N-Ethylmaleimide-sensitive Factor Attachment Protein α (αSNAP) Regulates Matrix Adhesion and Integrin Processing in Human Epithelial Cells*

Background: Vesicle trafficking plays important roles in regulating cell adhesion and motility. Results: Depletion of the membrane fusion protein, N-ethylmaleimide-sensitive factor attachment protein α (αSNAP), induced cell detachment, whereas overexpression of this vesicle regulator increased cell-matrix adhesion. Conclusion: αSNAP promotes cell-matrix adhesion by controlling integrin trafficking and assembly of focal adhesions. Significance: The αSNAP level may serve as a molecular rheostat controlling matrix adhesion and cell motility under normal conditions and in diseases. Integrin-based adhesion to the extracellular matrix (ECM) plays critical roles in controlling differentiation, survival, and motility of epithelial cells. Cells attach to the ECM via dynamic structures called focal adhesions (FA). FA undergo constant remodeling mediated by vesicle trafficking and fusion. A soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein α (αSNAP) is an essential mediator of membrane fusion; however, its roles in regulating ECM adhesion and cell motility remain unexplored. In this study, we found that siRNA-mediated knockdown of αSNAP induced detachment of intestinal epithelial cells, whereas overexpression of αSNAP increased ECM adhesion and inhibited cell invasion. Loss of αSNAP impaired Golgi-dependent glycosylation and trafficking of β1 integrin and decreased phosphorylation of focal adhesion kinase (FAK) and paxillin resulting in FA disassembly. These effects of αSNAP depletion on ECM adhesion were independent of apoptosis and NSF. In agreement with our previous reports that Golgi fragmentation mediates cellular effects of αSNAP knockdown, we found that either pharmacologic or genetic disruption of the Golgi recapitulated all the effects of αSNAP depletion on ECM adhesion. Furthermore, our data implicates β1 integrin, FAK, and paxillin in mediating the observed pro-adhesive effects of αSNAP. These results reveal novel roles for αSNAP in regulating ECM adhesion and motility of epithelial cells.


Integrin-based adhesion to the extracellular matrix (ECM) plays critical roles in controlling differentiation, survival, and motility of epithelial cells. Cells attach to the ECM via dynamic structures called focal adhesions (FA). FA undergo constant remodeling mediated by vesicle trafficking and fusion. A soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein ␣ (␣SNAP) is an essential mediator of membrane fusion; however, its roles in regulating ECM adhesion and cell motility remain unexplored. In this study, we found that siRNA-mediated knockdown of ␣SNAP induced detachment of intestinal epithelial cells, whereas overexpression of ␣SNAP increased ECM adhesion and inhibited cell invasion. Loss of ␣SNAP impaired
Golgi-dependent glycosylation and trafficking of ␤1 integrin and decreased phosphorylation of focal adhesion kinase (FAK) and paxillin resulting in FA disassembly. These effects of ␣SNAP depletion on ECM adhesion were independent of apoptosis and NSF. In agreement with our previous reports that Golgi fragmentation mediates cellular effects of ␣SNAP knockdown, we found that either pharmacologic or genetic disruption of the Golgi recapitulated all the effects of ␣SNAP depletion on ECM adhesion. Furthermore, our data implicates ␤1 integrin, FAK, and paxillin in mediating the observed pro-adhesive effects of ␣SNAP. These results reveal novel roles for ␣SNAP in regulating ECM adhesion and motility of epithelial cells.
Adhesion of epithelial cells to extracellular matrix (ECM) 2 is a key determinant of tissue integrity and morphogenesis. It allows formation of simple planar epithelial sheets as well as complex three-dimensional tubules, ducts, and glands (1,2). Adhesion to ECM is critical for a steady-state migration of epithelial cells in constantly self-renewing tissues such as intestinal mucosa (3,4). It also drives disease-related cell motility during wound healing in inflamed mucosa or invasion of metastatic tumor cells (5,6). Epithelial cells attach to ECM via specialized adhesion structures formed at their basal surface. Focal adhesions (FA) represent the most abundant and important type of ECM adhesions. FA-mediated adhesion is primarily dependent on integrins, a family of heterodimeric transmembrane receptors that are engaged in direct interactions with ECM components (7,8). Integrins are activated, clustered, and linked to the underlying cytoskeleton by a number of cytoplasmic scaffolding molecules including paxillin, vinculin, and talin (9,10). Additionally, FA are enriched in signaling molecules. Among these signaling molecules non-receptor kinases, such as members of the focal adhesion kinase (FAK) and Src families, are known to drive FA biogenesis by activating scaffolding proteins and stimulating actin polymerization at the ECM attachment sites (11,12).
FA are very dynamic structures that undergo a constant remodeling. A simplified model of FA dynamics in motile cells implies that FA become assembled and disassembled at the migrating leading edge and the trailing end of the cell, respectively (13,14). Molecular components of disassembled FA get delivered to the leading edge where they are reused during formation of new adhesion complexes. Such recycling of molecular constituents is important for the continuity of FA biogenesis and ECM adhesion. Vesicle trafficking is known to be a major mechanism of FA remodeling. It is required for endocytosis and recycling of FA proteins as well as for the delivery of newly synthesized FA components via exocytosis from the Golgi to the plasma membrane (15,16). Generally, exocytosis and recy-cling represent complex processes involving hierarchically organized cascades of vesicle tethering, docking, and fusion with the plasma membrane (17,18). Each step of this cascade is controlled by specific multiprotein machinery to ensure fidelity and efficiency of inter-membrane interactions. Membrane fusion is the final and rate-limiting step of this process resulting in either insertion of designated molecules into the plasma membrane or their release from the cells. Membrane fusion is mediated by the SNARE (soluble N-ethylmalemide-sensitive factor associated receptor) protein complex (19,20). Direct interactions of different SNARE proteins located on the vesicle and the target membrane bring two lipid bilayers in close proximity, driving their fusion. A number of previous studies implicated several SNARE proteins in regulation of integrin trafficking, FA assembly, and cell-matrix adhesion. Cell adhesion and motility appear to be orchestrated by SNAREs located in different cellular compartments including plasma membrane-resident syntaxins 3 and 4 and SNAP23 (21,22), endosomal vesicle-associated membrane proteins 3 (22)(23)(24), and Golgi/endosomal syntaxin 6 (24 -26).
Membrane fusion yields extremely stable cis-SNARE complexes that must be disassembled and reused for new fusion events. Disassembly of these post-fusion SNARE complexes requires an oligomeric ATPase, N-ethylmaleimide-sensitive factor (NSF) and its adaptors, soluble NSF-attachment proteins (SNAPs) (27,28). Ubiquitously expressed ␣SNAP and neuronal ␤SNAP are capable of interacting simultaneously with SNAREs and NSF via distinct binding sites (28 -30). These interactions result in NSF recruitment to the SNARE complex, stimulation of its ATPase activity, and transduction of conformational changes from NSF leading to disassembly of the SNARE cylinder. Overexpression of a dominant-negative NSF mutant or introduction of NSF-inhibiting peptides was shown to interrupt integrin trafficking and diminish ECM adhesion and migration of epithelial cells and leukocytes (31)(32)(33). Because ␣SNAP is a critical regulator of NSF activity, with a number of NSF-independent cellular functions (34 -37), it is essential to characterize its effects on ECM adhesion and cell motility. Surprisingly, this important question has not been previously addressed. In this study, we identified ␣SNAP as a critical positive regulator of ECM adhesion of human epithelial cells via multiple mechanisms that control FA assembly and Golgi-dependent maturation and trafficking of ␤1 integrin.
Cell Culture-SK-CO15 (a gift from Dr. E. Rodriguez-Boulan, Weill Medical College of Cornell University, NY) human colonic epithelial cells were cultured in DMEM supplemented by 10% fetal bovine serum as previously described (38,39). Cells were grown in standard T75 flasks and, for immunolabeling/ confocal microscopy experiments, were seeded on either collagen-coated, permeable polycarbonate filters 0.4-m pore size (Costar, Cambridge, MA) or collagen-coated coverslips. For biochemical experiments, the cells were seeded on 6-well plastic plates.
RNA Interference-Small interfering (si) RNA-mediated knockdown of ␣SNAP, NSF, GBF1, ␤1, and ␤4 integrins was carried out as previously described (35,40,42). Individual siRNA duplexes, GAAGGUGGCUGGUUACGCU (duplex 1) and CAGAGUUGGUGGACAUCGA (duplex 2; Dharmacon, Lafayette, CO), were used to down-regulate ␣SNAP expression, whereas knockdown of other targets was performed by using gene-specific siRNA pools purchased either from Dharmacon or Santa Cruz Biotechnology (Dallas, TX). A noncoding siRNA duplex-2 (Dharmacon) served as a control. SK-CO15 cells were transfected using DharmaFect1 in Opti-MEM I medium (Invitrogen) according to the manufacturer's protocol with a final siRNA concentration of 50 nM. Cells were harvested and analyzed 3-4 days after transfection.
Preparation of ␣SNAP-overexpressing Epithelial Cells-For rescue experiments, the coding sequence of bovine ␣SNAP (gift from Dr. Reinhard Jahn, Max Planck Institute for Biophysical Chemistry, Gottingen, Germany) was cloned in pLXSN retroviral vector (Clontech, Mountain View, CA) as a BamHI-EcoRI fragment. The bovine ␣SNAP transcript has 4 nucleotide mismatches in the corresponding sequences and is not targeted by siRNA duplexes 1 and 2; the duplexes used to deplete human ␣SNAP. For retroviral particle production, ProPak-A cells (ATCC, Manassas, VA) were transfected with either empty control vector or bovine ␣SNAP-containing vector using a Trans-IT 293 transfection reagent (MirusBio, Madison, WI). To harvest the viruses, infected cells grown to 75% confluence were incubated for 16 h at 37°C. Collected media was passed through a 0.45-m syringe filter and stored at Ϫ80°C. SK-CO15 cells plated at 30 -40% confluence were exposed overnight to viruses preincubated with 4 g/ml of Polybrene. Stable cell lines expressing either bovine ␣SNAP or their appropriate controls were selected with 1.2 g/ml of G418. For expression of wild-type (wt) ␣SNAP, its coding sequence was excised (BamHI and EcoRI, blunted by Klenow) and inserted into lentivector pWPT-GFP (obtained from Addgene, Boston, MA) at the BamHI site, which was also blunted by Klenow DNA polymerase. The site-directed mutagenesis of M105I and L294A was carried out using QuikChange Sitedirected Mutagenesis Kit according to the manufacturer's instructions (Agilent Technologies, Santa Clara, CA). To produce lentiviral supernatants, HEK293LT cells (6-cm dish) were co-transfected with 5 g of lentiplasmids, 4 g of psPAX2, and 1 g of pCMV-VSV-G plasmids using Lipofectamine 2000. Virus-containing supernatants were collected and combined from 36 to 60 h post-transfection and used to infect cells in the presence of 8 g/ml of Polybrene (Sigma).
Immunofluorescence Labeling and Image Analysis-Epithelial cell monolayers were fixed/permeabilized in 100% methanol for 20 min at Ϫ20°C. Fixed cells were blocked in HEPESbuffered Hanks' balanced salt solution (HBSS ϩ ) containing 1% bovine serum albumin (blocking buffer) for 60 min at room temperature and incubated for another 60 min with primary antibodies diluted in blocking buffer. Cells were then washed, incubated for 60 min with Alexa dye-conjugated secondary antibodies, rinsed with blocking buffer, and mounted on slides with ProLong Antifade medium (Invitrogen). Immunofluorescence labeled cell monolayers were examined with an Olympus FluoView 1000 confocal microscope (Olympus America, Center Valley, PA). The Alexa Fluor 488 and 568 signals were imaged sequentially in frame-interlace mode to eliminate cross-talk between channels. The images were processed using the Olympus FV10-ASW 2.0 Viewer software and Adobe Photoshop. Images shown are representative of at least 3 experiments, with multiple images taken per slide.
Cell Adhesion and Matrigel Invasion Assays-Cell adhesion assay was performed as previously described (43). Briefly, 24-well plates were coated with collagen I and washed with sterile PBS. Control and ␣SNAP overexpressing cells were trypsinized, harvested, counted with a hemocytometer, and resuspended in DMEM. Where indicated, cells were preincubated with either isotype control or MAB13 antibodies (5 g/ml) for 2 h (44). 1 ϫ 10 4 cells were added to each well and cells were allowed to adhere to the ECM for 30 min at 37°C. After incubation, unattached cells were removed and attached cells were fixed and labeled using a DIFF stain kit (IMEB Inc., San Marcos, CA). Adherent cells were photographed under bright field microscope.
The invasion assay was performed using commercially available BD Biocoat invasion chambers (BD Biosciences). Cells were trypsinized, resuspended in DMEM without serum, and added to the upper chamber at a concentration of 5 ϫ 10 4 cells per chamber. In the lower chamber complete growth medium containing 10% FBS as a chemoattractant was added and cells were allowed to migrate for 24 h at 37°C. Cells were fixed and stained using the same DIFF stain kit described above and nonmigrated cells were removed from the top of the membranes using cotton swabs. For both adhesion and invasion experiments, the number of cells was determined by manual counting in 15 randomly selected microscopic images obtained from 3 independent experiments.
Immunoblotting-Cells were homogenized in RIPA lysis buffer (20 mM Tris, 50 mM NaCl, 2 mM EDTA, 2 mM EGTA, 1% sodium deoxycholate, 1% Triton X-100, and 0.1% SDS, pH 7.4), containing a protease inhibitor mixture (1:100, Sigma) and phosphatase inhibitor mixtures 1 and 2 (both at 1:200, Sigma). Lysates were cleared by centrifugation (20 min at 14,000 ϫ g), diluted with 2ϫ SDS sample buffer, and boiled. SDS-polyacrylamide gel electrophoresis and immunoblotting were conducted by standard protocols with an equal amount of total protein (10 or 20 g) per lane. Protein expression was quantified by densitometry of three immunoblot images, each representing an independent experiment, with an Epson Perfection V500 photoscanner and ImageJ version 1.47 software (National Institute of Health, Bethesda, MD). Data were presented as normalized values assuming the expression levels in control siRNA-treated groups were at 100%. Statistical analysis was performed with row densitometric data using Microsoft Excel.
Real-time RT-PCR-RNA was isolated using an RNeasy mini kit and treated with DNase I prior to cDNA synthesis using an iSCRIPT cDNA synthesis kit (Bio-Rad). iTaq Universal SYBR Green supermix was used to perform real-time PCR on an Applied Biosystems ABI 9700HT REAL TIME PCR SYSTEM, using the following primers for human FAK: forward, GCCT-TATGACGAAATGCTGGGC and reverse, CCTGTCTT-CTGGACTCCATCCT. Cycle threshold values (C t ) were used to calculate the fold-change in target mRNA using the ⌬⌬C t method. The 18 S ribosomal RNA subunit was used as a reference gene.
Statistics-For Western blot quantification numerical values from individual experiments were pooled and expressed as mean Ϯ S.E. throughout. Obtained numbers were compared by two-tailed Student's t test, with statistical significance assumed at p Ͻ 0.05. Invasion and migration data were analyzed using Sigma-PLOT professional statistics software (Systat Software Inc., San Jose, CA). For analyses of variance, one-way analysis of variance with pairwise multiple tests was used for intergroup comparisons with p Ͻ 0.001.

Loss of ␣SNAP Expression Impairs ECM Adhesion of Human
Epithelial Cells-During our previous studies we made a serendipitous observation that loss of ␣SNAP expression caused a marked detachment of cultured human epithelial cells. Because this observation suggested a previously unrecognized role of ␣SNAP in regulating ECM adhesion, we decided to investigate molecular mechanisms that may determine poor adhesiveness of ␣SNAP-depleted epithelia. RNA interference (RNAi) was used to down-regulate ␣SNAP expression in SK-CO15 human intestinal epithelial cells along with a rescue approach involving overexpression of RNAi-resistant bovine ␣SNAP. Transfection with two different siRNA duplexes dramatically reduced the ␣SNAP protein level in control SK-CO15 human colonic epithelial cells (SK-neo) without affecting expression of this protein in bovine ␣SNAP-rescued cells (SK-␣SNAP; Fig. 1A). Consistent with our previous results (34), NSF expression was not modulated by the ␣SNAP level (Fig. 1A). Remarkably, depletion of ␣SNAP resulted in a substantial (up to 60%) detachment of SK-CO15 cells on day 3 post-transfection ( Fig. 1, B and C). Overexpression of bovine ␣SNAP prevented this cell detachment, indicating that the observed defects of ECM adhesion represent a specific consequence of decreased ␣SNAP expression. Beside SK-CO15 cells, loss of ␣SNAP decreased adhesiveness of HeLa, HEK-293, PC3, and HCT116 cells (data not shown), highlighting a generality of this response in different types of cultured epithelial cells.
Loss of ␣SNAP Disrupts Morphology of FA and Alters Processing of Their Major Molecular Constituents-Because FA are known to be the major structural determinants of epithelial cell attachment to ECM, we hypothesized that poor adhesiveness of ␣SNAP-depleted cells can be due to impaired FA assembly. To test this hypothesis, FA were visualized in control and ␣SNAPdepleted SK-CO15 cells by using immunolabeling of vinculin with subsequent confocal microscopy. In control cells, a signif-icant fraction of vinculin accumulated within large elongated basal clusters representing FA ( Fig. 2A, arrows). By contrast, no defined FA was detected in ␣SNAP-depleted cells where vinculin staining was diffusely distributed within the cytoplasm and cell membrane ( Fig. 2A, arrowheads). Immunoblotting analysis of major transmembrane, scaffolding, and signaling constituents of FA revealed two prominent effects of ␣SNAP depletion on molecular organization of FA. The first effect was a marked decrease in phosphorylation of paxillin and FAK accompanied by a significant expressional down-regulation of FAK (Fig. 2, B and C). The magnitude of FAK dephosphorylation (ϳ4-fold) was higher than the magnitude of total FAK down-regulation (ϳ2-fold). This difference indicates multiple mechanisms of FAK inactivation by ␣SNAP knockdown. Although the mechanism of reduced total FAK expression remains to be elucidated, our quantitative RT-PCR analysis ruled out decreased mRNA transcription (data not shown). The second effect of ␣SNAP depletion was noticeable defects in integrin maturation. Indeed, immunoblotting analysis of control SK-CO15 cells detected a major ␤1 integrin doublet (Fig. 2B) composed of a predominant slowly migrating mature protein and a faster migrating ␤1 integrin precursor (45). Remarkably, loss of ␣SNAP resulted in accumulation of ␤1 integrin precursor as manifested by a dramatic increase in the intensity of the lower band in the doublet (Fig. 2B). Likewise, ␣SNAP-depleted cells demonstrated significant accumulation of slowly migrating immature forms of ␤4 and ␣5 integrins (Fig. 2B). Immunofluorescence analysis of control SK-CO15 cells demonstrated predominant localization of ␤1 integrin on the lateral plasma membrane in the areas of cell-cell contacts (Fig. 3A, arrows). By contrast, loss of ␣SNAP resulted in diffuse intracellular distribution of ␤1 integrin (Fig. 3A, arrowheads). Interestingly, overexpression of siRNA-resistant bovine protein completely restored normal plasma membrane localization of ␤1 integrin and rescued defects of integrin maturation and phosphorylation of FA proteins in ␣SNAP-depleted SK-CO15 cells (Fig. 3, A and B). Together these results indicate that loss of ␣SNAP impairs maturation and plasma membrane delivery of integrins and interrupts signaling events essential for FA assembly. Loss of ECM Adhesion in ␣SNAP-depleted Cells Is Independent of Apoptosis-Recent studies by our group (35) and others (46) have found that down-regulation of ␣SNAP expression promotes epithelial cell apoptosis. One can therefore suggest that the observed loss of ECM adhesion in ␣SNAP-depleted cells represents a simple consequence of apoptotic cell death. To investigate this possibility, control and ␣SNAP siRNA-transfected SK-CO15 cells were treated with either vehicle or a pan-caspase inhibitor, Q-VAD-OPH (20 M). Caspase inhibition effectively prevented apoptosis in ␣SNAP-depleted cells as indicated by a complete suppression of poly(ADP-ribose) polymerase cleavage (Fig. 4A) and lack of caspase 3 activation (data not shown). However, inhibition of apoptosis did not alleviate abnormal maturation and mislocalization of ␤1 integrin caused by ␣SNAP knockdown (Fig. 4, A and B, arrowheads) and failed to prevent cell detachment (Fig. 4C).
␣SNAP Modulates Epithelial ECM Adhesion and Motility by NSF-independent Mechanisms-Because ␣SNAP depletion dramatically decreased ECM adhesion of epithelial cells, it is important to examine whether increased expression of this vesicle fusion regulator may have pro-adhesive effects. This question was addressed by expressing either wild-type ␣SNAP or its two known mutants in SK-CO15 cells using a lentiviral delivery system. The L294A mutant is unable to activate NSF and therefore has dominant-negative effects on SNARE-mediated exocytosis (47), whereas the M105I (hyh) mutant has increased affinity to another ␣SNAP target, AMP-activated protein kinase (37). SK-CO15 cells overexpressing wild-type ␣SNAP demonstrated an ϳ2-fold increase in adhesion to collagen I compared with non-transfected or control GFP-transfected cells (Fig. 5, A and B). Both L294A and M105I mutants acceler-ated ECM adhesion similarly to wild-type ␣SNAP (Fig. 5, A and  B). Interestingly, all overexpressed ␣SNAP proteins significantly increased levels of total and phosphorylated paxillin (Fig.  5, C and D). Such increased ECM adhesion of ␣SNAP-overexpressing cells was accompanied by ϳ3-fold reduction of their invasion in Matrigel (Fig. 5, E and F). It is noteworthy that anti-invasive effects of ␣SNAP mutants were similar to that of the wild-type protein (Fig. 5, E and F). We also investigated effects of ␣SNAP knockdown on epithelial cell motility and observed a complete block of Matrigel invasion in ␣SNAP-depleted SK-CO15 cells (data not shown). Similar effects of wildtype and L294A ␣SNAP on epithelial cell adhesion and migration suggest that these ␣SNAP activities may not depend on its major binding partner, NSF. To obtain more evidence supporting this hypothesis we examined whether depletion of NSF results in defects of epithelial cell adhesiveness recapitulating those of ␣SNAP knockdown. In concordance with our previous studies (34,35), NSF-specific siRNA markedly down-regulated expression of the targeted protein in SK-CO15 cells without decreasing the ␣SNAP level (Fig. 6A). Moreover, such NSF depletion did not result in significant dephosphorylation of paxillin and FAK (Fig. 6A), Interestingly, NSF depletion increased the relative amount of immature ␤1 integrin (Fig. 6A) and enhanced integrin accumulation within intracellular vesicles (Fig. 6B). However, in contrast to ␣SNAP knockdown, the majority of ␤1 integrin was able to reach the plasma membrane in NSF-depleted cells (Fig. 6B, arrows). Furthermore, loss of NSF failed to induce SK-CO15 cell detachment (Fig. 6C) thereby indicating a lack of dramatic effect on ECM adhesion. Collectively, our data suggests that disruption of NSF-mediated vesicle fusion cannot be a major mechanism of impaired FA assembly and defective ECM adhesion in ␣SNAP-depleted epithelial cells.

Disruption of the Golgi Recapitulates the Effects of ␣SNAP Knockdown on ECM Adhesion and Processing of FA Proteins-
Our recent studies documented dramatic fragmentation of the Golgi complex in ␣SNAP-depleted epithelial cells and demonstrated that Golgi disruption phenocopied many cellular effects of ␣SNAP knockdown (34 -36). Therefore it was next asked if Golgi abnormalities may also mediate diminished ECM adhesion caused by ␣SNAP depletion. Effect of Golgi fragmentation on SK-CO15 cell adhesion was investigated by using two known pharmacological Golgi disruptors, brefeldin A and Golgicide A (48,49). In agreement with our previously published results (35), both agents induced complete dispersion of the compact Golgi morphology visualized by either Giantin or GM130 immunolabeling (data not shown). Furthermore, 24 h treatment of SK-CO15 with either brefeldin A (1 M) or Golgicide A (50 M) caused a number of biochemical and phenotypic abnormalities including decreased expression and dephosphorylation of paxillin and FAK (Fig. 7, A and B), inhibition of ␤1 integrin maturation and trafficking (Fig. 7, A and C), and significant (up to 50%) cell detachment (Fig. 7D). All these alterations recapitulated the adhesion defects of ␣SNAP-depleted cells. Brefeldin A and Golgicide A are known to have one common molecular target, Golgi brefeldin factor 1 (GBF1), which serves as guanine nucleotide exchange factors for Golgi-resident ARF small GTPases (49 -51). Previously, we demonstrated that loss of ␣SNAP decreased GBF1 expression in SK-CO15 cells and that GBF1 knockdown mimicked major effects of ␣SNAP depletion on epithelial junctions and the autophagic flux (34,36). Therefore, we sought to investigate if loss of GBF1 can recapitulate disruption of ECM adhesion observed in ␣SNAP knockdown. Expression of GBF1 was dramatically down-regulated in SK-CO15 cells by RNAi (Fig. 7E) resulting in decreased levels of total and phosphorylated paxillin and dephosphorylation of FAK (Fig. 7, E and F). This was accompanied by defects of ␤1 integrin maturation (Fig. 7E), dramatic accumulation of ␤1 integrin within cytoplasmic vesicles (Fig. 7G, arrowheads), and significant (ϳ30%) cell detachment (Fig. 7H). All these changes closely resembled the defects in ECM adhesion observed in ␣SNAP-depleted epithelial cells. Because altered maturation of ␤1 integrin appeared to be a prominent consequence of ␣SNAP knockdown linked to Golgi fragmentation, we sought to investigate mechanisms underlying such integrin misprocessing. It is generally believed that the major ␤1 integrin doublet detected by immunoblotting reflects different glycosylated states of the protein with the upper (ϳ125 kDa) and lower (ϳ105 kDa) bands representing a fully glycosylated and incompletely glycosylated protein, respectively (45). Post-translational modification of ␤ integrin chains is a two-step process that involves initial N-glycosylation in the endoplasmic reticulum (ER) and further glycosylation by Golgi-resident enzymes (52,53). Because structure and function of the Golgi and the ER are interdependent, Golgi fragmentation in ␣SNAP-depleted cells may impair activity of glycosylation enzymes residing in both organelles. Consequently, we next sought to determine which step of ␤1 integrin glycosylation is impaired in ␣SNAP-depleted epithelial cells. Control SK-CO15 cells were treated with either a known inhibitor of ER-dependent N-glycosylation, tunicamycin (54), or inhibitors of Golgi-dependent glycosylation, DMJ and swainsonine (52,55). Tunicamycin (1 M) eliminated a typical ␤1 integrin doublet and resulted in accumulation of much faster migrating integrin bands at ϳ85 kDa (Fig.  8A) that corresponds to a core non-glycosylated ␤1 integrin polypeptide (45). Similar transformation of the ␤1 integrin doublet into the non-glycosylated core species was detected after treatment of control and ␣SNAP-depleted cell lysates with peptide N-glycosidase F removing all saccharide moieties (data not shown). Because the ␤1 integrin band at 125 kDa accumulated in the ␣SNAP-depleted epithelial cell was clearly distinct from the deglycosylated core protein, we concluded that ␣SNAP depletion did not affect initial N-glycosylation of ␤1 integrin at the ER. By contrast, inhibition of Golgi-dependent glycosylation by either swainsonine (50 M) or DMJ (50 M) resulted in a disappearance of mature ␤1 integrin and accumulation of its immature form similar to that observed in ␣SNAPdepleted cells (Fig. 8A). This suggests that loss of ␣SNAP impairs glycosylation of ␤1 integrin in the Golgi. Interestingly, only inhibition of ER-dependent N-glycosylation by tunicamycin dramatically attenuated expression and plasma membrane delivery of ␤1 integrin, and impaired ECM adhesion of SK-CO15 cells (Fig. 8, B and C). By contrast, swainsonine and DMJ treatment increased the amount of intracellular ␤1 integrin, but did not completely block its accumulation at the plasma membrane (Fig. 8B). Furthermore, inhibition of Golgi glycosylation had no effects on cell attachment to ECM (Fig. 8C). This data suggests that although loss of ␣SNAP impaired ␤1 integrin glycosylation in the Golgi, we observed that the defect of ␤1 integrin maturation may not be sufficient to diminish ECM adhesion of ␣SNAP-depleted epithelial cells.

␤1 Integrin and FA Proteins Mediate ␣SNAP-dependent Regulation of ECM Adhesion and Invasion-Because experiments with glycosylation inhibitors demonstrated that impaired ␤1
integrin maturation is not sufficient to explain the poor adhesiveness of ␣SNAP-depleted epithelial cells, we next sought to continue investigating the roles integrins play in our experimental conditions. Specifically, we asked two questions: (i) does inhibition of ␤1 or ␤4 integrin subunits recapitulate the effects of ␣SNAP knockdown on ECM adhesion and cell motility, and (ii) is accumulation of integrin precursors responsible for the loss of adhesiveness of ␣SNAP-depleted cells? Using RNAi, we down-regulated expression of ␤1 and ␤4 integrins either individually or in combination with ␣SNAP knockdown. Depletion of either ␤1 or ␤4 integrin in control SK-CO15 cells neither significantly altered phosphorylation of paxillin or FAK (Fig. 9A) nor induced spontaneous epithelial cell detachment (Fig.  9B). Likewise, co-knockdown of either ␤1 or ␤4 integrins with ␣SNAP failed to inhibit cell detachment (Fig. 9B). This finding argues against the roles of integrin precursors in these events. Interestingly, the matrix adhesion and Matrigel invasion assays showed that depletion of ␤1 integrin did not affect ECM attachment and invasion of control or ␣SNAP-overexpressing SK-CO15 cells (data now shown). These surprising results indicate that epithelial cells can compensate for the loss of individual integrin subunits. To avoid such compensation, we acutely blocked ␤1 integ-rin function by using the inhibitory MAB13 antibody (56). Compared with the isotope-specific control, MAB13 markedly inhibited ECM attachment and Matrigel invasion of control SK-CO15 cells (Fig. 9, C-E). Interestingly, the ␤1 integrin-inhibitory antibody reversed the hyper-adhesiveness of ␣SNAPoverexpressing SK-CO15 cells in parallel to reduced cell invasiveness (Fig. 9, C-E). Finally, we investigated if FAK and paxillin play roles in ␣SNAP-dependent regulation of ECM adhesion. SK-CO15 cells were exposed for 24 h to PF 431396 (20 M), a dual pharmacological inhibitor of FAK, and its ho- molog, PYK2 (57). This treatment did not induce cell apoptosis as indicated by negligible poly(ADP-ribose) polymerase cleavage and caspase 3 activation (Fig. 10A), and resulted in a noticeable dephosphorylation of FAK and paxillin (Fig. 10B). It is of note that FAK inhibition triggered significant detachment of SK-CO15 cells that was not blocked by caspase inhibition (Fig.  10C). To examine the role of paxillin, we down-regulated its expression in control and ␣SNAP-overexpressing SK-CO15 cells. Remarkably, paxillin knockdown completely reversed the increased ECM adhesion and accelerated invasion of ␣SNAP-overexpressing cells without affecting motility of control cells (Fig. 10, E and F). Together, this data reveals multiple mechanisms underlying the effects of ␣SNAP on epithelial EMC adhesion that involve activity of ␤1 integrin, FAK family kinases, and the FA scaffolding protein, paxillin.

DISCUSSION
ECM adhesion and motility of epithelial cells are regulated by elaborate vesicle trafficking/fusion machinery that is critical for assembly and remodeling of matrix adhesion complexes (58 - 60). The present study identifies novel functions of a key membrane fusion protein, ␣SNAP, in controlling ECM adhesion and invasion of human intestinal epithelial cells. Our results highlight ␣SNAP as an important positive regulator of ECM adhesion, because depletion of this protein leads to a marked cell detachment (Fig. 1B), whereas its overexpression dramatically enhances cell adhesiveness (Fig. 5). Although we provide the first direct evidence implicating ␣SNAP in regulating ECM adhesion, our findings are consistent with reported phenotypic consequences of decreased ␣SNAP expression in vivo. Thus, given the critical roles of ECM adhesion in tissue integrity and morphogenesis, it is predictable that loss of ␣SNAP expression should impair development and survival of multicellular organisms. Indeed previous genetic analysis demonstrated lethality of ␣SNAP deletion in Drosophila mutants (61). Furthermore, the so-called hyh mutation that decreases ␣SNAP expression in mice (62,63) was also shown to impair animal survival and development (64). Interestingly, homozygous hyh mice are characterized by progressive loss of neuroepithelium in brain ventricles (65,66), which is consistent with the weakening of ECM and cell-cell adhesions of ␣SNAP-depleted cells.
Another novel and important finding of this study is the role of ␣SNAP in regulating epithelial cell invasion. The observed relationship between the cellular level of ␣SNAP and cell invasiveness appears to be non-linear because both depletion and robust overexpression of this trafficking protein decreased cell invasion into Matrigel ( Fig. 5 and data not shown). It is well established that ECM adhesion is an important determinant of cell migration; however, relationships between these two processes are complex. Indeed, the most efficient cell migration occurs at intermediate attachment strength and both weak and strong matrix adhesions can inhibit cell motility (67,68). Therefore, the effects of diminished and enhanced ␣SNAP expression on cell motility can be explained by distinct cell adhesiveness when either loss of cell-matrix adhesion of ␣SNAP-deleted cells, or very strong attachment of ␣SNAPoverexpressing cells, impedes cell movement. This notion is supported by our paxillin depletion experiments that showed reversed hyperadhesiveness of ␣SNAP overexpressing cells and restored Matrigel invasiveness (Fig. 10, E and F).
Because ␣SNAP regulates a number of membrane trafficking events in different cellular compartments (69,70), it can affect ECM adhesion and motility via multiple mechanisms. Two major mechanisms suggested by this study include maturation and trafficking of ␤1 integrin and regulation of expression/activity of cytoplasmic FA proteins. Indeed, loss of ␣SNAP resulted in accumulation of incompletely glycosylated ␤1 integrin and intracellular retention of this ECM receptor (Figs. 2 and 3). The immature ␤1 integrin appeared to be initially N-glycosylated at the ER, but was not properly glycosylated at the Golgi (Fig. 8). Roles and regulation of integrin glycosylation at the Golgi remain poorly understood and only a few previous studies demonstrated modulation of this process by non-enzymatic mechanisms. For example, depletion of a guanine nucleotide exchange protein, BIG1, impaired Golgi-dependent glycosylation of ␤1 integrin in hepatocarcinoma cells (71), whereas such integrin maturation was enhanced in embryonic fibroblasts deficient in the catalytic subunits of the ␥-secretase complex (72). In these studies, the amount of fully glycosylated ␤1 integrin positively correlated with avidity of ECM adhesions. In contrast, pharmacological inhibition of Golgi-dependent ␤1 integrin glycosylation did not recapitulate defects of its targeting to the plasma membrane and decreased cell adhesiveness characteristics off ␣SNAP depletion (Fig. 8). This highlights a dual role of ␣SNAP in regulating ␤1 integrin maturation in the Golgi and its subsequent trafficking to the plasma membrane. The role of ␣SNAP in regulating ␤1 integrin activity on the cell surface is further supported by our finding that inhibitory anti-␤1 integrin antibody completely reversed enhanced ECM adhesion of ␣SNAP-overexpressing SK-CO15 cells and inhibited invasion of cells with different levels of ␣SNAP expression (Fig. 9, C-E). However, unlike ␣SNAP knockdown, inhibition of ␤1 integrin did not induce spontaneous detachment of control SK-CO15 cells (Fig. 9, A and B, and data not shown), which may have a methodological explanation. Thus, MAB13 inhibitory antibody may not be able to access and/or disrupt the established ␤1 integrin-ECM complexes. On the other hand, siRNA-mediated knockdown of ␤1 integrin can be functionally compensated by other integrins as has been reported under different experimental conditions (73).
Several lines of evidence in our study indicate that ␣SNAP regulates epithelial ECM adhesion by controlling assembly of FA. First, loss of ␣SNAP resulted in disappearance of FA along with inactivation (dephosphorylation) of essential FA proteins, FAK and paxillin (Fig. 2). Second, pharmacological inhibition of FAK family proteins mimicked the effects of ␣SNAP depletion on cell attachment (Fig. 9). Finally, overexpression of ␣SNAP increased the protein level and phosphorylation of paxillin (Fig.  5), whereas paxillin knockdown reversed the enhanced ECM adhesion and reduced invasiveness of ␣SNAP-overexpressing cells (Fig. 10). Interestingly, depletion of paxillin did not affect adhesion or invasion of control SK-CO15 cells. This data likely reflects a functional compensation by other members of the paxillin family, such as leupaxin or Hic-5 (74,75), and highlights the unique role of paxillin in the pro-adhesive activity of ␣SNAP. Overall, our results are consistent with previous studies that demonstrated the involvement of different SNAREs in FA assembly as well as FAK expression and activation (26,33). Although different FA proteins can readily shuttle between plasma membrane and intracellular compartments such as the nucleus (76), mechanisms of trafficking-dependent regulation of FAK and paxillin remain unknown. Loss of ␣SNAP can indirectly inhibit FAK and paxillin phosphorylation by decreasing the amount of activated integrins at the plasma membrane. . Inhibitory anti-␤1 integrin antibody, but not siRNA-mediated integrin depletion, inhibits epithelial ECM adhesion and invasion. A and B, SK-CO15 cells were transfected with control, ␤1 integrin, ␤4 integrin siRNAs, or co-transfected with siRNAs against these integrins and ␣SNAP. Expression of adhesion proteins and cell attachment to the substrate were examined 72 h post-transfection. C-E, control (SK-neo) and ␣SNAP-overexpressing (SK-␣SNAP) epithelial cells were preincubated with either ␤1 integrin-inhibitory antibody, MAB13, or the isotype-matched control antibody. Cell adhesion to collagen I (C) and invasion into Matrigel (D and E) was examined as described under "Experimental Procedures." Data are presented as the mean Ϯ S.E. (n ϭ 3); *, p Ͻ 0.001 compared with the isotype control antibody-treated cells.
Alternatively, ␣SNAP can directly regulate expression and trafficking of these FA proteins. This last mechanism is supported by reported associations of paxillin with known vesicle trafficking regulators, exocyst and GIT1 (77,78).
Interestingly, our data strongly suggests that ␣SNAP modulates epithelial ECM adhesion independently of its key binding partner, NSF. Indeed, the ␣SNAP L294A mutant, although unable to activate NSF, still significantly enhanced ECM adhesion and paxillin expression/phosphorylation similar to the wild-type protein (Fig. 5). Additionally, unlike ␣SNAP depletion, loss of NSF neither decreased FAK or paxillin phosphorylation nor diminished cell-matrix attachment (Fig. 6). These intriguing observations add to a growing body of evidence indicating that important cellular functions of ␣SNAP can be executed via NSF-independent mechanisms. Examples of such NSF-independent functions include regulation of apical plasma membrane trafficking (79) and integrity of apical junctions in polarized epithelial cells (34), as well as control of cell survival (35) and autophagy (36). A recent study has described a novel role of ␣SNAP in regulating mitochondrial biogenesis via direct interaction and dephosphorylation of AMP-activated protein kinase (37). However, such phosphatase activity is unlikely to be involved in modulation of epithelial ECM adhesion, where ␣SNAP demonstrates the opposite effect by promoting phosphorylation of FA proteins.
␣SNAP was previously identified as a member of two SNARE complexes containing syntaxin-5 and syntaxin-18 (80,81) that control anterograde and retrograde vesicle trafficking between the ER and Golgi, respectively (82). Also, according to cell-free vesicle reconstitution experiments, ␣SNAP can mediate both ER to Golgi and intra-Golgi vesicle transport (69,83). Because a constant bidirectional vesicle trafficking is essential for Golgi integrity, one can expect that loss of ␣SNAP should impair Golgi morphology and functions. Indeed, our recent studies demonstrate a marked fragmentation of the Golgi and decreased expression of Golgi-resident proteins such as GBF1 in ␣SNAP-depleted epithelial cells (34 -36). Similarly, a loss of function mutation of the brain-enriched ␤SNAP isoform was shown to trigger Golgi dispersion in zebrafish photoreceptors leading to photoreceptor degeneration (84). Importantly, we observed that Golgi fragmentation by either pharmacologic agents or GBF1 depletion phenocopied all effects of ␣SNAP knockdown on ECM adhesion including impaired processing and mislocalization of ␤1 integrin, decreased expression and phosphorylation of paxillin, and cell detachment (Fig. 7). This data highlights Golgi disruption as an important upstream event that mediates disruption of ECM adhesion in ␣SNAPdepleted epithelial cells. The Golgi-dependent mechanism can explain the severity of the effects of ␣SNAP knockdown, eventuating in epithelial cell detachment. Indeed, the Golgi is a crucial processing and sorting center for the majority of plasma membrane and secreted proteins. Its disruption would impair not only delivery and plasma membrane assembly of matrix adhesion complexes but also secretion of extracellular matrix proteins, thereby diminishing both receptor and ligand components of cell-matrix interactions.
Much remains to be learned about pathophysiologic implications of ␣SNAP-dependent alterations of cellular functions including ECM adhesion. No systematic studies of ␣SNAP tissue expression under pathologic states are available, although some reports demonstrated disease-related alterations in the level of this trafficking protein. For example, increased expression of ␣SNAP was observed in human colorectal carcinoma, which correlated with poor prognosis of the disease (85). Furthermore, loss of ␣SNAP expression was detected in the brain of Down syndrome fetuses (86) and in the striatum of patients with Huntington disease (87). Finally, some viral proteins reportedly down-regulate ␣SNAP levels in tumor cells, thereby sensitizing these cells to death-inducing stimuli (41). This data demonstrates an intriguing possibility that the ␣SNAP level can serve as a molecular rheostat determining cellular responses and behavior in different pathological conditions. Further studies are warranted to examine possible abnormalities and pathophysiologic roles of ␣SNAP in various human diseases such as neurodegeneration, inflammation, and cancer.