Subcellular targeting and cytoskeletal attachment ... - Semantic Scholar

3 downloads 433 Views 1MB Size Report
(SAPs), which includes SAP90/PSD-95, SAP102 and ... translocation of SAP90/PSD-95 to the plasma membrane of .... Spectrum or Adobe Photoshop. To disrupt ...
2365

Journal of Cell Science 111, 2365-2376 (1998) Printed in Great Britain © The Company of Biologists Limited 1998 JCS4561

Subcellular targeting and cytoskeletal attachment of SAP97 to the epithelial lateral membrane Hongju Wu, Susanne M. Reuver, Sven Kuhlendahl, Wook Joon Chung and Craig C. Garner* Department of Neurobiology, University of Alabama at Birmingham, 1719 Sixth Ave. S, Birmingham, AL 35294-0021, USA *Author for correspondence (E-mail: [email protected])

Accepted 15 June; published on WWW 30 July 1998

SUMMARY The synapse-associated protein SAP97 is a member of a novel family of cortical cytoskeletal proteins involved in the localization of ion channels at such membrane specializations as synaptic junctions. These multidomain proteins have binding sites for protein 4.1, GKAPs/SAPAPs, voltage- and ligand-gated ion channels and cell-adhesion molecules containing C-terminal T/SXV motifs. In this study, we evaluated the contribution of individual domains in SAP97 to its selective recruitment and attachment to the cortical cytoskeleton in epithelial cells. We find that the PDZ, SH3 and GK domains, as well as the I3 insert in SAP97, are not essential for subcellular

targeting, though both PDZ1-2 domains and the I3 insert affect the efficiency of localization. Instead, we show that the first 65 amino acid residues in SAP97, which are absent from SAP90/PSD-95 and SAP102, direct the selective subcellular localization and can mediate at least one point of attachment of SAP97 to the cytoskeleton assembled at sites of cell-cell contact. Our data demonstrate that it is the sequences unique to SAP97 that direct its subcellular targeting to the epithelial lateral membrane.

INTRODUCTION

homology 3 (SH3) region, and a carboxyl-terminal guanylate kinase (GK)-like domain (Fig. 2). These domains are sites of protein-protein interactions (reviewed by Garner and Kindler, 1996). For example, the PDZ domains bind with high affinity to the C-terminal peptide motif T/S-X-V/I (T/SXV-motif) in a number of proteins, including the voltage-gated Shaker K+ channels and the NR2 subunits of the NMDA receptor (reviewed by Sheng, 1996). In vivo, these interactions are required to cluster ion channels and cell adhesion molecules to synaptic junctions (Tejedor et al., 1997; Thomas et al., 1997a,b). SH3 domains are classical sites of protein-protein interaction first identified in src protein tyrosine kinases (Musacchio et al., 1992), though in SAPs, SH3 binding partners are not known. The GK domain in SAPs exhibits a striking sequence similarity to low molecular mass guanylate kinases, but it does not encode an active guanylate kinase (Kistner et al., 1995; Kuhlendahl et al., 1998). Instead, the GK domain has evolved to interact with a family of synaptic proteins called GKAPs, for guanylate kinase-associated proteins (Kim et al., 1997), or SAPAPs, for SAP90/PSD-95associated proteins (Takeuchi et al., 1997). Although both GKAPs/SAPAPs and channel proteins can promote the translocation of SAP90/PSD-95 to the plasma membrane of transfected fibroblasts (Takeuchi et al., 1997), the functional significance of these interactions for the localization of SAPs to sites of cell-cell contact is unclear. Studies on SAP97/hDlg in non-neuronal cells have provided the first clues to how SAPs may be recruited to specific

Members of the SAP97 family of synapse-associated proteins (SAPs), which includes SAP90/PSD-95, SAP102 and Chapsyn-110/PSD-93, have recently emerged as central players in the molecular organization of synapses (reviewed by Budnik, 1996; Garner and Kindler, 1996; Sheng, 1996). SAPs are scaffold proteins involved in linking ligand- and voltagegated ion channels to the cytoskeleton assembled at various membrane specializations (reviewed by Budnik, 1996; Garner and Kindler, 1996; Sheng, 1996). In neuronal cells, SAPs are differentially localized either to the pre- and/or postsynaptic sites of excitatory or inhibitory synapses, along unmyelinated axons or dendritic profiles (Brenman et al., 1996; Hunt et al., 1996; Kim et al., 1996; Kistner et al., 1993; Müller et al., 1995, 1996). In addition to the synaptic localization of SAP97 in rat brain, it, as well as its human homolog hDlg, are expressed in epithelial cells, where they are localized along the lateral membrane at cell-cell adhesion sites (Lue et al., 1996; Müller et al., 1995; Reuver and Garner, 1998). At present, the cellular mechanisms regulating the attachment of SAPs to the cortical cytoskeleton and their distinct spatial distribution are unknown. An alignment of the deduced amino acid sequences of SAP97/hDlg, SAP90/PSD-95, SAP102 and Chapsyn110/PSD-93 reveals the presence of five regions of high sequence homology flanked by unrelated regions. The homologous regions represent five domains that include three amino-terminal PDZ (PSD-95/DLG/ZO1) domains, an src

Key words: Actin, Cell-cell contact, Cytoskeleton, MAGUK, PDZ domain

2366 H. Wu and others membrane specializations. In epithelial cells, E-cadherinmediated cell-cell adhesion induces the recruitment of SAP97/hDlg to the plasma membrane and its association with the cortical cytoskeleton at cell-cell contact sites (Reuver and Garner, 1998). In vitro, specific isoforms of SAP97/hDlg, containing an I3 insert situated between the SH3 and GK domains, bind the erythrocyte cytoskeletal protein 4.1 via I3 and sequences situated between PDZ1 and PDZ2 (Lue et al., 1994, 1996). This has led to the hypothesis that these interactions are important for the subcellular targeting of SAP97/hDlg in epithelial cells (Lue et al., 1996). Clearly, the identification of regions in SAPs that direct their site-specific localization and attachment to the cortical cytoskeleton is fundamental to defining how their differential distribution at cell-cell contact sites is achieved. Here, we have analyzed the requirement of individual domains in SAP97 for its subcellular targeting and association with the cortical cytoskeleton at the lateral membrane in epithelial cells. The first 65 amino acid residues of SAP97 (S97N1-65) were found to direct the selective recruitment of SAP97 to sites of cell-cell contact, and the PDZ1-2 domains and I3 insert to affect the efficiency of localization. Furthermore, we show that S97N1-65 mediates at least one point of attachment of SAP97 to the cortical cytoskeleton at lateral membranes. Our data provide critical clues to the domains in SAPs that direct their site-specific localization. MATERIALS AND METHODS DNA construction Green fluorescent protein (GFP)-tagged SAP90/PSD-95, SAP97 and SAP102 were constructed by subcloning the coding sequences of SAP90/PSD-95, SAP97 and SAP102 (Kistner et al., 1993; Müller et al., 1995; 1996) into the pEGFP-C1 vector (Clontech) after being amplified by polymerase chain reaction (PCR) as described by the manufacturers (GIBCO-BRL). Sense oligonucleotide primers contained the start codon and an EcoRI site incorporated at the 5′ end (S-EcoRI). Antisense primers included the native stop codon and a KpnI site incorporated at the 5′ end (AS-KpnI). The SAP97 deletion mutants (∆S97N, ∆GK) and segments of SAP97 (S97N-PDZ1-3, PDZ1-3, PDZ1-2, SH3-I3-GK and GK) were constructed by PCR with pairs of primers flanking the region to be amplified. EcoR1 or KpnI sites were incorporated at the 5′ end of sense and antisense primers, respectively. p31f2, which contains an I3 insert, was used as the DNA template. The PCR products were all ligated into the pCRTM2.1 vector (Invitrogen, TA cloning kit) and then subcloned into the EcoRI and KpnI sites of the pEGFP-C1 vector or the pEGFP-C2 vector (Clontech). SAP97 deletion mutants (∆PDZ1, ∆PDZ1-2, ∆PDZ1-3, ∆I3 and ∆SH3) were constructed by PCR using primers that flanked the regions of SAP97 to be deleted. These primers contained a MluI site incorporated at the 5′ end and were used with either S-EcoRI or ASKpnI. The PCR products were cloned into pCRTM2.1 vector and then subcloned into pEGFP-C1 in a three-piece ligation. Chimeric molecules between SAP97 and SAP90/PSD-95 or SAP102 were also constructed by PCR. The S97N was amplified with a pair of primers in which an EcoRI site or MluI site were incorporated at the 5′ end of sense and antisense primers, respectively. The Cterminal segments of SAP90/PSD-95 or SAP102 were amplified by PCR with a pair of primers in which an MluI site or KpnI site were incorporated at the 5′ end of sense and antisense primers, respectively. All of the PCR products were cloned into pCRTM2.1 vector and subcloned into the EcoRI and KpnI sites of pEGFP-C1 (Clontech) as three-part ligations. The expression of all constructs was assessed by

western blot analysis of transiently transfected CACO-2 cells (see below). Cell culture, transfection, Latrunculin B treatment and Triton extraction of live cells The human colon carcinoma cell line CACO-2 obtained from the American Type Culture Collection (ATCC# HTB 37) was maintained in MEM supplemented with 20% fetal bovine serum (FBS), at 37°C, under a 5% CO2 atmosphere. GFP-tagged DNA constructs were transiently transfected into CACO-2 cells (grown to 60-70% confluency) using lipofectamine as described by the manufacturer (GIBCO-BRL). In brief, 2.1 µg DNA and 18 µl lipofectamine were applied to cells grown on 22×22 mm coverslips (in 3.5 cm dishes). Both DNA and lipofectamine were diluted into 150 µl Opti-MEM (GIBCO BRL), then mixed together and incubated at room temperature for 30-45 minutes to allow DNAlipid complexes to form. 600 µl Opti-MEM was added, mixed gently and then overlaid onto the Opti-MEM pre-washed cells. Cells were incubated for 5 hours at 37°C and then the medium replaced with culture medium. Cells were fixed with 3.7% formaldehyde 24 hours later. Coverslips were mounted with Vectashield mounting medium (Vector) and fluorescent images taken with a Nikon Diaphot 300 microscope equipped with a Photometrics CH250 CCD-camera. Digitally stored images were combined and displayed with IP lab Spectrum or Adobe Photoshop. To disrupt filamentous-actin (F-actin), 24 hours after transfection, 5 µM Latrunculin B (Calbiochem) in DMSO was added directly to the culture medium of transfected CACO-2 cultures and incubated for 5 minutes at 37°C. Control cells were treated with DMSO alone. Cultures were immediately fixed with 3.7% formaldehyde and processed for epifluorescence microscopy. To assess the association of GFP-tagged constructs with the cortical cytoskeleton, transfected cells were extracted with Triton X-100. In brief, cells were cultured and transfected in glass-bottom Microwells (MatTek Corporation). 24 hours after transfection, the cells were extracted by perfusion with 0.5% Triton X-100 in phosphate-buffered saline (PBS) under the microscope at room temperature. This allowed real-time imaging of small groups of transfected cells. Cells were initially perfused for 3 minutes with PBS, for 1.5 minutes with 0.5% Triton X-100 in PBS, and then with PBS again. Images of selected cells were taken prior to perfusion with 0.5% Triton X-100 in PBS and then at 1-minute intervals. After 10 minutes, the loss or retention of fluorescence at sites of cell-cell contact was examined in neighboring cells. When retained, fluorescence was still visible 30 minutes after extraction. Antibodies, immunoprecipitation and immunoblotting The mouse monoclonal antibodies mAb-197.4 and mAb-119 (produced by the hybridoma core facility at UAB) were generated against the N-terminal 163 and 119 amino acid residues of SAP97 and SAP102, respectively. The rabbit polyclonal antibodies rAb-63 and rAb-PDZ12 were generated against the N-terminal 63 amino acid residues or PDZ1-2 of SAP90/PSD-95 fused to glutathione-Stransferase (GST), respectively. The mouse polyclonal antibody AbGK was generated against the GK domain of SAP97 fused to GST. The GFP monoclonal antibody mAb-GFP was purchased from Clontech, the mouse monoclonal antibody anti-human PSD-95 from Upstate Biotechnology and rabbit Kv1.4 polyclonal from Alomone Labs. Immunoprecipitation was performed essentially as described in Kim and Sheng (1996). Transfected and nontransfected cells were homogenized in RIPA buffer (50 mM Tris, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) and incubated on ice for 1 hour to solubilize proteins. After centrifugation at 16,000 g for 30 minutes, the supernatant was precleared with BSAtreated Protein-A sepharose for 1 hour at 4°C, followed by incubation with antibody bound to Protein-A sepharose for 2 hours at 4°C.

Subcellular targeting of SAP97 2367 Immunoprecipitates were washed, separated by SDS-PAGE and analyzed by immunoblotting. For immunoblotting, transfected and nontransfected cells cultured in 60-mm dishes were scraped off with a rubber policeman in the presence of 1× SDS sample buffer. Samples were boiled for 5 minutes and separated on 10% SDS polyacrylamide gels, then transferred to nitrocellulose membrane (Micron Separations Inc.). Western blots were blocked with 5% skimmed milk, 0.5% Nonidet P-40 in Trisbuffered saline (TBS) (buffer A) for 30 minutes before incubating for 2 hours at room temperature with primary antibodies diluted 1:500 in TBS/3% FBS. After three washes in TBS and reblocking in buffer A, blots were incubated for 2 hours at room temperature with alkaline phosphatase-conjugated goat anti-mouse antibodies (Sigma) (dilution 1:2000). Membranes were washed twice with TBS, 0.5% Tween-20 and once with TBS. Immunoreactive bands were detected as described previously (Müller et al., 1995).

RESULTS Differential distribution of SAPs in epithelial cells In neuronal cells, SAPs are differentially distributed (See Garner and Kindler, 1996). In epithelial cells, SAP97 is localized along the lateral membrane at sites of cell-cell contact (Müller et al., 1995; Reuver and Garner, 1998). We therefore investigated whether SAP90/PSD-95 and/or SAP102 were also expressed in epithelial cells and whether this cell system could be used to define protein sequences directing the subcellular localization of SAPs. Western blots of CACO-2 cell extracts were stained with SAP-specific antibodies (Fig. 1A). Rabbit antibody against SAP90/PSD-95 detected a single band at 95 kDa with the same relative mobility as SAP90/PSD-95 immunoreactive bands present in rat brain extracts (Fig. 1A). Antibody against SAP102 reacted with a 102 kDa band in brain extracts but failed to detect a similar molecular mass protein in CACO-2 cell extracts (Fig. 1A). Antibodies against SAP97 recognized a doublet at 120 kDa in both CACO-2 and brain extracts (Fig. 1A). The expression of SAPs in CACO-2 cells was also assessed by immunofluorescence microscopy. Confluent monolayers of CACO-2 cells stained with antibody against SAP97 showed an intense fluorescence at sites of cell-cell contact (Fig. 1B). In contrast, cells stained with antibody against SAP90/PSD-95 exhibited a diffuse pattern (Fig. 1B). No immunofluorescence labeling was observed with the SAP102 antibody (not shown). These data indicate that while SAP90/PSD-95 is expressed along with SAP97 in CACO-2 cells, it is not co-recruited with SAP97 to sites of cell-cell contact in these cells. The ability of SAP97 to become localized to the lateral membrane in CACO-2 cells, in contrast to SAP90, was analyzed further by expressing recombinant SAP97 and SAP90/PSD-95 as well as SAP102 in CACO-2 cells. Recombinant proteins were distinguished from endogenously expressed SAPs by epitope tagging either with a 10 amino acid peptide from c-myc (Munro and Pelham, 1987) or with the green fluorescent protein (GFP). Similar to endogenous SAP90, GFP-tagged SAP90/PSD-95 and SAP102 (Fig. 1B) exhibited a diffuse cytoplasmic distribution in cells expressing both low and high levels of recombinant proteins. At sites of cell-cell contact between two high-expressing cells some increase in fluorescence was observed. However, since no increase was observed at sites of cell-cell contact with nontransfected cells, a non-localization phenotype was scored. In

Fig. 1. Expression and distribution of endogenous and GFP-tagged SAP90/PSD-95, SAP97 and SAP102 in the epithelial cell line CACO-2. (A) Western blots of rat brain homogenates (lanes 1, 4 and 7), total cellular extracts from untransfected CACO-2 cells (lanes 2, 5 and 8), or CACO-2 cells transfected with GFP-SAP90/PSD-95 (lane 3), GFP-SAP97 (lane 6) or GFP-SAP102 (lane 9). Antibodies against SAP90/PSD-95 (lanes 1-3), SAP97 (lanes 4-6) or SAP102 (lanes 7-9) were used to probe the blots. The asterisk marks a SAP97 breakdown product. (B) Confluent monolayers of CACO-2 cells were incubated with antibodies against SAP90/PSD-95, SAP97 or SAP102, followed by FITC-conjugated secondary antibodies, to reveal the distribution of endogenous SAP90/PSD-95 (endo SAP90), SAP97 (endo-SAP97) or SAP102 (not shown). The distributions of GFP-tagged SAP90/PSD-95, (GFP-SAP90), SAP97 (GFP-SAP97), SAP102 (GFP-SAP102) or GFP alone were assessed by GFP epifluorescence with FITC filters, 24 hours after transfection of semiconfluent CACO-2 cells. Note the diffuse pattern for both SAP90/PSD-95 and SAP102 compared to the lateral membrane localized pattern for SAP97.

2368 H. Wu and others contrast, GFP-tagged SAP97 became localized to the lateral membrane at sites of cell-cell contact between adjacent cells (Fig. 1B). Myc-tagged recombinant SAPs showed a similar distribution (data not shown). The presence of recombinant protein of the correct molecular mass was assessed by analyzing cell extracts of transfected cells on western blots (Fig. 1A). As expected, GFP-tagged SAP90/PSD-95, SAP97 and SAP102 were found to be approx. 27 kDa larger than the endogenous epithelial and/or brain proteins (Fig. 1A). These data show that SAP97, as compared to SAP90/PSD-95 and SAP102, is selectively localized at specific membrane regions in epithelial cells and that GFP-tagged recombinant proteins can be used to identify regions in SAPs that direct their subcellular distribution. The N-terminal domain directs the selective lateral membrane localization of SAP97 A high degree of sequence similarity between SAPs (Kistner et al., 1993; Müller et al., 1995; 1996) and the lateral membrane localization of SAP97 in CACO-2 cells, in contrast to the cytoplasmic distribution of SAP90/PSD-95 and SAP102, allowed a chimeric approach to be used to identify regions in

Fig. 2. Spatial distribution of chimeric SAP97/SAP90 and SAP97/SAP102 molecules in CACO-2 cells. (A) Schematic diagrams of the GFP-tagged SAP97, SAP90/PSD-95, SAP102 and chimeric molecules. The basic organization of the PDZ, SH3 and GK domains in each SAP as well as the I3 insert in SAP97 are indicated as gray boxes. Except for I3, the regions flanking boxes exhibit a low degree of sequence similarity (5-30%). The amino acid boundaries of SAP97 and SAP90 or SAP102 are indicated for each chimeric molecule. The localization efficiency was measured as the percentage (±s.d.) of transfected cells showing clear lateral membrane localization. Note that only transfected cells exhibiting a low to moderate level of protein were included in this analysis. Data shown are an average of three independent experiments with approximately 200 cells per experiment counted. (B) Spatial distribution of GFP-tagged SAP90/PSD-95, SAP102, S97N/S90PDZ1-GK and S97N/S102PDZ1-GK of transfected CACO-2 cells, assessed by GFP epifluorescence.

SAP97 that direct its subcellular localization. Initial chimeric proteins were made between SAP97 and SAP90/PSD-95 by replacing C-terminal segments of SAP97 with the analogous regions from SAP90/PSD-95. The cross-over points were selected at boundaries between conserved and unique sequences (Fig. 2A). All three GFP-tagged chimeras were found to localize to the epithelial lateral membrane in a pattern identical to endogenous and GFP-tagged SAP97 (Fig. 2B). Common to all three were the N-terminal 186 amino-acid residues of SAP97 that precede the PDZ1 domain (Fig. 2A), suggesting that this region (henceforth termed S97N) is important for the selective targeting of SAP97 in epithelial cells. This conclusion was supported by the ability of S97N to also direct the lateral membrane localization of SAP102 (Fig. 2B). PDZ1-2 and the I3 domain contribute to the efficient localization of SAP97 A comparison of the efficiency of lateral membrane localization of different S97-S90 and S97-S102 chimeras (Fig. 2A) revealed that S97N/S90PDZ1-GK (33%) and S97NPDZ2/S90PDZ3-GK (37%) were less effective at localization

Subcellular targeting of SAP97 2369 than S97N-I3/S90GK (65%), SAP97 (72%) or S97N/S102PDZ1-GK (60%). Localization efficiency was assessed by comparing the percentage of transfected cells showing clear lateral membrane localization. Only transfected cells expressing low to moderate levels were counted, since

localization could not be assessed unequivocally in highexpressing cells. These data suggest that one or more domains may play a role in mediating the efficient lateral membrane localization of SAP97. A series of GFP-tagged deletion constructs were therefore created (Fig. 3A). Deletion of PDZ3

Fig. 3. Localization efficiency of GFP-tagged SAP97 deletion mutants. (A) Schematic diagram of deletion constructs of SAP97 used in this study. The amino acid boundaries of each deletion are indicated. The localization efficiency was measured as the percentage (±s.d.) of transfected cells showing clear lateral membrane localization. ND, not determined, due to the susceptibility of expressed protein to degradation. Also summarized are the abilities of individual constructs to remain localized at sites of cell-cell contact (+) or to be extracted (−) 10 minutes after being treated with 0.5% Triton X-100 for 1.5 minutes. Similar results were obtain in three independent experiments in which GFP fluorescence in small groups of cells was monitored in real time during the first 5 minutes and on the remainder of the dish after 10 minutes. (B) Western blots of protein extracts from CACO-2 cells transfected with the GFP-tagged SAP97 deletion constructs. The molecular masses of individual GFP-fusion proteins were assayed by immunoblotting with either a monoclonal antibody against the S97N domain of SAP97 (mAb197.4), the GK domain of SAP97 (Ab-GK) or against GFP (mAb-GFP). The position of endogenous SAP97 is indicated with an arrow.

2370 H. Wu and others (∆PDZ3) from SAP97 had no effect on localization efficiency (70%). In contrast, deleting PDZ1-2 (∆PDZ1-2) or PDZ1-3 (∆PDZ1-3) caused a reduction of approx. 50% in the localization efficiency as compared to full-length SAP97 (Fig. 3A). This suggests that PDZ1-2 contributes to the efficient localization of SAP97. A specific role for PDZ1 could not be analyzed since transfection of ∆PDZ1 mutant constructs did not lead to the synthesis of full-length proteins (Fig. 3A,B). Removing the SH3 (∆SH3) or GK (∆GK) domains had no effect on the localization efficiency of GFP-tagged SAP97 (Figs 3A, 4), indicating that these two domains are dispensable for localization of SAP97. However, when the SH3 and GK domains were deleted together with intervening sequences (S97N-PDZ1-3), a significant effect on the localization efficiency (43%) was observed. These observations suggest that sequences situated between the SH3 and GK domains play a role in localization. This region of SAP97 has previously been shown to be a site of sequence variation that arises via alternative splicing (Lue et al., 1994; Müller et al., 1995). The GFP-SAP97 used in this study contains the I3 insert. When I3 was deleted (∆I3), a significant decrease in efficiency of localization (40%) was observed as compared to GFP-SAP97 (70%) (Fig. 3A). These data indicate that the I3 insert as well as PDZ1-2 play a role in the efficient localization of SAP97. However, the deletion of neither domain abolished lateral membrane localization of SAP97. We therefore deleted both PDZ1-3 and I3 (S97N-GK) to examine whether these domains act synergistically to localize SAP97 efficiently. This mutant localized with the same efficiency as ∆PDZ1-2, ∆PDZ1-3, ∆I3 and S97N-PDZ1-3 (Fig. 3A), indicating that there is no synergism between these regions. Furthermore, we examined whether either region had an innate ability to localize to the lateral membrane of CACO-2 cells. GFP-tagged PDZ1-2,

Fig. 4. Distribution of GFPtagged SAP97 deletion constructs. Semi-confluent monolayers of CACO-2 cells were transfected with GFPtagged SAP97 constructs as indicated and 24 hours later visualized by direct GFP fluorescence with FITC filter set. Note that selected images do not necessarily reflect localization efficiency, due to the nature of the transcient transfection assay. For localization efficiency of individual constructs, see Fig. 3. Interestingly, SH3-I3-GK exhibits a nuclear pattern.

PDZ1-3, GK and SH3-I3-GK all exhibited a diffuse pattern (Figs 3A, 4). Taken together, these data show that the PDZ1-2 and I3 domains appear to contribute to the localization efficiency of SAP97 at the epithelial lateral membrane, yet they do not contain the localization signals. The N-terminal segment of SAP97 is required for localization Significantly, all deletion constructs capable of localizing to the lateral membrane contained the amino-terminal 186 amino acid residues of SAP97. Furthermore, the addition of S97N to those constructs which by themselves could not localize led to a precise localization to the lateral membrane (Figs 3A, 4). To test whether this region is essential for the localization of SAP97, we deleted S97N (∆S97N) from GFP-SAP97. This ∆S97N mutation abolished the lateral membrane localization of SAP97 (8%) (Fig. 3A), demonstrating that this segment of SAP97 was critical for its subcellular targeting. We also examined whether S97N was sufficient to direct lateral membrane localization of SAP97. When S97N was fused either up- (S97N-GFP) or down- (GFP-S97N) stream of GFP, a diffuse pattern was observed, though proteins of the appropriate molecular mass were expressed. However, the addition of one half of PDZ1 (PDZ1/2) or glutathione Stransferase (GST) to GFP-S97N (S97N-PDZ1/2; S97N-GST, respectively) resulted in lateral membrane localization (Figs 3, 4). These data demonstrate that S97N is both necessary and sufficient to direct the subcellular localization of SAP97 in epithelial cells and that the presence of additional C-terminal sequences appears to help maintain it in an active conformation. The S97N domain can be divided into two regions (Fig. 5A). The first region (residues 1-104) (S97N1-104) is unique to

Subcellular targeting of SAP97 2371

Fig. 5. The first 65 amino acid residues in SAP97 direct its localization to the lateral membrane. (A) Alignment of the deduced N-terminal amino acid sequences preceding PDZ1 of SAP97, SAP90/PSD-95, SAP102 and Chapsyn-110/PSD-93. Identical amino acid residues are indicated by gray boxes. Cysteine residues involved in the multimerization of SAP90/PSD-95 and Chapsyn-110/PSD-93 as well as the two found in SAP97 are marked with asterisks. Arrows also label the start of the first PDZ domain in each protein as well as the boundaries of the S97N domain fused to the SAP97 GK domain. (B) Schematic diagram of S97N-GK deletion constructs. The amino acid boundaries of each deletion are indicated. Tabulated are the localization efficiency and stability of each construct to extraction by Triton X-100. (C) Distribution of S97N1-104-GK, S97N1-65-GK, S97N105186-GK and S97N66-186-GK in transfected CACO-2 cells analyzed by GFP fluorescence.

SAP97. The second (residues 105-186) (S97N105-186) shares about 60% similarity with the N-terminal regions of SAP90/PSD-95, SAP102 and Chapsyn-110/PSD-93 (Fig. 5A). This latter region in SAP90/PSD-95 and Chapsyn-110/PSD-93 has been shown to promote ion channel clustering through a head to head multimerization (Hsueh et al., 1997). In both SAPs, ion channel clustering in HEK293 cells (Hsueh et al., 1997; Topinka and Bredt, 1998) and visualization of multimers on non-reducing SDS gels (Hsueh et al., 1997) are mediated by a pair of cysteine residues situated in the N terminus. Given the importance of S97N in the selective localization of SAP97 to the epithelial lateral membrane, we examined whether this potential multimerization domain and/or the two cysteine residues (C66 and C73) in the S97N domain of SAP97 (Fig. 5A) play a role in the localization of SAP97 to the lateral membrane. These possibilities were initially investigated by examining whether removing amino acid residues 105-186, containing the putative multimerization domain, could block S97N-mediated localizaiton of SAP97. This was accomplished by attaching the GK domain as a stabilizer to S97N1-104 (S97N1-104-GK) or S97N105-186 (S97N105-186 -GK). Only

S97N1-104-GK became localized to the lateral membrane (Fig. 5B,C). These data demonstrate that sequences in S97N homologous to the multimerization domain in SAP90/PSD-95 are not essential for the lateral membrane localization of SAP97 and that the targeting signal is present in S97N1-104. Next, we tested whether SAP97 can lead to the formation of stable multimers in non-reducing conditions, as reported for SAP90/PSD-95 (Hsueh et al., 1997). Whereas high molecular mass immunoreactive bands (180 kDa), indicative of multimers of the 95 kDa SAP90/PSD-95, could be observed with recombinant SAP90 under non-reducing conditions, the mobility of SAP97 immuno-reactive bands remained at 120 kDa (data not shown), indicating that SAP97 does not form multimers that can be stabilized by disulfide bridges. To directly test whether the two cysteine residues (C66 and C73) are required for the localization of SAP97, two additional deletion molecules were constructed. The first, S97N1-65-GK, lacks these two residues while the second, S97N66-186-GK, contains them. Here, only S97N1-65-GK was localized to the epithelial lateral membrane (Fig. 5B,C). These data demonstrate that the localization of SAP97 is not mediated by

2372 H. Wu and others Fig. 6. Localization of the GFP-tagged S97N is not mediated by association with the endogenous SAP97. (A) Western blots of immunoprecipitates from transfected COS-7 cells. Extracts from transfected cells were immunoprecipitated either with mAb-GFP (lanes 2-4) or anti-Kv1.4 (lane 5) antibodies and western blot-probed with anti-SAP97 (mAb-1974) (lanes 1-4) or anti-SAP90 (rAb-63) (lane 5). Lane 1, total cell extracts from nontransfected COS 7 cells; lane 2, cell extracts from nontransfected cells; lane 3, cells transfected with GFP-tagged SAP97; lane 4, cells transfected with GFP-S97N-PDZ1/2; lane 5, cells doubly transfected with SAP90/PSD-95 and Kv1.4. The arrow marks the position of endogenous SAP97. (B) Western blots of GFP-tagged S97N1-65-GK transfected CACO-2 cells immunoprecipitated with rAb-PDZ1-2, which recognizes the PDZ1-2 domain of SAP97. Blots from supernatant (S) or immunoprecipitates (P) were incubated either with mAb-197.4 (SAP97) (upper panel) or mAb-GFP (S97N1-65-GK) (lower panel).

a mechanism that involves its amino-terminal cysteines or putative multimerization domain. As a final test of whether S97N-mediated localization could be due to the interaction of this domain with endogenous SAP97 in CACO-2 cells, immunoprecipitation experiments were conducted. In the first, GFP-tagged SAP97 and S97N-PDZ1/2 were transfected into COS-7 cells and cell extracts subjected to immunoprecipitation with GFP antibody followed by immuno-blotting with mAb197.4. Here only GFP-fusion proteins and not endogenous SAP97 could be detected in the immunoprecipitates (Fig. 6A). In the second experiment, CACO-2 cells were transfected with GFP-tagged S97N1-65-GK and cell extracts were immunoprecipitated with antibody rAb-PDZ12, which recognizes the endogenous SAP97 but not S97N1-65-GK. Only the endogenous SAP97 was detected in the immunoprecipitates (Fig. 6B). GFP-S97N1-65-GK stayed only in the supernatant. As a positive control, extracts of COS-7 cells doubly transfected with SAP90 and Kv1.4 were immunoprecipitated with Kv1.4 antibody. In western blots incubated with SAP90 antibody, a 95 kDa band is visible (Fig. 6A), indicating that conditions used for immunoprecipitation permit proteinprotein interactions to be maintained. Thus these data show that the localization of GFP S97N fusion proteins is not mediated by stable association with the endogenous SAP97 and that other proteins mediate its localization. S97N1-65 directs attachment of SAP97 to the lateral membrane A striking feature of SAP97 in epithelial cells is its stable association with the cortical cytoskeleton along the lateral membrane (Reuver and Garner, 1998). To identify which domains in SAP97 play a role in its attachment to the cortical cytoskeleton, two assays were employed. In the first, live cells were briefly extracted with 0.5% Triton X-100, a condition that extracts soluble and most integral membrane proteins but leaves the cortial cytoskeleton intact (Kaiser et al., 1989; Nelson et al., 1990; Nelson and Veshnock, 1986; 1987; Reuver and Garner, 1998). Under these conditions, GFP-tagged SAP97 (Fig. 7) as well as S97N-containing molecules, such as ∆PDZ1-3, ∆SH3, ∆I3, ∆GK, S97N-PDZ1-3, S97N-GK and S97N-PDZ1/2, were retained at sites of cell-cell contact, indicating that they were also associated with the Triton X-100resistant cortical cytoskeleton (Figs 3A, 7). Unexpectedly, constructs lacking S97N (∆S97N, GK, PDZ1-3 and SH3-I3GK) exhibited a cytoplasmic distribution and were readily

extracted by Triton X-100 perfusion (Fig. 3A), even though constructs like PDZ1-3 and SH3-I3-GK contain a protein 4.1 binding site. These data indicate that without S97N, protein 4.1 binding does not occur in vivo. In addition, we found that S97N1-65-GK was also retained at sites of cell-cell contact after Triton-X-100 extraction, yet S97N66-186-GK was easily removed (Figs 5B, 7), indicating that the first 65 amino acids provide an attachment point for SAP97 to the cortical cytoskeleton as well as a localization signal. As a second measure of the ability of S97N to mediate the association of SAP97 with the actin-cortical cytoskeleton, cells were treated with Latrunculin B, a drug that binds globular actin with high affinity and disrupts F-actin (Spector et al., 1989). Treatment of CACO-2 cells for 5 minutes with 5 µM Latrunculin B caused a profound disruption of cortical F-actin and a redistribution of endogenous SAP97 from sites of cellcell contact to an even cytoplasmic pattern (Reuver and Garner, 1998). GFP-tagged SAP97 as well as S97N1-65-GK (Fig. 8), S97N-PDZ1/2 and S97N-GST (data not shown) became diffuse in the cytoplasm after treatment with Latrunculin B, demonstrating that the interaction of S97N with the epithelial lateral membrane requires an intact cortical cytoskeleton. Taken together, these data show that S97N1-65 is not only required for the localization but also is an attachment point for SAP97 to the cortical cytoskeleton. DISCUSSION Selective subcellular targeting of SAP97 In this study, we have examined the contribution of individual domains to the localization of SAP97 to the epithelial lateral membrane. We show that amino acid residues preceding PDZ1 are required to direct the localization of SAP97 to sites of cellcell contact as well as to provide at least one point of attachment to the cortical cytoskeleton. This region alone can direct the lateral membrane localization of individual SAP97 domains as well as heterologous protein GST. Furthermore, it can provide a lateral membrane localization signal to two non-localized members of the SAP97 family: SAP90/PSD-95 and SAP102. Taken together, these findings indicate that the N terminus of SAP97 plays a critical role in the selective association of SAP97 with its specific membrane specializations. Our analysis of SAP97 revealed that the PDZ, SH3 and GK domains, as well as the I3 insert, do not contain functionally

Subcellular targeting of SAP97 2373 independent epithelial lateral membrane localization signals. Nonetheless, we observed that removal of either PDZ1-2 or I3 caused about a 50% decrease in localization efficiency. These regions do not appear to act synergistically, since no further decrease in efficiency was detected when both were deleted (S97N-GK). Presumably, their effect on localization is due to the ability of these regions to interact with components of the lateral membrane. This conclusion is consistent with recent in vitro studies on the human homologue of SAP97, hDlg, showing that PDZ1-2 and I3, but not PDZ3, SH3 or GK, have binding sites available to them at the lateral membrane of permeabilized epithelial cells (Lue et al., 1996). Potential lateral membrane binding partners for PDZ1-2 and I3 regions of SAP97/hDlg are non-erythroid isoforms of protein 4.1, which are present at the epithelial lateral membrane (Lue et al., 1994, 1996; Schischmanoff et al., 1997). In vitro, erythroid protein 4.1 binds the I3 insert of SAP97/hDlg as well as the hinge region between PDZ1 and PDZ2 (Lue et al., 1994, 1996; Marfatia et al., 1996). Nonetheless, data presented here demonstrate that these protein 4.1 binding sites are not critical

for directing the lateral membrane localization in epithelial cells but may facilitate the stable association of SAP97/hDlg with the lateral membrane. In erythrocytes, where SAP97/hDlg is also expressed (Marfatia et al., 1996) the situation is less clear. Here, protein 4.1 plays a critical role in maintaining membrane structural integrity via its stabilization of spectrin-actin linkage at the plasma membrane (Discher et al., 1993) and its association with glycophorin C, p55 and SAP97/hDlg (Ruff et al., 1991; Marfatia et al., 1995; 1996). In erythrocytes from patients with hereditary elliptocytosis lacking protein 4.1, neither glycophorin C, p55 nor SAP97/hDlg are present at the plasma membrane (Chishti et al., 1996; Marfatia et al., 1996). This suggests that protein 4.1 plays a vital role in maintaining the association of SAP97/hDlg with the erythrocyte plasma membrane. However, it remains unclear whether the loss of SAP97/hDlg from the plasma membrane is due to a general effect on cytoskeletal structure or to the loss of protein 4.1 as a binding site. At the epithelial septate junction in Drosophila, a different mechanism appears to be used to direct the localization of Dlg,

Fig. 7. Triton X-100 extraction of CACO-2 cells transfected with SAP97 deletion constructs. Small clusters of CACO-2 cells transfected with either GFP-tagged SAP97 (a, b), S97N-PDZ1/2 (c, d), S97N1-65-GK (e, f) or S97N66-186-GK (g, h) were imaged with FITC filter set either prior to (−TX-100) or after (+TX-100) perfusing cells for 1.5 minutes with 0.5% Triton X-100.

2374 H. Wu and others variations to occur at least at the I3 site (Lue et al., 1994; Müller et al., 1995; C. C. Garner, unpublished data), which in turn may provide an increased repertoire of binding partners at this position. Moreover, utilizing the unique sequences in SAPs, i.e. in the N terminus, for subcellular targeting rather than a conserved domain, such as PDZ2, provides a flexible system to differentially localize individual SAPs. Significantly, replacing the N-terminal domains of SAP102 and SAP90 with S97N, directed these SAPs to the epithelial lateral membrane. However, whereas S97N/S102PDZ1-GK localized with a similar efficiency to SAP97 (approx. 60%), only about 33% of the cells transfected with S97N-S90PDZ1GK exhibited a clear lateral membrane localization. This suggests that like SAP97/hDlg, SAP102 contains an element which acts to increase its localization efficiency that is missing in SAP90. Although we did not map this element in SAP102, there are several indications that it is situated between its SH3 and GK domains. The most compelling is that SAP102 contains inserts between its SH3 and GK domains which resemble those in some isoforms of SAP97/hDlg (Lue et al., 1994; Muller et al., 1995, 1996). In contrast, SAP90 lacks inserts at this site and localizes with a similar low efficiency to the ∆I3 deletion construct of SAP97 (42%).

Fig. 8. Effect of disassembling F-actin on the retention of GFPtagged SAP97 or S97N1-65-GK at the epithelial lateral membrane. CACO-2 cells grown on coverslips were transfected with either GFPtagged SAP97 or S97N1-65-GK. 24 hours later, cultures were treated with either 5 µM Latrunculin B (LaB) in DMSO or DMSO alone for 5 minutes at 37°C before being fixed with 3.7% formaldehyde. Cells were stained with rhodamine-conjugated phalloidin to detect F-actin. Latrunculin B caused F-actin, in stress fibers and at the lateral membrane, to breakdown into small clumps. Under these conditions, GFP-tagged SAP97 and S97N1-65-GK relocalized from the lateral membrane to the cytoplasm.

a Drosophila homolog of SAP97/hDlg (Hough et al., 1997). Here, two domains, PDZ2 and HOOK, were found to act in concert to localize Dlg to septate junctions. PDZ2 was shown to possess the septate junction targeting signal. The HOOK region, which is homologous to I3 and situated between the SH3 and GK domains, is required for association with the plasma membrane. The N terminus of Dlg, which is quite short (20 amino acid residues), is apparently non-essential for subcellular targeting (Hough et al., 1997). Given that PDZ1-2 and I3 are not necessary for subcellular targeting of SAP97, moving the targeting signal to the N terminus allows sequence

Attachment of SAP97 to the epithelial lateral membrane In a recent study, we have found that cell-cell adhesion between epithelial cells mediated by E-cadherin induces the translocation of SAP97/hDlg from cytoplasmic pools to the plasma membrane at sites of cell-cell contact (Reuver and Garner, 1998). The attachment of SAP97/hDlg to the lateral membrane was found to be mediated by its tight association with the cortical actin cytoskeleton, requiring an F-actin linkage to tether it to the E-cadherin adhesion complex (Reuver and Garner, 1998). In this study, we have sought to identify the regions that not only target SAP97/hDlg to the lateral plasma membrane but also mediate its association with the Triton X100-resistant cortical cytoskeleton, a feature characteristic of cortical cytoskeletal proteins (Luna and Hitt, 1992; Nelson et al., 1990; Reuver and Garner, 1998). Surprisingly, the putative protein 4.1 binding sites, I3 and hinge region between PDZ1 and PDZ2, were not required to maintain SAP97 at the lateral membrane of CACO-2 cells extracted with Triton X-100. Furthermore, the three PDZ domains as well as the SH3 and GK domains were found to be non-essential for the association of SAP97 with the Triton X-100-resistant cytoskeleton. Instead, this characteristic was found to be mediated by S97N. This conclusion is based on the ability of SAP97 deletion constructs containing S97N to (1) be retained at sites of cellcell contact after extraction with Triton X-100 and (2) be released into the cytoplasm after disrupting cortical actin with Latrunculin B. Interestingly, deletion analysis of S97N reveals that both the lateral membrane targeting signal and the region that mediates an association with the cytoskeletal are situated within the first 65 N-terminal amino acid residues of SAP97. This indicates that the receptor for S97N at the lateral membrane is likely to be a component of the cortical cytoskeleton or an integral membrane protein tightly tethered to this structure. The concept that the N termini of SAPs confer fundamental properties to this protein family is not restricted to SAP97.

Subcellular targeting of SAP97 2375 Recent studies on SAP90/PSD-95 and Chapsyn-110/PSD-93 have revealed that sequences preceeding their first PDZ domain (residues 1-60 (S90N) and 1-93 (C110N), respectively) mediate clustering of ion channels by these SAPs in transfected cells (Kim et al., 1996; Hsueh et al., 1997; Topinka and Bredt, 1998). Mechanistically, clustering is mediated by two properties of these N termini. The first is membrane attachment through palmitoylation of a pair of cysteine residues at C3 and C5 in S90N or C5 and C7 in C110N (Topinka and Bredt, 1998) (Fig. 5A). The second is multimerization (Hsueh et al., 1997; Topinka and Bredt, 1998), which can either be homo- or heteromer between SAP90/PSD-95 and Chapsyn-110/PSD-93 and is thus thought to be mediated by the region of shared homology between these two N termini (S90N1-60; C1101-93) (Fig. 5A). Interestingly, multimerization can be observed in cell extracts run on non-reducing SDS-gels (Hsueh et al., 1997), due to the labile nature of palmitoylation and the rapid formation of intramolecular disulfide bridges (Hsueh et al., 1997; Topinka and Bredt, 1998). Although SAP97 shares about 60% sequence homology to the multimerization domains in SAP90/PSD-95 and Chapsyn-110/PSD-93 (residues 100-186) and has two cysteine residues (C66 and C73) (Fig. 5A), it does not cluster ion channels in heterologous cells (Kim and Sheng, 1996) nor form heteromeric (Hsueh et al., 1997) or homomeric multimers, as evaluated on non-reducing SDS-gels. Furthermore, recombinant GFP-tagged SAP97 constructs could not be co-immunoprecipitated with endogenously expressed SAP97 in COS or CACO-2 cells. Thus, in contrast to SAP90/PSD-95 and Chapsyn-110/PSD-93, the N terminus of SAP97 does not appear to be involved in the formation of multimers. Instead, data presented here demonstrate that S97N, in particular residues 1-65, confers a novel property to SAP97 as a lateral membrane localization signal. Significantly, this signal lies outside of its putative multimerization domain, does not require its N-terminal cysteines and appears to interact with a component of the Triton X-100-resistant cortical actin cytoskeleton. In epithelial cells, four proteins have been characterized that share a similar domain structure to SAP97/hDlg. Three, ZO-1, ZO-2 and ZO-3, contain three PDZ domains, an SH3 domain, a guanylate kinase-like domain and a long C-terminal extension. They are exquisitely restricted in polarized epithelial to tight junctions, where they interact with each other with actin filaments and the cytoplasmic domain of occludin (Haskins et al., 1998). At present, tight junction targeting domains in this subfamily of membrane-associated guanylate kinase homologs (MAGUKs) have not been described, but given their distinct spatial distribution from SAP97 (Reuver and Garner, 1998), it appears unlikely that they interact with SAP97 or affect its localization along the lateral membrane. The fourth protein, CASK, contains a single PDZ domain, an SH3 domain, a GK domain and an N-terminal calmodulindependent protein kinase-like domain (CaMK), and is reported to exhibit a similar lateral membrane localization to SAP97 (Brecher Cohen et al., 1997). The mechanism underling the localization of CASK is not known. Interestingly, a short segment of CASK situated between residues 350-400 shares a moderate degree of sequence similarity to residues 30-55 in SAP97 (data not shown). Whether this is just a coincidence or represents a lateral membrane targeting signal in CASK remains to be tested.

Together, these data indicate that the N domains in SAPs confer unique fundamental properties to each SAP. In the cases of SAP90/PSD-95 and Chapsyn/PSD-93, the N domains appear to play a central role in the clustering of channels, a feature that is essential to achieve a high density of ligand and voltage-gated channels within the postsynaptic density (PSD). Alternatively, S97N appears to perform a primary role in the attachment of SAP97 to the cortical cytoskeleton. Based on its localization along the epithelial lateral membrane and along unmyelinated axons (Müller et al., 1995), this feature would allow the even distribution of channels and other proteins along the length of these membrane specializations. An important but unresolved issue is whether the N-domains in SAP97, SAP90/PSD-95, SAP102 and Chapsyn-110/PSD-93 also play a role in their differential localization in neurons or whether other regions are required. We would like to thank Stefan Kindler for critical reading of the manuscript. This work was supported by Keck Foundation (931360) and the National Institutes of Health (P50 HD32901; AG 12978-02, AG 06569-09).

REFERENCES Brenman, J. E., Christopherson, K. S., Craven, S. E., McGee, A. W. and Bredt, D. S. (1996). Cloning and characterization of postsynaptic density 93, a nitric oxide synthase interacting protein. J. Neurosci. 16, 7407-7415. Brecher Cohen, A. R., Cohen, D. W. and Anderson, J. M. (1997). CASK, the human homolog of the C. elegans protein LIN-2, interacts with syndecan-2 in hepatocytes. Mol. Biol. Cell 8, 177a. Budnik, V. (1996). Synapse maturation and structural plasticity at Drosophila neuromuscular junctions. Curr. Opin. Neurobiol. 6, 858-867. Chishti, A. H., Palek, J., Fisher, D., Maalouf, G. J. and Liu, S. C. (1996). Reduced invasion and growth of Plasmodium falciparum into elliptocytic red blood cells with a combined deficiency of protein 4.1, glycophorin C, and p55. Blood 87, 3462-3469. Discher, D., Parra, M., Conboy, J. G. and Mohandas, N. (1993). Mechanochemistry of the alternatively spliced spectrin-actin binding domain in membrane skeletal protein 4.1. J. Biol. Chem. 268, 7186-7195. Garner, C. C. and Kindler, S. (1996). Synaptic proteins and the assembly of synaptic junctions. Trends Cell Biol. 6, 429-433. Haskins, J., Gu, L., Wittchen, E. S., Hibbard, J. and Stevenson, B. R. (1998). ZO-3, a novel member of the MAGUK protein family found at the tight junction, interacts with ZO-1 and occludin. J. Cell Biol. 141, 199-208. Hough, C. D., Woods, D. F., Park, S. and Bryant, P. J. (1997). Organizing a functional junctional complex requires specific domains of the Drosophila MAGUK Disc large. Genes Dev. 11, 3242-3253. Hsueh, Y. P., Kim, E. and Sheng, M. (1997). Disulfide-linked head-to-head multimerization in the mechanism of ion channel clustering by PSD-95. Neuron 18, 803-814. Hunt, C. A., Schenker, L. J. and Kennedy, M. B. (1996). PSD-95 is associated with the postsynaptic density and not with the presynaptic membrane at forebrain synapses. J. Neurosci. 16, 1380-1388. Kaiser, H. W., O’Keefe, E. and Bennett, V. (1989). Adducin: Ca++-dependent association with sites of cell-cell contact. J. Cell Biol. 109, 557-569. Kim, E., Cho, K. O., Rothschild, A. and Sheng, M. (1996). Heteromultimerization and NMDA receptor-clustering activity of Chapsyn110, a member of the PSD-95 family of proteins. Neuron 17, 103-113. Kim, E., Naisbitt, S., Hsueh, Y. P., Rao, A., Rothschild, A., Craig, A. M. and Sheng, M. (1997). GKAP, a novel synaptic protein that interacts with the guanylate kinase-like domain of the PSD-95/SAP90 family of channel clustering molecules. J. Cell Biol. 136, 669-678. Kim, E. and Sheng, M. (1996). Differential K+ channel clustering activity of PSD-95 and SAP97, two related membrane-associated putative guanylate kinases. Neuropharm. 35, 993-1000. Kistner, U., Garner, C. C. and Linial, M. (1995). Nucleotide binding by the synapse associated protein SAP90. FEBS Lett. 359, 159-163. Kistner, U., Wenzel, B. M., Veh, R. W., Cases-Langhoff, C., Garner, A. M. Appeltauer, U., Voss, B., Gundelfinger, E. D. and Garner, C. C. (1993).

2376 H. Wu and others SAP90, a rat presynaptic protein related to the product of the Drosophila tumor suppressor gene dlg-A. J. Biol. Chem. 268, 4580-4583. Kuhlendahl, S., Spangenberg, O., Konrad, M., Kim, E. and Garner, C. C. (1998). Functional analysis of the guanylate kinase-like domain in the synapse-associated protein SAP97. Eur. J. Biochem. 252, 305-313. Lue, R. A., Brandin, E., Chan, E. P. and Branton, D. (1996). Two independent domains of hDlg are sufficient for subcellular targeting: the PDZ1-2 conformational unit and an alternatively spliced domain. J. Cell Biol. 135, 1125-1137. Lue, R. A., Marfatia, S. M., Branton, D. and Chishti, A. H. (1994). Cloning and characterization of hdlg: the human homologue of the Drosophila discs large tumor suppressor binds to protein 4.1. Proc. Nat. Acad. Sci. USA 91, 9818-9822. Luna, E. J. and Hitt, A. L. (1992). Cytoskeleton-plasma membrane interactions. Science 258, 955-964. Marfatia, S. M., Cabral, J. H., Lin, L., Hough, C., Bryant, P. J., Stolz, L. and Chishti, A. H. (1996). Modular organization of the PDZ domains in the human discs-large protein suggests a mechanism for coupling PDZ domain-binding proteins to ATP and the membrane cytoskeleton. J. Cell Biol. 135, 753-766. Marfatia, S. M., Leu, R. A., Branton, D. and Chishti, A. H. (1995). Identification of the protein 4.1 binding interface on glycophorin C and p55, a homologue of the Drosophila discs-large tumor suppressor protein. J. Biol. Chem. 270, 715-719. Musacchio, A., Gibson, T., Lehto, V. P. and Saraste, M. (1992). SH3 – an abundant protein domain in search of a function. FEBS Lett. 307, 55-61. Müller, B. M., Kistner, U., Kindler, S., Chung, W. J., Kuhlendahl, S., Fenster, S. D., Lau, L. F., Veh, R. W., Huganir, R. L., Gundelfinger, E. D. and Garner, C. C. (1996). SAP102, a novel postsynaptic protein that interacts with NMDA Receptor complexes in vivo. Neuron. 17, 255-265. Müller, B. M., Kistner, U., Veh, R. W., Cases-Langhoff, C., Becker, B., Gundelfinger, E. D. and Garner, C. C. (1995). Molecular characterization and spatial distribution of SAP97, a novel presynaptic protein homologous to SAP90 and the Drosophila discs-large tumor suppressor protein. J. Neurosci. 15, 2354-2366. Munro, S. and Pelham, H. R. B. (1987). A C-terminal signal prevents secretion of luminal ER proteins. Cell 48, 899-907. Nelson, W. J., Shore, E. M., Wang, A. Z. and Hammerton, R. W. (1990). Identification of a membrane-cytoskeletal complex containing the cell adhesion molecule uvomorulin (E-cadherin), ankyrin, and fodrin in MadinDarby canine kidney epithelial cells. J. Cell Biol. 110, 349-357.

Nelson, W. J. and Veshnock, P. J. (1986). Dynamics of membrane-skeleton (fodrin) organization during development of polarity in Madin-Darby canine kidney epithelial cells. J. Cell Biol. 103, 1751-1765. Nelson, W. J. and Veshnock, P. J. (1987). Modulation of fodrin (membrane skeleton) stability by cell-cell contact in Madin-Darby canine kidney epithelial cells. J. Cell Biol. 104, 1527-1537. Reuver, S. M. and Garner, C. C. (1998). E-cadherin mediated cell adhesion recruits SAP97 into the cortical cytoskeleton. J. Cell Sci. 111, 1071-1080. Ruff, P., Speicher, D. W. and Husain-Chishti, A. (1991). Molecular identification of a major palmitoylated erythrocyte membrane protein containing the src homology 3 motif. Proc. Nat. Acad. Sci. USA 88, 65956599. Schischmanoff, P. O., Yaswen, P., Parra, M. K., Lee, G., Chasis, J. A., Mohandas, N. and Conboy, J. G. (1997). Cell shape-dependent regulation of protein 4.1 alternative pre-mRNA splicing in mammary epithelial cells. J. Biol. Chem. 272, 10254-10259. Sheng, M. (1996). PDZs and receptor/channel clustering: rounding up the latest suspects [comment]. Neuron. 17, 575-578. Spector, I., Shochet, N. R., Blasberger, D. and Kashman, Y. (1989). Latrunculins – novel marine macrolides that disrupt microfilament organization and affect cell growth: I. Comparison with cytochalasin D. Cell Mot. Cyt. 13, 127-144. Takeuchi, M., Hata, Y., Hirao, K., Toyoda, A., Irie, M. and Takai, Y. (1997). SAPAPs. A family of PSD-95/SAP90-associated proteins localized at postsynaptic density. J. Biol. Chem. 272, 11943-11951. Tejedor, F. J., Bokhari, A., Rogero, O., Gorczyca, M., Zhang, J., Kim, E., Sheng, M. and Budnik, V. (1997). Essential role for dlg in synaptic clustering of Shaker K+ channels in vivo. J. Neurosci. 17, 152-159. Thomas, U., Kim, E., Kuhlendahl, S., Koh, Y. H., Gundelfinger, E. D., Sheng, M., Garner, C. C. and Budnik, V. (1997a). Synaptic clustering of the cell adhesion molecule fasciclin II by Discs-large and its role in the regulation of presynaptic structure. Neuron 19, 787-799. Thomas, U., Phannavong, B., Muller, B., Garner, C. C. and Gundelfinger, E. D. (1997b). Functional expression of rat synapse-associated proteins SAP97 and SAP102 in Drosophila dlg-1 mutants: effects on tumor suppression and synaptic bouton structure. Mech. Dev. 62, 161174. Topinka, J. R. and Bredt, D. S. (1998). N-terminal palmitoylation of PSD95 regulates association with cell membranes and interaction with K+ channel Kv1.4. Neuron 20, 125-134.