The Arf6 GTPase-activating Proteins ARAP2 and ACAP1 Define Distinct Endosomal Compartments That Regulate Integrin α5β1 Traffic*

Background: Arf6 has a number of distinct effects on trafficking of integrins. Results: Two Arf6 GAPs, ARAP2 and ACAP1, were distinctly localized and had differential effects on integrin trafficking and integrin adhesions. Conclusion: Arf6 effects on integrins are determined by associated Arf6 GAPs. Significance: Arf GAPs define control points of integrin traffic important to cellular behaviors related to invasion and metastasis in cancer. Arf6 and the Arf6 GTPase-activating protein (GAP) ACAP1 are established regulators of integrin traffic important to cell adhesion and migration. However, the function of Arf6 with ACAP1 cannot explain the range of Arf6 effects on integrin-based structures. We propose that Arf6 has different functions determined, in part, by the associated Arf GAP. We tested this idea by comparing the Arf6 GAPs ARAP2 and ACAP1. We found that ARAP2 and ACAP1 had opposing effects on apparent integrin β1 internalization. ARAP2 knockdown slowed, whereas ACAP1 knockdown accelerated, integrin β1 internalization. Integrin β1 association with adaptor protein containing a pleckstrin homology (PH) domain, phosphotyrosine-binding (PTB) domain, and leucine zipper motif (APPL)-positive endosomes and EEA1-positive endosomes was affected by ARAP2 knockdown and depended on ARAP2 GAP activity. ARAP2 formed a complex with APPL1 and colocalized with Arf6 and APPL in a compartment distinct from the Arf6/ACAP1 tubular recycling endosome. In addition, although ACAP1 and ARAP2 each colocalized with Arf6, they did not colocalize with each other and had opposing effects on focal adhesions (FAs). ARAP2 overexpression promoted large FAs, but ACAP1 overexpression reduced FAs. Taken together, the data support a model in which Arf6 has at least two sites of opposing action defined by distinct Arf6 GAPs.

Endocytic traffic of transmembrane proteins, including cell adhesion molecules such as integrins, is emerging as a critical pathway in cell migration and signaling related to cancer invasion and metastasis (1). Arf6, a GTPase of the Ras superfamily, is a central component of the regulatory machinery controlling the relevant endocytic trafficking (2)(3)(4). However, the trafficking pathways and the site of Arf6 action are now appreciated to be more complex than prevailing paradigms. The understand-ing of processes such as invasion and metastasis will be advanced by further description of the relationship between Arf6 and the range of itineraries possible for cargos such as integrins.
Distinct endocytic compartments and the connections between them are continuing to be discovered. In the prevailing paradigms, endocytosed cargos are translocated to a common sorting compartment from which they are directed toward the degradative pathway or either a fast or slow recycling pathway (5,6). This description is now understood to be an oversimplification, with recent descriptions of integrins being recycled from the degradative pathway (7) and other cargos bypassing the common sorting early endosome for direct transfer to the Arf6/Rab22 tubular recycling endosome (8). Therefore, the pre-early endosome may be a critical sorting station for determining the endocytic itinerary. One class of pre-early endosomes defined by the presence of adaptor protein containing PH 2 domain, phosphotyrosine-binding (PTB) domain, and leucine zipper motifs 1 and 2 (APPL1/2) has been found to traffic integrins (9).
An understanding of the mechanism of Arf6 function within the endocytic recycling compartment has remained elusive, as has been the effect of Arf6-dependent recycling of the cargo integrin on cellular structures and behaviors. Arf6 has been reported to accelerate the recycling of cargo, including transferrin and integrin ␣5␤1 (10 -12). However, it slows the recycling of integrin ␣v␤3 (12), the opposite of the effect on integrin ␣5␤1. In addition, Arf6 has been reported to increase the endocytosis of integrin ␣5␤1 (13). Adding to the complexity are the reported cellular effects of Arf6-dependent endocytic traffic. Focal adhesions (FAs) are structures that mediate the attachment of cells to the extracellular matrix. FAs contain clustered integrins that bind the extracellular matrix and a cytoplasmic plaque that connects the integrins to the actin cytoskeleton and is comprised of hundreds of structural and signaling proteins, including vinculin and paxillin (14). Arf6-induced endocytosis * This work was supported by the intramural program of the NCI, National
Integrin Internalization Assay-Integrin internalization and recycling were followed in complex with antibodies to integrin ␤1 and detected with secondary antibody for FACS (23) and immunofluorescence experiments. To study internalization, cells were incubated at 4°C with R-phycoerythrin (PE)-or FITC-conjugated anti-␤1 integrin (clones MEM101A and 12G10, respectively) antibody at 5 g/ml for 30 min. Cells were washed to remove unbound antibody. Internalization was initiated by adding medium (37°C). At each time point, cells were put on ice, washed with ice-cold PBS ϩ 0.2% BSA, and incubated at 4°C for 30 min with Alexa Fluor 647 anti-mouse antibody in cold PBS ϩ 0.2% BSA to detect the remaining cell surface primary antibody. Cells were washed, detached from plates using StemPro Accutase (Invitrogen), and analyzed using flow cytometry (BD FACSCalibur, BD Biosciences). The ratio of normalized fluorescence intensity from integrin antibody on the cell surface at each time point (signal from secondary Ab) and total surface integrin labeled by antibody at time 0 (signal from primary integrin Ab) was determined. For the internalization assay using MAB13 Ab (inactive ␤1 integrin) that was unconjugated to a fluorophore, all procedures were the same, except the signal was normalized to time 0.
Transferrin Recycling Assay-Cells were incubated for 30 min with 0.025 g/ml Transferrin-Alexa Fluor 488 (Invitrogen) in media. Cells were washed three times with ice-cold PBS and incubated in media at 37°C for the indicated times. Cells were transferred on ice, washed, and detached for FACS as described for integrin. The recycled transferrin was calculated as 100% minus the percentage of cell-associated signal at time 0.
EGFR Degradation-siRNA-treated HeLa cells were serum starved overnight in DMEMϩ0.1% FBS and treated with 100 ng/ml EGF for the times indicated. Cells were lysed in lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100 supplemented with protease and phosphatase inhibitor mixture), and 20 g of lysates were subjected to Western blotting using anti-EGFR antibody.
Immunoprecipitation and Western Blotting-Cells were lysed in lysis buffer. Proteins were immunoprecipitated using either anti-FLAG antibody conjugated to agarose or anti-GFP and anti-ARAP2 antibody precipitated with protein A/G-agarose. Immunoblots were processed by standard methods and visualized with enhanced chemiluminescence.
Immunofluorescence and Confocal Microscopy-Cells were plated on coverslips coated with 10 g/ml fibronectin and fixed in 4% paraformaldehyde. For AlF 4 Ϫ treatment, cells were incubated in media containing 30 mM NaF and 50 M AlCl 3 for 30 min before fixation. Fixed cells were incubated in 15 mM glycine for 10 min and 50 mM NH 4 Cl for 2 ϫ 10 min and then permeabilized and blocked with 0.2% saponin, 0.5% BSA, and 1% FBS in PBS for 20 min. Cells were incubated in primary antibody for 1 h, followed by secondary antibodies for 1 h, and mounted in DakoCytomation fluorescent mounting medium (Dako, Carpinteria, CA). Images for fixed cells were taken on a Zeiss LSM 510 attached to a Zeiss Axiovert 100 M with a 63 ϫ 1.4 numerical aperture plan Neofluar oil immersion lens (Carl Zeiss, Thornwood, NY). For FRAP, the cells were plated on a Lab-TekII four-chambered coverglass coated with fibronectin and visualized with a Zeiss Axiovert 200 microscope and PerkinElmer Life Sciences UltraView spinning disk confocal system using a 63 ϫ 1.4 numerical aperture plan Neofluar oil immersion objective. Cells were imaged in real time on a stage preheated to 37°C with a CO 2 chamber using an Orca-ERII camera (Hamamatsu, Bridgewater, NJ). ␣5-Integrin-EGFP-labeled adhesions were bleached using a 488-nm laser at full power. During fluorescence recovery, images were captured at 200-to 500-ms intervals for less than 5 min at a reduced laser intensity (40ϳ60%) to minimize undesired photobleaching. FRAP data were fit to exponential curves using GraphPad Prism (La Jolla, CA).
Image Analysis and Statistics-Focal adhesion number per cell was analyzed using ImageJ (National Institutes of Health, http://rsb.info.nih.gov/ij/) as described previously (24). For analysis of photobleaching recovery, the integrated fluorescence intensity was recorded in prebleach and recovery image series and normalized so that the prebleach value was 100%. The recovery halftime (t1 ⁄ 2 ) and mobile fraction were calculated using the normalized recovery values with the one-phase exponential association fit in GraphPad Prism (GraphPad Software).
For quantification of the colocalization of integrins with endosomal markers, z stack images of consecutive optical planes spaced by 0.5 m were acquired to cover the whole cell volume using confocal microscopy. Pearson's coefficient was determined using Imaris 7.4.0. The differences between multiple and two treatments were analyzed by one-way analysis of variance using Bonferroni's multiple comparison test and Student's t test, respectively, with p Ͻ 0.05 considered to be significant.

ARAP2 Affects Endocytosis and
Recycling of Integrins Differently Than ACAP1-Arf6⅐GTP can have opposing effects on integrin traffic and integrin-dependent cell behaviors. Arf6⅐GTP has been reported to accelerate integrin recycling with an increase in integrin-based cell adhesion (11,25,26), and it has also been reported to decrease cell adhesion because of increased integrin internalization, leading to a loss of FAs (13,27). These apparently contradictory results may be explained by binding of Arf6⅐GTP to different effectors and/or functioning of Arf6⅐GTP at multiple sites. One Arf6 GAP, ACAP1, not associated with FAs, promotes recycling of integrins. Three other bona fide Arf6 GAPs, ARAP2 and GIT1/2, reside in and regulate FAs, but their effects on integrin traffic have not been examined. Here we examined ARAP2 and ACAP1. We hypothesize that differential effects of Arf6 are mediated, at least in part, by interacting with structurally distinct Arf6 GAPs. In an initial test of this idea, we compared the effect of ACAP1 and ARAP2 knockdown on the net internalization rate of integrin ␤1, examined for 30 min. Integrins are found in at least two states: an active form, which has a high affinity for extracellular matrix ligands, and an inactive form that has a low affinity for extracellular matrix ligands. The two forms have different conformations and can be distinguished by binding to specific antibodies. We first focused on the active form of integrin ␤1. An antibody specific for active ␤1 was used to measure the endocytosis of integrin (23). Knockdown of ACAP1 reduced the surface levels of active integrin ␤1 by 13.7 Ϯ 0.02%. Total cellassociated active ␤1 integrin was also decreased by 17.1 Ϯ 0.06% (Fig. 1B). The ratio of surface to total integrin was not changed. Cells with reduced ACAP1 internalized a greater fraction of integrin ␤1 after 20 min of internalization. After 30 min, there was 34% greater internalization of integrin ␤1 compared with controls. This result is consistent with reduced recycling through the Rab11 (slow) recycling pathway that has been reported for ACAP1 knockdown (Fig. 1C) (22,28). ARAP2 knockdown was different from ACAP1 knockdown in two respects. First, ARAP2 knockdown cells had a greater content of integrin ␤1, whereas ACAP1 knockdown cells had a small decrease in integrin ␤1 content. Surface levels of active ␤1 integrin were increased ϳ2-fold in ARAP2 siRNA-treated HeLa (Fig. 1E) and MDA-MB-231 cells ( Fig. 2A). Total cellassociated active ␤1 integrin was also increased ϳ2-fold in ARAP2 knockdown cells (Figs. 1E and 2A). Consequently, the ratio of surface to total integrin was not affected (48 Ϯ 3.1% for control versus 44 Ϯ 7.1% for ARAP2 knockdown in HeLa cells and 14 Ϯ 1.7% versus 14 Ϯ 0.4% in MDA-MB-231 cells). The second difference between ARAP2 and ACAP1 knockdown was in the internalization of integrins. In ARAP2 knockdown cells, the net rate and extent of integrin ␤1 internalization were 50ϳ60% that of the control HeLa cells (Fig. 1F), and the difference from the control was observed within 5 min of initiating uptake. In MDA-MB-231 cells, ARAP2 knockdown also reduced the net internalization rate of integrin ␤1 (Fig. 2B). The effects of ARAP2 knockdown to increase integrin levels and to  slow the net rate of integrin internalization were independent of the activation state of integrin ␤1 because these results were also observed when using an antibody specific for inactive ␤1 or an antibody that does not discriminate between active and inactive ␤1 (Fig. 1, H-K).
The net rate of internalization is a balance of endocytosis and recycling. Therefore, a slower net rate of integrin internalization in ARAP2 knockdown cells could be due to reduced endocytosis or accelerated rapid recycling (29). In the presence of primaquine (PQ), which blocks recycling, therefore isolating endocytosis (30), the integrin endocytic rate in ARAP2 knockdown cells was indistinguishable from that in control cells (Fig.  1F). Therefore, the reduction in the net internalization rate in ARAP2 knockdown cells was likely due to accelerated recycling, in contrast to the effect of ACAP1 knockdown of slowing recycling (28). Because endocytosis occurred rapidly compared with the time resolution of the experiment, we cannot exclude an effect of ARAP2 on endocytosis on the basis of this experiment.
For further information on endocytosis, at least at FAs, we determined the exchange rate of integrin at FAs by FRAP analysis using ␣5-EGFP as a model integrin. ␣5-EGFP integrin was coexpressed with untagged ␤1 integrin in cells treated with control or ARAP2 siRNA. In Fig. 3A, example FRAP results for control and ARAP2 knockdown cells are shown. In Fig. 3B, the averages of t1 ⁄ 2 for recovery and mobile fraction determined from 38 traces for control cells and 39 traces for ARAP2 knockdown cells are shown. The fluorescence recovery time of ␣5-EGFP was shorter and the exchangeable fraction of ␣5-EGFP was larger in ARAP2 knockdown cells (Fig. 3). This result is consistent with more rapid endocytosis and recycling rates of integrin in ARAP2 knockdown cells.
We examined additional cargos that transit the endocytic pathway. Following endocytosis, most transferrin receptor is recycled. We found that neither ARAP2 knockdown nor ACAP1 knockdown had a detectable effect on transferrin recycling (Fig. 4, A and B). The result with ACAP1 is different from that reported previously (22), which could be the result of a difference in the cell strain used or methodology. We also examined EGFR. Following receptor occupancy and endocytosis, EGFR is primarily targeted to the lysosome for degradation. We measured the rate of EGFR degradation to determine the effects on this pathway (31). Different from integrins, neither the total EGFR level nor the EGFR degradation rate was affected by ARAP2 knockdown or ACAP1 knockdown (Fig. 4, C and D).
ARAP2 Affects the Transit of Integrins from APPL to EEA1 Endosomes-We characterized the intracellular itinerary of integrins in HeLa cells with reduced ARAP2 expression. APPL1/2 is recruited by Rab5 and Rab21 to pre-early endosomes that form immediately after clathrin-dependent endocytosis. Subsequently, EEA1 is recruited to form the canonical early endosome that sorts transmembrane cargo proteins to various endosomal compartments, including the tubular recycling endosome that is positive for Rab11 and Rab22 (5,32). ACAP1 has been reported to affect traffic of integrins through the tubular recycling compartment (11,22,28). To determine the step in the endocytic pathway affected by ARAP2, cells with reduced expression of ARAP2 were coimmunostained for ␤1 integrins and either endogenous APPL or endogenous EEA1 (Fig. 5). ␤1 integrins and APPL colocalized to a greater extent in HeLa cells with reduced ARAP2 than controls (Fig. 5, A and C). Similar results were obtained in the breast cancer cell line MDA-MB-231 (Fig. 5D). ␤1 integrin was also stained in control and ARAP2 knockdown cells expressing Rab5-GFP or Rab21-mCherry (Fig. 5E). Similar trends of more association of ␤1 integrin with these Rab5 family proteins were observed when ARAP2 was knocked down, although the data were not as clear, possibly because of the heterogeneity of the Rab5 compartment, which includes both APPL and EEA1 compartments. In contrast, the association of ␤1 integrins with EEA1 decreased in ARAP2 knockdown cells (Fig. 5, B and C). This reciprocal change in integrin association with APPL and EEA1 endosomes was observed for both active and inactive integrins (Fig. 5, C and  E). ␤1 integrins no longer accumulated in APPL endosomes when FLAG-tagged wild-type ARAP2 was overexpressed in ARAP2 knockdown cells. ArfGAP-dead FLAG-[R728K]-ARAP2 did not rescue the defect (Fig. 5, A and C). From EEA1 endosomes, integrins can be sorted to the late endosomes/lysosomes or recycling compartments containing Rab4 or Rab11 (5,33). In ARAP2 knockdown cells, colocalization of active ␤1 integrins with lysosomes (LAMP1) decreased (Fig. 5E), which may explain increased cellular levels of integrin (Fig. 1). We did   OCTOBER 31, 2014 • VOLUME 289 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 30241 not detect differences in integrin association with Rab4 or Rab11 in ARAP2 knockdown cells (Fig. 5E).

ARAP2 Defines an Arf6/APPL Endosome
ARAP2 also affected the cellular distribution of the APPL compartment. In both HeLa and MDA-MB-231 cells, APPL1positive structures are distributed throughout the cells (Fig. 5). A small fraction of the cells had the APPL1 signal restricted to the periphery. Reducing ARAP2 expression in HeLa (Fig. 5, F and G) and MDA-MB-231 cells (Fig. 5H) increased the percentage of cells with peripheral APPL endosomes, further supporting the idea that ARAP2 regulates the APPL compartment.
ARAP2 Associates with an APPL1/Arf6-positive Compartment Morphologically Distinct from the ACAP1 Compartment-We compared the cellular distributions of ARAP2 and ACAP1 relative to APPL and Arf6 in the absence and presence of AlF 4 Ϫ , which can form a stable membrane-associated Arf6⅐GDP⅐AlF 4 Ϫ ⅐ArfGAP complex (34). In the absence of AlF 4 Ϫ , GFP-APPL and FLAG-ARAP2 colocalized in punctate structures with Arf6-HA (Fig. 6A, Pearson's coefficient 0.20 Ϯ 0.061). In cells expressing Arf6-HA with FLAG-ARAP2 and GFP-APPL1 and treated with AlF 4 Ϫ , the three proteins colocalized in membrane protrusions (Fig. 6, A and B, Pearson's coefficient 0.24 Ϯ 0.051 for FLAG-ARAP2 and GFP-APPL1). In cells expressing FLAG-ACAP1, GFP-APPL1, and Arf6-HA and treated with AlF 4 Ϫ , ACAP1 colocalized with Arf6 in tubules that had terminal expansions enriched in GFP-APPL1 (Fig. 6B). The tubules were not observed in cells expressing ARAP2 instead of ACAP1. In cells coexpressing FLAG-ARAP2 and GFP-ACAP1, ARAP2 and ACAP1 did not colocalize in the tubular structures (Fig. 6C).
Examination of complex formation with APPL also revealed differences between ARAP2 and ACAP1. FLAG-ARAP2 and FLAG-ACAP1 were immunoprecipitated from lysates of cells expressing GFP-APPL1 and either FLAG-ARAP2 or FLAG-ACAP1 with an anti-FLAG antibody. GFP-APPL1 was detected in the precipitates from cells coexpressing FLAG-ARAP2 but not in precipitates from cells expressing GFP-APPL1 alone or GFP-APPL1 with ACAP1 (Fig. 6D). Similarly, when GFP-APPL1 was immunoprecipitated, more than 3% input of FLAG-ARAP2 was detected in the precipitates (Fig. 6E, note that the input signal is light, indicated by the asterisk in the second lane from the right). FLAG-ACAP1 was also present in these precipitates, but the precipitation was less efficient, with less than 3% of the input FLAG-ACAP1 present. Finally, endogenous APPL1 was also coimmunoprecipitated with endogenous ARAP2 (Fig. 6F), supporting the conclusion that ARAP2 associates with APPL1 in cells.
Cell morphology was differentially affected by ARAP2 and ACAP1. In the experiments determining whether ARAP2 and ACAP1 associated with different cellular structures, the cell morphology appeared to be related to the GAP that was expressed (Fig. 6). Furthermore, ACAP1 expression has been reported previously to prevent the formation of Arf6-induced protrusions (17). To determine whether ARAP2 also affects Arf6-dependent cell morphology, we compared cells expressing Arf6-HA alone or with either FLAG-ARAP2 or FLAG-ACAP1 and treated with AlF 4 Ϫ (Fig. 7). We classified Arf6-positive structures on the basis of morphology: ruffles, tubules, protrusions, and blebs (Fig. 7A). In the presence of AlF 4 Ϫ , Arf6-HA-expressing cells exhibited prominent membrane ruffles at the periphery and on the dorsal surface of the cell (Fig. 7, A and  B, Ruffles), which were rarely seen when ARAP2 or ACAP1 was coexpressed with Arf6-HA. Expression of ARAP2 and ACAP1
The tubules and protrusions induced by ACAP1 and ARAP2 were dependent on Arf6⅐GTP. Neither formed in AlF 4 Ϫ -treated cells expressing ACAP1 or ARAP2 with a dominant negative mutant of Arf6 ([N122I]Arf6-HA) (Fig. 7, C-E). We conclude that the formation of protrusions and blebs depend on ARAP2 together with active Arf6, whereas tubules depend on active Arf6 with ACAP1.   OCTOBER 31, 2014 • VOLUME 289 • NUMBER 44

ARAP2 Defines an Arf6/APPL Endosome
FAs using vinculin as a marker. ARAP2 overexpression resulted in fewer small adhesions and more large adhesions (Ͼ3 m 2 ) (Fig. 8, A and B). In contrast, ACAP1 overexpression reduced all adhesions (Fig. 8, A and B). The expression levels of the recombinant ARAP2 were estimated to be 200-to 500-fold higher than that of endogenous ARAP2 (Fig. 8C). However, ARAPs are highly regulated (35), which would limit extreme effects of overexpression. In fact, ARAP2 knockdown has been reported previously to reduce FA size and number (36), consistent with our result in which overexpression resulted in larger FAs. Neither ARAP2 nor ACAP1 overexpression had a significant effect on total or surface active integrins (Fig. 8D). However, the overexpressed ARAP2 and ACAP1 associated with specific Arf6 compartments and had distinct effects on FAs. Therefore, the effect on the morphology of Arf6 structures and FAs cannot be explained by different amount of integrins and is not likely due to a nonspecific effect of the Arf GAP on Arf6 function. Rather, our results are consistent with the idea that a reduction in recycling by ARAP2 overexpression and accelerated recycling by ACAP1 overexpression may explain the effects on FAs because rapid recycling of ␣5␤1 has been reported to destabilize adhesions (12). We conclude that ARAP2 and ACAP1 are Arf6 GAPs that associate with different Arf6-positive endosomal compartments and have different effects on integrin traffic and opposing effects on FA morphology.

DISCUSSION
We examined two Arf6 GAPs to understand the opposing effects of Arf6 on integrins and integrin-dependent cellular processes. Arf6 and Arf6 exchange factors have been reported to affect both integrin endocytosis and exocytosis (11)(12)(13)26), leading to the idea that Arf6 may affect integrin traffic at two distinct sites. We found that ARAP2 and ACAP1, two bona fide Arf6 GAPs, localize to morphologically distinct compartments and differentially affect integrin traffic and FAs. The Arf GAPs may, at least in part, define the Arf6 compartment and action.
The precise relationship of the ARAP2/APPL/Arf6 compartment within the complex traffic pathways of integrins remains to be fully characterized. Integrins are endocytosed into vesicles that fuse with a pre-early endosome. Some pre-early endosomes contain APPL. From the pre-early endosome, integrins are transported to the early endosome. Consistent with the accumulation of integrin in the APPL compartment in ARAP2 knockdown cells, ARAP2 may promote pre-early to early endosome traffic (Fig. 9, route A). From the early endosome, integrin is transported to either the degradative pathway or the recycling pathway and may be recycled from either (1,33). A block in pre-early endosome to early endosome transport in ARAP2 knockdown cells would explain the result of decreased transit through the degradative pathway and increased total integrin Ϫ were examined. A higher magnification is shown in insets, and arrows indicate structures where colocalization was detected. Scale bars ϭ 10 m. D and E, differences in ARAP2-APPL1 and ACAP1-APPL1 complex formation. Lysates from HeLa cells expressing FLAG-ARAP2 or FLAG-ACAP1 alone and in combination with GFP-APPL1 were immunoprecipitated (IP) through the FLAG epitope or GFP followed by immunoblotting using anti-GFP and anti-FLAG antibodies. The asterisk indicates the band representing FLAG-ARAP2 input in double-transfected cells. F, association of endogenous ARAP2 with APPL1. Endogenous ARAP2 was immunoprecipitated from HeLa lysates, followed by anti-APPL immunoblotting.
levels. The accelerated recycling rate may be the result of transport directly from the pre-early endosome to the tubular recycling compartment (6,8) or transport from the APPL endosome to the plasma membrane (Fig. 9, routes B and C). It is plausible that ARAP2 prevents transport from pre-early endosome to tubular recycling endosome or to the plasma membranes and that knockdown of ARAP2 would, therefore, accel-erate transport through these two routes (Fig. 9, routes B and  C). Alternatively, with a block in transport of integrin from the pre-early endosome to the early endosome, the tubular recycling compartment or direct transport to the plasma membrane may be a default pathway for transport. Regardless of the specific path, as far as we are aware, this is the first report of an Arf GAP regulating the APPL compartment. Arf6 function in a pre-early endosome has been described previously in reports examining the effects of expressing constitutively active Arf6 (37). Arf6 associated with two compartments. One was the recycling endosome that also contained ACAP and Rab11 (4,38). The other compartment was enriched in phosphatidylinositol 4,5-bisphosphate and trapped cargos (37,39). The structure was assumed to have formed early during endocytosis. A plausible connection is that this phosphatidylinositol 4,5-bisphosphate-rich Arf6 compartment matures into the APPL/ARAP2/Arf6 compartment.
The mechanism responsible for the morphological differences in the Arf6-positive compartments in cells expressing ARAP2 or ACAP1 and treated with AlF 4 Ϫ remains elusive. ARAP2 or ACAP1 could generate the structures via an unique domain. For example, the Bin/Amphiphysin/Rvs domain of ACAP1 could induce the curvature necessary for the tubules to form. Alternatively, the inability to hydrolyze GTP on Arf6 because the GAP is sequestered in the Arf6⅐GDP⅐AlF4 Ϫ ⅐GAP complex stabilizes or expands the specific GAP compartment.
The association of ARAP2 with an APPL/Arf6-positive compartment that regulates integrin traffic links several previous studies. APPL, which associates with very early endosomes, has been implicated in FA dynamics through its interaction with GAIP-interacting protein C terminus, member 1 (GIPC1, which binds to integrins (9)) to control integrin traffic and as an element of signaling that regulates FA dynamics (40). Arf6 has been reported to control integrin recycling and adhesion (11,13,29) and has been found to function downstream of synde-can4 to direct the itinerary of specific integrins to control FA dynamics (12). Increased Arf6⅐GTP levels accelerated ␣5␤1 integrin recycling, destabilizing FAs, whereas decreased Arf6⅐ GTP levels increased ␣v␤3 recycling, stabilizing FAs. ARAP2 could respond to cellular signals to regulate Arf6 and APPL in an endosomal compartment that controls the integrin traffic itinerary, therefore providing a single model explaining these reports.
Cargo sorting may be a common feature of Arf-dependent trafficking steps. We found that ARAP2 specifically affects the trafficking of ␤1 integrin without effects on EGFR or transferrin. Similarly, ACAP1 is thought to regulate the traffic of specific cargos from the recycling endosome (10,41). A role for ArfGAP1, functioning with a class 1 or 2 Arf, in the cargo sorting in Golgi-to-endoplasmic reticulum trafficking has also been described (42)(43)(44), and ArfGAP3, also with a class 1 or 2 Arf, is postulated to affect the traffic of specific cargo between the early and late endosomes (45). Arf1 and Arf5 have also been reported to affect FAs and integrin trafficking (23,36). Particular Arf GAPs may provide the cargo specificity necessary to navigate the complex endocytic pathways. Particular Arf GAPs may also determine specific intracellular itineraries. ARAP2 and ACAP1 handle integrin ␣5␤1 in common but affect different compartments with different consequences.
Thirty-one genes encode Arf GAPs in humans, many with multiple splice variants (15). In many cases, a clear Arf specificity for Arf GAPs has not been able to be determined. It is possible that the Arf GAPs may be promiscuous in this regard. The "wobble" in the Arf may provide additional Arf/Arf GAP pairs for fine regulation of membrane traffic, with both the specific Arf and the specific GAP determining the particular effect.
ARAP2 may also have a central role in signaling from the APPL/Arf6 compartment. Phosphoinositides may be affected through regulation of Arf6, which activates phosphatidylinositol phosphate kinase (46), and by ARAP binding to SHIP2, a phosphoinositol 5Јphosphatase (47). ARAP2 also affects Rac1⅐GTP levels and is a target of Rho (18,36). Given that ARAP2 integrates a number of signals, we envision ARAP2 to be a key regulatory component of a signaling and membrane traffic station defined by the presence of APPL and dependence on Arf6. A plausible model is that ARAP2 and ACAP1 respond differentially to cellular signals to control the integrin trafficking necessary for specific effects on cell migration.
In summary, we identify a unique Arf6/APPL/ARAP2 endosomal compartment that is distinct from the ACAP1 endosomal compartment. We propose that, in addition to sorting specific cargos, particular endocytic itineraries are mediated by Active α5β1 Inactive α5β1 Arf6•GTP FIGURE 9. Model for the sites of action of ARAP2 and ACAP1 in the endocytic pathway. One Arf6 GAP, ARAP2, associates with APPL1 and controls the traffic of integrins from APPL endosomes. ARAP2 may promote integrin traffic from APPL to EEA1 early endosomes (route A) or prevent integrin transport from APPL directly to tubular recycling endosomes (route B) or to the plasma membrane (route C), leading to reduced ␣5␤1 recycling. In contrast, another Arf6 GAP, ACAP1, localizes to tubular recycling endosomes and promotes ␣5␤1 recycling. RE, recycling endosome; Ly, lysosome; LE/MVB, late endosome/multivesicular body.
distinct compartments defined, at least in part, by particular Arf/Arf GAP pairs (Fig. 9).