A PDZ Protein Regulates the Distribution of the Transmembrane Semaphorin, M-SemF*

M-SemF is a membrane-associated, neurally enriched member of the semaphorin family of axon guidance signals. We considered whether the cytoplasmic domain of M-SemF might possess a signaling function and/or might control the distribution of M-SemF on the cell surface. We identify a PDZ-containing neural protein as an M-SemF cytoplasmic domain-associated protein (SEMCAP-1). SEMCAP-2 is a closely related nonneuronal protein. SEMCAP-1 has recently also been identified as GIPC, by virtue of its interaction with the RGS protein GAIP in vitro (De Vries, L., Lou, X., Zhao, G., Zheng, B., and Farquhar, M. G. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 12340–12345). Expression studies support the notion that SEMCAP-1(GIPC) interacts with M-SemF, but not GAIP, in brain. Lung SEMCAP-2 and SEMCAP-1(GIPC) are potential partners for both GAIP and M-SemF. The protein interaction requires the single PDZ domain of SEMCAP-1(GIPC) and the carboxyl-terminal four residues of M-SemF, ESSV. While SEMCAP-1(GIPC) also interacts with SemC, it does not interact with other proteins containing a class I PDZ binding motif, nor does M-SemF interact with other class I PDZ proteins. Co-expression of SEMCAP-1(GIPC) induces the redistribution of dispersed M-SemF into detergent-resistant aggregates in HEK293 cells. Thus, SEMCAP-1(GIPC) appears to regulate the subcellular distribution of M-SemF in brain, and SEMCAPs could link M-SemF to G protein signal transduction pathways.

The creation and refinement of highly specific neuronal connections is essential for the function of vertebrate nervous systems. The molecular mechanisms guiding axonal extension and synapse formation are becoming better understood. Pioneering studies demonstrated the general importance of neurotrophic factors and cell adhesion molecules in supporting neurite outgrowth (1,2). More recent studies have defined additional pathway-specific roles for the netrins, ephrins, and semaphorins (3).
Semaphorins are an expanding family of glycoproteins expressed in the nervous system and in many somatic tissues. There are more than 20 semaphorins reported at the present time (4). While they may have diverse functions ranging from vasculogenesis (5,6) to neoplastic transformation (7) to immune modulation (8), their role in axon repulsion is best documented. Members of the semaphorin family feature an extracellular Sema domain of 500 amino acid residues with multiple conserved cysteine residues (9). Collapsin-1 (semaphorin III/ D), a secreted semaphorin, was initially purified from chick brain as a potent inducer of dorsal root ganglion neuronal growth cone collapse (10). SemIII/D binds to neuropilin-1 on the axonal growth cone surface (11,12) and inhibits growth cone motility through a signal transduction cascade, which may involve heterotrimeric G proteins, collapsin response mediator proteins, and Rac1 (13)(14)(15)(16). Analyses of mice with targeted SemIII(D) gene deletions have verified the role of the protein as a repulsive signal for selected sensory neurons in the peripheral nervous system (5,17).
Phylogenetic analysis has divided semaphorins into at least six groups. Group II and group III are secreted semaphorins. Group VI semaphorins associate with plasma membrane via a GPI linkage (18). Groups I, IV, and V, which represent about half of known semaphorins, contain transmembrane and cytoplasmic domains. Since they are integral membrane proteins, these transmembrane semaphorins are thought to provide contact-mediated guidance cues to developing neurons. There is only limited evidence concerning specific physiologic roles for transmembrane semaphorins. Antibody blockade of G-Sema I, a grasshopper transmembrane semaphorin, interferes with Ti1 neuron axonal pathfinding in the limb bud (19). Genetic analysis of its Drosophila homologue, D-Sema I, suggests that it may also function as a repulsive signal for selected motor and central nervous system axons (20).
Although expression of a secreted derivative of D-Sema I with a truncated tail partially complements a loss of function mutation of D-Sema I (20), several recent reports suggest that the cytoplasmic domains of transmembrane semaphorins may contribute to their biological functions. M-SemaG (h-CD100), which regulates B lymphocyte differentiation, associates with a serine kinase through its cytoplasmic domain (21). Mouse Se-maVIb forms a complex with c-Src through a proline-rich intracellular domain (22). These data raise the possibility that the transmembrane semaphorins may function as receptors, as well as ligands, and transduce extracellular binding events to cytoplasmic changes. Alternatively, the cytoplasmic domains may control the spatial distribution and hence ligand activity of transmembrane semaphorins. The activity of secreted collapsin (SemIII/D) requires dimerization and proteolytic activation (23,24). The clustering of ephrins due to membrane anchorage is critical for the activation of Eph receptors and their biological activities (25).
M-SemF is a transmembrane semaphorin of unknown function identified by its homology in the Sema domain (26). The predicted cytoplasmic portion of M-SemF, 146 amino acids in length, contains a proline-rich region that includes the consen-sus sequence for interaction with Src homology 3 domains as well as cytoskeletal proteins. In this report, we demonstrate that the cytoplasmic domain of M-SemF associates specifically with two related intracellular proteins containing single PDZ domains. We isolated these proteins as M-SemF cytoplasmic domain-associated proteins (SEMCAP-1 and SEMCAP-2). 1 SEMCAP-1 is identical to GIPC, a protein interacting with the RGS protein GAIP (27). Thus, SEMCAP(GIPC) may regulate the distribution of M-SemF and link a semaphorin to G protein signaling cascades.

EXPERIMENTAL PROCEDURES
The MATCHMAKER two-hybrid assay system and mouse E17 cDNA pGAD10 library were purchased from CLONTECH Laboratories, Inc. (Palo Alto, CA). General chemicals of molecular biology grade were obtained from Sigma, J. T. Baker Inc., or American Bioanalytical (Natick, MA). Reagents for tissue culture were from Life Technologies, Inc. Reagents used in yeast and bacteria culture were from Difco. Restriction enzymes were obtained from New England Biolabs (Beverly, MA), and thermophilic DNA polymerase (Taq) was from Promega (Madison, WI). 9E10 ascites (M-5546), anti-bIII tubulin monoclonal antibody (T-8660), fluorescein isothiocyanate-conjugated IgG fraction of goat anti-rabbit IgG (F-6005), and tetramethylrhodamine isothiocyanate-conjugated IgG fraction of goat anti-mouse IgG (T-7782) were from Sigma. The IMAGE Consortium expressed sequence tag cDNA clones were from Genome Systems Inc. (St. Louis, MO).
Yeast Two-hybrid Analysis-The procedure for yeast two-hybrid analysis has been described previously (29). In brief, yeast reporter strain Hf7C was sequentially transformed with pGBT-SEMF and a mouse E17 cDNA library cloned in the pGAD10 vector. Colonies that grew under triple selection (Leu Ϫ , Trp Ϫ , His Ϫ ) conditions were assayed for ␤-galactosidase activity. Library-derived plasmids from LacZ-positive yeast cells were recovered and purified for further analyses. Liquid ␤-galactosidase activities were assayed with doubly transformed SFY526 yeast reporter cells using o-nitrophenyl ␤-D-galactopyranoside as the substrate. The DNA sequence analysis of isolated clones was carried out utilizing the fluorescently conjugated dideoxynucleotide termination method with a thermal cycling protocol on automatic DNA sequencers (Applied Biosystems model 377) at the W. M. Keck Biotechnology Laboratory of Yale University.
RNA Isolation and Northern Analysis-Cellular RNA was extracted from indicated murine tissues using the acid guanidium thiocyanate/ phenol/chloroform method (30). Twenty micrograms of total RNA were separated on 1% agarose-formaldehyde gel and transferred to Hybond N ϩ nylon membrane (Amersham Pharmacia Biotech) by capillary elution with 20ϫ SSC. SEMCAPs and M-SemF cDNA probes were prepared by random priming with the Klenow fragment of DNA polymerase in the presence of [␣-32 P]dCTP. Hybridization was performed as described previously (31). Washed nylon membranes were exposed to Kodak BioMax film for 12-36 h with a Kodak BioMax MS intensifying screen at Ϫ80°C.
In Vitro Binding-Glutathione S-transferase (GST), GST-SEMF, and GST-SEMF mutant fusion proteins were expressed in bacteria and purified on glutathione-agarose beads as described (32). Recombinant SEMCAP proteins were transiently expressed in HEK293 cells by calcium phosphate transfection using 20 g each of pCMVMycSEMCAP-1(GIPC), pcDNA-SEMCAP-1(GIPC), and pcDNA-SEMCAP-2, respectively. Forty eight hours following transfection, cells were resuspended in 1-3 ml of Dulbecco's phosphate-buffered saline (DPBS, pH 7.4) plus 1 mM phenylmethylsulfonyl fluoride and 10 M leupeptin. The cytosol was collected after three freeze-thaw cycles in liquid nitrogen and centrifuged at 13,000 rpm (30,000 ϫ g) for 20 min at 4°C. Aliquots (300 l) of cytosol were incubated with 100 g of the designated fusion proteins immobilized on 20-l glutathione-agarose beads for 4 h at 4°C. The beads were washed five times with 1 ml of DPBS containing 0.1% Triton X-100. The bound proteins were eluted in 100 l of SDS sample buffer, and 20-l aliquots were used for immunoblot analyses. Brain cytosol was prepared by homogenizing mouse brain in 2.5 ml of DPBS/g of tissue followed by centrifuging at 100,000 ϫ g for 1 h at 4°C. The supernatant brain cytosol was collected and utilized for the in vitro binding assay as described above.
Antisera Preparation-The antisera against M-SemF were prepared using GST-SEMF as the immunogen. Rabbits were initially immunized with 100 g of GST-SEMF fusion protein mixed with complete Freund's adjuvant and were boosted with 100 g of GST-SEMF in incomplete Freund's adjuvant at 3-6-week intervals. For anti-SEMCAP-1(GIPC) antisera, the rabbits were immunized with purified MycSEMCAP-1(GIPC) fusion proteins from transfected HEK293 cells with high level expression of the fusion protein. With high level expression, the fusion protein forms a 1 M NaCl and 1% Triton X-100-insoluble complex. The pellet from centrifugation in NaCl/Triton X-100 at 30,000 ϫ g for 15 min contained MycSEMCAP-1(GIPC) fusion protein of Ͼ95% purity as judged by Coomassie Blue staining of SDS-PAGE. Such pellets were sonicated in DPBS and used for the immunization using the same protocol as for GST-M-SEMF. The immunization service was performed at the Yale Animal Resource Center.
Immunoblotting-The immunoblotting procedure was carried out as described by Wang and Strittmatter (29). The anti-M-SemF antisera were used at a 1:2500 dilution. The anti-SEMCAP-1(GIPC) antisera were used at a 1:2000 dilution. Immunoreactive proteins were visualized using TMB peroxidase substrate reagents (Vector), after incubation with horseradish peroxidase-conjugated goat anti-rabbit IgG and horseradish peroxidase-conjugated rabbit anti-mouse IgG.
with the anti-M-SemF antiserum and 9E10 monoclonal antibody.
Indirect Immunofluorescence Microscopy-Transfected HEK293T cells or rat E15 cortical cells (33) were cultured on coverslips for 48 h and then fixed with 4% paraformaldehyde in phosphate buffer (100 mM sodium phosphate, pH 7.4, 150 mM NaCl) for 1 h at room temperature. Fixed cells were treated sequentially with permeabilization solution (0.1% Triton X-100 in phosphate buffer) and blocking solution (1% bovine serum albumin in phosphate buffer) and then were incubated with primary antibodies in blocking solution overnight at 4°C, followed by a 1-h incubation with secondary antibodies at room temperature. After each antibody incubation, cells were washed four times in phosphate buffer containing 0.1% bovine serum albumin for 15 min. Anti-M-SemF serum, anti-SEMCAP-1(GIPC) serum, and 9E10 ascites fluid were each used at 1:1000 dilution. The secondary antibodies, fluorescein isothiocyanate-conjugated IgG fraction of goat anti-rabbit IgG and tetramethylrhodamine isothiocyanate-conjugated IgG fraction of goatanti-mouse IgG, were diluted 1:100. After antibody incubations, coverslips were rinsed in distilled water and mounted on glass slides. Immunofluorescence was captured using either a CCD camera (Astrocam TE3) or a laser scanning confocal image system (MRC600, Bio-Rad), and digital images were processed with Adobe Photoshop software.
Triton X-100 Extraction of Membranes-Brains from 2-day-old rats (1.25 g) were homogenized in a glass homogenizer with 3 ml of DPBS. Membrane fractions were collected by centrifuging at 100,000 ϫ g for 1 h at 4°C. Brain membranes were rehomogenized in 3 ml of DPBS and collected by centrifugation. Washed membranes were homogenized again in 12 ml of DPBS, and Triton X-100 (TX-100) was added slowly to a final concentration of 1%, followed by a 1-h incubation at 4°C. Insoluble membrane fractions were collected by centrifugation and re-extracted with 1% TX-100 in DPBS. The final membrane pellet was resuspended in DPBS and frozen at Ϫ80°C. For transfected cell membranes, HEK293T cell membranes were treated with DPBS containing 1% TX-100 for 1 h on ice. Particulate fractions were collected by centrifuging at 13,000 rpm (30,000 ϫ g) for 15 min at 4°C. The resulting membrane pellet was sonicated in DPBS containing 1% TX-100 and incubated for 1 h on ice. Insoluble membrane fractions were collected by centrifugation as described above.
SEMCAP-1(GIPC) Overlay Assay-Brain membrane proteins and 293 cell lysates were separated by SDS-PAGE and transferred to nitrocellulose membrane. The membrane was incubated with 5% dry milk in TBS (20 mM Tris, pH 7.6, 137 mM NaCl, and 0.1% Tween 20) for 1 h at room temperature and then incubated with purified MycSEMCAP-1 fusion protein (8 g/ml) that had been preincubated with a 1:2500 dilution of anti-SEMCAP antiserum for 2 h at room temperature. The bound MycSEMCAP-1-antibody complexes were detected using horseradish peroxidase-conjugated goat anti-rabbit IgG and visualized with TMB peroxidase substrate reagents.

RESULTS
Expression of M-SemF-To explore possible functions of M-SemF, we analyzed its mRNA and protein expression pattern in the mouse. Northern analysis reveals that M-SemF mRNA is expressed primarily in the nervous system. The mRNA is detected in lung at much lower levels, whereas expression in kidney, liver, heart, and skeletal muscle is almost undetectable (Fig. 1A). The expression of M-SemF is first apparent at E12 and persists through adulthood at nearly constant levels. This raises the possibility that M-SemF participates both in axon guidance during development and in synaptic plasticity of the adult central nervous system.
An antiserum generated against a GST/M-SemF cytoplasmic tail fusion protein specifically recognizes an M r 100,000 protein in M-SemF-expressing HEK293T cell membranes but not in control 293T cell membranes (Fig. 1B). The same protein species is detected by immunoblot analysis of brain membranes. Coincident with the mRNA distribution, the expression of M-SemF protein is neurally restricted and continues into adulthood. Any protein expected to interact with M-SemF should exhibit an overlapping pattern of expression.
Isolation of SEMCAPs-To search for cellular proteins that may interact with the cytoplasmic domain of M-SemF, we employed a yeast two-hybrid screen of a E17 mouse cDNA library with a fusion protein expressing the GAL4 DNA binding domain and the cytoplasmic tail of M-SemF (residues 689 -834) as bait. From 8 ϫ 10 6 colonies, 22 clones were obtained that activate the expression of a ␤-galactosidase reporter gene. Two clones were false positives that encoded transcription factors interacting with the GAL4 DNA-binding domain itself. Fifteen of the 22 clones derived from the same cDNA and interacted specifically with the M-SemF cytoplasmic domain ( Fig. 2A). The other five specifically interacting clones were derived from a single cDNA encoding a novel muscle-specific PDZ protein. 2 The four most variable inserts from the 15 clones were sequenced, the longest insert being 1.7 kb. All four of the sequences contain a polyadenylate tail and have 5Ј-ends within 150 bases of one another. A single extended open reading frame was identified and was in frame with the GAL4 transactivation domain. The interacting protein is designated SEMCAP-1. The SEMCAP-1 cDNA encodes an open reading frame of 333 amino acids. The longest clone has 27 base pairs of 5Ј-untranslated region, and the shortest clone is missing the first 40 amino acids from the conceptual sequence. A BLAST search of nucleic acid data bases using the predicted amino acid sequence identified a matching partial sequence derived from a human cDNA that interacts with the carboxyl-terminal portion of the transactivator (Tax) protein of human T cell leukemia virus type 1. Data base searches after this manuscript was in preparation revealed that SEMCAP-1 is identical to GIPC, a protein identified by virtue of its interaction with the RGS protein, GAIP (27). The search identified two predicted open reading frames of Caenorhabditis elegans genes, C35D10.2 and F44D12.6, with significant homology. We also found several overlapping clones of mouse expressed sequence tags whose sequences were highly homologous to SEM-CAP-1(GIPC). From two of those expressed sequence tag clones (IMAGE consortium, AA266091 and AA107535), we obtained a 1.35-kb cDNA. The predicted 314-amino acid residue polypeptide interacts specifically with the cytoplasmic tail of M-SemF in yeast two-hybrid analysis ( Fig. 2A). Thus, this protein is designated SEMCAP-2.
The amino acid sequences of SEMCAP-1(GIPC) and SEM-CAP-2 share 62% identity, whereas those of C35D10.2 and F44D12.6 share 86% identity with one another and 52% similarity with SEMCAP-1(GIPC) (Fig. 2B). Homology is distributed throughout these four proteins, but the most variable region is the amino terminus, and the two C. elegans gene products have extended carboxyl-terminal regions of about 20 amino acids. In addition, an internal segment of about 70 amino acids displays homology with sequences from a variety of proteins including CASK, PSD-95, syntrophin, and ZO-1. Further analysis reveals that this 70-amino acid segment aligns with the PDZ domains of those proteins. PDZ domains are known to interact with the cytoplasmic tail of transmembrane proteins and localize such proteins to sites of specialized cell-cell contacts such as PSD and zona occludens (34,35). The interaction requires conserved amino acid residues at the carboxyl terminus of the transmembrane protein. We examined the sequence of M-SemF and found that the last four amino acid residues are Glu-Ser-Ser-Val, which agrees with the consensus sequence of the class I PDZ domain binding motif: X-(Ser/Thr)-X-Val (36). This suggests that the cytoplasmic tail of M-SemF may associate with cellular protein(s) through a PDZ domain. When the PDZ domains from the four predicted SEMCAP proteins were compared with those from PSD-95 and CASK, they displayed only 10 -20% sequence identity. However, conservation is high for residues most critical in the formation of the carboxylate binding loop (hydrophobic residue-Gly-hydrophobic residue) and for the histidine residue in helix B (Fig. 2C) (37,38).
SEMCAP-1(GIPC) also possesses weak sequence similarity to acyl carrier proteins (residues 264 -320) (27), but no other protein modules are identifiable in the SEMCAPs. There is no recognizable transmembrane domain or nuclear localization signals; therefore, SEMCAPs are predicted to be cytosolic proteins. The partial human cDNA of Tax interacting protein clone 2 (TIP-2) is 98% identical to SEMCAP-1(GIPC) from amino acid 78 to the end (39). Thus, TIP-2 is a human homolog of SEM-CAP-1(GIPC). TIP-2 binding to Tax requires the class I PDZ binding motif, Glu-Thr-Glu-Val, at the carboxyl terminus of Tax.  Characterization of SEMCAP Expression-If the SEMCAPs are to interact with M-SemF in vivo, then they must be coexpressed in the same cells. To examine the potential of SEM-CAPs for physiological interaction with M-SemF, the mRNA and protein distribution of SEMCAP-1(GIPC) and SEMCAP-2 were determined. Northern analysis of SEMCAP-1(GIPC) reveals a single mRNA species of 1.8 kb, whereas SEMCAP-2 Northern blotting shows the presence of a single 1.5-kb transcript (Fig. 3A). Thus, it is very likely that we have obtained clones containing full-length cDNAs for both SEMCAP-1(GIPC) and SEMCAP-2. SEMCAP-1(GIPC) mRNA is most abundant in the brain; less prominent in kidney and lung; and essentially undetectable in heart, liver, and skeletal muscle of adult mice. In contrast, the SEMCAP-2 transcript is present in kidney and lung but is undetectable in the brain, heart, liver, and skeletal muscle of adult mice. SEMCAP-1(GIPC) mRNA is expressed as early as E8, and expression persists throughout embryonic and adult stages (Fig. 3B).
To characterize the SEMCAP proteins, three rabbit antisera against full-length SEMCAP-1(GIPC) protein were prepared. Each antiserum recognizes a 39-kDa protein in HEK293 cells transfected with SEMCAP-1(GIPC) cDNA. The antisera also detect a 37-kDa protein in SEMCAP-2 cDNA-transfected cells, while no immunoreactive protein is detected in control cells (Fig. 3C). In P1 mouse tissue extracts, a 39-kDa protein is detected by the anti-SEMCAP sera. The 39-kDa SEMCAPimmunoreactive protein, likely to be SEMCAP-1(GIPC), is most abundant in brain, kidney, and lung with less expression in heart, liver, and skeletal muscle (Fig. 3C). In addition, a 37-kDa protein is found in kidney and lung. The apparent molecular weight of this protein suggests that it is SEMCAP-2. In adult mice, the expression of 39-kDa SEMCAP-1(GIPC) is restricted exclusively to brain and lung (Fig. 3C). In adult lung, the 37-kDa immunoreactive protein is likely to be SEMCAP-2. In adult kidney, the sera recognize a major 35-kDa protein.
Since SEMCAP-2 but not SEMCAP-1(GIPC) mRNA is highly expressed in the adult kidney, the 35-kDa species might reflect an adult kidney-selective post-translational modification of SEMCAP-2.
We conclude that the SEMCAP-1(GIPC) and SEMCAP-2 mRNAs are full-length and account for the distribution of SEMCAP proteins. Furthermore, the SEMCAP-1(GIPC) protein is expressed in a pattern consistent with an in vivo interaction with M-SemF in the brain. In contrast, GAIP expression is not detectable in brain but is highest in liver, muscle, and lung (40). Thus, brain SEMCAP-1(GIPC) is not a potential partner for GAIP. In the lung, all four proteins (GAIP, M-SemF, SEMCAP-1(GIPC), and SEMCAP-2) are expressed and might interact. The majority of GAIP in liver and muscle tissue does not have access to SEMCAP-1(GIPC) or SEMCAP-2.
SEMCAPs Interact with the Carboxyl Terminus of M-SemF-An in vitro binding assay was performed to determine whether SEMCAP-1(GIPC) interacts directly with the cytoplasmic tail of M-SemF. GST-M-SemF fusion proteins were expressed in E. coli and immobilized on glutathione-agarose beads. Cell lysates from HEK293 cells transiently expressing MycSEMCAP-1(GIPC) were incubated with immobilized GST-M-SemF fusion proteins. SEMCAP-1(GIPC) interaction with GST-M-SemF fusion proteins was monitored by immunoblot analysis of proteins bound to the agarose beads. No interaction is observed with GST leader sequence, but MycSEMCAP-1(GIPC) binds strongly to wild type GST-SEMF (Fig. 4A) 1(GIPC). Thus, the in vitro binding experiments confirm the direct interaction between the cytoplasmic tail of M-SemF and SEMCAP-1(GIPC). As expected for a class I PDZ protein, SEM-CAP-1(GIPC) binding depends on the extreme carboxyl terminus of M-SemF.
To examine whether endogenous SEMCAP-1(GIPC) protein could bind to the cytoplasmic tail of M-SemF, brain cytosol was incubated with immobilized GST-SEMF and GST-SEMF-831* fusion proteins. Association of brain SEMCAP-1(GIPC) with M-SemF is observed by anti-SEMCAP immunoblot of retained proteins (Fig. 5A). The truncated form of SemF exhibits no brain SEMCAP-1(GIPC) binding. Thus, endogenous brain SEMCAP-1(GIPC) binds selectively to the cytoplasmic tail of M-SemF and requires the presence of the terminal 3 amino acid residues.
Specificity of SEMCAP Interaction with M-SemF-The presence of the PDZ binding motif at the cytoplasmic tail of M-SemF prompted us to examine the intracellular domains of other transmembrane semaphorins for similar sequences. Of published semaphorin sequences, only SemC contains a PDZ binding motif, Asp-Ser-Val-Val, at its carboxyl terminus (41). Both SemC and M-SemF belong to semaphorin group IV. To determine if SemC would interact with SEMCAPs, the cytoplasmic domain of SemC was analyzed with SEMCAP-1(GIPC) or SEMCAP-2 in the yeast two-hybrid system. We also examined the cytoplasmic tail from SemB, a type IV semaphorin with no predicted PDZ binding motif at the carboxyl end, as a control. The SemB tail does not interact with either SEMCAP-1(GIPC) or SEMCAP-2. In contrast, both SEMCAP-1(GIPC) and SEMCAP-2 associate with the SemC tail (Fig. 4B). In the yeast two-hybrid analysis, SemC does not bind as avidly to SEMCAP-1(GIPC) or SEMCAP-2 as does M-SemF.
The similarity among proteins carrying class I PDZ binding motifs raises the possibility that SEMCAP-1(GIPC) and SEM-CAP-2 will bind to other proteins with PDZ binding sequences although the yeast two-hybrid screen did not yield any known PDZ proteins. To examine this possibility, SEMCAPs were coexpressed with the cytoplasmic tail from NMDA glutamate receptor type 2C (NR2C, carboxyl terminus Glu-Ser-Glu-Val). Despite the similarity in the carboxyl termini of NR2C and M-SemF, neither SEMCAP-1(GIPC) nor SEMCAP-2 interacts with NR2C (Fig. 4B). The PDZ domain of PSD-95 is known to associate with NR2C in vivo (42). While PSD-95 binds strongly to NR2C in the yeast two-hybrid system, there is no interaction of PSD-95 with SemC or M-SemF.
Major SEMCAP-1(GIPC)-binding Proteins in Brain-We considered whether the SEMCAP-1(GIPC) proteins utilized in the yeast two-hybrid analysis are the major SEMCAP-1(GIPC)-SemC, or NR2C along with the candidate binding partners, full-length SEMCAP-1(GIPC), SEMCAP-2, or the PDZ domains of PSD-95. Colonies growing on double selection plates were isolated and cultured in selective liquid medium. ␤-Galactosidase activity was determined as described in the legend to Fig. 2. C, SEMCAP-1(GIPC) and SEMCAP-2 are both capable of interacting with GAIP. SFY526 yeast cells were cotransformed with pGBT and pGAD vectors expressing full-length GAIP, the cytoplasmic domain of M-SemF, SEMCAP-1(GIPC), and SEMCAP-2 as indicated. Colonies growing on double selection plates were isolated, and ␤-galactosidase activity was determined using a liquid assay as described above. binding proteins in brain. Brain extracts were separated by SDS-PAGE and then overlaid with SEMCAP-1(GIPC) (Fig. 6). M-SemF expressed in HEK293T cells supports SEMCAP-1 binding in this overlay assay. The two most prominent SEM-CAP-1(GIPC)-binding proteins in brain membranes exhibit apparent molecular masses of 100 and 33 kDa. The higher molecular mass binding protein precisely co-migrates with M-SemF immunoreactivity from brain. The identity of the 33-kDa band is not clear, but it is notable that the mass is nearly identical to that reported for GAIP in nonneuronal tissues (43). Thus, these assays are consistent with M-SemF serving as one of the major binding sites for brain SEMCAP-1(GIPC). Some PDZ proteins are capable of forming microscopically visible clusters in heterologous cells and shifting their binding partners into these clusters (44). We examined the distribution of M-SemF and SEMCAP-1(GIPC) in HEK293 cells. When expressed alone, M-SemF is distributed in a uniform fashion on the cell surface (Fig. 7A). In contrast, overexpression of Myc-SEMCAP-1(GIPC) results in large aggregates of Myc-immunoreactive protein. When the two proteins are co-expressed, M-SemF localization is dramatically altered to resemble the SEMCAP-1(GIPC) pattern. The clusters of M-SemF antigen in double transfected cells are precisely coincident with SEM-CAP-1(GIPC) clusters. It is noteworthy that co-expression of GAIP and SEMCAP-1(GIPC) does not lead to co-clustering (27), indicating that GAIP may have other binding partners in these cells or that the lower affinity of the GAIP/SEMCAP-1(GIPC) interaction as compared with the M-SemF/SEMCAP-1(GIPC) interaction does not support clustering.

SEMCAP-1(GIPC) Clusters M-SemF in Transfected Cells-
The size of the SEMCAP-1(GIPC) clusters in 293 cells suggested the creation of a unique subcellular domain by SEM-CAP-1(GIPC). Since clustering by the PDZ protein PSD-95 is thought to responsible in part for the existence of the detergent-insoluble nature of the postsynaptic density, we considered whether SEMCAP-1(GIPC) and co-expressed M-SemF exist in a state not dissociable by nonionic detergents. Immunoblot analysis of membrane and detergent-extracted insoluble fraction from 293 cells indicates that detergent insolubility is a correlate of the SEMCAP-1(GIPC) clusters (Fig. 7B). In the absence of SEMCAP-1(GIPC), M-SemF is solubilized by the addition of Triton X-100, but coexpression of SEMCAP-1(GIPC) results in the redistribution of a majority of M-SemF to a detergent-insoluble fraction.
The detergent-insoluble nature of M-SemF/SEMCAP-  6. Major SEMCAP-1(GIPC)-binding proteins in brain. Detection of SEMCAP-1-binding proteins in brain membrane fraction using an overlay assay is shown. Fifty micrograms of 100,000 ϫ g brain membrane pellet from 2-day-old rats (lanes 1-3) 5). The reactive proteins were detected with peroxidase-conjugated goat anti-rabbit IgG and followed by TMB substrate reagent. MycSEMCAP-1 fusion proteins detects two major brain membrane-associated proteins at 33 and 100 kDa. The 100-kDa protein comigrates with endogenous and recombinant M-SemF.
1(GIPC) clusters in HEK293T cells might correspond to a physiologically relevant state of these proteins or be a condition created by high level expression in a nonneuronal cell. Brain membranes were fractionated by Triton X-100 extraction and analyzed for the presence of immunoreactivity. While a significant portion (about 60%) of SEMCAP-1(GIPC) is cytosolic, particulate brain SEMCAP-1(GIPC) is not extracted by Triton X-100 (Fig. 7B). Similarly, a proportion of M-SemF immunoreactivity in the brain cannot be solubilized with Triton X-100. These data raise the possibility that M-SemF and SEMCAP-1(GIPC) comprise part of the postsynaptic density. However, this is highly unlikely, since PSD fractions highly enriched for known PSD proteins are not enriched in either M-SemF or SEMCAP-1(GIPC) (data not shown). Instead, these proteins appear to associate with a detergent-resistant structure with properties similar to that observed in co-transfected HEK293T cells.
To further define the endogenous distribution of these proteins, we performed an immunohistochemical study of embryonic cortical neurons. SEMCAP-1(GIPC) immunoreactivity is detected in neurons, and within cell soma discrete intensely stained puncta are visualized (Fig. 8). Although not quite as large, these aggregates appear similar to those observed in transfected HEK293T cells (Fig. 7A). Similar immunohistochemical analysis of M-SemF localization in cortical neurons was not technically feasible with the available antisera. Together, these HEK293T and neuronal studies demonstrate that SEMCAP-1(GIPC) can physically associate in clusters with M-SemF in cells and alter M-SemF microscopic distribution and biochemical properties. fractionation. Thus, it appears likely that at least a fraction of brain M-SemF exists in a complex with SEMCAP-1(GIPC). Several factors indicate that this interaction is relatively specific: 1) the glutamate receptor subunit NR2C does not interact with SEMCAP-1(GIPC); 2) PSD-95 does not interact with M-SemF; and 3) none of the many known PDZ proteins were isolated in the yeast two-hybrid screen with M-SemF as bait. However, there is some promiscuity in this SEMCAP/semaphorin interaction in that SEMCAP-2 has a similar binding profile to SEMCAP-1(GIPC) and that SemC binds nearly as well as M-SemF to SEMCAP-1(GIPC). The SEMCAP family of proteins is highly conserved through evolution from worm to mouse. There is no information now available on the distribution or binding partners of the two C. elegans members of the SEMCAP family.
Since SEMCAP-2 is expressed in the kidney and lung, but not the brain, it cannot be a major binding partner for neuronal M-SemF. The expression of both M-SemF and SEMCAP-2 in lung suggests that they may form a lung-specific complex in vivo.
Additional Binding Partners for SEMCAP-1(GIPC) and SEMCAP-2-It remains possible that other binding partners for SEMCAPs exist in particular cell types. It is not unusual for a PDZ domain to interact with multiple binding partners, and vice versa. For example, the PDZ domain in neuronal nitricoxide synthase associates with PSD-95 and CAPON (45,46). The third PDZ domain of PSD-95 interacts with CRIPT and neuroligin (47,48). Additional SEMCAP-1(GIPC) or SEM-CAP-2 interactions may occur in discrete cell types and may contribute to the assembly of physiological relevant multiprotein complexes. In the brain, SEMCAP-1(GIPC) is capable of interacting with an unidentified 33-kDa protein. The in vitro binding studies of Farquhar and colleagues (27) identified SEMCAP-1(GIPC) as a GAIP-interacting protein. We show here that both SEMCAP-1 and SEMCAP-2 can interact with GAIP. The tissue distribution studies raise the possibility that all four proteins (M-SemF, GAIP, SEMCAP-1(GIPC), and SEMCAP-2) could interact in lung. In the brain, only SEM-CAP-1(GIPC) and M-SemF are expressed, not SEMCAP-2 or GAIP. It is an intriguing possibility that the 33-kDa SEMCAP-1(GIPC)-binding protein in brain extracts is a GAIP-related protein. The major sites of GAIP expression, liver and muscle (40), exhibit neither SEMCAP-1(GIPC) nor SEMCAP-2 expression, indicating that the majority of GAIP cannot be complexed with the two PDZ domain proteins studied here. The PDZ domain of human SEMCAP-1(GIPC) (TIP-2) can also bind the carboxyl terminus of the transactivator protein (Tax) of human T cell leukemia virus type 1 in vitro (39). This pathologic interaction may or may not be relevant in vivo. We have not specifically measured the level of SEMCAP-1(GIPC) expression in lymphocytes.
Binding Is Mediated by a PDZ Domain Type Interaction-The interactions between the semaphorin cytoplasmic tails and SEMCAPs are likely to be mediated by the PDZ domain for several reasons. Both M-SemF and SemC carry the class I PDZ binding motif in the carboxyl termini. The sequence requirements in the M-SemF carboxyl terminus for interacting with SEMCAPs are very similar to those characterized in other PDZ domain associations. The PDZ domain is a polypeptide module of about 100 amino acids that mediates protein/protein interactions. Based upon the consensus binding sequences at the carboxyl termini, PDZ domains are divided into two classes. Class I PDZ domains selectively bind to the sequence motif Ser/Thr-X-Val, and class II PDZ domains favor Phe/Tyr-X-Ala/ Phe/Val (36). The crystal structures of PDZ domains from two class I families and one class II family have been determined (37,38). Although the primary sequences show only limited similarity, the secondary structures of different PDZ domains are highly conserved, containing a core of six ␤-sheets (␤A-␤F) and two ␣-helices (␣A and ␣B). When the amino acid sequences encoding PDZ domains from SEMCAP-1(GIPC) and SEM-CAP-2 are aligned with those from PSD-95 and CASK, it is found that the amino acid residues forming the binding pockets are well conserved. Both SEMCAP-1(GIPC) and SEMCAP-2 encode class 1A PDZ domains that have a His residue at the initiation of ␣-helix B. The specificity of SEMCAP binding to variable amino acids at the carboxyl end could also be rationalized from the results of the structural analysis. A Leu residue at position ␤B1 of the carboxylate binding pocket in SEM-CAP-1(GIPC) and SEMCAP-2 instead of the residue Phe in PSD-95 may render the flexibility to accommodate different amino acids at the terminal position of the PDZ binding consensus.
Clusters of M-SemF/SEMCAP-1(GIPC)-The binding of M-SemF to SEMCAP-1(GIPC) appears to create an aggregated complex that is resistant to nonionic detergent extraction. Since PDZ proteins are frequently found at specialized cell contacts, the biochemical properties of the M-SemF-SEMCAP-1(GIPC) complex raise the possibility that they are part of a specialized cellular domain in the brain. We have excluded the possibility that M-SemF and SEMCAP-1(GIPC) are selectively localized to PSDs, but we have not determined the nature of M-SemF/SEMCAP-1(GIPC) clusters in the brain. It is conceivable that they define a heretofore unrecognized structure in the developing and adult nervous system. Immunohistologic analysis at the ultrastructural level may be required to explore this issue.
PDZ domain proteins are best known for their localization to cell-cell junctions in epithelial cells and neurons, where they bind to cytoplasmic domains of channels or receptors and coordinate the assembly of multiprotein complexes. Such PDZ proteins generally have more than one PDZ domain as well as other recognized functional domains, such as Src homology 3, guanylate kinase, CaM kinase, or tyrosine phosphatase domains. These multivalent, multiprotein complexes provide both a structural basis and a signaling function for the regulation of specific physiological activities. In contrast, SEMCAP-1(GIPC) and SEMCAP-2 have only a single PDZ domain, and no other protein interacting domain is identified. However, since SEMCAP-1(GIPC) is capable of forming aggregates in the absence of M-SemF, it is possible that the single PDZ domain of SEMCAP-1(GIPC) in a multivalent aggregate might assemble multiple associated proteins. The mechanism of SEMCAP-1(GIPC) aggregation is unclear. Neither SEMCAP-1(GIPC) nor SEMCAP-2 protein interacts with itself or one another other in the yeast two-hybrid system (data not shown).
Biologic Functions of M-SemF-The function of M-SemF remains ill defined, although other semaphorin family members are known to function as axonal guidance signals in the developing nervous system. The identification of SEMCAP-1(GIPC) as an intracellular M-SemF binding partner is consistent with a role for M-SemF as a ligand functioning in cell surface communication. We have found no proteins with known signal transduction activity associated with the cytoplasmic tail of M-SemF. However, if SEMCAP-1(GIPC) or SEMCAP-2 were to link M-SemF to the RGS protein GAIP in lung, it is conceivable that this could underlie receptor-type function.
It is tempting to speculate that M-SemF clustering by SEM-CAP-1(GIPC) is required for biologic activity as a ligand for receptors on adjacent cells directing axonal growth in the nervous system. The function of another axonal guidance molecule, fasciclin II, depends upon co-expression with the PDZ protein, Dlg (49,50). Fasciclin II is a member of the extensively studied immunoglobulin superfamily of homophilic cell adhesion molecules. Unless Dlg is co-expressed with fasciclin II, the latter protein is not found at developing neuromuscular junctions, and these junctions are functionally defective. Our results indicate that at least some members of the semaphorin family of axonal guidance signals will require similar PDZ-dependent localization and clustering for activity. Model studies of clustered and unclustered signal transduction cascades in prokaryotes have emphasized the dramatic effects that oligomerization can have for ligand receptor signaling complexes (51).
It is also of interest that both M-SemF and SEMCAP-1(GIPC) are expressed at high levels in the mature nervous system. It is conceivable that this complex plays a role in synaptic plasticity of the adult brain. Further studies defining the subcellular localization and perturbing the expression of M-SemF and SEMCAP-1(GIPC) will illuminate these issues.