ADP-ribosylation Factor-like GTPase ARFRP1 Is Required for Trans-Golgi to Plasma Membrane Trafficking of E-cadherin*

ADP-ribosylation factor-related protein 1 (ARFRP1) plays a specific role in Golgi function controlling recruitment of GRIP domain proteins and ARL1 to the trans-Golgi. Deletion of the mouse Arfrp1 gene causes embryonic lethality during early gastrulation, because epiblast cells detach from the ectodermal cell layer and do not differentiate to mesodermal tissue. Here we show that in Arfrp1-/- embryos E-cadherin is mistargeted to intracellular compartments, whereas in control embryos it is present at the cell surface of trophectodermal and ectodermal cells. In enterocytes of intestine-specific Arfrp1 null mutants (Arfrp1vil-/-), E-cadherin is associated with intracellular membranes, partially colocalizing with the cis-Golgi marker GM130 or with punctae close to the cell surface. In contrast, in control enterocytes E-cadherin is exclusively located in the lateral membranes. In addition, ARL1 is dislocated from Golgi membranes to the cytosol of Arfrp1vil-/- enterocytes. Depletion of endogenous ARFRP1 by RNA interference leads to a dislocation of E-cadherin from the cell surface in HeLa cells and to a reduced cell aggregation in Ltk-Ecad cells. ARFRP1 was coimmunoprecipitated in a complex with E-cadherin, α-catenin, β-catenin, γ-catenin, and p120ctn from lysates of Madin-Darby canine kidney cells stably expressing myc-ARFRP1. These data indicate that knock-out of Arfrp1 disrupts the trafficking of E-cadherin through the Golgi and suggest an essential role of the GTPase in trans-Golgi network function.

GTPases of the ADP-ribosylation factor (ARF) 3 family operate as molecular switches in the regulation of vesicular traffick-ing and organelle structure (1,2). The ARF family includes three different groups of proteins, the ARFs, the ARLs (ARFlike proteins), and the secretion-associated Ras-related proteins. ARF-related protein 1 (ARFRP1) is a 25-kDa GTPase and member of the ARL family (2,3). In contrast to other ARFs and ARLs, ARFRP1 can hydrolyze GTP in the absence of a GTPaseactivating protein and lacks the N-myristoylation site (glycine 2), which is required for membrane association (3). For the closest relative of ARFRP1, the yeast Arl3p protein, it was shown recently that membrane association is mediated via acetylation of the N-terminal methionine residue (4,5). ARFRP1 interacts with the Sec7 domain of the ARF-specific guanine nucleotide exchange factor cytohesin 1 in a GTP-dependent manner. This interaction resulted in the inhibition of the ARF/Sec7-dependent activation of phospholipase D in vitro and in vivo (6).
We and others have recently shown that ARFRP1 as well as its yeast ortholog Arl3p specifically control targeting of ARL1 and its effector Golgin-245 to the trans-Golgi (7)(8)(9)(10). GTPbound ARFRP1 (ARFRP1-Q79L mutant) was associated with Golgi membranes and colocalized with ARL1. In contrast, the guanine nucleotide exchange defective ARFRP1 mutant (ARFRP1-T31N) clustered within the cytosol. Expression of ARFRP1-T31N or depletion of endogenous ARFRP1 by RNA interference disrupted the Golgi association of ARL1 and the GRIP domain protein Golgin-245 and altered the distribution of a trans-Golgi network (TGN) marker, syntaxin 6 indicating that ARFRP1 plays an important role for TGN structure and function (10).
Deletion of Arfrp1 in mice resulted in embryonic lethality (11). Arfrp1 Ϫ/Ϫ blastocysts implanted in vivo and formed egg cylinder-stage embryos that appeared normal until day 5. During early gastrulation (at day 6 -6.5), Arfrp1 Ϫ/Ϫ embryos exhibited profound alterations of the distal part of the egg cylinder. Rounded pyknotic cells within this area were only loosely attached to the ectodermal cell layer, and some apoptotic cells were found in the proamniotic cavity. This observation suggested that ARFRP1 plays a critical role in processes during early gastrulation such as adhesion-dependent morphogenesis, cytoskeletal reorganization, and/or development of cell polarity (11).
Specific contacts of cells to the extracellular matrix and to neighboring cells are fundamental for embryogenesis, survival, and wound repair. Cadherins represent a large family of cell-cell adhesion proteins that play crucial roles in tissue patterning, cellular growth control, and in the regulation of cell shape and migration (12)(13)(14). Changes in cadherin expression are associated with numerous developmental events such as epithelial-mesenchymal transitions, e.g. during gastrulation each member of the family exhibits a specific spatial and temporal expression pattern (15). E-cadherin, the prototypical member of the classic cadherin family, is a major component of epithelial adherens junctions, where it mediates cell-cell adhesion through calcium-dependent, homophilic binding between molecules on adjacent cells (13,16,17). At the adherens junction, E-cadherin is bound to catenins with ␤-catenin attached to the cytoplasmic domain of E-cadherin and ␣-catenin associated with ␤-catenin. In contrast to previous models, ␣-catenin does not directly link the cadherin-catenin complex to the actin cytoskeleton (18,19). Recently, EPLIN/Lima-1 was identified as a missing link between the cadherin-catenin complex and the actin cytoskeleton (20). p120 ctn binds to a juxtamembrane site in the cytoplasmic tail of E-cadherin (21), and several roles of p120 ctn modulating cadherin function have been discussed in the literature (22). p120 ctn is implicated to be involved in exocytosis, endocytosis, and turnover of cadherins (23)(24)(25)(26).
In this study, we tested the hypothesis that ARFRP1 modulates cadherin-mediated adhesion processes. We find ARFRP1 in a complex with E-cadherin, ␤-catenin, ␣-catenin, ␥-catenin, and p120 ctn . Additional data suggest that ARFRP1 is required for cell surface localization of E-cadherin because in the absence of ARFRP1, E-cadherin is dislocated from the plasma membrane, and cell adhesion is markedly reduced in vivo and in vitro.
Cell Culture and Transient Transfection-HeLa cells were cultured in minimum essential medium with Earle's salts plus 10% (v/v) FCS. For aggregation assays, Ltk Ϫ Ecad cells were grown in Dulbecco's modified Eagle's medium (DMEM) high glucose in the presence of 10% (v/v) FCS, 100 units/ml penicillin, and 100 g/ml streptomycin at 5% CO 2 . Transient transfections of cells were performed with Lipofectamine 2000 (Invitrogen) according to manufacturer's protocols.
MDCK T23 Cells and Stable Transfection of Myc-Arfrp1-The MDCK T23 cell line, which stably expresses the tetracycline-repressible transactivator, was described earlier (31) and was kindly provided by Prof. Keith E. Mostov (Department of Anatomy, University of California, San Francisco). MDCK T23 cells were maintained in DMEM with 10% FCS supplemented with the necessary antibiotics and cultured under continuous presence of 40 ng/ml doxycycline. The medium was renewed every 48 h. N-terminal Myc tag was fused to the mouse Arfrp1 open reading frame by PCR and cloned into the pTRE2hyg vector (Clontech). Transfection of MDCK T23 cells was performed in 6-well plates with Lipofectamine 2000 according to the manufacturer's instructions. Twenty four hours after transfection, cells were reseeded into 10-cm dishes, and selection of transfected cells was achieved with 40 ng/ml doxycycline and 300 g/ml hygromycin in DMEM. After selection for 12 days, surviving colonies were isolated with the use of cloning rings and expanded in 48 wells. At confluency, cells from each of the surviving clones were split and maintained in the presence or absence of doxycycline. ARFRP1 expression was assessed by immunofluorescence microscopy 48 h after removal of doxycycline. Clones positive for ARFRP1 expression were expanded, and inducible expression was confirmed by Western blot analysis.
Immunocytochemistry and Indirect Immunofluorescence Microscopy-At the indicated time points, cells were washed with PBS and fixed with methanol (Ϫ20°C for 10 min). Cells were washed with PBS, blocked with PBS, 0.1% (v/v) Tween 20:5% (v/v) normal goat serum for 20 min at room temperature and incubated with primary antibodies in antibody diluent (Dako, Glostrup, Denmark) for 1 h at room temperature. After extensive washing with PBS, 0.1% (v/v) Tween 20, cells were incubated with Alexa Fluor 488-or Alexa Fluor 546-conjugated secondary antibodies in antibody diluent at room temperature for 30 min. After washing with PBS, 0.1% (v/v) Tween 20, cells were mounted in fluorescent mounting medium (Dako) and analyzed with a Leica TCS SP2 Laser Scan inverted microscope. We scanned the cells sequentially with an argonkrypton laser (488 nm) to excite the Alexa 488 dye, and with a helium-neon laser (543 nm) to excite the Alexa 546 dye. The spectral detector recorded light emission at 510 -560 and 580 -660 nm, respectively. We processed images of 1024 ϫ 1024 pixels with Adobe Photoshop CS (version 8.0.1).
Generation of Intestine-specific Arfrp1 Null Mutants (Arfrp1 vilϪ/Ϫ Mice)-For tissue-specific disruption of Arfrp1, we used the Cre/loxP recombination system and generated Arfrp1 flox/flox mice in which exons 2 and 4 of Arfrp1 were flanked with loxP sites. The targeting vector also contained pGKneo/HSVtk cassette (Neo/tk) with a third loxP site that was introduced between exon 4 and 5. It was electroporated into embryonic stem (ES) cells that were screened for homologous recombination. A homologous recombined ES cell clone containing the targeted allele was retransfected with pIC-Cre to generate ES cell clones carrying the floxed Arfrp1 allele. One ES cell clone was injected into blastocysts, which were subsequently transferred into a day 2.5 pseudopregnant female C57BL/6 mouse. Male chimeric mice were mated with C57BL/6 females to generate Arfrp1 flox/ϩ mice. Arfrp1 flox/ϩ mice were backcrossed with C57BL/6 three times. Intestinespecific Arfrp1 null mutants (Arfrp1 vilϪ/Ϫ ) were generated by intercrossing Arfrp1 flox/flox with transgenic mice that express Cre recombinase under the control of the villin promotor/enhancer (villin-Cre) (32). The animals were housed in a controlled environment (20 Ϯ 2°C, 12:12-h light/dark cycle) and had free access to water and standard chow diet. All animal experiments were approved by the Ethics Committee of the Ministry of Agriculture, Nutrition, and Forestry (State of Brandenburg, Germany).
RNA Preparation and First Strand cDNA Synthesis-Total RNA from different tissues of the mice was extracted, and cDNA synthesis was performed as described previously (33).
Quantitative RT-PCR-Quantitative real time PCR analysis (qRT-PCR) was performed using the Applied Biosystems 7300 real time pcr System as described previously (33). For the determination of Arfrp1 mRNA levels in ileum and colon, a TaqMan gene expression assay was used (Arfrp1 E6_E7, Mm00513004_m1). Data were normalized (34), and a ␤-actin expression assay (Mm00607939_s1) was used as endogenous control.
Knockdown of Endogenous ARFRP1 by shRNA Interference-The mammalian expression vector, pSUPER.basic (OligoEngine), was used for expression of shRNA targeting human ARFRP1 in HeLa cells. A gene-specific insert defining a 19-nucleotide sequence corresponding to nucleotides 691-709 (GTGGATGGTGAAGTGT-GTC, GenBank TM accession number NM_003224.2, ARFRP1-shRNA) was separated by a 9-nucleotide noncomplementary loop sequence (TTCAAGAGA) from the reverse complement of the same 19-nucleotide sequence. Both sequences were subcloned into the BglII and HindIII sites of the pSUPER vector and referred to as pSUPER-ARFRP1. HeLa cells were transfected with pSUPER or pSUPER-ARFRP1 and processed for Western blot analysis or immunofluorescence after 4 -8 days of incubation.
Adhesion Assay-For cell aggregation assays, 10 5 Ltk Ϫ Ecad cells were transfected with siCONTROL TM nontargeting siRNA, ARFRP1-specific siRNAs (si-a-ARFRP1 and si-b-ARFRP1), and mutated siRNAs (scrambled-a and scrambled-b) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocols. After 3 days, cell aggregation assays were performed as described previously (37). At the time points 0 and 45 min, two 10-l aliquots were removed, and cells/cell aggregates were photographed using an Olympus BX-60 microscope with an Achrostigmat objective (10ϫ magnification, 0.25 numerical aperture). JPEG images were generated with the Soft Imaging System Color View 12 and the analysis 3.0 software. After counting of particles (cells and cell aggregate), the aggregation index was calculated according to Nagafuchi and Takeichi (41) as follows: where N 0 is the total particle number at t ϭ 0 min, and N t is the particle number after an incubation period of 45 min. Mean values Ϯ S.D. of four independent aggregation assays are presented.

Distribution of E-cadherin Is Altered in Arfrp1
Ϫ/Ϫ Embryos-We have previously shown that deletion of Arfrp1 in mice results in embryonic lethality because of the failure of differentiating epiblast cells to form the mesoderm. Epiblast cells of Arfrp1 Ϫ/Ϫ embryos detached from the embryonic ectoderm, consistent with a defect in the regulation of cell-cell adhesion (22). Because changes in cadherin expression coincide with gastrulation (38), we analyzed the expression and distribution of E-cadherin and N-cadherin in control and Arfrp1 Ϫ/Ϫ embryos between ED 5.0 and 6.5.
In ED 5.0 control embryos (Fig.  1), a weak ARFRP1 expression is detected in ectodermal cells, and E-cadherin is located at the lateral membrane of trophectodermal epithelial cells (arrow in Fig. 1) and at the surface of ectodermal cells. In contrast, in ED 5.0 Arfrp1 Ϫ/Ϫ embryos, no E-cadherin staining was visible in the trophoblast, and only a punctate pattern was detected in the embryonic ectoderm. The phenotype observed at ED 5.0 was even more pronounced at ED 6.0. At this stage, control embryos show a regular E-cadherin staining at the surface of each cell, although it is present predominantly in intracellular, aggregate-like structures in the Arfrp1 Ϫ/Ϫ embryos. No N-cadherin staining was detectable in ED 5.0 and ED 6.0 embryos as described previously (data not shown, see Ref. 39). To test whether other plasma membrane proteins were affected in Arfrp1 Ϫ/Ϫ embryos, we stained the glucose transporter GLUT1. GLUT1 was detected at the cell surface of both control and Arfrp1 Ϫ/Ϫ embryos (supplemental Fig. 1A). In addition, the plasma membrane marker Na ϩ /K ϩ -ATPase, an integral membrane protein complex, was detected at the cell surface of the Arfrp1 Ϫ/Ϫ embryos (supplemental Fig. 1B). Retention of E-cadherin in the Golgi of Arfrp1 Ϫ/Ϫ Intestinal Epithelial Cells-The intestinal epithelium is characterized by rapid cellular turnover with continuous proliferation, cellular migration, differentiation, and polarization (40). Here, E-cadherin, together with integrins, plays an important role for the development and maintenance of normal intestinal epithelial architecture and is required for complex cell-cell interactions.
As shown in Fig. 3, E-cadherin was localized in the lateral membrane of the cell surface of crypts and villi in control Arfrp1 flox/flox mice. In contrast, in Arfrp1 vilϪ/Ϫ mice we detected E-cadherin also in intracellular compartments (arrows in Fig. 3A) and as punctae (arrowheads in Fig. 3A) close to the plasma membrane. The expression of E-cadherin as detected by Western blotting (Fig. 3B) was not altered in ileum and colon of Arfrp1 vilϪ/Ϫ mice.

Defective Trans-Golgi Organization Is Associated with Altered E-cadherin Distribution in Intestinal Enterocytes in the
Absence of Arfrp1-Because we and others have previously shown that ARFRP1 controls targeting of ARL1 to Golgi membranes (7,8,10), we analyzed the ARL1 distribution in intestinal epithelial cells of control and Arfrp1 vilϪ/Ϫ mice. In Arfrp1 flox/flox mice, ARL1 was associated with Golgi membranes (arrows in Fig. 4A, left panel) and was also present in the cytosol (arrowheads in Fig. 4A, left panel). In contrast, in Arfrp1 knock-out mice ARL1 was exclusively located in the cytosol of epithelial cells. Consistent with the intense staining of ARL1 in knock-out tissue, the ARL1 protein in total lysates of ileum (Fig.  4A, lower panel) was more abundant in Arfrp1 vilϪ/Ϫ mice than in controls.
We next stained sections of ileum with specific Golgi markers. The trans-Golgi marker TGN38 (Fig. 4B) detected the Golgi in all control cells, whereas no staining was observed in intestinal sections of Arfrp1 vilϪ/Ϫ mice. In contrast, the Golgi complex marker 58K stained the Golgi in epithelial cells of control and Arfrp1 vilϪ/Ϫ mice (supplemental Fig. 2). The cis-Golgi could be visualized with the anti-GM130 antibody in both Arfrp1 flox/flox and Arfrp1 vilϪ/Ϫ cells. However, the pattern of GM130 staining differed somehow in the Arfrp1 vilϪ/Ϫ cells where the cis-Golgi appeared to be broader. Interestingly, costaining with the anti-E-cadherin antibody demonstrated that the intracellular E-cadherin was partially colocalized with GM130 (arrows in Fig. 4C), consistent with the conclusion that E-cadherin is retained in the Golgi.
To examine whether ARFRP1 also regulates the trafficking of catenins through the Golgi, we stained sections of intestine of control and Arfrp1 vilϪ/Ϫ mice with an anti-␤-catenin antibody. ␤-Catenin was detected at the cell surface of the epithelial cells of Arfrp1 flox/flox mice. In cells of Arfrp1 vilϪ/Ϫ ileum, ␤-catenin was partially located intracellularly (arrows in supplemental Fig. 3). In contrast, other cell surface proteins, e.g. the apical  OCTOBER 3, 2008 • VOLUME 283 • NUMBER 40 protein dipeptidyl peptidase 4 and the lateral protein GLUT2, showed no differences in their subcellular localization between cells from control and Arfrp1 vilϪ/Ϫ mice (Fig. 4D).

ARFRP1 Is Essential for Correct Targeting of E-cadherin in HeLa
Cells-The results shown in Figs. 1  and 3 suggested that the correct targeting of the cell-adhesion molecule E-cadherin requires the presence of ARFRP1. To further support this conclusion, we studied the cellular localization of E-cadherin in HeLa cells in which expression of ARFRP1 was suppressed by an ARFRP1-specific shRNA construct (Fig. 5A). The same shRNA construct was previously used to demonstrate that Golgi association of ARL1 and its effector Golgin-245 is disrupted in cells lacking ARFRP1 (10).
Indeed, in ARFRP1 knockdown cells identified by a diffuse ARL1 staining (10), no E-cadherin was detected at the plasma membrane (Fig. 5B). Only in cells with Golgiassociated ARL1 was cell surface staining of E-cadherin detected (see arrows in lower panel of Fig. 5B). In contrast to the results obtained by in vivo knock-out of Arfrp1 (Figs. 2-4), E-cadherin protein levels were markedly reduced by suppression of Arfrp1 expression (Fig. 5,  upper panel).
ARFRP1-regulated Targeting of E-cadherin Is Independent of ARL1-To elucidate whether the ARFRP1-dependent recruitment of ARL1 to the trans-Golgi is required for the correct targeting of E-cadherin to the cell surface, we inhibited ARL1 expression in HeLa cells as described earlier (28) and stained them for E-cadherin. No difference in E-cadherin localization was detectable in control and ARL1 knockdown cells (supplemental Fig. 4), indicating that the association of ARL1 and its effector Golgin-245 with Golgi membranes is not required for the correct E-cadherin localization at the plasma membrane.
ARFRP1 Is Associated with the E-cadherin-Catenin Protein Complex-The observation that ARFRP1 expression is required for cell sur-face targeting of E-cadherin suggests an association of ARFRP1 with the E-cadherin-catenin complex. To test this assumption, we used MDCK cells stably expressing Myctagged ARFRP1 (myc-ARFRP1) in the absence of doxycycline (Tet-Off system). Fig. 6A demonstrates that expression of myc-ARFRP1 begins 2 days after cultivation of MDCK cells in the absence of doxycycline. The expression reaches a maximum at day 4 and stays stable until day 10 (Fig. 6A). ARFRP1 was immunoprecipitated with anti-Myc antibody, and the associated proteins were detected by Western blot analyses. As shown in Fig. 6B, E-cadherin, ␣-catenin, ␤-catenin, ␥-catenin, and p120 ctn coimmunoprecipitated with ARFRP1. In contrast, IQGAP1 was not coimmunoprecipitated with the anti-Myc antibody (Fig. 6B). These observations suggest that ARFRP1, by interacting with the E-cadherin-catenin-p120 ctn complex, is involved in correct localization of E-cadherin to the cell surface. To analyze in which cellular compartment the interaction of ARFRP1 with the E-cadherin complex occurs, we stained MDCK cells overexpressing Myc-tagged ARFRP1 for E-cadherin with the anti-gp84 antibody and ARFRP1 with the anti-Myc antibody. As shown in Fig. 6C, E-cadherin is predominantly located at the cell surface, and is only partially present in intracellular membranes where it colocalizes with ARFRP1 (arrows in Fig. 6C). In contrast, myc-ARFRP1 is predominantly located in intracellular membranes and the cytosol and only to a small part at the cell surface (arrowheads in Fig. 6C).
Reduced E-cadherin-mediated Adhesion in ARFRP1 Knockdown Cells-To analyze whether E-cadherin-mediated adhesion was affected by knockdown of Arfrp1, we used Ltk Ϫ Ecad cells that are mouse fibroblasts (Ltk Ϫ ) stably expressing E-cadherin (37,41). Arfrp1 expression was inhibited by siRNA (Fig. 7, C and D), and cell aggregation assays were performed. Cells were trypsinized to generate single cell suspension and were allowed to aggregate under constant agitation for 45 min. Cell aggregates were counted before and after the incubation period. In untransfected cells or in cells transfected with a nontargeting siRNA, Ltk Ϫ Ecad cells form aggregates. In con-

. Inhibition of ARFRP1 expression in HeLa cells impairs cell surface localization of E-cadherin.
A, ARFRP1 expression was down-regulated in HeLa cells by transfection of pSUPER-ARFRP1, and protein lysates were prepared 4, 6, or 8 days after transfection. Expression of ARFRP1, E-cadherin and ␣-tubulin as a loading control were detected by Western blot analyses as described under "Experimental Procedures." B, after 4 days, cells were fixed with methanol and stained for ARL1 with an affinity-purified polyclonal anti-ARL1 antibody in combination with an Alexa488-conjugated secondary antibody. E-cadherin was stained with the anti-gp84 antibody in combination with an Alexa546-conjugated secondary antibody. Immunofluorescence was analyzed by confocal laser scanning microscopy as described under "Experimental Procedures." Arrows in the lower panel depict a cell with normal ARL1 localization showing cell surface staining of E-cadherin.
trast, in cells transfected with ARFRP1-siRNAs, a marked reduction in aggregation was observed, resulting in a 50 -60% reduction of the aggregation index (Fig. 7, A and B). To show the specificity of this effect, cells were transfected with siRNA oligonucleotides with 3 bases of the ARFRP1-specific siRNAs mutated (scrambled siRNA). Cell aggregation was unaltered in these cells. Furthermore, Arfrp1-specific siRNA only suppressed mRNA expression of Arfrp1, whereas mRNA levels of E-cadherin were not altered. These data confirm that ARFRP1 is required for localization of functional E-cadherin at the cell surface, and is therefore involved in regulation of E-cadherin-mediated cell-cell adhesion. It should be noted that in Arfrp1-depleted Ltk Ϫ Ecad cells, E-cadherin protein levels were reduced (Fig.  7C), suggesting that incorrect targeting of E-cadherin results in enhanced degradation in these cells. In contrast, E-cadherin expression was not reduced in Arfrp1 vilϪ/Ϫ intestinal cells (Fig. 3B), and E-cadherin did not colocalize with the lysosomal marker LAMP1 (supplemental Fig. 5).

DISCUSSION
Here we demonstrate that ARFRP1 is required for cell surface localization of E-cadherin and that it is associated with the E-cadherincatenin complex. First, as early as at day 5.0, Arfrp1 Ϫ/Ϫ embryos exhibit an abnormal subcellular distribution of E-cadherin, which appeared retained in intracellular compartments (Fig. 1A). Second, in epithelial cells of intestine-specific Arfrp1 knock-out (Arfrp1 vilϪ/Ϫ ) mice, E-cadherin was associated with intracellular membranes, partially colocalizing with the cis-Golgi (Figs.  3 and 4). Third, shRNA-mediated knockdown of Arfrp1 in HeLa cells resulted in a loss of E-cadherin from the cell surface (Fig. 5). Fourth, E-cadherin and its binding partners ␣-catenin, ␤-catenin, ␥-catenin, and p120 ctn coimmunoprecipitated with ARFRP1 from lysates of MDCK cells overexpressing myc-ARFRP1 (Fig. 6). Finally, E-cadherin-mediated adhesion was affected by knockdown of Arfrp1 in Ltk Ϫ Ecad cells (Fig. 7).
After its synthesis, E-cadherin is translocated from the trans-Golgi network to the cell surface for incorporation, together with catenins, into adherens-junction complexes (22,42). The partial colocalization of E-cadherin with a cis-Golgi marker in Arfrp1 knockout cells suggests that ARFRP1 is involved in the transport of E-cadherin from the Golgi apparatus to the cell surface. We also detected ␤-catenin partially located in intracellular membranes of Arfrp1 vilϪ/Ϫ epithelial cells (supplemental Fig. 3) indicating that mistargeting of E-cadherin also affects its interaction partner ␤-catenin.
Based on data in vitro data from cultured cells transfected with ARFRP1 constructs, we and others have previously suggested that the GTPase is required for targeting of ARL1 and of its effector Golgin-245 to the trans-Golgi network (7)(8)(9)(10)43). The present data provide solid proof for this conclusion in showing that ARL1 dissociated from Golgi membranes to the cytosol in an in vivo knockdown of Arfrp1 (Arfrp1 vilϪ/Ϫ epithelial cells, Fig. 4A). Furthermore, the trans-Golgi marker TGN38 was undetectable in the Arfrp1 vilϪ/Ϫ intestine (Fig. 4B), indicating that ARFRP1 is required for the correct organization of the trans-Golgi.
The observation of marked alterations of the trans-Golgi network raises the question whether ARFRP1 plays a general role in organizing the trans-Golgi, and thereby affects E-cadherin targeting, or whether ARFRP1 specifically regulates E-cadherin trafficking through the TGN. Three findings support the conclusion that ARFRP1 specifically modulates the transport of E-cadherin from the Golgi to the plasma membrane as follows.
1) ARFRP1 interacts with the E-cadherin-catenin complex. 2) Knockdown of the trans-Golgi protein ARL1, which is also required for the organization of the trans-Golgi, did not result in a mistargeting of E-cadherin in HeLa cells (supplemental Fig. 4) and in A431 cells (data not shown). 3) Knock-out of Arfrp1 did not dislocate other cell surface proteins such as the ubiquitously expressed glucose transporter GLUT1 (supplemental Fig.  1A) or the Na ϩ /K ϩ -ATPase (supplemental Fig. 1B) in intracellular membranes.
Interestingly, removal of the ARL1 effector Golgin-97 from Golgi membranes by overexpression of GRIP domains or depletion of Golgin-97 by siRNA resulted in an inhibition of E-cadherin targeting (44). This finding supports the hypothesis that specific properties of the TGN are required for sufficient and correct targeting of E-cadherin. However, because depletion of ARL1 failed to modify the E-cadherin distribution in HeLa cells (supplemental Fig. 4), we can exclude that ARL1 acts downstream of ARFRP1 in controlling E-cadherin localization.
The investigation of the functional relevance of ARFRP1 for E-cadherin cell surface localization demonstrated a markedly reduced adhesiveness in Ltk Ϫ Ecad cells in which ARFRP1 was depleted by siRNA. We chose this model because here cell adhesion is mediated by E-cadherin only. The impaired ability of cell-cell adhesion in the absence of ARFRP1 in this system may be ascribed to incorrect sorting of E-cadherin. In contrast, intestinal epithelial cells express other cell adhesion proteins such as LI-cadherin (45) and integrins (46). In fact, the intestinal epithelium of Arfrp1 Ϫ/Ϫ mice does not exhibit visible adhesion defects despite the dislocation of E-cadherin (Fig. 3). However, as demonstrated in Figs. 1 and 3, deletion of Arfrp1 in embryos or the intestinal epithelium did not result in a complete loss of E-cadherin from the cell surface suggesting that the transport of E-cadherin from the TGN to the plasma membrane is not completely disrupted but markedly impaired.
In both in vivo systems, Arfrp1 Ϫ/Ϫ embryos (ED 6.0) and intestine of Arfrp1 vilϪ/Ϫ mice, we detected an altered subcellular distribution of E-cadherin but no reduction of E-cadherin protein levels in comparison with wild-type controls. In con-FIGURE 7. Depletion of ARFRP1 impairs E-cadherin-mediated cell aggregation. Ltk Ϫ Ecad cells were transfected as indicated without siRNA (control), nontarget siRNA (si-control), two different ARFRP1-specific siRNAs (si-a-ARFRP1 and si-b-ARFRP1), and two mutated ARFRP1-specific siRNAs (scrambled-a and scrambled-b). Single cell suspensions were allowed to aggregate for the indicated times. A, microscopic examination of Ltk Ϫ cells stably expressing E-cadherin (Ltk Ϫ Ecad) treated with different ARFRP1-directed or control siRNAs at t ϭ 0 min and at t ϭ 45 min. B, quantification of single cells and cell aggregates represented as the aggregation index. Aggregation index was calculated according to Nagafuchi and Takeichi (41) as A i ϭ (N 0 Ϫ N t )/N 0 , where N 0 is the total particle number at t ϭ 0 min and N t is the particle number after an incubation period of 45 min. Bars represent mean values (ϮS.D.) of four independent experiments with two samples counted at each time point. C, expression of ARFRP1 and E-cadherin in transfected Ltk Ϫ Ecad was analyzed by Western blotting with ␤-actin detected as a loading control. D, expression of ARFRP1 and E-cadherin mRNA was detected by qRT-PCR as described under "Experimental Procedures." OCTOBER 3, 2008 • VOLUME 283 • NUMBER 40 trast, knockdown approaches in HeLa and Ltk Ϫ cells showed decreased E-cadherin protein levels in Arfrp1-depleted cells. Because the RNA levels of E-cadherin were not affected, we suggest that a mistargeting or impaired processing of E-cadherin results in its elevated degradation in some systems.

ARFRP1 Regulates E-cadherin Localization
In addition to ARFRP1, several other GTPases modulate the exocytotic and endocytotic trafficking of E-cadherin (22,47). Rac1 regulates endocytosis and trafficking of E-cadherin to the cell surface during epithelial morphogenesis (48). Wang et al. (49) demonstrated that expression of dominant-negative Rac1 and Cdc42 led to the accumulation of E-cadherin at a distinct post-Golgi step before E-cadherin interacts with p120 ctn . In addition, expression of Rab5 (50) or ARF6 (51) mutants can block endocytosis of E-cadherin.
In summary, our data provide evidence that ARFRP1 plays an important role for the correct cell surface targeting of E-cadherin. This finding suggests that ARFRP1 is thereby involved in the regulation of the specific spatio-temporal expression pattern of E-cadherin during early embryogenesis, which is essential for morphological events such as gastrulation, neurulation, cardiogenesis, and somitogenesis (15,52,53).