STOP-like Protein 21 Is a Novel Member of the STOP Family, Revealing a Golgi Localization of STOP Proteins*

Neuronal microtubules are stabilized by two calmodulin-regulated microtubule-associated proteins, E-STOP and N-STOP, which when suppressed in mice induce severe synaptic and behavioral deficits. Here we show that mature neurons also contain a 21-kDa STOP-like protein, SL21, which shares calmodulin-binding and microtubule-stabilizing homology domains with STOP proteins. Accordingly, in different biochemical or cellular assays, SL21 has calmodulin binding and microtubule stabilizing activity. However, in cultured hippocampal neurons, SL21 antibodies principally stain the somatic Golgi and punctate Golgi material in neurites. In cycling cells, transfected SL21 decorates microtubules when expressed at high levels but is otherwise principally visible at the Golgi. The Golgi targeting of SL21 depends on the presence of cysteine residues located within the SL21 N-terminal domain, suggesting that Golgi targeting may require SL21 palmitoylation. Accordingly we find that SL21 is palmitoylated in vivo. N-STOP and E-STOP, which contain the Golgi targeting sequences present in SL21, also display distinct Golgi staining when expressed at low level in cycling cells. Thus neuronal proteins of the STOP family have the capacity to associate with Golgi material, which could be important for STOP synaptic functions.

Neurons contain abundant subpopulations of stable microtubules that resist depolymerizing conditions such as exposure to the cold. This property is due to microtubule association with E-and N-STOP 2 , two neuronal calmodulinbinding and calmodulin-regulated proteins (1)(2)(3). STOP proteins are important for synaptic function and STOP null mice present defects in both short-and long-term synaptic plasticity, associated with severe behavioral and neurotrans-mitter deficits (1,4). At the ultrastructural level, the main signature of STOP suppression is a dramatic depletion of synaptic vesicle pools in glutamatergic neurons.
E-and N-STOP arise from mRNA splicing of a single gene (5). These STOP proteins contain modular motifs (3) that are either bifunctional (calmodulin-binding and microtubule-stabilizing sequences, Mc and Mn modules, respectively) or are comprised of calmodulin-binding sequences, unrelated to microtubule-stabilizing sequences.
We have previously described a number of STOP splicing variants, which are principally present in non-neuronal cells (6), but for a long time, STOPs seemed to be unique with no STOP-related proteins. However, strong evidence for a 21-kDa STOP-like protein (SL21) has arisen from data base searches (3). SL21 contains a 35-amino acid (aa) N-terminal stretch that shares 83% homology with the N-terminal part of E-or N-STOP and contains a calmodulin-binding motif. A second domain of SL21, spanning 24 aa, shares 71% homology with a microtubule-stabilizing Mn module of E-and N-STOP (3,7).
Here, we demonstrate that SL21 is a neuronal protein, strictly expressed in the postnatal brain. We show that SL21 has microtubule stabilizing activity similar to E-and N-STOP. However, surprisingly, endogenous SL21 co-localizes with the somatic Golgi apparatus and with dendritic Golgi structures in cultured hippocampal neurons. We show that the N-terminal part of SL21 comprises a Golgi-targeting sequence with critical cysteine residues, which sustain SL21 palmitoylation. The SL21 Golgi targeting sequence is also present and functional in E-and N-STOP. Thus, in light of studies indicating that the presence of Golgi material in neurites and synapses may be important for synaptic plasticity (8), Golgi interactions with STOP could be important for STOP function in synapses.

EXPERIMENTAL PROCEDURES
Sequence Analysis-Mouse N-STOP protein sequence (GenBank TM accession number CAA75930) was submitted to the BLAST program to identify STOP-related proteins. GenBank TM accession numbers of these proteins are CAA63762 for rat N-STOP, NP_149052 for human N-STOP, XP_508647 for chimpanzee N-STOP, XP_596995 for cow N-STOP, NP_941001 for mouse SL21, XP_344040 for rat SL21, AAH06434 for human SL21, XP_516902 for chimpanzee SL21, and XP_608343 for cow SL21. Protein sequence alignments were performed using the multiple alignment software ClustalW (9).
Plasmid Constructs-On the basis of the mouse SL21 cDNA sequence (GenBank TM accession number BY727771), the entire open reading frame of mouse SL21 was amplified by PCR. We used mouse brain Marathon Ready cDNA (BD Biosciences) and the Advantage-GC2 polymerase mix (BD Biosciences) with primers containing BglII extensions. The resulting PCR product was first cloned into pCR2.1-TOPO (Invitrogen). For expression in mammalian cells, the complete coding sequence was cloned in pSG5 vector (Stratagene) and in pcDNA3.1(Ϫ)/Myc-His-A vector (Invitrogen) to be fused with the myc epitope DNA. For production and purification in the bacterial system, the complete coding sequence was cloned into pGex-4T3 vector (Amersham Biosciences) to fuse GST after the last amino acids of SL21. The two SL21 deletion mutants, obtained by PCR, correspond to Met ϩ aa 35-191 (SL21⌬2-34) and to aa 1-122 ϩ aa 146 -191 (SL21⌬Mn3) according to the numbering of mouse SL21 protein. The replacement of Cys 5 , Cys 10 , and Cys 11 by glycine residues (SL21-C(5/10/11)G) was obtained by PCR, with a degenerate 5Ј-oligonucleotide (5Ј-AgA TCT ATg gCg Tgg CCC ggC ATC AgC Cgg CTA ggC ggC CTg gCC-3Ј). The STOP coding sequence of LNt⌬Mn1Mn2, corresponding to aa 1-123 ϩ Ala ϩ aa 139 -161 ϩ Ile-Gln ϩ aa 175-225 according to the numbering of rat N-STOP, was amplified by PCR and then subcloned into the pcDNA3.1(Ϫ)/Myc-His-A vector to be fused with the myc epitope DNA. N-STOP plasmid was described previously (2, 10). The STOP coding sequence of N-STOP⌬2-19, corresponding to Met ϩ aa 20 -952 according to the numbering of rat N-STOP, was amplified by PCR and then subcloned into the pSG5 vector. All of the SL21 and STOP constructs were confirmed by DNA sequencing.
Production of SL21-GST-Plasmid pGex-4T3 containing SL21 cDNA was transformed in One Shot BL21 (DEA3) Star cells (Invitrogen) according to the manufacturer's instructions. Transformed cells expressing SL21-GST were grown at 37°C to an A 600 of 0.6, and induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h at 37°C. Cells were pelleted, resuspended in PBS containing 1% Triton X-100, protease inhibitors (Complete Mixture tablets, Roche Applied Science), and sonicated for 1.5 min before centrifugation (17,000 ϫ g for 10 min). The supernatant was incubated with glutathione-Sepharose beads. These beads were placed in columns and successively washed with PBS containing 1% Triton X-100, then with PBS, and finally with 50 mM Tris, pH 8.0. The GST fusion proteins were eluted with 10 mM glutathione in 50 mM Tris, pH 8.0.
Western Blot Analysis-Tissues from mice were homogenized in a 100 mM K-Pipes, 1 mM EGTA, and 1 mM MgCl 2 buffer, pH 6.65. The homogenates were centrifuged at 200,000 ϫ g for 30 min at 4°C. Supernatants were then harvested and analyzed by SDS-PAGE (11). For immunoblotting, proteins were subsequently transferred onto nitrocellulose membranes and processed as described previously (12).
Sedimentation of Proteins with Microtubules-All proteins were preclarified at 150,000 ϫ g for 15 min in a TL-100 Ultracentrifuge at 4°C prior to the start of experiments. STOP proteins were purified as described in Pirollet et al. (13). A microtubule-binding assay was performed as in Masson and Kreis (14) using taxol-stabilized microtubules (4 M) as substrates and a 60% glycerol cushion.
Assay of Calmodulin Binding to Immobilized Peptide-Immobilized peptide arrays corresponding to the human SL21 were produced according to Frank (15) with an ABIMED ASP 222 automated SPOT robot. The protein sequence was subdivided into 15-mer peptides with an overlap of 12 aa. The peptides were synthesized spotwise as an array on a specially manufactured cellulose membrane (AIMS Scientific Products GmbH, Braunschweig, Germany). The peptides are permanently linked to the membrane through their carboxyl terminal aa via a polyethylene glycol spacer. The synthesis procedure followed exactly that of Frank and Overwin (16). Overlay binding assay were performed as described elsewhere (3), using 35 Slabeled calmodulin. Cell Culture-Hippocampal cell cultures were prepared as described previously (17). Briefly, hippocampi brain tissue from E18.5 (embryonic day 18.5) mice were removed and digested in 0.25% trypsin in Hepes-buffered Hanks' balanced salt solution at 37°C for 15 min. After manual dissociation, cells were plated at a concentration of 5,000 -15,000 cells/cm 2 on poly-L-lysine (Sigma)-coated coverslips, in DMEM-10% fetal bovine serum. One hour after plating, the medium was changed to DMEM containing B27 and N2 supplements (Invitrogen). HeLa and NIH 3T3 cells were cultured in RPMI-Glutamax or DMEM-Glutamax supplement, respectively, with 10% fetal bovine serum and 1% penicillin/streptomycin (Invitrogen).
Transient Transfection and Analysis of Microtubule Stability-Exponentially growing HeLa or NIH 3T3 cells were transfected with the cDNAs described above using FuGENE 6 (Roche Diagnostics) according to the manufacturer's instructions. Transfected cells were either directly processed for immunofluorescence or exposed to cold temperatures (30 min on ice) or to nocodazole (20 M for 30 min). Cells were then permeabilized in lysis buffer (30 mM Pipes, 1 mM EGTA, 1 mM MgCl 2 , 10% glycerol, 0.1% Triton X-100, pH 6.75) for 1 min and processed for immunofluorescence.
Immunofluorescence-Hippocampal neurons were then either permeabilized in OPT buffer (80 mM Pipes, pH 6.7, 1 mM EGTA, 1 mM MgCl 2 , 0.5% Triton X-100, 10% glycerol) at 35°C for 2 min before fixation or were fixed directly for 10 min in methanol at Ϫ20°C. Cells were then incubated with primary antibodies for 45 min in PBS-Tween 0.5% and with secondary antibodies for 40 min. Cells were analyzed with a laser confocal microscope (TCS-SP2, Leica) or with an inverted microscope Axioscop 50 (Zeiss) controlled by Metaview (Universal Imaging, Downingtown, PA). Images were digitalized using a Coolsnap ES camera (Roper Scientific).
[ 3 H]Palmitate Labeling and Immunoprecipitation-HeLa cells were either nontransfected or transfected with SL21-myc or SL21 C(5/10/11)G-myc cDNAs. One day after transfection, cells were labeled for 3 h with 500 Ci of [ 3 H]palmitic acid (Amersham Biosciences). The radioactive palmitic acid was completely dried under nitrogen, resuspended in 12.5 l of Me 2 SO, and complemented with 1 ml of RPMI containing 0,5% defatted bovine serum albumin (Sigma). Cells were washed and scraped with PBS. After centrifugation at 12,000 ϫ g for 2 min, cells were lysed in IP buffer (50 mM Tris, 150 mM NaCl, 0,05% deoxycholate, 1% Triton X-100, 10% glycerol, pH 8.0) in the presence of protease inhibitors (Complete Mixture tablets, Roche Applied Science) and 5 mM CaCl 2 . After centrifugation of the cell lysate at 12,000 ϫ g for 5 min at 4°C, the supernatant was supplement with 5 mM EGTA and incubated for 2 h at 4°C FIGURE 2. Calmodulin-binding properties of SL21. SL21 calmodulinbinding motifs are identified using immobilized peptide array and calmodulin-agarose binding of SL21. A, 35 S-labeled calmodulin overlay of a membrane containing SL21 immobilized peptide array. The overlapping peptides (15-mers, with an overlap of 12 aa) were numbered from the amino-terminal to the carboxyl-terminal residues of SL21. Numbers correspond to the first peptide of each line. Four peptide clusters interacting with 35 S-labeled calmodulin were detected by autoradiography. B, quantitative analysis of [ 35 S]calmodulin binding to mouse SL21 peptides. The radioactive signals observed in A were quantified using a phosphorimaging device, and results were plotted. Signal values are in arbitrary units (A.U.). The four peaks of maximum radioactivity signal correspond to SL21 aa 6 -20 (peak 1), aa 87-101 (peak 2), aa 141-155 (peak 3), and aa 177-191 (peak 4). C, calmodulin-binding sites SLCam1 to SLCam4, corresponding to peaks 1-4 in B, are reported on the schematic representation of SL21 (asterisks), and their sequences are given. D, binding of SL21-GST to immobilized calmodulin-agarose. Recombinant SL21-GST was produced in bacteria and loaded onto a calmodulin-agarose column in the presence (left panel) or absence (right panel) of 1 mM Ca 2ϩ . The column was washed, and proteins were subsequently eluted with EGTA-containing buffer. Equal aliquots of the loaded sample (L), flow-through fraction (F ), and EGTA elution fractions 1-5 were analyzed for SL21-GST content using SDS-PAGE followed by immunoblot analysis with SL21 polyclonal antibody 3315. with 25 l of myc antibody (Santa Cruz Biotechnology) preincubated for 2 h at 4°C with 30 l of protein G-Sepharose (Amersham Biosciences). After centrifugation at 12,000 ϫ g for 2 min, the immunoprecipitates were washed five times in IP buffer for 10 min at 4°C. Samples were separated by SDS-PAGE, and gels were fixed for 30 min in 50% methanol-10% acetic acid and then treated for 30 min with Amplify solution (Amersham Biosciences) to improve the radioactive signal. The gels were dried and exposed on autoradiographic films for 10 days at Ϫ80°C.

RESULTS
STOP Homology Domains in SL21-E-and N-STOP contains two classes of bifunctional calmodulin-binding (Mc) and -stabilizing modules (Mn) (Fig. 1A). E-and N-STOP also contain additional calmodulin-binding sequences unrelated to Mn or Mc modules (Fig. 1A). The first 35 aa of SL21 share 83% homology with the N-terminal aa of E-and N-STOP, which comprises, in STOPs, the functional calmodulin-binding motif called Cam1 (3). A second domain of SL21, spanning 24 aa, shares 71% homology with the E-and N-STOP bifunctional calmodulin-binding (Cam5) and microtubule-stabilizing Mn3 module (3). The homologous domains of N-STOP and SL21 are aligned (Fig. 1B).
Calmodulin-binding Properties of SL21-All of the STOP calmodulin-binding sites including Cam1 and Cam5 were initially identified as calmodulin-binding motifs on N-STOP immobilized peptide arrays (3). A similar assay with SL21 peptides (Fig. 2, A and B) showed that the STOP Cam1 and Cam5 homologous motifs of SL21 (SL-Cam1 and SL-Cam3, respectively), also bound [ 35 S]calmodulin (Fig. 2, B and C, peaks 1 and 3) together with two other apparently weaker calmodulin-binding motifs, SL-Cam2 and SL-Cam4, peaks 2 and 4 respectively (Fig. 2, B and C).
Calmodulin-binding motifs in N-STOP have been shown to bind to immobilized Ca 2ϩ -calmodulin in vitro (3). Accordingly, four independents experiments consistently showed that recombinant SL21-GST bound to calmodulin column, as illus-    Fig. 2D. The binding was strictly dependent on the presence of calcium as shown by the release of SL21 from the calmodulin column by EGTA buffer (Fig. 2D) and by the nonbinding of SL21 to calmodulin column in the absence of calcium (Fig. 2D). The binding of SL21-GST to calmodulin is mediated by SL21, as GST alone was unable to bind to immobilized calmodulin (data not shown). Thus, SL21 behaves as a bona fide calmodulin-binding protein in vitro.
Microtubule Binding and Stabilizing Activity of SL21-In E-and N-STOP, Mn modules mediate STOP association with microtubules with resulting inhibition of microtubule depolymerization upon exposure to the cold or to nocodazole (3). In agreement with the presence of an Mn module sequence in SL21, SL21 co-sedimented with taxol-stabilized microtubules in standard microtubule binding assays while remaining in the soluble fraction in the absence of microtubules (Fig. 3A). In the same microtubule binding assay, N-STOP co-sedimented with microtubules, whereas GST remained in the supernatant. In quantitative experiments, various amount of tubulin were polymerized and then incubated with SL21-GST at 2 M. Then, polymerized tubulin was pelleted, and SL21 content was analyzed on immunoblots, in both supernatants and pellets (Fig. 3B). The concentrations of tubulin at which approximately half of SL21 cosedimented with microtubules was ϳ5 M (Fig. 3B). In additional experiments, SL21-GST at various concentrations was mixed with polymerized tubulin (10 M). Then, microtubules were pelleted, and SL21 content was analyzed on immunoblots in both supernatants and pellets (Fig. 3C). In this experiment, SL21 at 4 M was found both in the supernatant and in the pellet, whereas at concentration of SL21, ranging from 2 to 0.5 M, SL21 was only found associated with microtubules. Altogether these results showed that SL21 behaves as a microtubule-binding protein in vitro.
We used HeLa cells for microtubule-binding and -stabilization tests in vivo. HeLa cells are devoid of STOP, of SL21, and of cold-stable or nocodazole-resistant microtubules (Fig. 3D). When transfected in HeLa cells, a SL21 mutant (SL21⌬2-34) lacking the N-terminal domain and containing the Mn module decorated cytoplasmic microtubules and induced both microtubule cold stability and microtubule resistance to nocodazole (Fig. 3E), compatible with the functionality of the Mn module of SL21 in vivo. As shown below, the behavior of the full-length SL21 in cells was more complex than the mutant lacking the N-terminal domain (SL21⌬2-34) due to the interfering influence of the N-terminal sequence.
Tissue Distribution and Cellular Localization of Endogenous SL21-We used Western blot analysis of mouse tissues with SL21 polyclonal antibody 3315 to determine the distribution pattern of SL21. SL21 was only detected in brain tissue (Fig. 4A). In tissue extracts from newborn brain, SL21 expression was low at P0 and increased after P10 (Fig. 4A). We then tested the presence of SL21 in various neuronal cell lines, using both immunoblot analysis and immunofluorescence microscopy. SL21 was not detected in neuronal cell lines (N2A, PC12, and NG108) whether cultured in the presence or absence of nerve growth factor (data not shown). In contrast, SL21 was present in neurons in primary cultures but absent from glial cells (data not shown). Interestingly, in differentiating hippocampal cultured neurons, SL21 antibodies principally stained a juxtanuclear structure corresponding to the somatic Golgi apparatus, as demonstrated by co-localization with the cis-Golgi marker GS28 (Fig. 4B). At late stages of neuronal differentiation in vitro (16 DIV), in addition to the somatic Golgi, both SL21 and GS28 antibodies yielded a punctuated staining of neuritic extensions (Fig. 4B), corresponding to staining of neurite vesicular Golgi (19). A distinct microtubule staining was also visible after Triton-soluble protein extraction (Fig. 4D). Disruption of the somatic Golgi with brefeldin A treatment in differentiated neurons induced a dramatic redistribution of SL21 and GS28 stainings, both of which became diffuse (Fig. 4C). These results indicate that SL21 is specifically expressed in differentiated neuronal cells and that it localizes to microtubules, to the somatic Golgi, and to Golgi material traveling throughout neurites.
Dual Localization of Transfected SL21 in NIH 3T3 Cells Resulting from Microtubule-and Golgi-targeting Sequences in SL21-We used NIH 3T3 cells, which have a well organized somatic Golgi, and SL21 mutants for further analysis of SL21 localization and domain structure. When expressed in NIH 3T3 cells, the apparent localization of SL21 varied as a function of its expression level. At low levels of expression, SL21 was principally detectable on the Golgi (Fig. 5A). At higher levels of expression, there was also distinct SL21 staining of cytoplasmic microtubules (Fig. 5B). The deletion of the Mn module (SL21⌬Mn) suppressed the microtubule association of SL21 at all expression levels, while preserving Golgi localization (Fig. 5C). Conversely, at all levels of expression in NIH 3T3 cells, a deletion mutant of SL21 lacking 34 N-terminal amino acids (SL21⌬2-34) associated with microtubules (data not shown), as in the case of HeLa cells (Fig. 3F ), with no Golgi localization. In additional microtubule stabilization assays in HeLa cells, we found that native SL21 expressed at high levels and the SL21⌬2-34 deletion mutant had microtubule stabilizing activity, whereas the SL21 mutant SL21⌬Mn, which showed exclusive Golgi localization, had no detectable effect on microtubule stability (not shown). Altogether, these results indicate that the targeting of SL21 to microtubules depends on the pres-  Fig. 1A. E, incorporation of [ 3 H]palmitate into SL21-myc and SL21-C(5/10/11)G-myc mutant. HeLa cells overexpressing SL21-myc or SL21-C(5/10/11)G-myc mutant were incubated with [ 3 H]palmitate, and proteins were immunoprecipitated with anti-myc antibody. Immunoprecipitated proteins were separated by SDS-PAGE and subjected to autoradiography or Western blot using SL21 polyclonal antibody 3315. Untransfected cells were used as control.
ence of the Mn module and that SL21 comprises a Golgi-targeting sequence located in the N-terminal domain of the protein.
Palmitoylation of SL21-Palmitoylation is often required for protein association with membranes and has been observed previously in neuronal proteins associated with tubulin and with the Golgi, such as SCG10 (20). Palmitoylation occurs at cysteine residues surrounded with basic residues (21). Mouse SL21 contains three cysteine residues, all located in the Golgitargeting domain (positions 5, 10 and 11) and surrounded by basic residues (Fig. 6A). To assess whether these cysteines were involved in the localization of SL21 to the Golgi apparatus, we produced cDNAs encoding SL21 mutants in which Cys 5 , Cys 10 , and Cys 11 were replaced by glycine residues. These constructs were then transfected in NIH 3T3 cells, and the Golgi localization of the mutants was examined. In contrast to SL21, which localizes to the Golgi, the SL21 mutant where all the three cysteine residues were replaced (SL21-C(5/10/11)G), did not localize to the Golgi but co-localized with microtubules (Fig. 5D).
These results indicate a requirement of SL21 N-terminal Cys 5 , Cys 10 , and Cys 11 for proper localization of SL21 to Golgi apparatus. To test SL21 palmitoylation directly, HeLa cells were incubated with [ 3 H]palmitate with or without prior transfection with SL21-myc or SL21-C(5/10/11)G-myc cDNA. Then, SL21-myc and SL21-C(5/10/11)G-myc were immunoprecipitated from cell extracts, and the immunoprecipitated proteins were analyzed both by Western blot, using SL21 polyclonal antibody 3315, and by autoradiography (Fig. 5E). As shown in Fig. 5E, SL21 incorporated 3 H, whereas SL21-C(5/10/11)G mutant did not, demonstrating that SL21 is indeed palmitoylated in cells and showing the crucial role of Cys 5 , Cys 10 , and Cys 11 in this process. These results were obtained for three independent experiments, and similar results were observed when using SL21-GFP instead of SL21-myc (data not shown).
Golgi Targeting of N-STOP-The N-terminal Golgi-targeting sequence of SL21, which is conserved among mammals, is also present at the N terminus of N-STOP (Fig. 6A). Accordingly, a 225-aa N-terminal fragment of N-STOP containing the Golgitargeting sequence and deleted for microtubule-binding modules Mn1 and Mn2 (LNt⌬Mn1Mn2) was uniformly addressed to the Golgi when transfected in NIH 3T3 cells (Fig. 6B). Despite the presence of such a functional Golgi-targeting sequence at their N terminus, E-and N-STOP have not been detected at the Golgi in previous studies, which could be due either to a dominant influence of the STOP microtubule-targeting sequences (Mn and Mc modules) or to unfavorable experimental procedures. To test these possibilities, we reexamined N-STOP localization in various cell types and conditions. In neurons, STOP antibody yielded a bright staining of the whole cell body, making a possible STOP localization at the Golgi difficult to detect (2). Interestingly, in NIH 3T3 cells, transfected N-STOP staining was principally detectable at the Golgi at low levels of expression (Fig. 6C), with both Golgi and microtubule staining being present at high expression levels (Fig. 6D). Moreover, a transfected N-STOP mutant (N-STOP⌬2-19) lacking the Golgi targeting sequence failed to associate with the Golgi (Fig. 6E). Altogether, these results indicate that the ability to localize to Golgi material is a shared property of STOP and STOP-like proteins, because of the presence of a Golgi-targeting sequence in their shared N-terminal domain.

DISCUSSION
In this study, SL21 and neuronal STOPs (E-STOP and N-STOP) emerge as a class of microtubule-and calmodulinbinding proteins, sharing both a Golgi-targeting sequence and microtubule-stabilizing modules. We note that the STOP Golgi-targeting sequence is only present in E-and N-STOP and is absent from all of the other known STOP variants (6). As are E-and N-STOP, SL21 is specific to neurons, and both SL21 and N-STOP are expressed only in the postnatal brain. Despite such similarities in structure and activity, there is an apparent difference in the cellular distribution of endogenous SL21 and N-STOP in neurons, with SL21 preferentially associated with the somatic Golgi and N-STOP with microtubules. However both SL21 and N-STOP are detected principally at the Golgi when expressed at low levels in cells, whereas they are also visible on cytoplasmic microtubules at high levels of expression. The apparent differences between endogenous SL21 and N-STOP localization in neurons may thus reflect differences in expression levels, with N-STOP probably much more abundant than SL21. Superimposed cell regulations may also affect SL21 or STOP localization. The Golgi targeting of SL21 and of STOP most likely depends on palmitoylation, which is a regulated and reversible covalent protein modification (22,23). Also, the microtubule binding activity of Mn modules can be inhibited by calmodulin (24) and, in the case of STOP, through phosphorylation, opening the possibility of a protein shift from microtubules to other cell compartments (25).
Why should neurons contain specific microtubule-associated proteins with Golgi binding activity? The somatic Golgi in neurons has an organization similar to that observed in nonneuronal cells, although the Golgi orients toward the longest dendrite, and this Golgi polarity precedes the asymmetric dendrite growth (26). Additionally there is evidence for the presence of Golgi material along neurites. Previous ultrastructural studies have reported membranes analogous to Golgi cisternae within the spines of distal dendrites (27,28). Moreover many neurons possess both somatic Golgi and discrete, discontinuous Golgi-type structures ("Golgi outposts") located far into the dendrites. These Golgi outposts are mobiles structures and are positioned to serve particular dendrites regions or sets of syn- Among the strongly conserved amino acids, the cysteine residues 5, 10, and 11 are boxed in red. Numbers correspond to aa position. B, NIH 3T3 cells transfected with a cDNA encoding STOP fragment LNt⌬Mn1Mn2-myc and double-stained for Golgi (giantin polyclonal antibody) and LNt⌬Mn1Mn2 (myc mAb). C and D, NIH 3T3 cells transfected with a cDNA encoding N-STOP and expressing either low (C ) or high levels (D) of N-STOP. Cells were double-stained for STOP (polyclonal antibody 23C) and for either Golgi (GM130 mAb) or microtubule (␣-tubulin mAb). E, NIH 3T3 cells transfected with a cDNA encoding N-STOP⌬2-19 and double-stained for STOP (polyclonal antibody 23C) and microtubule (␣-tubulin mAb). Schematic representations of proteins are as in Fig. 1A. Bar, 20 m. apses, where they may be important for synaptic plasticity (8,19). The ability of Golgi binding may therefore be important for the function of N-STOP in synaptic plasticity (1). Additionally, palmitoylation may allow STOP interaction with vesicular or membrane material other than the Golgi (29) and may thereby be central to the dramatic effects of STOP suppression on the synaptic vesicle density (1). Moreover, palmitoylation has specifically been shown to play a role in neuronal protein trafficking and in the clustering of receptors and the associated scaffolding proteins at synapses (29,30).
According to the present study, SL21 and STOP have extensively overlapping activities. Yet, STOP suppression is not compensated by SL21 in STOP-deficient mice, and we have not detected any obvious modification in SL21 expression (data not shown). What is the utility of SL21 in the presence of STOP? We have performed SL21 small interfering RNA (siRNA) experiments in hippocampal neurons, with inconclusive results (data not shown). It is likely that, as for N-STOP, SL21 function will be revealed by its suppression in whole animals, where complex aspects of synaptic function can be investigated. STOP or SL21 null animals may offer exciting models in which to test the role of the Golgi in integrated brain functions.