembranous processing of DCC and the potential functional significance of this processing event.
First, we demonstrate that in cultured mammalian cells, the production of the ␥-secretase-generated cytoplasmic derivative of DCC, termed DCC-ICD, is inhibited by a highly potent inhibitor of ␥-secretase activity, or by the expression of the dominant-negative D385A PS1 mutant, but not by mutations at the conserved valine residues in the transmembrane domain of DCC. The PS-dependent ␥-secretase generated DCC-ICD fragments that are highly labile and only detected by the addition of the proteosomal inhibitor, lactacystin. The identity of the amino termini of these fragments is not known. However, it seems somewhat unlikely that the fragments are all generated by intramembranous cleavage events as the difference in molecular weights of the fragments are too broad. Thus, we sus-   1, 5, and 9), 50 nM compound E (lanes 2, 6, and 10), 10 M lactacystin (lanes 3, 7, and 11), or both compound E and lactacystin (lanes 4, 8, and 12). Detergent lysates were analyzed by immunoblotting with the 9E10 antibody. FL, full-length DCC; ␣, membranetethered ␣ fragment; ␥, DCC-ICD fragments generated by ␥-secretase; caspase, COOH-terminal half of DCC-ICD fragment generated by caspase. ␣ and DCC-ICD fragments are also marked by asterisks and circles, respectively. pect that differing levels of posttranslational modifications such as phosphorylation or ubiquitination contribute to the differing mobilities of the fragments.
Second, DCC appears to be a bona fide substrate for an ectodomain shedding activity in vivo by virtue of the fact that a membrane tethered ␣ derivative accumulates to high levels in spinal cords of mice with genetic ablations of PS1. Because we have failed to detect DCC-ICD in spinal cords of wild type embryos, we can only hypothesize that the ␣ fragment is a substrate for ␥-secretase. Indeed, the absence of detectable DCC-ICD in wild type embryos and accumulation of DCC ␣ derivatives in spinal cords from PS1-deficient embryos is very similar to our earlier findings that APP-derived AICD is undetectable in brains of wild type embryos, while ␣and ␤-secretase-generated APP-CTFs accumulate in brains of PS1deficient embryos (3). Interestingly, our preliminary immunocytochemical studies have revealed that commissural axon projections in the developing spinal cord are highly disorganized. It is not clear whether the disorganization of commissural axons is a reflection of defects in ␥-secretase processing of DCC, alone, or deficits in intramembranous processing of other axonal molecules involved in migration and cell adhesion, including nectin-1␣ and N-cadherins.
Third, we examined the hypothesis that the ␥-secretasegenerated DCC-ICD fragments are translocated to the nucleus and activate transcription of a reporter gene. We show that a chimera in which the Gal4 DNA-binding domain is inserted between the DCC transmembrane and cytoplasmic domains is competent to activate transcription of a promoter containing Gal4 binding sites and that the transcriptional activation by DCC was largely blocked by ␥-secretase inhibitors. These results indicate that the intracellular domain of DCC acts as a transcriptional activator in mammalian cells and that ␥-secretase plays an essential role in the regulation of DCC-dependent transcription. At present, the identity of the factors that regulate translocation and transcriptional activity of DCC-ICD is not known. Future efforts to characterize these factors and the downstream target genes may offer new insights into the signaling pathways responsible for mediating DCC-regulated axonal guidance during development and cell cycle regulation in neoplasias.