γ-Protocadherins, Presenilin-mediated Release of C-terminal Fragment Promotes Locus Expression*

γ-Protocadherins (γ-pcdhs) are type I membrane-spanning glycoproteins, widely expressed in the mammal and required for survival. These cell adhesion molecules are expressed from a complex locus comprising 22 functional variable exons arranged in tandem, each encoding extracellular, transmembrane and intracellular sequence, and three exons for an invariant C-terminal domain (γ-ICD). However, the signaling mechanisms that lie downstream of γ-pcdhs have not been elucidated. Here we report that γ-pcdhs are subject to presenilin-dependent intramembrane cleavage (PS-IP), accompanied by shedding of the extracellular domain. The cleaved intracellular domain (γ-ICD) translocates to the cell nucleus and was detected in subsets of cortical neurons. Notably, gene-targeted mice lacking functional γ-ICD sequence showed severely reduced γ-pcdh mRNA levels and neonatal lethality. Most importantly, inhibition of γ-secretase decreased γ-pcdh locus expression. Luciferase reporter assays demonstrated that γ-pcdh promoter activity is increased by γ-ICD. These results reveal an intracellular signaling mechanism for γ-pcdhs and identify a novel vital target for the γ-secretase complex.

␥-Protocadherins (␥-pcdhs) are type I membranespanning glycoproteins, widely expressed in the mammal and required for survival. These cell adhesion molecules are expressed from a complex locus comprising 22 functional variable exons arranged in tandem, each encoding extracellular, transmembrane and intracellular sequence, and three exons for an invariant Cterminal domain (␥-ICD). However, the signaling mechanisms that lie downstream of ␥-pcdhs have not been elucidated. Here we report that ␥-pcdhs are subject to presenilin-dependent intramembrane cleavage (PS-IP), accompanied by shedding of the extracellular domain. The cleaved intracellular domain (␥-ICD) translocates to the cell nucleus and was detected in subsets of cortical neurons. Notably, gene-targeted mice lacking functional ␥-ICD sequence showed severely reduced ␥-pcdh mRNA levels and neonatal lethality. Most importantly, inhibition of ␥-secretase decreased ␥-pcdh locus expression. Luciferase reporter assays demonstrated that ␥-pcdh promoter activity is increased by ␥-ICD. These results reveal an intracellular signaling mechanism for ␥-pcdhs and identify a novel vital target for the ␥-secretase complex.
Cadherins represent a large superfamily of transmembrane glycoproteins, sharing common structural features, but exhibiting differential adhesive binding specificities (1). Characterized best are the so-called classic cadherins, including E-, R-, N-, and P-cadherin, and the remainder forms subgroupings collectively referred to as protocadherins (pcdhs). 1 Classic cadherins are characterized by an ectodomain consisting of 5 extracellular cadherin-like (EC) repeats, followed by a single transmembrane domain and a highly conserved C-terminal cytoplasmic region. Classic cadherins engage mostly in homophilic adhesive binding triggered by Ca 2ϩ ions intercalating between the EC domains to produce a rigidified, rod-like ectodomain (2). Intracellularly, classic cadherins associate with different proteins, including catenins, that regulate their adhesive properties or mediate downstream signaling (3), which is important for establishing brain structure and connectivity including synaptic structure, function and plasticity (4 -9). Besides their well documented signaling function via catenins, E-and N-cadherin have been shown to undergo presenilin dependent intramembrane proteolysis (PS-IP), which consists of matrix protease-mediated cleavage of the ectodomain and the characteristic release of a soluble cytoplasmic domain by the activity of the ␥-secretase complex and its catalytic constituents presenilin-1/2 (10,11). The released cytoplasmic cadherin domains can assume signaling function in the cytoplasm and/or nucleus, in analogy to the cytoplasmic domains of prominent targets of ␥-secretase, Notch, APP, ErbB-4, and SREBP-1 (12)(13)(14)(15).
In comparison to classic cadherins, pcdhs differ in the number of EC domains and have divergent intracellular domains. Recently, Ͼ50 novel type-I transmembrane pcdhs have been described (16,17), of which most, if not all, are expressed in brain. They are thought to participate in specific synaptic connections based on their structural similarities to cell adhesion molecules and their synaptic localization (18 -21). As a characteristic feature these pcdhs harbor six EC domains related to those of classic cadherins, but contain different cytoplasmic domains. In human and mouse they are arranged in three physically linked clusters, termed ␣, ␤, and ␥, located on mouse chromosome 18 (human chromosome 5). Each cluster contains multiple tandemly arranged functional variable exons (14 for ␣, 22 each for ␤ and ␥), each encoding the extracellular sequence, the transmembrane region and part of the intracellular domain. Transcription of each ␣and ␥-pcdh variable exon is initiated at sequences immediately upstream of the translational start site (22,23). Notably, each variable ␥-pcdh exon contains its own promoter/enhancer, including a conserved sequence element (CSE) essential for proper transcription, ϳ200-bp upstream of the translational start site (22). In addition to the variable exon array, the ␣and ␥-pcdh, but not the ␤-pcdh, locus harbor three exons 3Ј of the variable exon cluster, which together encode invariant cytoplasmic domains (␣-ICD, ␥-ICD) (16,17). Thus, cis-splicing of the cluster transcripts produces ␣and ␥-pcdh isoforms with distinct adhesive properties but identical cluster specific intracellular domains.
The ␥-pcdh locus represents the best studied of the three subclusters of pcdhs. Initial investigations revealed that the ␥-pcdh genes are abundantly transcribed in neurons. Importantly, an intact ␥-pcdh locus is required, as demonstrated by neonatal death of mice from locus deletion (19). Global aspects of neurogenesis, including neuronal migration, axonal outgrowth and synapse formation appeared unaffected, but specific neuronal populations, especially spinal interneurons, died at late stages of development, indicating a critical requirement for ␥-pcdh-mediated signaling for their survival. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
As yet, the mode and mechanism of ␥-pcdh-mediated signaling are unknown. An intriguing aspect is certainly provided by the ␥-ICD, which as integral part of all 22 functionally expressed ␥-pcdhs is likely to interact and provide cross-talk with cellular signaling pathways. However, no proteins interacting with the ␥-ICD have been described. Here we provide evidence that the ␥-ICD can be released by PS-IP from ␥-pcdhs inserted in the plasma membrane and translocates to the nucleus where the ␥-pcdh locus appears to be a potential target for ␥-ICDmediated transcriptional enhancement by as yet unknown mechanisms. This autoregulatory aspect of ␥-pcdh expression may contribute to unexpectedly low ␥-pcdh levels in mice heterozygous for a ␥-pcdh locus lacking ␥-ICD exon sequence.
In Situ Hybridization-Embryos or whole brains from E14/P42 C57-BL/6 mice were frozen on dry ice. Saggital or coronal sections (20 m) prepared by cryostat were mounted on glass slides. In situ hybridization was performed as described (29). The anti-␥-C2 oligonucleotide, 5Ј-GT-CAAACTGGTTGTTGGGCCAGGTGCCGGTGTCATCGCCATTTTG-3Ј, was 3Ј-labeled with [ 35 S]dATP and terminal deoxynucleotidyl transferase (Roche Applied Science). After hybridization and washing in 0.5ϫ SSC at 60°C for 5 min, glass slides were exposed to KODAK BioMax MR-1 film for ϳ48 h.
Immunohistochemistry-Fresh brains from newborn mice (n ϭ 3 for wild type and knock out) were immersion-fixed in 4% paraformaldehyde (PFA). Adult mice were transcardially perfused with 4% PFA. Brain sections (50 -100 m) were preincubated in 5% normal goat serum (NGS; Sigma) for 2 h at room temperature and transferred to the polyclonal anti-␥-ICD antibody solution containing 0.5% Triton X-100 and 1% NGS in phosphate-buffered saline (PBS). For controls, the primary antibody was preabsorbed with 0.4 g of GST-␥-ICD protein/l antibody for 2 h at 4°C and centrifuged for 10 min at 13 krpm, or was omitted from the incubations. Sections were incubated in primary antibody solution for 48 h at 4°C, washed 10 min in PBS, and incubated in biotinylated secondary antibody (goat anti-rabbit IgG 1:500; (Vector)) for 2 h at room temperature. After washing in PBS (10 min), the sections were incubated in streptavidin-biotin peroxidase complex (ABC standard kit, Vector) for 1-2 h at room temperature. Immunolabeling was revealed by the glucose oxidase-diaminobenzidine (GOD-DAB) method. Intensification of the GOD-DAB reaction product was by osmium intensification (30,31). Vibratome sections were washed in PBS, mounted on Superfrost (Menzel) glass slides, dried, dehydrated, and coverslipped from Xylene.
Western blots, Immunofluorescence, and Co-immunoprecipitations-For Western blot (WB) analysis cultured cells or mouse brains were solubilized in 25 mM HEPES, pH 7.5, 150 mM NaCl, protease inhibitor mixture (Roche Applied Science), containing 1% Triton X-100. SDS-PAGE and blotting were performed using standard conditions. ␥-ICD rabbit polyclonal antiserum was used in 1:200 (fixed cells) or 1:500 (WB) dilutions, all other antisera were diluted according to manufacturer's protocols. Immunofluorescence was performed using standard protocols. To visualize nuclei, cells were treated with 4Ј6-diamidino-2phenylindole (DAPI, 1:5000; Sigma) for 5 min at room temperature. Mounted sections were examined in a Zeiss fluorescence microscope. Quantification of Western blots was performed with ImageJ software. For co-immunoprecipitation COS-1 cells were cotransfected with vectors for Flag ␥-pcdh-B1 and EYFP ␥-pcdh-A1 or GFP ␥-pcdh-B1 or EYFP ␥pcdh-A1⌬CT or myc ␥-ICD. All subsequent operations were at 4°C. Cell lysates were precleared with 25 l of immobilized protein A (Immuno-Pure®ϩ, Pierce) for 30 min, and ϳ400 l of (protein concentration ϳ1 g/l) were incubated with ϳ5 g of antibody (AFP or Myc) overnight. Precipitates were recovered after 4 h of incubation with 25 l of beadimmobilized protein A. Beads were washed 4ϫ with 25 mM HEPES, pH 7.5, 150 mM NaCl, 0.5% Triton X-100, and bound material was eluted at 99°C in 20 l of 3ϫ sample buffer.
Real Time PCR-An ABIPRISM 7000 Sequence Detection System (PE Applied Biosystems, Foster City, CA) was used for cDNA quantification. Total RNA was isolated from HEK293 cells cotransfected with vectors for GFP and myc ␥-ICD or myc ␣-ICD or from pharmacologically treated SH-SY5Y cells. RNAs were DNase I-treated and reverse-transcribed. About 1-2 g of the resulting cDNA was analyzed in a reaction volume of 25 l (on 96-well plates), using the TaqMan PCR Core reagent kit (PE Applied Biosystems). cDNA of each cell culture dish was analyzed in triplicate. Primers and probes for real time PCR were obtained from PE Applied Biosystems, designed using the manufacturer's programs. The primer and probe sequences located in the ␥-pcdh 3Ј-untranslated region were: Probe-UTR3-VIC, 5Ј-TTGCTACCAAGCC-TCTT-3Ј; RP-UTR5, 5Ј-TCCCCAGGGCGTTGG-3Ј; FP-UTR3, 5Ј-GCT-GGATTTAGGGAGGGCA-3Ј. GFP probe and primers: 6-FAM, 5Ј-CTA-TAACTCCCACAATGTG-3Ј; FP-GFP, 5Ј-TCGGCCACAAGTTGGAATA-CA-3Ј; RP-GFP, 5Ј-TGCTTGTCGGCCATGATGT-3Ј, (for the pTR-UF2 plasmid, described in Ref. 32), GAPDH primers (assay on demand) and probe were designed by the company and used according to their protocols (www.appliedbiosystems.com). The detected amount of ␥-pcdh transcripts was subsequently normalized for that of GAPDH transcripts and for transfection efficiency, assessed by the amount of GFP transcripts.
Cell Lines, Transfections, and Treatments-COS-1, HEK293, SH-SY5Y cells and dko fibroblasts (33,34) were grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum and 50 mg/l penicillin/ streptomycin in 5% CO 2 at 37°C. Cells were transfected with recombinant plasmids by the Ca 2ϩ -phosphate method, supplied with fresh medium 20 h after transfection and grown for another 24 h before harvesting. For treatment with monensin (final concentration, 83 M) and brefeldin A (0.3 mM final) (both from Sigma), the agents were given with the fresh medium supply for 24 h before cell harvesting. L-685,458 (7 M) and DAPT (10 M) (CalBiochem) were added 12 h after the change to fresh medium, and cells were harvested 12 h later. TAPI (40 M) (CalBiochem) treatment was similar, except exposure to cells was 10 h.
Transactivation Assays-HEK293 cells grown in 6-well plates with ϳ5 ϫ 10 5 cells per well were cotransfected with reporter constructs, and constructs for ␤-galactosidase (Promega) and one of several activators indicated in Fig. 6. Transactivation of reporter constructs was measured in lysates from ϳ8 ϫ 10 4 cotransfected HEK293 cells according to the manufacturer's protocol (firefly luciferase system, Promega) 30 h after transfection in a luminometer (Macherey and Nagel). To evaluate individual transfection efficiencies for subsequent normalization, X-Gal-positive cells were counted in parallel wells. Luciferase activities were determined in a Beckman luminometer. Cotransfections were tested in triplicate in at least two independent experiments. The transactivation potential of myc ␥-ICD was determined by comparing the amount of relative light units in parallel transfections with ␥-pcdh-A1⌬CT or pRK5-myc ␣-ICD as control. For promoter control we used p0.4XPluci, containing the core promoter of the NR2C gene (28).
Splicing from variable to constant region exons was determined by RT-PCR with sense oligonucleotide primers generic for variable region exons A1-12, or specific sense primers for variable exons A8 (not shown) and B4, in combination with an antisense primer within ␥-C3. The amplicons, resolved on agarose gels (Fig. 1C) and checked by DNA sequencing (not shown), revealed that the pgk-neo insertion did not cause skipping of ␥-C1 or ␥-C2, and that loss of ␥-C1 resulted in splicing from variable exons to ␥-C2, with no detectable band corresponding to a splice directly to ␥-C3. Moreover, the skewed ratio of the two RT-PCR fragments, respectively containing or lacking ␥-C1 exon sequence, from RNA of heterozygous ϩ/⌬C1 mice (Fig. 1C) documented that spliced transcripts from the ⌬C1 allele were in much lower abundance than those from the wild-type ␥-pcdh allele.
Quantification by Northern analysis from poly(A) ϩ RNA with four probes for variable and conserved exon sequences showed that the amount of the ϳ4.4-kb ␥-pcdh mRNA was in ϩ/neo mice little over, and in ϩ/⌬C1 significantly below, the 50% expected from the intact wild-type allele present in these genotypes (Fig. 2, A and B). This indicated that both modified alleles (neo, ⌬C1) were severely impaired in mRNA production. Moreover, the analysis revealed that the wild-type ␥-pcdh allele in ϩ/⌬C1 mice produced subnormal mRNA amounts. In neo/neo mice, ␥-pcdh mRNA was reduced to ϳ20% of wild-type levels; the intronic pgk-neo insertion may have resulted in reduced splicing efficacy of the allelic neo transcripts. In ⌬C1/ ⌬C1 mice, ␥-pcdh mRNA lacking ␥-C1 sequence was Ͻ10% of wild-type levels.
We also evaluated ␥-pcdh protein levels in the different genotypes. The ␥-pcdh specific ϳ114 kDa protein band reacting on Western blots with our antibody showed relative to wild type an intensity of ϳ60% in ϩ/neo mice, and ϳ10% in neo/neo mice (Fig. 2, C and D), in good correspondence to the respective mRNA levels. Thus, low ␥-pcdh levels might underlie the perinatal death of neo/neo mice. Most notably, the Western analysis of brain lysates from ϩ/⌬C1 mice revealed steady-state ␥-pcdh levels of ϳ25% of wild type, even though twice as much should be generated by the remaining functional wild-type allele alone (Fig. 2, C and D). Apparently, such low ␥-pcdh levels suffice for the survival. We finally evaluated possible effects of the modified ␥-pcdh alleles on mRNA levels of the physically proximal, paralogous ␣-pcdh locus, but found none (data not shown).
N-and C-terminal ␥-pcdh Portions Localize Differentially in Cells-In situ hybridization with an oligonucleotide probe to ␥-C2 on coronal brain sections of adult wild-type mice and on saggital sections of embryonic day 14 embryos documented prominent neuronal ␥-pcdh transcript distribution in CNS (Fig.  3A). Staining of forebrain sections of adult mice from different brain regions with our anti-␥-ICD antibody revealed that subsets of neurons showed strong immunoreactivity, with most neurons having stained nuclei (Fig. 3B, inset). The nuclear staining was not apparent in sections from newborn mice (Fig.  3C, upper panel). The specificity of staining was demonstrated by competition with excess ␥-ICD, omitting of the primary antibody (Fig. 3B, right panel, inset), and in ⌬C1/⌬C1 newborns lacking ␥-ICD (Fig. 3C, lower panel).
To trace the cellular distribution of the ␥-ICD in a system amenable to pharmacological intervention, we constructed a vector for N-terminal fusion of EYFP to ␥-pcdh-A1 ( EYFP ␥-pcdh-A1) and expressed the fusion protein transiently in COS-1 cells. Anti-GFP antibody staining on nonpermeabilized cells showed fluorescent protein clusters on the cell surface, clearly indicating that ectopically expressed EYFP ␥-pcdh-A1 was efficiently inserted into the plasma membrane (Fig. 3D, panel 1). Direct EYFP-mediated fluorescence was exclusively detected in the cytoplasm, with little or no fluorescent signal visible in the nucleus (Fig. 3D, panel 2), showing that the N-terminal portion of EYFP ␥-pcdh-A1 is absent in the nuclear compartment. However, antibody staining against the ␥-ICD, after permeabilizing transfected COS-1 cells and use of a fluorescein isothiocyanatelabeled secondary antibody, revealed strong fluorescence in both, nucleus and cytoplasm (Fig. 3D, panel 3), as further confirmed by counterstaining with DAPI (not shown). This concurs with the endogenous ␥-ICD signal in mouse brain sections.
Detecting N-and C-terminal portions of ␥-pcdhs in different subcellular compartments could reflect proteolytic processing of ␥-pcdhs and nuclear translocation of ␥-ICD, which is predicted to contain a bipartite nuclear localization sequence (Fig.  1B). To test if ␥-ICD gets into the nucleus, we expressed it as Myc-fusion ( myc ␥-ICD) ectopically in COS-1 cells. Indeed, recruitment of myc ␥-ICD into the nucleus was evinced by the strong nuclear signal generated by both anti-Myc and anti-␥-ICD antibodies (Fig. 3D, panels 4 and 5). Moreover, staining of untransfected cells, photographed with longer exposure time, documented nuclear localization also of the endogenous ␥-ICD (Fig. 3D, panel 6).
␥-pcdhs Undergo Intramembrane Proteolysis by ␥-Secretase-A likely candidate for proteolytic processing of ␥-pcdhs leading to intracellular release of ␥-ICD sequence is the ␥-secretase complex, which has recently been implicated in the processing of a growing number of membrane proteins (35)(36)(37). To test if ␥-secretase is involved in ␥-pcdh processing, we treated COS-1 cells transiently expressing EYFP ␥-pcdh-A1 with specific ␥-secretase inhibitors (L-685,458 and DAPT) and monitored the accumulation of a cytoplasmic membrane-bound cleavage intermediate, which should appear following ectodomain shedding and ␥-secretase inhibition (Fig. 4A). Both substances efficiently inhibited cleavage of transiently transfected APP2Z (38), a well-described target of the ␥-secretase complex ( Supplementary Fig. 1). In case of transiently expressed EYFP ␥- Filled ovals beneath open boxes indicate (from left to right) variable exons ␥-A2, ␥-A3, ␥-B1, and ␥-C4, the promoters of which were tested for transactivation. Open ovals in neo allele indicate the extent of the targeting construct (XhoI-BstBI) for murine ES cells. Positions of probes (solid lines below ϩ allele) and restriction sites used for Southern blot analysis are indicated. Arrows (above ϩ allele) mark primers used for PCR-based genotyping. B, amino acid sequence of ␥-ICD encoded by exons ␥-C1, ␥-C2 (gray shading), and ␥-C3. The short frameshifted sequence IPKW resulting from splicing of allelic ␥-⌬C1 transcripts from ␥-VAR to ␥-C2 is shown above ␥-C2. The putative bipartite nuclear localization signal (NLS) is underlined. C, RT-PCR products for the different genotypes with the generic sense primer for ␥-VAR exons A1-12 (left) or the specific sense primer for exon B4 (right) in combination with the antisense primer C3 on ␥-C3. Note the marked intensity difference in the two DNA fragments in the ϩ/⌬C1 lanes, indicating a large imbalance in mRNA levels from the two alleles. M, molecular weight marker; Ϫ, no cDNA. pcdh-A1, Western analysis with our antibody against the ␥-ICD showed significant accumulation of a ϳ26 kDa (␥-26 kDa) band by inhibition of ␥-secretase (Fig. 4A). Moreover, an antibody-reactive ϳ20 kDa (␥-20 kDa) fragment, consistently present at lower levels than the ␥-26 kDa band, was conspicuously absent in cells treated with the ␥-secretase inhibitors. This fragment is of the expected size for the intracellular ␥-pcdh portion released by ␥-secretase (Fig. 4A). When transfected cells were treated for 24 h with brefeldin A, a compound disassembling the Golgi apparatus, or monensin, which prevents transfer of proteins from Golgi to plasma membrane, no cleavage products of EYFP ␥-pcdh-A1 were detected in Western blots ( Supplementary Fig. 1), indicating that ␥-pcdh insertion into the cell membrane is mandatory for processing by ␥-secretase.
As ␥-secretase requires prior extracellular protease cleavage, resulting in shedding the substrate's ectodomain, we collected the medium of EYFP ␥-pdch-A1 expressing COS-1 cells and probed Western-blotted medium protein with antibody against the Nterminal EYFP moiety of the fusion protein. This analysis revealed two shedded cleavage products, with the larger one efficiently blocked by TAPI, a matrix metalloprotease inhibitor. A different enzyme may release the smaller shedded product. In any event, the significant decrease in the amount of the ␥-26 kDa and ␥-20 kDa fragments from lysates of TAPI-treated relative to untreated cells (Fig. 4A) demonstrated the influence of matrix metalloprotease cleavage on ␥-secretase efficacy.
The catalytic components of the ␥-secretase complex are the presenilins (39 -41). To support the pharmacological evidence for ␥-pcdh processing by ␥-secretase, we transiently transfected presenilin-1/2 double knockout (dko) cells (33,34) with vector for EYFP ␥-pcdh-A1. Subsequent Western analysis with our antibody for ␥-ICD revealed accumulation of the presumptive membranebound ␥-26 kDa fragment and absence of the ␥-20-kDa fragment FIG. 2. Expression of ␥-pcdh alleles with altered constant region. A, Northern blot analyses of poly(A) ϩ RNA from neonatal brains of the different genotypes. Blots were hybridized with a generic variable region probe (␥-VAR) and probes for ␥-C1, ␥-C2, and ␥-C2C3. RNA quantities were evaluated with a cyclophilin probe (CyP). B, quantification of relative transcript abundance from Northern blots. Relative signal intensities are plotted versus genotype, with values for wild type set at 100%. Shown are averaged results (mean Ϯ S.D.) from five mice for each genotype from at least three independent experiments. C, Western blot analysis of protein lysates from neonatal brains of the different genotypes. The anti-␥-ICD antibody detected the ␥-pcdh specific band. Asterisks indicate unspecific bands detected by the antibody. D, quantification of relative protein amounts from Western blots. Relative signal intensities are plotted versus genotype, with values for wild type set at 100% (mean Ϯ S.D.). Shown are averaged results from five mice for each genotype from at least three independent experiments. Note that the low protein levels detected in ϩ/⌬C1 mice are in congruence with the quantification of the transcript levels by Northern blots. (Fig. 4A). In accordance, the subcellular distribution of the ␥-ICD was strikingly affected. EYFP ␥-pcdh-A1 expressing dko cells, unlike COS-1 cells, showed ␥-ICD specific signal exclusively in cytoplasm (Fig. 4B, left and middle panels). However, dko cells transiently expressing myc ␥-ICD exhibited intense nuclear localization of this protein, demonstrating that nuclear import was not impaired in these cells (Fig. 4B, right panel). Similar results were obtained for GFP ␥-pcdh-B1 (data not shown).
Collectively, these data indicate that ␥-pcdhs in the plasma membrane are initially cleaved by a TAPI-sensitive matrix metalloprotease and presumably by other matrix proteases, leading to shedding of the ectodomain, and are additionally processed by the ␥-secretase complex to release a soluble cytoplasmic fragment containing ␥-ICD sequence, which translocates to the nucleus. These events are defining features of PS-IP. ␥-ICD Promotes ␥-pcdh Locus Expression-Nuclear import of ␥-ICD following processing of ␥-pcdhs by PS-IP suggests that this invariant protein domain may be involved in regulating transcription, in analogy to the Notch paradigm (12). We surmised that target genes might include the ␥-pcdh locus, given that overexpressing in HEK293 cells myc ␥-ICD but not myc ␣-ICD, the structurally unrelated invariant C-terminal domain of ␣-pcdhs (16,18), significantly increased ␥-pcdh levels (Fig. 5A, left panel).
We further investigated the effect of ␥-secretase inhibition on ␥-pcdh locus expression. Human SH-SY5Y cells, which re-sponded to pharmacological intervention better than HEK293 cells, were exposed to the ␥-secretase inhibitor DAPT and compared in ␥-pcdh expression to mock (Me 2 SO)-treated cells. Protein lysates showed on Western blots a reduction in endogenous ␥-pcdh levels of ϳ35% (Fig. 5B, left panel). Real-time PCR quantification indicated a similar reduction in endogenous ␥-pcdh transcripts after inhibiting ␥-secretase (Fig. 5B, right  panel). We conclude that ␥-pcdh locus expression is responsive to ␥-secretase-mediated release of the ␥-ICD.
Accordingly, for the other promoter constructs, EYFP ␥-pcdh-A1⌬CT overexpression served as the control to obtain the de- gree of induction by ␥-ICD (Fig. 6B). Results showed that all promoters were induced to similar degrees (Ͼ20-fold induction of luciferase activity) by ␥-ICD, although the activity levels of the promoters differed widely (Fig. 6B). We further observed that deleting the CSE from the ␥-pcdh-A2 promoter did not significantly affect the transcriptional activation by ␥-ICD, although promoter activity dropped ϳ20-fold, in congruence with published data (22). Lastly, no significant transcriptional enhancement by ␥-ICD was obtained with a luciferase reporter construct containing the core promoter sequence of the NR2C gene (28). We conclude that the ␥-ICD has the potential to enhance transcription of ␥-pcdh promoters, for which the CSE is not a prerequisite.

DISCUSSION
The ␥-pcdh locus is essential for survival, as evidenced by the neonatal death of gene-targeted mice lacking it (23) or the first exon for ␥-ICD (this work). Despite the importance of this locus and its widespread expression in most if not all tissues, the physiological functions of ␥-pcdhs remain largely unknown. As members of the cadherin superfamily, an involvement in cellcell interaction during development, postnatal and adult stages is predicted, but specific cell interactions dependent on subsets of ␥-pcdhs have not been determined. In fact, the phenotypes observed in the mutants (19) rather suggest a role for ␥-pcdhs in signaling for cell survival than in patterning.
The most striking feature of ␥-pcdh locus expression, and one shared with the ␣-pcdh locus, is the incorporation into all ␥-pcdh isoforms of an invariant cytoplasmic C-terminal domain (␥-ICD), which is likely to exert class specific signaling. The present study focused on this functional connotation. Mice were generated carrying ␥-pcdh loci in which ␥-C1 was deleted (⌬C1 alleles), in the expectation that this exon deletion would lead to the expression of C-terminally truncated ␥-pcdh variants, rendering the ␥-pcdhs similar to the ␤-pcdhs, which lack a constant intracellular domain (16,17). However, the ⌬C1 allele and its precursor, the neo allele, were severely expressionattenuated, as judged by RT-PCR and Northern analysis of the transcripts in hetero-and homozygous mutant mice. Most strikingly, wild-type ␥-pcdh transcript and protein levels dropped to 25% in ϩ/⌬C1 heterozygous mice instead of the 50% expected from the presence of the wild-type allele. This attenuation might reflect the scarcity of C-terminally intact ␥-pcdhs in the context of an autoregulatory loop for ␥-ICDmediated locus expression. Unfortunately, no expression data on heterozygous ␥-pcdh locus knockouts were reported (23), precluding comparison with ϩ/⌬C1 mice. The attenuated expression might also reflect negative impact of products from the modified allele on the expression of the wild-type allele. The low ␥-pcdh levels did however not cause systemic haploinsufficiency, as judged from the robust appearance of ϩ/⌬C1 mice.
These results prompted us to study the fate and cellular location of the ␥-ICD. We raised an antibody to ␥-ICD and observed that subsets of neurons in sections from adult mouse brain prominently express ␥-pcdhs and notably also exhibit nuclear staining. Nuclear localization of ␥-ICD and the disproportionately reduced ␥-pcdh expression in ϩ/⌬C1 mice led us to investigate signaling by ␥-ICD in cultured cells. Cells were transfected with various ␥-pcdh constructs and subjected to pharmacological treatments to interfere with cellular pathways. The combined results made a strong case for ␥-pcdhs undergoing PS-IP, with ␥-secretase inhibition modifying ␥-pcdh locus expression. Several classic cadherins are also subject to PS-IP (10, 11), but no nuclear translocation has been described.
The molecular mass of the putative C-terminal fragment (ϳ20 kDa) of ␥-pcdhs suggests that cleavage by presenilins occur toward the C-terminal end of the variable exon sequences, within or shortly C-terminal of the transmembrane region. Hence, cleavage might occur within a conserved sequence stretch among the variable ␥-pcdh exons. Prediction of this site is unfortunately not possible, as the presenilin cleavage sites among the known targets lack sequence conservation (42). Such lack of sequence specificity has led to the view that presenilins may initiate the degradation of transmembrane proteins, in addition to initiating specific signaling events (42). The increased ␥-pcdh locus expression by recombinantly expressed ␥-ICD and ␥-secretase inhibition indeed suggests that ␥-pcdh processing by PS-IP affects ␥-pcdh locus expression. We can however not distinguish between direct and indirect transcriptional enhancement by ␥-ICD, nor between repressor and activator interaction of ␥-ICD. Moreover, additional changes in gene expression are likely to be produced by ␥-ICD, which should be exposed by expression profiling in suitable paradigms.
Our study has revealed a vital target for ␥-secretase. Death of presenilin-1/2 double knockout (dko) mice has been attrib- FIG. 5. ␥-ICD promotes ␥-pcdh locus expression. A, left panel, extracts from transiently transfected HEK293 cells with empty vector and vectors for EYFP ␥-pcdh⌬CT, myc ␣-ICD, or myc ␥-ICD probed on Western blots with anti-␥-ICD and anti-GAPDH antibodies. Black arrowheads point at endogenous ␥-pcdh and GAPDH bands. Shown below the Westerns are averaged relative signal intensities (mean Ϯ S.D.) of endogenous ␥-pcdh from four independent experiments normalized to GAPDH expression; the endogenous ␥-pcdh-specific signal after myc ␥-ICD transfection is set at 100%. Right panel, graphic representation of relative transcript amounts in percent from the endogenous ␥-pcdh locus evaluated by real-time PCR on cDNA from HEK293 cells transfected with vectors for myc ␥-pcdh-A1⌬CT, myc ␣-ICD or myc ␥-ICD. myc ␣-ICD and myc ␥-pcdh-A1⌬CT had statistically similar activation potential, which was set to 100%. Shown are data (mean Ϯ S.D.) from 12 independent experiments, each in triplicate. B, left panel, Western of extracts from SH-SY5Y cells treated with Me 2 SO (DMSO) or DAPT, probed with anti-␥-ICD and normalized to GAPDH. Black arrowheads point at endogenous ␥-pcdh and GAPDH bands. Plotted below the lanes are relative intensities (mean Ϯ S.D.) of endogenous ␥-pcdh from four independent experiments; the signal intensity of endogenous ␥-pcdh signal after Me 2 SO treatment is 100%. Right panel, relative transcript amount in percent from the endogenous ␥-pcdh locus evaluated by real-time PCR of cDNAs from SH-SY5Y cells treated with Me 2 SO or DAPT. Shown are data (mean Ϯ S.E.) from 12 independent experiments carried out in triplicate. *, p Ͻ 0.02 by Student's t test.
uted to loss of Notch signaling (43), and deficiency in presenilin-1 alone leads to neonatal death presumably for the same molecular reason (44). However, the role of presenilinmediated ␥-pcdh processing in the context of survival should now also be considered. A forebrain specific presenilin-1/2 dko showed massive cortical shrinkage (45), which might, at least in part, stem from impaired ␥-pcdh processing (46). The importance of ␥-secretase processing of ␥-pcdhs for proper development may ultimately become apparent from mice carrying a mutation in ␥-pcdhs that precludes ␥-secretase cleavage.
To date, the clinically most interesting target of ␥-secretase is APP. Aberrant ␥-secretase cleavage, generated by mutations in presenilins or APP, accounts for the majority of familial Alzheimer cases (47). It is worth pointing out that neuronal apoptosis, a hallmark of late Alzheimer's disease, is prominent in ␥-pcdh knockout mice (19). The premature death of these mice precludes determining if lack of ␥-pcdh expression can also result in neuronal apoptosis in the adult cerebral cortex. An answer to this intriguing issue will hopefully come from a genetically controlled ablation of ␥-pcdhs in the adult brain. Such a genetic manipulation could produce pathology related to Alzheimer's disease, thus revealing ␥-pcdhs as relevant targets for clinical research.
Our study provides evidence that ␥-secretase affects ␥-pcdh locus expression, which appears to be mediated, at least in part, by transactivation of individual variable exon promotors via ␥-ICD. It will be interesting to see if this feedback loop is regulated, e.g. by homophilic or heterophilic interactions (48) between ␥-pcdhs of contacting cells. Conformational changes in the complexed transmembrane glycoprotein itself or, as shown for E-cadherin (49), zipper-like motifs, which help form dense cell-cell adhesive structures, could result in inaccessibility to matrix proteases and thus ␥-secretase. Hence, unbound ␥-pcdhs might be preferentially cleaved, thereby signaling for an increase in ␥-pcdhs to strengthen cell connections. FIG. 6. Transactivation of ␥-pcdh promoters by myc ␥-ICD. A, graphic representation of transactivation at the ␥-A2P promoter. HEK293 cells were cotransfected with the luciferase reporter construct containing ␥-A2P and a vector for EYFP ␥-pcdh-A1⌬CT, EYFP ␥-pcdh-A1, myc ␣-ICD, or myc ␥-ICD. The difference in relative light units between cotransfections of EYFP ␥-pcdh-A1⌬CT/␥-A2P and myc ␥-ICD/␥-A2P reflects the maximum induction (Ͼ20-fold). All experiments were performed in triplicate and repeated at least once. B, analysis of transactivation potential at individual 1-kb ␥-VAR promoters (see Fig. 1A) and an unrelated NR2C promoter by luciferase reporter assays. HEK293 cells were cotransfected with vectors for the reporters and for EYFP ␥-pcdh-A1⌬CT, myc ␥-ICD or myc ␣-ICD (second row from top only) and subjected to luciferase assays. Promoter fragments are shown as solid black lines. The black oval indicates the position of the CSE sequence element. Transcriptional transactivation is plotted as fold induction calculated as difference between transactivation potential of EYFP ␥-pcdh-A1⌬CT to myc ␥-ICD. Maximum mean relative light units (mean Ϯ S.D.) between different promoters differed widely.