The intracellular domain of teneurin-2 has a nuclear function and represses zic-1-mediated transcription

Studies in Drosophila have revealed the existence of two members of a new protein family, namely, Ten-a and Ten-m(Baumgartner and Chiquet-Ehrismann,1993; Baumgartner et al.,1994; Minet and Chiquet-Ehrismann, 2000; Fascetti and Baumgartner,2002). Ten-m, also known as odz(Levine et al., 1994), is a pair-rule gene (Baumgartner et al.,1994). This is surprising, as Ten-m is a cell-surface or secreted protein, and all other pair-rule genes are transcription factors(Lawrence and Struhl, 1996). Baumgartner et al. proposed that Ten-m acts as a pair-rule gene by binding to a receptor (Baumgartner et al.,1994), which in turn transmits the extracellular signal into the nucleus. Mutational analysis by Baumgartner et al., indicated that Ten-m initiates a signal transduction cascade via or in concert with opareceptor (Baumgartner et al.,1994), another pair-rule gene that encodes a zinc finger transcription factor.

During the later stages of development, Ten-a and Ten-m/Odz are predominantly expressed in the nervous system(Levine et al., 1997; Minet et al., 1999; Fascetti and Baumgartner,2002). The predominant neuronal expression is conserved in the vertebrate homologues ten-m1, 2, 3 and 4 in the mouse(Oohashi et al., 1999; Ben-Zur et al., 2000),neurestin in the rat (Otaki and Firestein,1999) and ten-m3 and ten-m4 in zebrafish(Mieda et al., 1999).

Most of the functional studies have been performed on the avian ten-m family members. Three family members have been described in the chicken so far and have been termed teneurin-1 (Minet et al., 1999), teneurin-2 (Rubin et al., 1999) and teneurin-4(Tucker et al., 2000). Teneurin-2 is a type II transmembrane protein containing a furin cleavage site in the extracellular domain (Rubin et al.,1999). Both teneurin-1 and -2 promote neurite outgrowth in vitro(Minet et al., 1999; Rubin et al., 1999). Teneurin-2 also acts as a homophilic adhesion protein and may play a role in the specification of neuronal circuits in the developing visual system(Rubin et al., 2002). In addition to being found in the nervous system, teneurin-2 and -4 are expressed in two important organizing centers of limb development: the apical ectodermal ridge and the zone of polarizing activity, respectively(Tucker et al., 2001; Tucker et al., 2000).

As all members of the teneurin family are type II transmembrane proteins(Rubin et al., 1999; Feng et al., 2002), one potential scenario by which such membrane-spanning proteins can fulfill their role as signaling molecules is by a mechanism recently described as regulated intramembrane proteolysis (RIP) (reviewed in Brown et al., 2000). RIP is a two-step mechanism that leads to the cleavage of transmembrane proteins at and in the lipid bilayer. The cleavage and release of the extracellular or intraluminal parts of the protein is a prerequisite for a second cleavage,which leads to the separation of the intracellular part from the membrane. The latter takes place within the transmembrane domain. The resulting soluble intracellular part translocates to the nucleus, where it participates in transcription. RIP was first proposed as a signaling model by which the sterol regulatory element binding protein (SREBP) regulates lipid metabolism(Brown and Goldstein, 1997). It is now known to control diverse cellular and developmental processes(Brown et al., 2000). The study of Notch, another protein exerting function by this mechanism, was crucial to discover important features of RIP(Chan and Jan, 1998). Also Ire1 (Niwa et al., 1999) and ATF6 (Haze et al., 1999), both of which are involved in the unfolded secretory protein pathway (endoplasmatic reticulum stress response), signal through RIP. Amyloid precursor protein(APP), which is thought to be involved in the Alzheimer's disease, is a prominent example of this mechanism (Haass and De Strooper, 1999; Ebinu and Yankner, 2002). Not only does proteolysis of APP lead to the accumulation of the toxic APP peptide underlying Alzheimer disease, but RIP may be part of normal APP signaling (Gao and Pimplikar, 2001). The most recently recognized and least described examples of RIP include CD44(Okamoto et al., 2001), ErbB-4(Ni et al., 2001; Lee et al., 2002), luman(Raggo et al., 2002) and E-cadherin (Marambaud et al.,2002). These diverse examples of RIP could well be just the tip of the iceberg of a large group of transmembrane proteins undergoing proteolytic cleavage to initiate a signal transduction cascade.

It was the aim of the present work to determine whether a similar proteolytic mechanism is responsible for the signaling by teneurins, thus reconciling the enigma of Drosophila ten-m being a pair-rule gene and a bona fide transcription regulator despite its cell-surface location. We found that indeed the intracellular domain of teneurin-2 can be released from the cell membrane and that it translocates to the nucleus where it is able to influence the transcription activity of zic, a vertebrate homologue of the Drosophila Opa (Yokota et al.,1996).

The following primary antibodies were used anti-teneurin-2(Rubin et al., 1999), anti-VSV(affinity-purified peptide antibody, from Andrew Matus, Friedrich Miescher-Institute), anti-FLAG (M2, Stratagene), anti-Gal4 (DGB RK5C1, Santa Cruz), anti-PML (PG-M3, Santa Cruz), anti-myc (c-Myc 9E10, Santa Cruz) and anti-HA (12CA5, Roche). Secondary antibodies used were Alexa594- and Alexa488-conjugated goat anti-mouse and goat anti-rabbit IgG (all from Molecular Probes), horseradish-peroxidase-coupled anti-mouse and anti-rabbit IgG (Soccochim).

The following DNA constructs were used: pFR-luc (luciferase reporter plasmid; Stratagene), pSV-β-Galactosidase (Promega), pBD-NFκB(encodes BDAD), pCMV-AD and pCMV-BD (Stratagene), p-CMX-PML and p-CMX-PML-RAR(expression plasmids encoding PML or PML-RAR fusion protein, respectively,kindly provided by Ronald M. Evans, San Diego)(Kakizuka et al., 1991; Lin and Evans, 2000), pEF-zic1(expression plasmid encoding myc-tagged zic-1, a generous gift from Jun Aruga,Saitama, Japan) (Aruga et al.,1996), pXP2-APOE189 (luciferase reporter plasmid under the control of an apolipoprotein E promoter; kindly provided by Francisco Zafra, Madrid) (Salero et al.,2001).

Eight different teneurin-2 constructs were used in this study(Fig. 1). They are named according to the teneurin-2 protein domains contained within their coding regions. Two of them (constructs TE and TEY) were described before(Rubin et al., 2002). Construct I represents the soluble intracellular domain of teneurin-2. It encodes the first 372 amino acids of the teneurin-2 sequence as described previously (Rubin et al.,1999), followed by a VSV tag for detection. Construct IT contains,in addition to the intracellular domain, amino acids 373-406, including the membrane-spanning domain and 10 extracellular amino acids followed by an HA tag for detection. In four constructs teneurin-2 was coupled to the Gal4-binding domain (BD) and the NFκB activation domain (AD), generating BDAD-teneurin-2 fusion proteins (see Fig. 1). These constructs were cloned by multiple PCR. The product of the PCRs comprised bases 675-1118 of pCMV-BD, coding for BD, bases 703-1267 of pCMV-AD coding for AD and bases 1-630 of teneurin-2 coding for the first 210 amino acids of the intracellular domain of teneurin-2 until the BlpI site. These fragments were connected by the method of SOE(Horton, 1995), and the resulting construct was cloned into the BamHI/BlpI site of the preexisting pcDNA3 vectors containing teneurin-2 constructs of different lengths (Rubin et al.,2002).

HT1080 fibrosarcoma and COS-7 green monkey kidney cells were routinely maintained in DMEM medium supplemented with 10% FCS. For transient transfections, the cells were seeded in six-well plates or 35 mm dishes containing four internal wells (Greiner). 12 hours later they were transfected with the indicated expression vectors (1 μg of each vector) by using FUGENE-6 (3 μl, 6 μl or 9 μl for one, two to three or four different plasmids, respectively; Roche). 24 hours after transfection the cells were rinsed in PBS and processed for either measuring luciferase andβ-galactosidase activities, western blotting or immunofluorescence.

Where indicated the cells were treated with the following substances at least 5 hours after transfection: ALLN (25 μg/ml;N-acetyl-leu-leu-norleu-AL; Sigma); tunicamycin (2 μg/ml; Sigma) or lactacystin (10 μM; Sigma) for 4 or 8 hours prior to harvesting.

Construct I was subcloned into the ecdyson-inducible expression vector pIND(Invitrogen) and transfected into EcR-293 cells (Invitrogen) according to the supplier's manual, resulting in the cell line EcR-293-I. Clones were tested for the inducible expression of construct I after the addition of increasing concentrations of ponasterone (1-10 μg/ml; Invitrogen) by immunoblotting using anti-VSV antibodies.

Clones of HT1080 cells stably expressing TEY (TEY cells) or TE (TE cells)on their surfaces, respectively, have been described previously(Rubin et al., 2002).

The cells were lysed by adding reporter lysis buffer (Promega). Appropriate dilutions of the lysed cell suspension were then pipetted into Microlite™ luciferase plates (Dynex Technologies), and the luciferase activity was measured in a Microlumat (LB96P, EG+G Berthold) by automatic injection of luciferin substrate solution (2 mM luciferin, 100 mM ATP in 250 mM glycin pH 7.8, 150 mM MgSO₄). All luciferase activities were normalized with respect to the transfection efficiency by co-transfecting aβ-galactosidase vector. To determine β-galactosidase activity the diluted cell suspensions were incubated with the substrate solution (4.5 mM 2-nitrophenyl-b-D-galactopyranoside in 0.2 M Na-phosphate, 2 mM MgCl₂, 0.1 mM β-mercaptoethanol) for 30 minutes at 37°C. To stop the enzymatic reaction, 1 M Na₂CO₃ was added and the OD was measured at 405 nm in a microplate reader (BioRad).

Teneurin-2 constructs I, IT and ITE were extracted by adding SDS sample buffer containing 20% β-mercaptoethanol directly to the cells. Extraction of the nuclear constructs BDAD and BDAD-I was achieved by performing nuclear fractionation. The transfected cells were harvested by scraping off the cell layer in lysis buffer [10 mM HEPES pH 7.5, 0.5% triton X-100, 300 mM sucrose,100 mM NaCl, 2 mM MgCl₂, protease inhibitors (Complete™,Roche Diagnostics)] on ice and subsequent centrifugation for 10 minutes at 420 g in an Eppendorf centrifuge. The resulting pellet was resuspended in lysis buffer and centrifuged again. The final pellet was then dissolved in SDS sample buffer containing 20% β-mercaptoethanol, 6 M urea and protease inhibitors (Complete™). Before loading on an 8% SDS-PAGE gel, DTT was added to a final concentration of 10 mM.

The transmembrane constructs BDAD-ITE and BDAD-ITEY were extracted from the cells by the following procedure. The cells were extracted on ice by a hypotonic buffer (2 mM Na-phosphate buffer pH 7.5, 20 mM KCl, 1 mMβ-mercaptoethanol), scraped off and centrifuged for 10 minutes at 6800 g at 4°C in an Eppendorf centrifuge. The resulting pellet was reconstituted in detergent buffer [50 mM Tris pH 8, 1% NP40, 150 mM NaCl,5 mM EDTA, 6 M urea, protease inhibitors (Complete™)], incubated for 20 minutes at 37°C and centrifuged for 10 minutes at 17,900 g. SDS sample buffer containing 20% β-mercaptoethanol, 6 M urea and protease inhibitors (Complete™) was added to the supernatant and incubated for 1 hour at 52°C. After DTT was added (10 mM), the samples were loaded on a 6% SDS-PAGE gel.

The gels were transferred to PVDF membranes. The proteins were detected by anti-Gal4 antibody (BDAD and BDAD-I) or by anti-teneurin-2 serum (BDAD-ITE and BDAD-ITEY), horseradish-peroxidase-coupled secondary antibodies and ECL SuperSignal® (Pierce).

The cells grown on 35 mm four-well staining dishes (Greiner) were fixed with 4% PFA for 30 minutes at room temperature and, where indicated,permeabilized with 0.1% triton X-100 for 5 minutes. Incubation with primary antibodies was performed for 60 minutes and that with secondary antibodies for 30 minutes, both at room temperature, and the cells were washed in PBS after each incubation. Finally, the specimens were mounted in Moviol and examined and photographed using a Zeiss Axiophot microscope (Carl Zeiss Ltd.) connected to a 3CCD camera (Sony).

Pilot results with a yeast two-hybrid assay indicated transcriptional activity for the intracellular part of teneurin-2 (data not shown). Consistent with this we found the intracellular domain of teneurin-2 (referred to as construct I, see schematic representation of constructs in Fig. 1) to be translocated to the nucleus if transfected into HT1080 cells. To our surprise, transfection of construct I did not lead to a uniform nuclear accumulation but its expression instead was confined to discrete spots within the nucleus(Fig. 2b,d). This staining pattern contrasted with that of the transmembrane teneurin-2 construct IT,which accumulated on the cell surface (Fig. 2a). The nuclear localization coincided with a very similar punctate pattern obtained by staining for PML protein (promyelocytic leukemia protein) (reviewed in Seeler and Dejean,1999) and may represent nuclear bodies, termed promyelocytic oncogenic domains (PODs) or PML bodies(Fig. 2e). Double immunofluorescence staining of construct I transfected cells did show substantial, but not complete, overlap of PML with the intracellular domain of teneurin-2 (Fig. 2d-f). Since PML bodies are involved in a number of functions associated with transcriptional control (Zhong et al.,2000), it was of interest to determine whether there was a genuine colocalisation of teneurin-2 and PML in nuclear bodies. We therefore cotransfected construct I with an expression plasmid encoding PML on the one hand and with PML-RAR (PML-retinoic acid receptor fusion protein) on the other. Transfection of PML into cells containing endogenous PML protein leads to a massive enlargement of the nuclear bodies(Doucas et al., 1996). In contrast, transfection of PML-RAR should result in a destruction of the PML body architecture (Dyck et al.,1994; Mu et al.,1994). We did indeed observe these effects, as illustrated in Fig. 2h,k,l. Interestingly, for teneurin-2 I an equivalent staining pattern was detected after cotransfection with PML, and I was pulled into the enlarged PML bodies(Fig. 2g-i). Furthermore,destruction of the PML body architecture after transfection with PML-RAR also changed the expression pattern of cotransfected I, which was no longer accumulated in discrete spots in the nucleus but seemed to be expressed homogeneously throughout the cells (Fig. 2j). These results indicate that the intracellular domain of teneurin-2 accumulates within nuclear bodies, thus supporting our previous hypothesis that the intracellular domain of teneurin-2 is involved in transcriptional regulation.

In Drosophila, ten-m was postulated to modulate the activity of Opa protein (Baumgartner et al.,1994). It was therefore of interest to investigate whether the zinc finger transcription factor zic, a vertebrate homologue of Opa, would influence or would be influenced by the intracellular domain of teneurin-2.

When both proteins were expressed in COS-7 cells by transient transfections we observed a marked downregulation of the intracellular domain I of teneurin-2 compared with its usual expression level(Fig. 3A). On the other hand,co-transfection of the two constructs did not have an effect on the level of zic (Fig. 3C). In contrast to its effect on teneurin-2, zic did not downregulate another co-transfected transcription factor BDAD (Gal4 DNA-binding domain fused to the NFκB activation domain) in an analogous analysis(Fig. 3B,D). The zic-induced downregulation of the intracellular domain I of teneurin-2 was counteracted by the addition of the proteasome inhibitor lactacystin(Fig. 3A). Thus the nuclear intracellular domain of teneurin-2 seems to be subject to degradation by the proteasome pathway. By immunofluorescence staining of the transfected cells we observed that zic-transfected cells revealed a relatively diffuse nuclear staining (Fig. 3E,F) and in nuclei containing high amounts of zic protein, the punctate staining of teneurin-2 I disappeared and became diffuse(Fig. 3F). Thus, the presence of zic prevented the association of the teneurin-2 intracellular domain with PML bodies and made it amenable to proteasome-mediated degradation.

To examine the potential effect of the teneurin-2 I on the transcriptional activity of zic, stably transfected EcR-293 cell lines were produced. In these EcR-I cells teneurin-2 I was only expressed upon addition of ponasterone(Fig. 4A). EcR-I cells were transiently transfected with zic and a luciferase reporter construct under the control of the apolipoprotein E (ApoE) promotor known to be activated by zic(Salero et al., 2001). Whereas the ApoE-luciferase reporter construct alone did not show any activity, the presence of zic led to a dramatic increase in luciferase acitivity(Fig. 4B). After the induction of teneurin-2 I by ponasterone we observed a marked reduction in the expression level of the reporter gene only in EcR-I cells(Fig. 4B) and not in EcR control cells (Fig. 4C). This result suggested that the intracellular domain of teneurin-2 did have an inhibiting effect on the transcriptional activity of zic, and this effect was more pronounced in the presence of the proteasome inhibitor ALLN, which stabilizes teneurin-2 I (Fig. 4A).

To be a functional regulator of transcription, wild-type transmembane teneurin-2 would have to be specifically cleaved in or at the plasma membrane,possibly upon a signal by ligand binding. In turn its intracellular part must be released and translocated to the nucleus in a manner similar to that established for proteins regulated by RIP (reviewed in Brown et al., 2000; Ebinu and Yankner, 2002). To test this hypothesis we developed a sensitive method to detect the released intracellular domain of teneurin-2 in the nucleus. We expressed fusion proteins of full-length teneurin-2 (or of smaller transmembrane versions truncated in their extracellular domain) fused to a Gal4 DNA-binding domain(BD) and a NFκB activation domain (AD; see Fig. 1). If cleavage and translocation to the nucleus occurred, BDAD-I could be detected by binding to specific Gal4 recognition sequences in the promotor of the cotransfected luciferase reporter plasmid, and subsequent initiation of luciferase gene expression activated by AD could be monitored.

Fig. 5 illustrates the correct expression of the transfected fusion proteins. BDAD and BDAD-I,serving as positive controls in this experiment, were detectable on a western blot of nuclear extracts by anti-Gal4 antibodies(Fig. 5A), and their accumulation in the nucleus was confirmed by immunofluorescence staining of permeabilised cells (Fig. 5B). At the same time BDAD-ITE and BDAD-ITEY could be identified as part of the plasma membrane by western blots of membrane fractions and by immunofluorescence of non-permeabilized cells(Fig. 5A,B).

For analysis of the luciferase activity induced by the teneurin-2 fusion constructs, HT1080 cells were cotransfected with the respective BDAD-teneurin-2 constructs, the luciferase reporter plasmid, as well as aβ-galactosidase construct for normalization of transfection efficiencies. As displayed in Fig. 6A,BDAD-ITE, BDAD-IT and BDAD-I did indeed lead to an induction of luciferase activity above the negative control (BD construct). However, BDAD-ITEY, being the largest fusion protein, did not lead to a significantly enhanced activity. This might partly be explained by the fact that the larger the transmembrane construct the lower the expression level. Alternatively, cleavage of the shorter constructs might be constitutive, whereas cleavage of the full-length construct might have to be specifically induced. This is the case for Notch:processing of Notch expressed on the cell surface is specifically activated by binding to its ligand Delta (Kidd et al.,1998; Schroeter et al.,1998; Struhl and Adachi,1998). Teneurin-2 has recently been shown to bind homophilically by its extracellular domain (Rubin et al.,2002). We therefore speculated that this interaction could induce cleavage and translocation of the intracellular domain of the BDAD-teneurin-2 fusion proteins, which in turn would be represented by enhanced luciferase activities.

To test whether homophilic interaction of teneurin-2 represents a signal for cleavage of its intracellular domain we transfected BDAD-ITEY and BDAD-IT into HT-1080 cells (HT control) or clones that constitutively express the entire extracellular domain TEY on their surface (TEY cells). The luciferase activity produced from the co-transfected reporter plasmid was then compared(Fig. 6B). As can be seen in Fig. 6B, luciferase activity obtained after transfection of BDAD-ITEY into the TEY cells was 2.6-fold higher than after transfection into HT control cells, whereas the opposite was the case for the BDAD-IT construct, which gave lower luciferase activities in the TEY cells than in the HT control cells, with a ratio of 0.4 fold(Fig. 6B). Thus, we see a 6.5-fold induction of luciferase activity of BDAD-ITEY in TEY cells compared with BDAD-IT-transfected cells. When we did the same experiment comparing transfection of BDAD-ITEY with BDAD-IT into TE cells versus HT control cells we did not see any difference between the two transfected constructs and the fold induction remained 1.0 (Fig. 6C; Table 1). Each of these experiments was repeated five times with consistent results(Table 1). We, therefore,conclude that interactions through the C-terminal half of the extracellular domain are able to stimulate the release of the intracellular domain of teneurin-2, as indicated in the model presented in Fig. 6C.

The induction of luciferase activity following transfection of BDAD-ITEY could again be markedly upregulated by the addition of protease inhibitors,such as ALLN and lactacystin (Fig. 7A). Thus, also the cleaved intracellular domain is subject to rapid degradation by the proteasome pathway. This was confirmed when the ITE protein of transfected cells was analysed on a western blot. The addition of ALLN led to the stabilization of two particular cleavage products, of which one matched the size of the entire intracellular domain and one was a fragment thereof (Fig. 7B, arrow and arrowhead).

Taken together, we conclude that the activity of the luciferase reporter gene originated from cleavage of the BDAD-teneurin-2 fusion proteins at (or in the vicinity of) the membrane. However, cleavage of full-length teneurin-2 led to a significant induction of the luciferase gene only when processing was upregulated by homophilic binding of the extracellular C-terminal part of teneurin-2. Furthermore, the cleaved intracellular domain is subject to rapid degradation by the proteasome pathway.

In this paper we show that teneurin-2 is a membrane-bound transcription regulator. This could explain the pair-rule phenotype of ten-m mutants in Drosophila, where ten-m was shown to be epistatic to opa and to regulate the expression of downstream target genes such as prd, slpor gsb (Baumgartner et al.,1994). The authors postulated that ten-m may influence the activity of opa. For this reason we investigated whether such an interaction could be shown for the respective vertebrate homologues. The zinc finger transcription factors zic-1 to -4 have been proposed to be the vertebrate homologues of opa (Aruga et al.,1996). It was shown that zic-1 and -2 are able to activate transcription from the apolipoprotein E promoter(Salero et al., 2001). We therefore tested whether teneurin-2 would influence the activation by zic-1 of a luciferase gene controlled by the ApoE promoter. We found that teneurin-2 repressed the zic-1 activity. At the same time the presence of zic-1 influenced the subnuclear localization of teneurin-2. Although teneurin-2 alone was present in PML nuclear bodies it became diffusely distributed in the nucleus in the presence of zic-1. This diffuse distribution coincided with a rapid degradation of teneurin-2. Thus on the one hand zic-1 causes downregulation of teneurin-2 expression and on the other hand teneurin-2 reduced zic-1-mediated transcriptional activation of the apolipoprotein E promotor.

The genetics in Drosophila imply that ten-m and opa interact to induce the transcription of the same downstream target genes(Baumgartner et al., 1994). Furthermore it is known that opa is expressed throughout the segments and ten-m only in the part of the segment where the downstream genes are induced(Baumgartner et al., 1994). On the basis of our results we therefore speculate that opa could act as a transcriptional repressor and that the repressor function is interfered with by ten-m. Thus, both ten-m and opa are required to determine localized transcription of their target genes in segmental stripes. Also, in vertebrates, functional interaction in vivo of zic proteins with teneurin family members is not unlikely, since they may be co-expressed in many tissues and both appear to be involved not only in regulating neuronal development but possibly also limb pattern formation (Aruga et al., 2002a; Rubin et al.,2002; Nagai et al.,1997; Tucker et al.,2001).

For the transmembrane protein teneurin-2 to function as a transcription regulator the release of the intracellular domain is indispensable. To date,two types of proteolytic mechanisms have been shown to account for such a release of the intracellular domains involved in transcription control, namely regulated intramembranous proteolysis RIP by either γ-secretase or S2P(site-2-protease) or regulated ubiquitin/proteasome-dependent processing RUP(Hoppe et al., 2001). In the case of teneurin-2, RUP is unlikely to be responsible for the cleavage since proteasome inhibitors enhance the presence of the cleaved intracellular domain. However, levels of the intracellular domain of teneurin-2 seem to be tightly controlled by degradation through the proteasome pathway. Under conditions where the intracellular domain of teneurin-2 is localized in PML bodies the teneurin-2 protein is stable. In contrast, conditions leading to a diffuse nuclear expression, as is the case for the teneurin-2 I BDAD fusion protein or the soluble intracellular domain in the presence of zic, result in the degradation of the protein, which can be inhibited by the proteasome inhibitors ALLN and lactacystin. These findings are intriguing in the light of recent discoveries concerning the regulation of transcription by ubiquitin,which causes a rapid turnover of the ubiquitinylated transcription factors(Molinari et al., 1999; Salghetti et al., 2001; Conaway et al., 2002). Interestingly, one of the first proteins discovered to function through RIP,SREBP, is also subject to rapid degradation by the ubiquitin-proteasome pathway (Hirano et al., 2001). The same is true for Notch, which is the most intensely studied example of a membrane-anchored transcription factor functioning through RIP. Cleaved fragments of Notch could not be identified at first owing to rapid downregulation in proteasomes simultaneously with a low sensitivity in the detection methods (Chan and Jan,1998; Annaert and De Strooper,1999; Kopan and Goate,2000; Schroeter et al.,1998). The same problem might be responsible for our inability to directly detect the cleaved intracellular domain of teneurin-2 in the nucleus by immunohistochemistry.

Cleavage of Notch is induced by interaction with its heterophilic ligand Delta (for a review, see Artavanis-Tsakonas et al., 1999). In the case of teneurin-2 we found that homophilic interactions can induce the release of the intracellular domain. This is interesting considering our previous observation that neurons making up functional circuits in the avian visual system express the same type of teneurin molecule (Rubin et al.,2002). Therefore, promotion of the release of the intracellular domain of teneurin-2 could be the mechanistic basis by which growing axons realize that they have reached a proper target, namely another neuron expressing teneurin-2. Since homophilic interaction leads to only a moderate activation of the release of the intracellular domain, we cannot exclude the possibility that more potent mechanisms exist by interaction with yet-to-be identified heterophilic ligands. The released intracellular domain could then turn on a gene expression program to stabilize the connection to differentiate and to build synapses. In this respect teneurin-2 would counteract the action of zic, which by itself was shown to promote the expansion of neuronal progenitors (Aruga et al.,2002b). This would fit our present observation that transcription from at least one zic target gene, namely ApoE, is downregulated by teneurin-2.

We are very grateful to Ronald M. Evans, Jun Aruga and Francisco Zafra for the generous provision of plasmids, and we would like to thank François Lehembre for helpful suggestions. We also would like to thank Marianne Brown-Luedi for help with the western blots of teneurin-2.