CALEB/NGC Interacts with the Golgi-associated Protein PIST*

CALEB/NGC is a neural member of the epidermal growth factor protein family expressed in axon and synapse-rich areas of the nervous system and shown to be important for neurite formation. It can bind to the extracellular matrix proteins tenascin-R and tenascin-C. Here we show that CALEB/NGC interacts with the Golgi-associated protein PIST. PIST was originally described as an interaction partner of the small GTPase TC10 and was then found to be Golgi-associated by binding to syntaxin-6 and to be important for the transport of frizzled proteins and the cystic fibrosis transmembrane conductance regulator to the plasma membrane. In addition, PIST was demonstrated to be involved in autophagy and linked to processes of neurodegeneration. CALEB/NGC interacts with PIST in the yeast two-hybrid system. This interaction can be confirmed by co-immunoprecipitations and co-localization studies. The juxtamembrane cytoplasmic peptide segment of CALEB/NGC, highly conserved during evolution, mediates the binding to PIST. CALEB/NGC co-localizes with PIST in the Golgi apparatus of transfected COS7 cells and in Golgi-derived vesicles after brefeldin A or nocodazole treatment. Co-localization studies in primary hippocampal cells and analysis of Purkinje cells of colchicine-treated rats, serving as an in vivo model system to block microtubule-dependent transport processes, support the view that PIST is an interaction partner of CALEB/NGC and implicate that this interaction may play a role in the intracellular transport of CALEB/NGC.

CALEB/NGC (chicken acidic leucine-rich EGF 1 -like domaincontaining brain protein/neuroglycan C) is a neural transmembrane protein, which was originally discovered in a screen for novel molecules implicated in cell differentiation processes of the nervous system (1)(2)(3). It is expressed in axon and synapserich areas of the nervous system and was shown to be important for neurite formation (1,4). CALEB/NGC is a type I transmembrane protein, the extracellular part of which contains the following domains. A highly acidic peptide segment is able to mediate binding to the extracellular matrix proteins tenascin-R and tenascin-C (5). An EGF-like domain located near the transmembrane region is structurally very similar to the corresponding domains of members of the EGF family of transmembrane growth and differentiation factors like EGF itself, neuregulin, and TGF-␣ (6,7). Many functions can be attributed to the members of this protein family, and it is also well established that these proteins serve as ligands for ErbB receptor tyrosine kinases. Currently, it is unclear whether CALEB/NGC may also be able to stimulate ErbB receptors. Whereas the functions of the members of the EGF family of transmembrane growth and differentiation factors as ligands for ErbB receptor tyrosine kinases had been examined in detail, several studies in the last years focused on the functions of the cytoplasmic domains of these transmembrane proteins. For example, it was found that proteolytic cleavage of the extracellular domain of transmembrane neuregulins is regulated by their cytoplasmic tails (8). Activated release of membraneanchored TGF-␣ was also shown to be dependent on the intracellular domain of this protein, and the carboxyl-terminal valine residue of TGF-␣ was recognized as important for this processing (9,10). Moreover, mutants of TGF-␣ lacking the carboxyl-terminal valine were present at substantially reduced levels at the cell surface, indicating that the C terminus of TGF-␣ might be important for intracellular trafficking of the transmembrane precursor. Indeed, two different PDZ domain proteins, TACIP18/syntenin-1 and p59/GRASP55, were discovered, which bind to the C terminus of pro-TGF-␣ and affect its targeting to the cell surface (11,12). GRASP55 was formerly shown to play a role in the stacking of the Golgi apparatus, and syntenin-1 had been published as an interaction partner for the C termini of several transmembrane proteins including syndecans, class B ephrins, EpH A7, neurexins, the anion exchanger AE2, neurofascin, different glutamate receptor subtypes, schwannomin, and the protein-tyrosine phosphatase (13)(14)(15)(16)(17)(18)(19)(20)(21).
Little is known about the transport of other members of the EGF family of transmembrane proteins to the cell surface. We wanted to address this issue to the transmembrane protein CALEB/NGC. Large parts of the cytoplasmic domain of this protein are highly conserved between chicks, mice, rats, and humans (5). So far, neither any interaction partner nor any function can be attributed to the intracellular part of CALEB/ NGC. In this study, we identified the Golgi-asssociated protein PIST (PDZ domain protein interacting specifically with TC10; also called FIG (fused in glioblastoma), GOPC (Golgi-associated PDZ and coiled-coil motif-containing protein), or CAL (CFTR-associated ligand)) (22)(23)(24)(25) as an interaction partner for CALEB/NGC. PIST was originally described as a PDZ domain protein, which binds to the small GTPase TC10 (22). Then it was shown to be Golgi-associated via binding to the Q-SNARE (Q-soluble NSF attachment protein receptor) protein syntaxin-6 (23). Further studies reported that PIST binds to the cell surface proteins frizzled 5 and frizzled 8, to the cystic fibrosis transmembrane conductance regulator (CFTR), and to the chloride channel ClC-3B and regulates the expression of these proteins at the plasma membrane (24 -26). Recent work demonstrated a novel isoform of PIST to be involved in auto-phagy and neurodegeneration by linking the ␦2 glutamate receptor to the autophagy regulator protein Beclin 1 (27).
In contrast to the transmembrane proteins frizzled 5, frizzled 8, CFTR, and ClC-3B, where the corresponding C termini bind to the PDZ domain of PIST, in the case of CALEB/NGC, the juxtamembrane cytoplasmic peptide segment mediates the interaction with this Golgi-associated protein. Co-immunoprecipitations from transfected HEK293 cells confirmed the results obtained by yeast two-hybrid analysis. We present data about a co-localization of CALEB/NGC and PIST in the Golgi apparatus of transfected COS7 cells and in Golgi-derived vesicles after brefeldin A or nocodazole treatment. Analysis of primary and transfected hippocampal cells and the results of our experiments with an in vivo model system to block microtubule-dependent transport in neurons support the view that the interaction of CALEB/NGC and PIST may play a role for the intracellular transport of CALEB/NGC from the soma of neurons into the processes.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening and Mapping of the Interaction of CALEB/NGC and PIST-A PCR fragment encoding the total cytoplasmic domain of CALEB/NGC was amplified from human cDNA (Marathon-Ready, Clontech) with the oligonucleotides 5Ј-CACGAATTCCAA-GAAGCTCTACCTGCTCAAGACG-3Ј (hCyt1) and 5Ј-CACGGATCCGA-CTCTGTACAAGAGGAGATAATGT-3Ј (hCyt2). The PCR fragment was cloned into the bait vector pAS2-1 (Clontech) using EcoRI and BamHI restriction sites, and the resulting construct was transformed into the yeast strain AH109 (Clontech) (28). The yeast strain Y187 pretransformed with a cDNA library from mouse brain cloned into the vector pACT2 (Clontech) was then mated with the AH109 strain that contains the bait construct according to the manufacturer's protocol. Considering mating efficiency and titer of the library, 10.7 ϫ 10 6 clones were screened. 23 His ϩ clones were isolated from selective medium lacking leucine, tryptophan, and histidine supplemented with 5 mM 3-amino-1,2,4-triazole. 19 of these clones developed a blue color when tested for expression of the MEL1 gene. Two clones were found to encode the full-length sequence of PIST as described (22).
Proteins, Peptides, and Antibodies-Polyclonal antibodies from rabbits were generated against a bacterially expressed recombinant fusion protein of the maltose-binding protein and full-length PIST. These antibodies were affinity-purified using the recombinant fusion protein and used at a final concentration of 1 g/ml. Polyclonal antibodies to PIST raised in guinea pigs were a gift from Wolf Wente and Dr. Hansjü rgen Kreienkamp. The peptide used for precipitating recombi-nant PIST from transfected COS7 cells and native PIST from brain extracts contains the sequence of the cell membrane transduction domain of the human immunodeficiency virus type I Tat protein and the CALEB/NGC-derived PIST-binding sequence, FFAKKLYLLKTENT-KLRKT. The control peptide had the amino acid sequence GGLEDENG derived from the extracellular region of CALEB/NGC.
The monoclonal antibodies M2 to the FLAG epitope, P5D4 to the VSV epitope, CDF-4 to Golgin-97, and 35 to GM130 were from Sigma, Roche Applied Science, Molecular Probes, Inc. (Eugene, OR), and BD Biosciences, respectively, and were used according to the manufacturers' instructions. The monoclonal antibody directed to CALEB/NGC was also from BD Biosciences and was used at a final concentration of 2 g/ml.
Cell Culture and Transfections-HEK293 and COS7 cells (American Type Culture Collection) were maintained in Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum and the antibiotics penicillin and streptomycin at 37°C in a 5% CO 2 environment. Transfections were performed with FuGene 6 (Roche Applied Science) or Polyfect (Qiagen) according to the manufacturers' instructions or using the DEAE-dextran method as described (29,30). The following expression plasmids were used. A cDNA fragment encoding full-length PIST was cloned into the expression vector pVM6 (Roche Applied Science) to encode PIST containing an amino-terminal VSV tag and a carboxylterminal His tag; a cDNA fragment encoding full-length CALEB/NGCb (according to the mouse sequence mCALEBb) was cloned into the vector p3ϫFLAG-Myc-CMV-25 (Sigma) to encode full-length CALEB/NGCb with an amino-terminal 3ϫ FLAG tag and a carboxyl-terminal Myc tag. In a similar manner, a cDNA fragment encoding full-length CALEBa-80 (the shorter isoform of the chicken ortholog of CALEB/ NGC) (5) was cloned into the vector p3ϫFLAG-Myc-CMV-25. The CALEB/NGC-derived construct 449 contains an amino-terminal 3ϫ FLAG epitope followed by the transmembrane and the cytoplasmic domain of CALEB/NGCb (mouse sequence) and by a carboxyl-terminal Myc epitope. The construct CCEX comprises the cytoplasmic domain of the longer isoform of CALEB/NGC (mouse sequence) without the juxtamembrane PIST-binding sequence. This construct contains an aminoterminal Myc tag. For immunocytochemistry the COS7 cells were fixed 2 days after transfection with 4% (w/v) paraformaldehyde in PBS for 20 min at room temperature. Nocodazole (Sigma) was added to the cells as indicated at a final concentration of 10 g/ml and incubated for 4 h in regular media with 1% fetal calf serum before the cells were washed in PBS and fixed as described above. Brefeldin A (Sigma) was added to the cells as indicated at a final concentration of 10 g/ml. After an incubation of 30 min to 1 h, the cells were washed with PBS and immediately fixed.
Primary hippocampal cells were prepared from embryonic day 19 rat embryos as described with slight modifications (31). The cells were cultured on poly-L-lysine-coated glass coverslips in neurobasal A medium supplemented with B27 (Invitrogen) (32), glutamine, and the antibiotics penicillin and streptomycin. These cells were transfected with LipofectAMINE 2000 (Invitrogen) or Effectene (Qiagen) according to the manufacturers' instructions.
Immunocytochemistry-After fixation, the cells were blocked in PBS supplemented with 5% fetal calf serum and permeabilized in the same solution containing 0.2% Triton X-100. The incubations with the primary antibodies as indicated were performed for 1 h at room temperature in blocking solution. After four quick washes with PBS with 0.1% Triton X-100, the cells were incubated with the secondary antibodies in blocking solution for 1 h at room temperature. As secondary antibodies, we used Cy3-conjugated goat anti-mouse, Cy3-conjugated goat antiguinea pig (both from Dianova), and Alexa488-conjugated goat antirabbit (Molecular Probes) according to the manufacturers' instructions. After extensive washes, the cells were mounted in PBS/glycerol (1 ϩ 1) or glycerol gelatin (Sigma). Analysis was done with a Leica Aristoplan fluorescence microscope. Images were captured with a CCD camera (Hamamatsu) and analyzed with the Openlab software (Improvision).
Immunoprecipitations, SDS-PAGE, and Immunoblots-Transiently transfected HEK293 cells were homogenized 2 days after transfection in extraction buffer (Tris-buffered saline containing 1% Triton X-100, 2 mM CaCl 2 , and the protease inhibitors phenylmethylsulfonyl fluoride, pepstatin, aprotinin, and leupeptin). Unlysed cells and insoluble cellular debris was removed by centrifugation at 14,000 rpm for 20 min at 4°C. The cell lysates were incubated overnight with an antibody directed to the VSV tag. The antibodies were collected by an incubation with protein A/G Plus-Sepharose (Santa Cruz Biotechnology) for 3 h at 4°C with constant agitation. The Sepharose beads were washed several times with extraction buffer and prepared for SDS-PAGE. For precipitation experiments with the immobilized peptides described above, detergent extracts of transfected HEK293 cells or detergent extracts of adult rat brain were incubated with the peptides immobilized to Nhydroxysuccinimide-activated Sepharose (Amersham Biosciences) for 4 h at 4°C with constant agitation. The Sepharose beads were washed several times with extraction buffer and prepared for SDS-PAGE. Proteins were separated by SDS-PAGE and analyzed by Western blot as described (1) with the indicated antibodies.
Immunohistochemistry-Male Wistar rats (n ϭ 6, 200 -230 g; Tierzucht Schönwalde, Germany) were anesthetized with chloral hydrate and transcardially perfused with Tyrode's solution followed by Zamboni's fixative containing 4% paraformaldehyde and 0.2% picric acid in 0.1 M phosphate buffer, pH 7.4. In a separate set of experiments, rats (n ϭ 4) received an intracerebroventricular injection of 10 g of colchicine 48 h before vascular perfusion (33). Control animals received an equal volume of vehicle injection (n ϭ 4). Brains and spinal cords were rapidly dissected and postfixed in the same fixative for 2 h at room temperature. For all animal procedures, ethical approval was sought before the experiments according to the requirements of the German National Act on the Use of Experimental Animals. Tissue was cryoprotected by immersion in 30% sucrose for 72 h at 4°C before sectioning using a freezing microtome. Free-floating sections (40 -50 m) were washed in phosphate-buffered saline and incubated in phosphate-buffered saline supplemented with 3% normal goat serum and 0.3% Triton X-100 for 1 h to block unspecific staining. Staining of the tissue sections was performed with a monoclonal antibody directed to CALEB/NGC (BD Biosciences) and an affinity-purified antibody directed to recombinant PIST. Images were captured as described above.

CALEB/NGC Interacts with PIST in the Yeast Two-hybrid
System-CALEB/NGC is a neural transmembrane protein, which is predominantly expressed in axon and synapse-rich areas in the nervous system. In the chicken retina, for example, it is expressed in the inner and outer plexiform layers, where most of the synapses are located, and the optic fiber layer, where the axons of the ganglion cells grow toward the optic fissure (1,4). CALEB/NGC is hardly detected in the inner and outer nuclear layers, where the cell bodies of most neurons of the retina reside. This expression pattern suggests that the CALEB/NGC protein must be targeted to the plasma membrane after its synthesis and has to be transported from the cell body to the processes of the neurons. In order to search for candidate proteins that may be involved in the trafficking of CALEB/NGC, we performed a yeast two-hybrid screen with the intracellular domain of the longer isoform of CALEB/NGC as a bait and a pretransformed library from mouse brain. After screening of 10.7 ϫ 10 6 clones, we obtained 19 positive clones, two of which encoded the full-length sequence of PIST as described (22). PIST is a protein with a predicted molecular mass of 49.8 kDa, which contains two putative coiled-coil domains and one PDZ domain (Fig. 1A). So far, five transmembrane proteins are known to be able to bind to the PDZ domain of PIST. Beclin 1 and syntaxin-6 were demonstrated to interact with the coiled-coil domains (23,27). We wanted to examine which domains of PIST are necessary for binding CALEB/NGC. Several constructs of PIST were tested in the yeast two-hybrid system, but only a few of them could bind to the intracellular domain of CALEB/NGC (Fig. 1A).
This means that the integrity of the PIST protein is important for the interaction with CALEB/NGC to occur. Only the amino-terminal or the very last carboxyl-terminal part of PIST can be deleted without completely destroying the ability to bind to CALEB/NGC. A second isoform of PIST was reported, which contains 8 additional amino acids at the beginning of the carboxyl-terminal coiled-coil domain and which is predominantly expressed in the nervous system (27). We tested this isoform in the yeast two-hybrid system and found it to bind to CALEB/ NGC, too (data not shown). Two isoforms of CALEB/NGC are known, which differ in their cytoplasmic region (5). The longer isoform CALEB/NGCb has one peptide segment of 27 amino acids in addition to the shorter isoform CALEB/NGCa with respect to the species mice, rats, and humans (Fig. 1B). We found that both isoforms are able to bind to PIST in the yeast two-hybrid system. When testing several other constructs, the juxtamembrane cytoplasmic peptide segment of CALEB/NGC could be determined as essential for binding PIST (Fig. 1B).
FIG. 1. PIST binds to CALEB/NGC in the yeast two-hybrid system. A, full-length PIST containing the two coiled-coil domains (CC) and the PDZ domain binds to the cytoplasmic peptide segment of CALEB/NGC in the yeast two-hybrid system. Yeast colony growth was determined in tryptophan/leucine-deficient medium (WL; transformation control) or medium that additionally lacks histidine (WL/H) or histidine and adenine (WL/HA). The ranges of yeast colony numbers in the corresponding selection medium after 4 -7 days of incubation are shown. Data are obtained from one representative out of two to four independent experiments. Yeast colony growth indicating binding of PIST constructs and CALEB/NGC constructs could be detected in the presence of a weak (histidine synthesis; H) or even a strong (adenine synthesis; A) selection marker. No colony growth on these selective media could be detected in the negative controls with bait or prey vectors only (data not shown). Of all constructs examined, only fulllength PIST, the construct m11f2, and construct Pi-29 bind to CALEB/ NGC, which indicates that the integrity of the PIST protein seems to be important for binding to CALEB/NGC. B, as a bait for the yeast twohybrid screen, the cytoplasmic domain of the longer isoform of human CALEB/NGC (hCALEBb; the additional peptide is indicated by a black box) was used. However, not only this domain is able to bind to PIST, but also the cytoplasmic peptide segment of the shorter isoform of human CALEB/NGC (hCALEBa) and a construct that contains only the juxtamembrane cytoplasmic peptide segment of CALEB/NGC (CCmx). In contrast, two constructs that lack this peptide segment (CC1 and CCex), do not bind to PIST. TM, transmembrane domain; bg, background, indicating growth of very small (Ͻ0.5-mm) yeast colonies after a long (longer than 7-10-day) incubation time.
Taken together, we found that the intracellular peptide segment of CALEB/NGC, which is located near the transmembrane region and which is highly conserved between all species examined so far, binds to the coiled-coil and PDZ domains containing PIST protein in the yeast two-hybrid system.
PIST Associates with CALEB/NGC in HEK293 Cells and Can Be Precipitated by the CALEB-derived PIST-binding Peptide from Transfected HEK Cells and from Brain Tissue-To further analyze the interaction between CALEB/NGC and PIST, we co-transfected HEK293 cells with expression plasmids encoding tagged versions of CALEB/NGC and PIST, respectively. We prepared detergent extracts of these transfected cells and performed immunoprecipitations directed to the VSVtagged PIST. PIST was precipitated (Fig. 2B), and, in addition, the FLAG-tagged CALEB/NGC could be co-precipitated ( Fig.  2A, lane 2). PIST was not precipitated from HEK293 cells, which had been transfected only with a CALEB/NGC-encoding plasmid but not with plasmid coding for PIST (Fig. 2D); nor could CALEB/NGC be co-precipitated from these cells with anti-VSV antibodies (Fig. 2C, lane 2).
The results of the mapping analysis with the yeast twohybrid system suggest that the juxtamembrane cytoplasmic peptide segment of CALEB/NGC mediates the interaction with PIST. To test this further, we used an immobilized peptide containing the amino acid sequence of the juxtamembrane peptide segment of CALEB/NGC (FFAKKLYLLKTENT-KLRKT; see "Experimental Procedures") to perform affinity purification from detergent extracts of COS7 cells expressing VSV-tagged PIST. A short peptide with an amino acid sequence derived from the extracellular part of CALEB/NGC was used as a control. The result of this experiment shows that the VSVtagged PIST can be isolated from the detergent extracts with the juxtamembrane peptide segment of CALEB/NGC (Fig. 2E, lane 2) but not with the control peptide (Fig. 2E, lane 3). To analyze whether it could be possible to precipitate PIST with the CALEB-derived PIST-binding peptide directly from brain tissue, we prepared polyclonal antibodies directed to recombinant PIST. These antibodies recognize PIST in a detergent extract of adult rat brain (Fig. 2F, lane 1). After using the immobilized PIST-binding peptide for affinity purification, we could detect PIST in the elution fraction (Fig. 2F, lane 2). No PIST was detected when working with the control peptide (Fig.  2F, lane 3). Collectively, the biochemical analysis of the interaction between CALEB/NGC and PIST indicates that the juxtamembrane intracellular peptide segment of CALEB/NGC is sufficient for the interaction with PIST to occur.
CALEB/NGC Co-localizes with PIST in COS7 Cells-The data obtained by yeast two-hybrid analysis and co-immunoprecipitations support the view that CALEB/NGC interacts with PIST. We next wanted to examine whether and where CALEB/ NGC and PIST co-localize within cells. COS7 cells were cotransfected with plasmids encoding tagged versions of CALEB/ NGC and PIST, respectively. We used for these experiments four different CALEB/NGC constructs. In one case, the fulllength longer isoform of CALEB/NGC from mouse (designated mCALEBb in Fig. 3D) was co-expressed with PIST. PIST expression can be detected in intracellular structures surrounding the nucleus (Fig. 3B). When being co-expressed together with PIST, a significant portion of CALEB/NGC co-localizes with PIST near the nucleus in intracellular structures reminiscent of the Golgi apparatus (Fig. 3, A and C, arrowheads). Our biochemical data indicate that the juxtamembrane cytoplasmic peptide segment of CALEB/NGC is necessary and sufficient for binding PIST. One isoform of chicken CALEB/NGC (designated CALEBa-80 in Fig. 3H) has, in contrast to all other isoforms in all species known so far, a very short intracellular domain, which, nevertheless, contains the PIST-binding peptide segment. We expressed this isoform of CALEB/NGC together with PIST and found it to co-localize (Fig. 3G) in similar intracellular areas as described above. To demonstrate that the extracellular region of CALEB/NGC is not important for the interaction with PIST, we co-expressed PIST together with a construct of CALEB/NGC, which lacks the extracellular part but contains the transmembrane region and the intracellular domain (designated 449 in Fig. 3L). Once again a strong colocalization can be seen (Fig. 3K). However, when co-expressing PIST together with a CALEB/NGC-derived construct that comprises large parts of the intracellular domain but lacks the PIST-binding juxtamembrane peptide segment (designated CCEX in Fig. 3P), no co-localization can be detected (Fig. 3O). The results of these co-localization studies are in strong accordance with the data of the biochemical interaction assays.
CALEB/NGC Interacts with PIST in the Golgi Apparatus-CALEB/NGC and PIST co-localizes in COS7 cells in intracellular structures around the nucleus, which are similar or identical to the Golgi apparatus. To verify this, we co-expressed CALEB/NGC together with PIST and examined the co-localization of CALEB/NGC and the Golgi markers GM130 and Golgin-97 (34,35). As can be seen in Fig. 4, C and F, CALEB/ NGC strongly co-localizes with these Golgi markers when coexpressed with PIST. PIST is known to be expressed in many tissues (22, 23, 25) and many cell lines (24). We stained COS7

FIG. 2. PIST associates with CALEB/NGC in HEK293 cells and can be precipitated with the CALEB/NGC-derived juxtamembrane PIST-binding peptide.
HEK293 cells were co-transfected with plasmids encoding FLAG-tagged CALEB/NGC and VSV-tagged PIST. Detergent extracts were prepared (Sol, lane 1), and immunoprecipitations (IP, lane 2) with antibodies to the VSV tag were performed. A, when precipitating VSV-tagged PIST, FLAG-tagged CALEB/NGC is co-precipitated (arrowheads). B, control to show the precipitated VSVtagged PIST. C, if the HEK293 cells were only transfected with plasmids encoding FLAG-tagged CALEB/NGC and detergent extracts that contain FLAG-tagged CALEB/NGC were prepared (lane 1), no CALEB/ NGC is co-precipitated after an immunoprecipitation (lane 2) with antibodies to the VSV tag. D, control to show that no VSV-tagged PIST is precipitated in the experiment described in C. E, detergent extracts were prepared from COS7 cells that had been transfected with plasmids encoding VSV-tagged PIST (lane 1). Using an immobilized peptide that contains the sequence of the juxtamembrane cytoplasmic peptide segment of CALEB/NGC, PIST is precipitated from the detergent extracts (P, lane 2). No PIST is precipitated when a peptide is used, the sequence of which is derived from the extracellular region of CALEB/NGC (Co, lane 3). F, endogenous PIST from detergent extracts of adult rat brain (lane 1) can be precipitated with the CALEB/NGC-derived PIST-binding peptide (P, lane 2) but not with the control peptide (lane 3). The asterisks in A-D indicate the heavy chains of the antibodies used for immunoprecipitations. cells for endogenously expressed PIST with an affinity-purified polyclonal antibody and found it to be localized near the nucleus in a region that is similar or identical to the Golgi apparatus (Fig. 4H). After expression of FLAG-tagged CALEB/NGC (Fig. 4G), a co-localization of CALEB/NGC with the endogenous PIST can be observed in or near the Golgi apparatus (Fig. 4I).
It is well known that the integrity of the Golgi apparatus depends on the microtubule network (36,37). Therefore, after destroying the microtubule network, the Golgi apparatus disperses into small Golgi-derived vesicles. We examined the interaction between CALEB/NGC and PIST after disruption of the Golgi apparatus due to depolymerization of the microtubules by nocodazole. We found that these two proteins colocalize in vesicular structures even after the destruction of the Golgi apparatus (Fig. 5C). A similar co-localization of CALEB/ NGC and endogenously expressed PIST in Golgi-derived vesicles after destruction of the Golgi apparatus by nocodazole can be observed (data not shown). To substantiate that these vesicular structures indeed are Golgi-derived vesicles, we performed a double staining of CALEB/NGC and the Golgi marker Golgin-97 after disruption of the Golgi apparatus by nocodazole. A strong co-localization of CALEB/NGC with Golgin-97 can be detected (Fig. 5F). In addition to nocodazole, we used another drug, brefeldin A, to reversibly destroy the Golgi apparatus (38,39). As was seen with nocodazole, CALEB/NGC co-localizes with the Golgi marker GM130 in Golgi-derived vesicles after dispersal of the Golgi apparatus (Fig. 5I). In summary, our results support the view that CALEB/NGC interacts with PIST in the Golgi apparatus and in Golgi-derived vesicles after destruction of the Golgi apparatus with nocodazole or brefeldin A.
CALEB/NGC Associates with PIST in Primary Hippocampal Cells-CALEB/NGC is a transmembrane protein, the expression of which had been postulated to be restricted to the brain (2). PIST, on the other hand, was demonstrated to be expressed in many cell types and tissues (22)(23)(24). In addition, it was published that PIST regulates the transport of several transmembrane proteins to the plasma membrane. We had shown that PIST and CALEB/NGC co-localize in the Golgi apparatus and in Golgi-derived vesicles of co-transfected heterologous cells. The Golgi apparatus is of very high importance for the maturation and the transport of proteins to the plasma membrane. We were interested in whether PIST could be involved in the intracellular transport of CALEB/NGC in neural cells. To explore this, we first examined whether CALEB/NGC associates with PIST in primary cells derived from brain. We established a primary hippocampal cell culture and co-transfected these cells with tagged versions of CALEB/NGC and PIST. The cells were fixed and analyzed by indirect immunofluorescence 24 -48 h after transfection. We found a strong co-localization of CALEB/NGC with PIST in large (Fig. 6C, small arrowheads) or small (Fig. 6F, small arrowheads) vesiclelike or granule-like structures in the main processes of hippocampal neurons. The images in Fig. 6, A-C, give the impression of a street of vesicle-like structures that begins in the cell body and extends into the main neuronal processes. Next we examined hippocampal cells that had been transfected only with plasmids encoding CALEB/NGC. In this case we analyzed the cells 5 h after transfection and found CALEB/NGC to be located at the poles of the neurons and in the large diameter segments of the neurites emanating from the poles of the cells (Fig. 6G). When looking for endogenously expressed PIST, we found it to be present in the soma and the neurites (Fig. 6H). A co-localization can be detected at the poles of the cells and in the large diameter segments of the neurites (Fig. 6I). To examine the endogenous expression of both CALEB/NGC and PIST in hippocampal cells, we used antibodies directed to the cytoplasmic peptide segment of CALEB/NGC and to recombinant PIST (a gift from Wolf Wente and Hansjü rgen Kreienkamp). To our surprise, we could not detect a significant co-localization of CALEB/NGC and PIST in many primary hippocampal cells at different times of development (data not shown). However, when analyzing hippocampal cells in a very early stage of differentiation (5 h in vitro), a few cells showed a strong colocalization of CALEB/NGC and PIST at one pole of the cells and in the major neurite (Fig. 6L). Taken together, in primary FIG. 3. CALEB/NGC co-localizes with PIST in COS7 cells. COS7 cells were cotransfected with plasmids encoding FLAGtagged constructs of CALEB/NGC and VSVtagged PIST. The CALEB/NGC-derived constructs were visualized by indirect immunofluorescence with antibodies directed to the FLAG epitope and Alexa488-conjugated secondary antibodies (A, E, I, and M). Indirect PIST staining was performed with antibodies to the VSV tag and Cy3-conjugated secondary antibodies (B, F, J, N). The overlays of the CALEB/NGC and PIST stainings are shown in C, G, K, and O. Schematic drawings of the CALEB/NGC-derived constructs, which were co-expressed together with full-length PIST, are depicted in D, H, L, and P. PIST strongly co-localizes with the longer isoform of mouse CALEB/ NGC (denoted mCALEBb in D), with the shorter isoform of chicken CALEB/NGC (denoted CALEBa-80 in H), and with the CALEB/NGC-derived construct 449 (denoted 449 in L), which contains only the transmembrane region and the cytoplasmic domain of mouse CALEB/NGC. The arrowheads in C, G, and K indicate these colocalizations. No co-localization can be detected (O) when PIST is co-expressed together with the CALEB/NGC-derived construct CCEX (denoted CCEX in P), which contains most of the intracellular part of mouse CALEB/NGC but lacks the juxtamembrane peptide segment. Bar, 20 m.
hippocampal cell culture, a co-localization that implicates an association of CALEB/NGC and PIST can be detected in native as well as transfected cells at the poles of the cells and in the primary processes.
CALEB/NGC Co-localizes with PIST in the Somata of Purkinje Cells of Colchicine-treated Rats-Our data concerning the interaction between CALEB/NGC and PIST in primary hippocampal cells suggest that PIST might play a role in the maturation of CALEB/NGC or in the transport of CALEB/NGC from the cell soma to the primary neurites. CALEB/NGC had been shown to be strongly expressed in axon and dendritecontaining layers in the brain (1,4,40). In contrast, many layers containing cell bodies are only weakly decorated by CALEB/NGC staining. For example, CALEB/NGC is present in the retina predominantly in the plexiform layers and in the optic fiber layer but only in very minor amounts in the nuclear layers and in the ganglion cell layer, where most of the cell bodies are located. Therefore, it seems to be that CALEB/NGC must be transported from the cell bodies into the processes. We wanted to analyze this aspect by using an in vivo model to block microtubule-dependent transport. Adult rats were treated with colchicine, a drug that binds to tubulin subunits and prevents the polymerization of microtubules (33,37). These rats were analyzed in comparison with control rats that did not receive any drug. The focus was on the cell bodies of Purkinje cells in the cerebellum. As can be seen in a control section of an adult rat cerebellum, CALEB/NGC is expressed in the molecular layer (ML) and in the granular cell layer (GCL) but is absent from the cell bodies of the Purkinje cells in the Purkinje cell layer (Fig. 7D, white arrows). On the other hand, PIST is strongly expressed not only in the molecular layer and in the granular cell layer but also in the Purkinje cell layer (Fig. 7E, white arrows). It is obvious that no co-localization can be observed in the Purkinje cell layer (Fig. 7F, white arrows). However, in a cerebellar section of an adult rat that had been treated with colchicine, a clear staining of the Purkinje cell layer by an antibody directed to CALEB/NGC can be detected (Fig. 7A, white arrows). In a higher magnification, it is recognizable that CALEB/NGC is located in vesicle-like or granulelike structures in the cell bodies of the Purkinje cells (Fig. 7A,  inset). As can be seen in the control, PIST is also expressed in the Purkinje cell layer (Fig. 7B, white arrows). Interestingly, in the somata of Purkinje cells of colchicine-treated rats, PIST is concentrated in the same vesicle-like structures as CALEB/ NGC (Fig. 7B, inset). In these vesicle-like structures, a strong co-localization of CALEB/NGC and PIST can be observed (Fig.  7C, inset).
In summary, these data show that CALEB/NGC and PIST do not associate in the somata of Purkinje cells of adult rat cerebellum, but both proteins do co-localize in vesicular-like structures in the Purkinje cells after in vivo inhibition of microtubuledependent transport processes by colchicine. DISCUSSION CALEB/NGC is a neural transmembrane protein with high structural similarity to members of the EGF family of transmembrane growth and differentiation factors. It is expressed in the nervous system, in particular in axon and synapse-rich regions. It is striking that the cell bodies of neurons in several layers of the brain are only weakly decorated by antibodies directed to CALEB/NGC or not at all, although the processes of these cells contain large amounts of this protein. This implies that CALEB/NGC must be transported with high efficiency to the processes of neural cells and must be prevented from returning to the somata. Little is known about the intracellular transport of members of the EGF family of transmembrane growth and differentiation factors, but a few studies had been presented. One of them reported that the PDZ protein TACIP18/syntenin-1 is involved in the targeting of pro-TGF-␣ to the cell surface. The interaction of pro-TGF-␣ with p59/ GRASP55 was also suggested to play a role in the maturation and transport of this member of the EGF family. In both cases, the carboxyl terminus of pro-TGF-␣ was demonstrated to be important for the binding to the PDZ domains of TACIP18/ syntenin-1 or p59/GRASP55 and for the targeting to the cell surface, too.
We were interested in proteins, which may be necessary for the targeting of CALEB/NGC to the plasma membrane and for the sorting of this molecule to axons and dendrites of neurons. A yeast two-hybrid screen was performed in order to identify intracellular interaction partners of CALEB/NGC presumably involved in these processes. We found the PDZ domain containing protein PIST, also called FIG, GOPC, or CAL, to interact with a cytoplasmic peptide segment of CALEB/NGC in the yeast two-hybrid system. We showed that the integrity of the whole PIST protein is important for the interaction with CALEB/NGC. This is in contrast to several other PIST-binding transmembrane proteins as frizzled 5, frizzled 8, the CFTR, the chloride channel ClC-3B, and the ␦2 glutamate receptor. All of these proteins bind via their carboxyl-terminal PDZ interaction motifs to the PDZ domain of PIST. We could show that not the carboxyl-terminal but the juxtamembrane cytoplasmic peptide segment of CALEB/NGC is necessary and sufficient for binding PIST. This was demonstrated in the yeast two-hybrid system and was confirmed by affinity precipitation of PIST from transfected COS7 cells with the CALEB/NGC-derived PIST-binding peptide. Moreover, with this peptide it was possible to precipitate native PIST from detergent extracts of rat brain. The interaction between CALEB/NGC and PIST could also be corroborated by co-immunoprecipitations from transfected HEK293 cells. So far, it was not possible to perform co-immunoprecipitations from brain tissue. This may be due to the fact that the interaction between CALEB/NGC and PIST seems to be important for neural cells only in a narrow time window as explained below.
In heterologous cells expressing both CALEB/NGC and PIST, a strong co-localization of these two proteins were found in the Golgi apparatus and in Golgi-derived vesicles after brefeldin A or nocodazole treatment. This co-localization could also be detected when using different CALEB/NGC isoforms or other constructs that contain the juxtamembrane cytoplasmic peptide segment of CALEB/NGC but not with CALEB/NGCderived constructs, which lack the CALEB/NGC-derived PISTbinding peptide. The other PIST-binding transmembrane proteins frizzled 5, frizzled 8, CFTR and ClC-3B were also shown to interact with PIST in the Golgi apparatus or in Golgi-derived vesicles. However, the consequences of these interactions were described as different for these proteins. Whereas PIST was described as favoring the retention of CFTR and ClC-3B within the cell (25,26), which means to inhibit the transport to the plasma membrane, the same protein was suggested to support the transport of frizzled 5 and frizzled 8 to the plasma membrane (24). Yet another study claims that a neuronal isoform of PIST binds to the ␦2 glutamate receptor and to Beclin 1 and synergizes with the latter protein to induce autophagy (27). The functional consequences of the interaction between PIST and the small GTPase TC10 are currently not known (22). The results of these studies suggest that PIST may fulfill different cellular functions depending on the molecular context, but most or even all of these functions are connected with the Golgi apparatus. This interpretation is also supported by the analysis of PIST/GOPC-deficient mice that show a lack of acrosome formation, which is dependent on the fusion of Golgi-derived vesicles (41). Preliminary data from our laboratory do not support a role of PIST in the general transport of CALEB/NGC to the plasma membrane in heterologous cells. 2 However, we could demonstrate that CALEB/NGC strongly co-localizes with PIST in co-transfected primary hippocampal cells in vesicularlike or granule-like structures, which extend into the main neurites. When only overexpressing CALEB/NGC in primary hippocampal cells, we could detect a strong co-localization of the exogenously applied CALEB/NGC with endogenously expressed PIST at the poles of neurons and in the initial segments of main processes. This co-localization pattern can be seen, however, only in early phases of differentiation in culture. The same is true for the co-localization of endogenous CALEB/ NGC with endogenous PIST. In this case, both proteins are found to associate at the poles of neural cells and in the main processes too. These findings raise the possibility that PIST may play a role in the transport of CALEB/NGC from the cell bodies to the neurites and may contribute to the exclusion of CALEB/NGC from the somata of several classes of neurons during later stages of development, when CALEB/NGC is almost exclusively localized to axons and dendrites. If this is true, then one will expect a strong co-localization of CALEB/ NGC and PIST in the somata of neurons after disturbing the intracellular transport into axons and dendrites. We examined this in an in vivo model system for destroying microtubule-dependent transport processes in neurons. We found, indeed, a strong co-localization of CALEB/NGC and PIST in vesicle-like structures in the somata of Purkinje cells of colchicine-treated rats. Compared with colchicine-treated rats, CALEB/NGC can be hardly detected in the cell bodies of Purkinje cells of untreated animals.
In summary, we report an interaction of the neural EGF family member CALEB/NGC with the Golgi-associated protein PIST. The juxtamembrane cytoplasmic peptide segment of CALEB/NGC mediates this interaction. Our data obtained from studies with heterologous cells, with primary hippocampal cells, and with colchicine-treated rats are in line with the interpretation that PIST may be involved in the transport of CALEB/NGC to the processes of neurons. However, currently we cannot exclude the possibility that the interaction of CALEB/NGC with PIST could also be necessary for other, most likely Golgi-dependent, cellular processes. Future analysis of the localization of CALEB/NGC in neural tissue of PIST/ GOPC-deficient mice may help to clarify this issue.