GIT1 is a multidomain protein that is thought to function as an integrator of signaling pathways controlling vesicle trafficking, adhesion and cytoskeletal organization. It regulates ARF GTPases and has binding domains for paxillin and PIX, which is a PAK-binding protein and an exchange factor for Rac. We show that GIT1 cycles between at least three distinct subcellular compartments, including adhesion-like structures, the leading edge and cytoplasmic complexes. The cytoplasmic structures, which also contain paxillin, PAK and PIX, do not detectably co-localize with endosomal Golgi or membrane markers, suggesting that they represent a novel supramolecular complex. The GIT1 cytoplasmic complexes are motile and tended to move toward the cell periphery where they joined existing adhesions. In retracting regions of the cells, the GIT1 complexes moved away from the disassembling adhesions toward the cell body. Using deletion mutants, we have identified domains that target GIT1 to each of the compartments. Localization to adhesions and the leading edge requires the paxillin-binding domain, which comprises the C-terminal 140 residues (cGIT1), whereas targeting to the cytoplasmic complexes requires the central region that contains ankyrin repeats and the PIX-binding domain. Expression of GIT1 or cGIT, but not nGIT1 in which the paxillin-binding domain is deleted, increases the rate of migration and the size and number of protrusions. The latter are inhibited when GIT1 is co-expressed with a kinase-dead PAK, suggesting that the GIT1 interaction with PAK is required for enhanced migration and protrusive activity. Furthermore, GIT1 targets constitutively activated PAK to adhesions and the leading edge via its interaction with paxillin. Since expression of cGIT targets endogenous GIT1 to the leading edge, it appears that the leading edge is the location of GIT1 responsible for these activities. Thus, GIT1 is a component of a motile, multimolecular complex that traffics a set of signaling components to specific locations in the cell where they regulate localized activities.
Cell migration is essential in many biological processes, including embryogenesis, inflammation, wound healing and tumor metastasis ( Lauffenburger and Horwitz, 1996). The migratory process requires the coordinated activation and targeting of both structural and signaling molecules. As migration is initiated, protrusions are extended at the leading edge and new adhesions are formed. The assembly and disassembly of actin is proposed to drive the formation and extension of the lamellipodia. Active myosin-based motors may generate the contractile forces necessary to translocate the cell body forward ( Lauffenburger and Horwitz, 1996). At the rear, adhesions must be released for the cell to move forward. Although many processes take place during cell migration, dynamic changes in polarized actin structures are believed to be among the most crucial.
Members of the Rho family of small GTPases, including Rho, Rac and Cdc42, are key regulators of actin dynamics and cell motility. Rho promotes the assembly of stress fibers and focal adhesions (Ridley and Hall, 1992). Rac induces membrane ruffling and the formation of lamellipodia, whereas Cdc42 regulates filopodial extension (Ridley and Hall, 1992; Nobes and Hall, 1995). These small GTPases cycle between an inactive (GDP-bound) state and an active (GTP-bound) state. The nucleotide state of these molecules is regulated by guanine nucleotide exchange factors (GEFs), which facilitate the exchange of GDP for GTP, and GTPase-activating proteins (GAPs) that promote GTP hydrolysis. In the active state, these proteins can interact with downstream targets or effector molecules to elicit a biological response.
One of the best characterized effectors of the Rho family of GTPases is p21-activated kinase (PAK). PAK is a serine/threonine kinase that binds to the active form of Rac or Cdc42, resulting in a conformational change in PAK, autophosphorylation of the molecule at several sites and subsequent activation ( Manser et al., 1994; Lim et al., 1996; Lei et al., 2000). The effects of PAK may be, at least partially, controlled by regulation of its subcellular localization. In resting cells, PAK is distributed throughout the cytoplasm; however, when the cells are stimulated, PAK is targeted to focal adhesions and membrane ruffles ( Dharmawardhane et al., 1997). Moreover, it has been demonstrated that it is the activated fraction of PAK that localizes to the leading edge and focal adhesions of migrating fibroblasts ( Sells et al., 2000). As PAK lacks a focal adhesion targeting domain, it is probably recruited to focal adhesions and membrane protrusions through interaction with binding partners. The PAK-binding protein, PIX, may be involved in the recruitment of PAK ( Manser et al., 1998), alternatively, the adaptor protein, Nck, may function in this capacity ( Bagrodia and Cerione, 1999; Bokoch et al., 1996). However, the mechanisms that mediate PAK targeting to focal adhesions and membrane ruffles are not well defined.
There is a growing body of evidence implicating PAK in reorganization of the actin cytoskeleton. Although the mechanism(s) by which PAK regulates actin dynamics is not well understood, it has been suggested that PAK may mediate these effects by phosphorylation of LIM kinase ( Edwards et al., 1999). This results in the inactivation of the actin depolymerizing protein, cofilin, and stabilization of actin polymers ( Edwards et al., 1999). PAK can also alter the phosphorylation state of myosin-II. Expression of constitutively activated PAK resulted in decreased activity of myosin light chain kinase and decreased phosphorylation of the regulatory myosin light chain (MLC) ( Sanders et al., 1999). By contrast, in NIH3T3 cells, activated PAK promoted phosphorylation of MLC and localization of the phosphorylated protein to the lamellipodia (Sells et al., 1999). Finally, PAK may affect the actin cytoskeleton through regulation of the Raf-Mek-ERK pathway ( Bagrodia et al., 1995).
Paxillin is a multidomain scaffolding protein that functions in the recruitment of both signaling and structural molecules to focal adhesions ( Turner, 2000). This protein comprises multiple SH3- and SH2-binding domains, four LIM domains and five leucine-rich motifs (LD motifs) ( Turner, 2000). Members of the ARF family of small GTPases, including GRK interactor 1 (GIT1) and paxillin kinase linker (PKL), interact with paxillin through its LD4 motif ( West et al., 2001; Zhao et al., 2000; Turner et al., 1999). Although the functional significance of the GIT1-paxillin interaction is not well defined, overexpression of GIT1 in fibroblasts has been reported to sequester paxillin from focal complexes ( Zhao et al., 2000). Furthermore, expression of another GIT1 family member, PAG3, inhibited paxillin recruitment to focal contact in an ARF-GAP-dependent manner ( Kondo et al., 2000). These proteins may function in linking PAK to paxillin through a paxillin-GIT1-PIX-PAK complex and in this manner localize PAK to paxillin-containing adhesions.
In this study, we show that GIT1 localizes to three distinct subcellular compartments, including focal adhesions, cytoplasmic complexes and membrane protrusions. Paxillin, PIX and PAK all colocalized with GIT1 in these cytoplasmic complexes, suggesting that these structures might represent a multiprotein signaling module. The cytoplasmic complexes are motile and appear to be involved in delivery of components to and from adhesions as they form and breakdown. Recruitment of GIT1 to focal adhesions and the leading edge of the lamellipodia requires the paxillin-binding domain. Expression of GIT1 or its C-terminal 140 amino acids, which contains the paxillin-binding site, increases the rate of migration and the size and frequency of protrusions. The GIT1-induced protrusions are inhibited by expression of kinase-dead PAK, suggesting that PAK association with GIT1 is necessary for the enhanced migration and protrusions. GIT1 targets constitutively activated PAK to focal adhesions and the leading edge through its interaction with paxillin. Expression of cGIT1 targets endogenous GIT1 to the leading edge. Thus, the leading edge appears to be the location of GIT1 responsible for its effects on protrusion and migration.
Materials and Methods
Dulbecco's modified media (DMEM) and lipofectamine transfection reagent were from Gibco (Grand Island, NY). G418 was from Sigma (St Louis, MO). Enhanced chemiluminescence (ECL) detection system was from Amersham Life Sciences (Buckinghamshire, UK). Effectene was from Qiagen (Valencia, CA) and Fugene 6 from Boehringer Mannheim (Germany). Polyacrylamide was from Bio-Rad laboratories (Hercules, CA). Cy-3-labelled transferrin was from Jackson Laboratoires (West Grove, PA), FM4-64 dye and BODIPY-Ceramide (Golgi marker) were from Molecular Probes (Eugene, OR). All PAK constructs were kindly provided by Jonathan Chernoff (Fox Chase Cancer Center, Philadelphia, PA).
The following antibodies were used for western blotting, immunoprecipitation and immunofluorescence. We prepared an affinity-purified rabbit polyclonal antibody raised against the central 140 amino acids of GIT1, the region that contains the least homology to any known GIT1 family members. The peptide was cloned into pGEX-2T (Promega, Madison, WI). The GST fusion product was purified from E. coli using the manufacturer's protocol and used for immunization. The antisera was produced by Covance (Denver, PA), and the antibody was purified by affinity chromatography as previously described (Koff et al., 1992) with our modifications. The antibody did not recognize GIT1's close homologue PKL in cells stably expressing GFP-PKL. Transferrin receptor B65.3 mouse monoclocal antibody was a kind gift from Sam Green (U. of Virginia) and Ian Trowbridge (Salk Institute, LaJolla, CA). Anti-FAK 2A7 was from Upstate Biotechnology (Lake Placid, NY). Monoclonal antibodies for paxillin, GIT1 and NCK were from Transduction Laboratories (Lexington, KY). HA monoclonal antibody was from Boehringer Mannheim. Anti-c-Myc 9E10, anti-GFP, B2 and anti-phosphotyrosine PY20 were from Santa Cruz Biotechnology (Santa Cruz, CA). FLAG M2 monoclonal antibody was from Stratagene (La Jolla, CA). Vinculin monoclonal antibody was from Sigma. Phosphopaxillin PY31 and PY118 antibodies were from Biosource International (Camarillo, CA). Rhodamine-conjugated phalloidin and the GFP polyclonal antibody A-11122 were from Molecular Probes. HRP-conjugated secondary sheep anti-mouse IgG and donkey antirabbit IgG antibodies were from Amersham. Rhodamine-conjugated sheep anti-mouse IgG, FITC-conjugated sheep anti-mouse IgG and FITC-conjugated goat anti-rabbit IgG were from ICN Biochemical Division (Aurora, OH).
Isolation of GIT1 cDNA
We isolated a 1.5 kb cDNA encoding the C-terminal 140 amino acids of GIT1 from a GFP-fusion cDNA library as previously described ( Manabe et al., 2000). Briefly, a cDNA library derived from human fibrosarcoma HT1080 cells was cloned into pCIneoEGFP ( Fujii et al., 1999). Individual cDNAs were purified and transfected into REF52 cells using Effectene according to the manufacturer's protocol.
For isolation of full-length GIT1, we amplified the 5′ GIT1 cDNA from a human fetal brain library (Clontech) using the following primers: 5′-AAGGATCCGTCGACATGTCCCGAAAGGGGCCGCG-3′ (forward) and 5′-TTGAATTCGCGGCCGCCTACTCTTTGCCC-AGCTCTAGAAACC-3′ (reverse). The amplified cDNA was sequenced and cloned into pCIneoEGFP-cGIT1. A cDNA expression vector for full-length GIT1 fused with FLAG at its C-terminus (Wt-GIT1-FLAG) was prepared as follows. A cDNA encoding C-terminal GIT1 without the stop codon was generated by PCR using the following primers: 5′-AGGTTTCTAGAGCTGGGC-3′ (forward) and 5′-NotI-Stop-FLAG-SalI-CTGCTTCTTCTCTCGGG-3′ (reverse). The amplified cDNA was excised with XbaI and NotI and cloned into pEGFP-N1 (Clontech) with the sequence encoding the N-terminus of GIT1. For the construction of the C-terminal deletion mutant of GIT1 (nGIT1), the PCR-amplified 5′ GIT1 cDNA described above was cleaved with SalI and NotI and cloned into the SalI and NotI site of pCIneohEGFP. For construction of a GFP-GIT1 chimera lacking a central part of GIT1 (MiniGIT1), a HindIII fragment encoding the central part of GIT1 was excised and the resulting plasmid was religated.
Cell culture and transfection
CHO-K1, REF-52, HT1080 and A431 cells were maintained in DMEM supplemented with 10% FBS, 4 mM L-glutamine, 1 mM sodium pyruvate, non-essential amino acids, penicillin and streptomycin. For most experiments involving microscopy, cells were plated on either 12 mm glass cover slips or dishes prepared for microscopic observation as previously described ( Palecek et al., 1996). The coverslips and dishes were coated overnight at 4°C with 2-10 μg/ml fibronectin (Fn). Cells were transfected with either Lipofectamine (CHOK1 cells) or Fugene 6 (REF-52) and incubated for 24 to 72 hours. To prepare CHO-K1 cells stably expressing our GIT1 constructs, cells were maintained in serum containing medium with 1 mg/ml G418 for 7 to 10 days after transfection. Cells were FACS sorted by the level of expression, and those with a medium expression level were selected for experiments. As confirmed by western blotting, the GFP-fused GIT1 in stably expressing cells was of the expected molecular mass indicating that it was not cleaved.
Immunoprecipitation and western blotting
Polyclonal GFP antibody was used to immunoprecipitate GFP-GIT1 from the stable CHOK1 transfectants. Cells were grown to 80-90% confluency, washed with ice-cold PBS and then lysed with ice-cold 10 mM Tris-HCl (pH 7.6), 50 mM NaCl, 1% NP-40 and 10% glycerol, 1 mM DTT, 10 μg/ml aprotonin, 1 μg/ml pepstatin A, 1mM benzamidine, 10 μg/ml leupeptin and 0.2 mM Pefabloc (Boehringer Mannheim). Cell lysates were incubated on ice for 30 minutes and clarified by centrifugation (13,000 g for 15 minutes). Protein concentrations were determined by BCA assay, and equivalent amounts of the lysates were precleared with 6 μg of rabbit IgG for 30 minutes followed by incubation with 6 μg of a GFP-specific polyclonal antibody for 1 hour at 4°C. Complexes were incubated with protein A sepharose overnight and washed four times with ice-cold lysis buffer. The immunoprecipitates were subjected to SDS-PAGE on 10% slabs, transferred to nitrocellulose and detected by western blot analysis.
For immunofluorescence, cells were grown to 50-70% confluency on glass coverslips or glass-bottomed 35 mm plates, washed with PBS once, fixed with 2% formaldehyde for 15 minutes, incubated with 0.15 M glycine for 10 minutes to stop fixation and permeabilized with 0.2% (v/v) Triton X-100 for 5 minutes at room temperature. After each step, cells were washed with PBS two or three times. Cells were then blocked with 5% normal goat serum in PBS for 1 hour at room temperature. This blocking buffer was also used for the antibody dilution. Primary antibodies or rhodamine-linked phalloidin were applied for 1 hour and FITC or rhodamine-conjugated secondary antibody for 40 minutes. Slips were mounted on slides with Vectashield mounting media (Vector Laboratories, Burlingame, CA) and visualized using rhodamine/Cy-3, EGFP or FITC filters.
For staining with dyes, CHO-K1 cells were plated at 30-50% confluency on 35mm glass bottomed plates in complete DMEM medium and allowed to spread overnight. All DMEM F-12 media used for staining (washes and dye dilution) was serum and phenolred free. For transferrin staining, cells were washed three times with DMEM F-12 media and incubated for 7 minutes with 4 μg/ml Cy-3 transferrin at 37°C. Cells were rinsed three times with PBS, fixed with 2% formaldehyde and visualized. After fixation, untransfected cells were stained with a GIT1-specific antibody followed by the FITC-conjugated rabbit IgG secondary antibody as described above and visualized. For FM4-64, cells were washed three times with DMEM F-12 and incubated in DMEM F12 with 1 μg/ml FM4-64 for 10, 30, 45, 60, 120, 180 minutes or overnight at 37°C. Cells were rinsed three times with PBS and used for microscopy. For BODIPY-Ceramide staining, cells were incubated in DMEM F-12 containing the dye at 1 μg/ml at 4°C for 30 minutes, washed three times and treated with dye free DMEM F-12 for another 30 minutes at 37°C and fixed.
Microscopy and image processing
For fluorescence, cells were visualized with a Nikon TE 300 conventional fluorescence microscope (Melville, NY). For EGFP fluorescence, an Endow GFP filter cube (ex. HQ470/40, em. HQ525/50, Q495LP, dichroic mirror) was used (Chroma, Brattleboro, VT). Rhodamine, FM4-64, Cy3 Transferrin and BODIPY-Ceramide were visualized using a rhodamin/TRITC cube (ex. BP520-550, barrier filter BA580IF, dichroic mirror DM565). AMCA fluorescence, for vinculin staining of PAK/GIT1 expressing cells, was visualized using a set of filters: ex. D360/40, em. D460/50. Both TRITC and AMCA filters were from Chroma. Individual images or z-stacks (for deconvolution) were collected using a cool CCD camera (Hamamatsu Photonics, Bridgewater, NJ) and processed with Isee software (Inovision, Durham, NC). Confocal images were taken on Zeiss Axiovert 100 confocal microscope (Carl Zeiss GmbH, Jena, Germany) with excitation and emission filters for FITC and TXRed/Cy3 and processed using Zeiss LSM Microsystem software. Images were analyzed using NIH Image software. For migration speed, the cell centroid was tracked. The average speed for the cell was then determined by computing the average net displacement of the cell centroid divided by the time interval at each time point. For protrusiveness analysis, the cells were outlined at two time points separated by 10 minutes; the two images were thresholded and then subtracted to estimate the new area. The area measurements were calibrated using a micrometer scale.
GIT1 localizes to distinct cellular compartments
To search for novel regulators of cell migration, cytoskeletal organization and adhesion assembly, we subcloned a human cDNA library, derived from the HT1080 fibrosarcoma cell line, into a GFP-tagged expression plasmid, pCINeo-GFP, and screened for clones that localize to focal adhesions or cytoskeletal structures ( Fujii et al., 1999). One of the clones localized prominently in adhesion-like structures that costained with paxillin. It encoded the 140 C-terminal residues of GIT1 ( Fig. 1B,C), which had been previously identified as a regulator of β2-adrenergic receptor endocytosis (Premont et al., 1998). To further examine the protein, we cloned its full-length cDNA into pCINeo-GFP and stably or transiently expressed the resulting GFP fusion protein in CHOK1 cells. Cells were sorted by flow cytometry into three expression categories: low, medium and high expressors. Western blot analysis showed that the level of ectopic GIT1 in the medium expressors was comparable to that of the endogenous protein. Thus, the medium expressors were used for most studies.
In CHOK1 cells expressing medium levels of the fusion protein, GIT1 localized in paxillin- and vinculin-containing adhesions and at the leading edge in membrane ruffles ( Fig. 1E). When GIT1 was expressed in REF-52, WI38 and NIH3T3 cells, the localization patterns were comparable to those observed in CHOK1 cells. The majority of cells expressing ectopic GIT1 showed cytoplasmic structures ( Fig. 1E). GIT1 tagged with FLAG or fused to the N-terminus of GFP showed localization patterns similar to GFP-GIT1, suggesting that the GFP fusion did not affect localization (data not shown).
To examine the localization of the endogenous protein, we raised antibodies against a GST fusion product consisting of the central 130 residues of GIT1 because this region showed the least homology to other GIT family members. When lysates from CHOK1 cells expressing GFP-GIT1 or GFP-PKL were analyzed on western blots using this antibody, a single prominent band was detected that corresponded to either endogenous GIT1 or GFP-GIT1 but not to the highly homologous GFP-PKL (data not shown). These results suggest that the antibody was specific for GIT1 and does not cross-react with other GIT family members, such as PKL. Pre-incubation of the antibody with the antigenic peptide inhibited the immunoblotting (data not shown). The endogenous GIT1 localized both to adhesion-like structures and small complexes throughout the cytoplasm ( Fig. 1D,H); neither of which were observed when the cells were stained with non-immune serum or antibody pre-incubated with the inhibitory peptide. These observations suggest that the cytoplasmic localization of GFP-GIT1 is analogous to that of the endogenous protein and is not a result of overexpression at least at medium expression levels.
To examine the domains responsible for GIT1 localization in various compartments, we prepared three GIT1 deletion mutants, N-, C- and `mini'-GIT1, tagged with GFP ( Fig. 1A). nGIT1 comprises 630 residues beginning at the N-terminus and including the ARF-GAP domain, ankyrin repeats and a putative PIX-interaction domain. The cGIT1 mutant encodes the C-terminal 130 residues, which contains the paxillin-binding domain. Mini-GIT1 has the central 477 residues deleted, but it retains the ARF-GAP domain and paxillin-binding region ( Zhao et al., 2000). Both wild-type GIT1 ( Fig. 1E), cGIT1 ( Fig. 1C) and mini-GIT1 ( Fig. 1G) localized to adhesions, whereas nGIT1 ( Fig. 1F) did not, confirming that the adhesion targeting domain resides at the C-terminal end. In contrast, both nGIT1 and wild-type GIT1 localized in cytoplasmic complexes, whereas c- and mini-GIT1 mutants did not. These observations point to the central region as the targeting domain for the cytoplasmic complexes. Finally, both cGIT1 and mini-GIT1 localized in ruffles whereas nGIT1 did not, implicating the adhesion targeting domain, possibly through its interaction with paxillin, in the localization to the leading edge. Thus, localization of GIT1 to three distinct subcellular compartments is mediated by discrete domains within the GIT1 protein.
The nature of the GIT1 cytoplasmic complexes
To determine whether the GIT1-containing cytoplasmic complexes are vesicles, we probed CHO K1 cells expressing GFP-GIT1 as well as untransfected cells with several markers for vesicular compartments. The markers included (a) endosomal markers — Cy-3-labeled transferrin and a transferrin receptor antibody, (b) markers for bulk membrane internalization — FM 4-64 dye and membrane-bound palmitoylated GFP and (c) two Golgi markers— mannosidase-2 antibody and BODIPY-ceramide dye. After staining, the cells were examined by confocal or deconvolution microscopy ( Fig. 2).
As GIT1 and other ARF-GAP-containing proteins are involved in regulation of endocytosis ( Moss and Vaughan, 1998; Premont et al., 1998), we asked whether GFP-GIT1 cytoplasmic complexes were of endocytic origin using Cy-3-labeled transferrin and transferrin receptor antibody as markers. While we observed some co-localization of GIT1 with the transferrin receptor in the perinuclear region, it did not co-localize with the numerous small complexes that resided outside this region. In addition, there was no detectable colocalization of GFP-GIT1 with transferrin.
ARF1, one of the ARFs for which GIT1 serves as a GAP, is reported to reside in the Golgi apparatus and regulate intra-Golgi and Golgi-ER trafficking ( Moss and Vaughan, 1998). In addition, PAG3, a distant GIT1 relative, is also reported to reside in the Golgi apparatus ( Kondo et al., 2000). Therefore, we probed for colocalization between GIT1 complexes and two Golgi markers — BODIPY-ceramide and a mannosidase-2 antibody. BODIPY-ceramide dye accumulates predominantly in the Golgi, but also shows a secondary localization in other organelles such as lysosomes ( Pagano et al., 1991), whereas the mannosidase-2 antibody is highly Golgi specific ( Baron et al., 1990). In spite of relatively strong labeling after a 30 minute incubation with BODIPY-ceramide, GFP-GIT1 complexes did not detectably co-localize with it. The absence of clear co-localization was also observed using the anti-mannosidase-2 antibody marker, which produced a typical perinuclear ring-like Golgi staining, but did not co-localize with GFP-GIT1.
Since the GIT1 complexes did not colocalize robustly with any of the compartment specific markers, we compared their localization with FM4-64 dye, a bulk membrane internalization marker. The cells were allowed to internalize the dye for various time intervals, from 5 minutes to overnight, to ensure saturation labeling. Although the FM4-64 signal was strong throughout the cell, there was no detectable co-localization with the GIT1 complexes. Furthermore, we could not detect co-localization of GIT1 cytoplasmic structures with membrane-bound palmitoylated GFP, which indicated that the GIT1 complexes are not membrane enclosed.
These data suggest that the majority of the cytoplasmic structures formed by GIT1 do not reside prominently in either endosomal or Golgi compartments. In addition, GFP-GIT1 localization does not resemble that of aggresomes, which are aggregates of overexpressed protein, reportedly formed by several GFP-fused molecules. These aggregates are less numerous and morphologically different from what we observed in the case of GIT1 ( Garcia-Mata et al., 1999).
As the GIT1 containing structures did not appear to be vesicular, we asked whether these cytoplasmic complexes contain signaling or structural components. Paxillin, PIX and PAK co-localized with GIT1 in the cytoplasmic complexes ( Fig. 3A,B,E-H). Although the majority of the GIT1 complexes contained PAK, we could not detect PAK in all of these structures. FAK, vinculin, α-actinin, ARF1, ARF6, Rac and the α5 integrin subunit were not detected in the GIT1 cytoplasmic complexes (data not shown).
To determine which domain of PAK was necessary for its localization in the cytoplasmic complexes, we co-expressed mutant forms of PAK with GIT1. Kinase-dead, Rac- and Nck-binding-deficient PAK co-localized with GIT1 in the complexes (data not shown); however, PIX-binding-deficient PAK was not observed in the cytoplasmic complexes with GIT1 ( Fig. 3C,D). This indicates that GIT1 interaction with PAK via PIX is essential for PAK localization in the complexes.
GIT1 cytoplasmic complexes are motile
The presence of signaling components in the GIT1-containing cytoplasmic complexes suggested that they might serve as motile signaling modules. This prompted an investigation of the dynamics of the small complexes that reside throughout the cell. When cells were viewed 2-4 hours after plating with time-lapse microscopy, GIT1 appeared in clusters of small discrete complexes near the cell periphery ( Fig. 4A). The GIT1 complexes co-localized with paxillin and vinculin ( Fig. 4B,C), which indicated their association with adhesions. The small GIT1 complexes were motile and tended to move toward the cell periphery and join a group of similar structures ( Fig. 4A) (see Movie 1 at jcs.biologists.org ). When a cell protrusion retracted, some GIT1 complexes would disassemble into discrete structures, which moved away from the retracting lamella ( Fig. 5) (see Movie 2 at jcs.biologists.org ). These complexes moved at an average rate of 30 μm/hour toward the cell center. In retracting regions of the cell, these GIT1 complexes co-localized with paxillin in fibroblasts expressing CFP-GIT1 and YFP-paxillin (data not shown).
When cells were allowed to adhere for longer periods of time, the number of GIT1 cytoplasmic structures decreased significantly. A corresponding increase in GIT1 co-localization with paxillin or vinculin-containing adhesions was observed. We hypothesized that GIT1 may be cycling between the small cytoplasmic complexes and adhesion-like structures. During cell detachment, GIT1 may move from the adhesions into the cytoplasm, whereas when it is needed in adhesions, it would then be delivered there in the same manner. Consistent with our hypothesis, in migrating cells GIT1 emanated from the rear of the cell in complexes and moved toward the cell body ( Fig. 6) (see Movie 3 at jcs.biologists.org ).
GIT1 localization affects cell migration
The localization to adhesion-like structures and the leading edge suggested that GIT1 might regulate cell migration. To address this, we estimated migration speeds from time-lapse observations of cells constitutively expressing GIT1 or its mutants. Both wild-type GIT1 and cGIT1 increased the migration speeds of CHOK1 cells when compared with cells expressing GFP alone ( Fig. 7A), whereas nGIT1 did not affect migration. For example, less than 40% of GEP or GFP-nGIT1 transfected cells migrated at a rate of at least 25 μm/hour; this fraction increased to over 80% for GIT1 expressors and almost 60% for cGIT1 expressors. Accompanying the effects of GIT1 on migration, we found that both wild-type and cGIT1 also increased the size and rate of formation of protrusions relative to GFP alone ( Fig. 7B-E).
Thus, it appears that the C-terminal portion of GIT1, which contains targeting domains for adhesions and the leading edge and binds to paxillin, is sufficient for promoting increased migration and protrusion formation. The remaining domains do not localize to adhesions or membrane ruffles and do not affect migration. This observation is surprising since the adhesion-targeting domain has no known activities other than binding to paxillin, whereas the other domains have activities that include an ARF-GAP and PIX binding, both of which might affect migration. It suggests, therefore, that ectopic expression of cGIT1 might act by competing with endogenous GIT1 for adhesion-binding sites and promoting its relocalization to a site where it is active. When cells were transfected with cGIT1, the endogenous GIT1 became more prominent in the membrane ruffles at the leading edge compared with the cells expressing GFP alone ( Fig. 7F-I).
GIT1 localizes PAK to focal adhesions and the leading edge of lamellipodia
PAK, an effector of Rac, regulates migration through its effects on the formation of protrusions. It also interacts with PIX, a binding partner of GIT1, and the active form localizes in adhesions and protrusions, suggesting an attractive mechanism for the effects of GIT1 on migration. To determine whether PAK mediates the effects of GIT1 on migration, we co-expressed GFP-GIT1 with a kinase-dead mutant of PAK in CHOK1 cells. When kinase-dead PAK was co-expressed with GIT1, the number and size of protrusions was dramatically reduced when compared with GIT1 alone ( Fig. 8A,B). When the number of protrusions was quantified, a three-fold decrease was observed in cells expressing kinase-dead PAK and GIT1 as compared with GIT1 alone (data not shown). These results suggest that the effects of GIT1 are mediated by PAK.
We then asked whether GIT1 was involved in the targeting of PAK, which has been reported to localize to adhesions and the leading edge ( Zhao et al., 2000). When expressed ectopically, a constitutively active mutant of PAK localized to both adhesions and the leading edge. This localization was significantly enhanced by co-expression with GIT1 ( Fig. 8C-E). In cells co-expressing GIT1 and constitutively active PAK, approximately 90% of the cells showed prominent PAK localization to adhesions and the leading edge, whereas this prominent localization was observed in only 30% of the cells co-expressing active PAK and GFP. Wild-type PAK, when co-expressed with GIT1, also localized to adhesions and the leading edge, although not as prominently as the constitutively active form. In addition, the kinase-dead form of PAK did not localize to adhesions or the leading edge irrespective of the presence of ectopic GIT1. When PAK is co-expressed with nGIT1 ( Fig. 8F-H) or mini-GIT1 ( Fig. 8I-K), increased PAK localization to vinculin-containing adhesions or the leading edge was not observed. This suggests that GIT1 targeting of active PAK to these regions is dependent on its interactions with both paxillin and PIX.
Rac controls GIT1 localization
Rac is another well known regulator of protrusion formation and migration ( Hall, 1998) that might mediate the effects of GIT1. Rac is regulated by PIX, a putative exchange factor that also binds to GIT1. Consonant with this hypothesis, we found that GIT1-induced protrusions were significantly reduced by co-expression of a dominant-negative mutant of Rac (N17-Rac) ( Fig. 9). When cells were transfected with an activated mutant of Rac, V12-Rac, we observed a dramatic increase in GIT1 localization to adhesions and to the edge of lamellipodia and a decrease in the number of GIT1 cytoplasmic complexes ( Fig. 9). This effect could be seen in most cells co-expressing GIT1, either constitutively or transiently, and the observations with V12-Rac were especially obvious in CHO-K1 cells because they do not normally form large substrate adhesions. When we expressed V12-Rac in parental CHO-K1 cells and probed for endogenous GIT1, we saw similar results, that is, the effect was detectable in greater than 90% of the transfected cells. Interestingly, an increase in the number, and to a lesser extent, the size, of vinculin containing adhesion was also observed in cells expressing V12-Rac ( Fig. 9). These observations suggest that the intracellular distribution of GIT1 between cytoplasmic structures and adhesions is regulated by Rac. In addition, Rac is necessary for the formation of GIT1-induced protrusions. The localization of nGIT1, a mutant lacking the paxillin-binding domain, in cytoplasmic complexes, but not in adhesions, was unaffected by V12-Rac, suggesting that the paxillin-binding domain was essential for adhesion targeting of GIT1 by V12-Rac.
GIT1 is rapidly emerging as a scaffolding protein that has multiple binding domains for several different molecules, including ARFs, PIX and FAK ( Turner et al., 2001). Despite a growing interest in the GIT family of proteins, its function in integrating signaling pathways that control cell adhesion and migration is not well defined. The goal of our study was to determine the mechanisms by which GIT1 regulates these cellular processes. Our results show that GIT1-containing cytoplasmic complexes can cycle between at least three distinct intracellular compartments including adhesions, the membrane at the leading edge and cytoplasmic complexes. The central region of GIT1 determines targeting to cytoplasmic complexes and interaction with PAK, whereas the C-terminus, which binds to paxillin, is necessary for localization to adhesive complexes and the leading edge. These cytoplasmic structures function as modules that deliver signaling molecules, such as PAK, to adhesions and the leading edge and as a pool of GIT1 complexes derived from the breakdown of adhesions. GIT1 localization appears to determine its functional effects. For example, GIT1 enhances migration and protrusive activity when it resides at the leading edge of the cell.
A GIT1-related protein, p95-APP1, has been reported to co-localize with the transferrin receptor and an endosomal marker, Lucifer yellow (DiCesare, 2000). In contrast, we were unable to detect significant co-localization of GIT1 complexes with either transferrin or its receptor. Furthermore, ARF6 andα 5 integrin, which reside in endosomal vesicles, did not co-localize with GIT1 complexes. The experimental methods of DiCesare et al. differed from ours in several respects. First, their study utilized conventional fluorescence microscopy, while we used deconvolution and confocal microscopy to remove out of focus light. Second, p95-APP1 was expressed transiently, while we selected stably expressing cells in which ectopic GIT1 levels were comparable to that of the endogenous protein. However, a more likely explanation for the discrepancy is that GIT1-related proteins can localize differently. This is supported by a study ( Kondo et al., 2000) demonstrating that PAG3, a GIT1-related protein, localizes in the Golgi, a site in which neither GIT1 nor p95-APP1 reside. These differences in location will probaby be critical to the functions of the GIT1-related proteins once they become more apparent.
As our study indicated that the GIT1-containing cytoplasmic structures were not vesicular, we hypothesized that GIT1 might serve as a scaffold to bring together molecules to form a signaling module. Structural and signaling molecules, including FAK, vinculin, α-actinin, ARF1, AFR6, Rac andα 5 integrin, were not detected in the cytoplasmic GIT complexes. However, paxillin, PIX and PAK were present in these cytoplasmic complexes. The localization of PAK with GIT1 complexes and the enhanced adhesion targeting of this molecule by GIT1 suggests that these cytoplasmic structures may target PAK to adhesions. Since PAK lacks an adhesion-targeting domain, its binding partners probably function in its recruitment to adhesions. The guanine nucleotide exchange factor PIX could be involved in PAK localization to adhesion; alternatively, the adaptor protein Nck could function in this capacity ( Sells et al., 2000; Bagrodia and Cerione, 1999; Bokoch et al., 1996). Our results show that PIX co-localizes with GIT1 in the cytoplasmic structures; however, we did not observe Nck in these complexes. In addition, a PAK mutant deficient in PIX binding could not be detected in GIT1 complexes. Thus, PIX most likely recruits PAK to these signaling complexes, which are targeted to adhesions and the membrane at the leading edge of the lamellipodia.
Our results show that expression of wild-type GIT1 significantly increased migration and the rate of protrusion formation in CHOK1 cells. We speculated that GIT1 mediates these effects by targeting PAK to adhesions and the leading edge of the lamellipodia. Active PAK may direct cellular movement by regulating reorganization of the actin cytoskeleton at the leading edge of the cell ( Dharmawardhane et al., 1997; Sells et al., 2000; Sells et al., 1997; Kiosses et al., 1999). Furthermore, PAK could regulate cell migration by altering the disassembly of actin stress fibers and adhesions ( Sells et al., 1997; Manser et al., 1997; Zhao et al., 2000). Consistent with our hypothesis, when kinase-dead PAK was co-expressed with GIT1, the number and size of protrusions was significantly reduced, suggesting that the effects of GIT1 are dependent upon PAK.
Interestingly, cGIT1 promoted migration, whereas nGIT, which contained the ARF-GAP and PIX-binding domain, had no effect. A previous study reported that the C-terminus of GIT1 had no effect on cell movement; however, the C-terminus of GIT1, which was targeted to the membrane by myristoylation, increased migration ( Zhao et al., 2000). This implies that membrane targeting of C-terminal GIT1 is necessary to stimulate migration. By contrast, our study suggests that expression of the C-terminus of GIT1, in the absence of myristoylation, can promote migration. As cGIT1 neither co-immunoprecipitated with PAK nor contains a putative PIX-PAK interaction domain, the enhanced migration is probably not due to the targeting of PAK. To try to address the mechanism, we asked whether ectopic expression of cGIT1 affects the distribution of endogenous GIT1. When cells were transfected with cGIT1 and stained with the GIT1 antibody, we observed increased localization of endogenous GIT1 at the leading edge of the lamellipodia. This result suggested that cGIT1 competes with endogenous GIT1 for adhesion-binding sites and promotes its localization to the leading edge of the cells.
GIT1 co-localized with paxillin in adhesions and cytoplasmic complexes. As small GTP-binding proteins, such as ARF1, participate in recruitment of paxillin to adhesions ( Norman et al., 1998), we asked whether GIT1 may be involved in cycling of paxillin between these two compartments. Two members of the GIT family, PAG3/PAPα and GIT2-short, appear to mediate the subcellular localization of paxillin to perinuclear areas through the ARF-GAP domain ( Kondo et al., 2000; Mazaki et al., 2001). In our study, when ectopic GIT1 was expressed in CHO K1 cells, we did not observe an increase in paxillin in adhesions. Rather GIT1 decreased the localization of paxillin to adhesions and increased its recruitment to the GIT1-containing cytoplasmic complexes. When we stained the GIT1-expressing cells with phospho-paxillin antibody, the active form of paxillin (tyrosine-phosphorylated) was detected in adhesions, but not in the cytoplasmic complexes. This suggests that GIT1 complexes sequester paxillin in its inactive conformation in the cytoplasmic pool. Once paxillin is recruited to adhesions, it may function to target GIT1 to this subcellular compartment. Consistent with our hypothesis, the kinetics of GIT1 localization indicated that paxillin was recruited to adhesions before GIT1 (R.-I.M. and A.F.H., unpublished). In addition, when paxillin was ectopically expressed, we observed an increase in GIT1 in adhesions. Furthermore, the N-terminus of GIT1 was not recruited to adhesions, indicating that the paxillin-binding domain, but not the ARF-GAP domain, is necessary for its localization.
The small GTPase Rac induces the formation of lamellipodia and initiates the development of focal complexes in this region ( Rottner et al., 1999; Hall, 1998). In our study, GIT1-induced protrusions were decreased in cells expressing dominant-negative Rac. This indicates that the formation of GIT1 protrusions is regulated by the Rac pathway. In cells expressing constitutively active Rac, we observed a significant increase in GIT1 localization to adhesions and the leading edge of the lamellipodia and a decrease in cytoplasmic structures. However, Rac did not localize to the GIT1 cytoplasmic complexes. Thus, Rac can regulate the intracellular distribution of GIT1, but its localization to the cytoplasmic complexes is not necessary for the targeting of GIT1. As localization of nGIT1 was unaffected by expression of constitutively active Rac, paxillin may be involved in Rac-induced localization of GIT1 to adhesions.
On the basis of our observations, we propose the following working model. GIT1 complexes cycle between cytoplasmic pools, adhesions and the leading edge. The paxillin-binding domain is necessary for GIT1 localization to adhesions, suggesting that paxillin recruits GIT1 to these structures. Consistent with this, paxillin localizes to adhesions prior to GIT1 recruitment (R.-I.M. and A.F.H., unpublished). GIT1 may then deliver PAK and PIX to adhesions and membrane protrusions. At the leading edge, PAK can regulate actin polymerization. Additionally, PAK targeting to adhesions may regulate their turnover and thereby modulate cell migration. PIX, in this location, might also serve to activate Rac owing to its activity as an exchange factor. As adhesions break down at the rear of a migrating cell, the GIT1 cytoplasmic complexes move from the adhesions toward the cell body. These GIT1 signaling modules are then available for targeting to adhesions and the membrane at the leading edge of the lamellipodia. In this manner, GIT1, through its function as an adaptor protein, can regulate protrusive activity and cell migration. It has recently been shown that GIT1 is tyrosine phosphorylated in an adhesion-dependent manner ( Bagrodia et al., 1999). It is tempting to speculate that phosphorylation of GIT1 may serve as a regulatory mechanism, perhaps by altering the conformation of this protein, and thus affecting its subcellular localization and/or binding partners. Rac may also play a role in regulating the distribution of this molecule among the compartments.
This work was supported by NIH Grant GM23244 and the University of Virginia Cancer Center. D.J.W. was supported by National Institutes of Health postdoctoral training grant HD07528-01. We would like to thank Jonathan Chernoff and Chris Turner for reagents and suggestions. We are especially grateful to James Casanova for helpful discussion and advice. We would also like to extend our gratitude to Karen Donais for excellent help with the figures.
Movies available on-line
- Accepted January 23, 2002.
- © The Company of Biologists Limited 2002