A mechanism for differential sorting of the planar cell polarity proteins Frizzled6 and Vangl2 at the trans-Golgi network

In planar cell polarity (PCP), the epithelial cells are polarized along the plane of the cell surface perpendicular to the classical apical–basal axis, a process mediated by several conserved signaling receptors. Two PCP-signaling proteins, VANGL planar cell polarity protein 2 (Vangl2) and Frizzled6 (Fzd6), are located asymmetrically on opposite boundaries of the cell. Examining sorting of these two proteins at the trans-Golgi network (TGN), we demonstrated previously that the GTP-binding protein ADP-ribosylation factor–related protein 1 (Arfrp1) and the clathrin-associated adaptor protein complex 1 (AP-1) are required for Vangl2 transport from the TGN. In contrast, TGN export of Frizzled6 does not depend on Arfrp1 or AP-1. Here, to further investigate the TGN sorting process in mammalian cells, we reconstituted release of Vangl2 and Frizzled6 from the TGN into vesicles in vitro. Immunoblotting of released vesicles indicated that Vangl2 and Frizzled6 exit the TGN in separate compartments. Knockdown analysis revealed that a clathrin adaptor, epsinR, regulates TGN export of Frizzled6 but not of Vangl2. Protein interaction analysis suggested that epsinR forms a stable complex with clathrin and that this complex interacts with a conserved polybasic motif in the Frizzled6 cytosolic domain to package Frizzled6 into transport vesicles. Moreover, we found that Frizzled6–epsinR binding dissociates epsinR from AP-1, which may separate these two cargo adaptors from each other to perform distinct cargo-sorting functions. Our results suggest that Vangl2 and Frizzled6 are packaged into separate vesicles that are regulated by different clathrin adaptors at the TGN, which may contribute to their asymmetric localizations.

Epithelial tissues are polarized along the apical-basal axis such that the apical plasma membrane and the basolateral plasma membrane are enriched with distinct groups of proteins to perform different physiological functions. In addition to the apical-basal polarity, many epithelial tissues are polarized along the tissue plane on an axis perpendicular to the classical apical-basal axis, a phenomenon referred to as planar cell polarity (PCP). 5 The establishment of PCP is regulated by an evolutionarily conserved set of signaling molecules such as Vangl2 and Frizzled6 (Fzd6) (1). Vangl2 is essential for neural tube closure, cardiac development, and orientation of stereociliary bundles in the cochlea (2)(3)(4)(5). Fzd6 controls hair patterning in the skin and regulates nail and claw formation (6,7). Double knockout of Fzd3 and Fzd6 in mice causes severe defects in neural tube closure and planar polarity of inner-ear sensory hair cells (8).
During PCP signaling, Vangl2 and Frizzleds are asymmetrically localized on opposing cell boundaries (1). The asymmetrically localized PCP proteins are proposed to recruit distinct downstream effectors to drive PCP-mediated processes (9). The mechanisms mediating this asymmetric distribution are poorly understood. One view is that the PCP asymmetry is regulated by intercellular interactions between PCP proteins across cellular boundaries (1,10). Additional signaling pathways, including the Wg/Wnt4 pathway and the Dachsous/Fat/ Four-jointed pathway, are proposed to provide long-range directional information for establishing the PCP asymmetry (11,12). Besides these, evidence from live imaging analysis demonstrates that vesicles containing Frizzled and vesicles containing Dishevelled, another PCP protein, are preferentially transported to the distal boundaries in Drosophila wing (13,14), suggesting that polarized intracellular trafficking contributes to establishing PCP asymmetry. Although PCP and apicalbasal polarity are controlled by distinct molecular mechanisms, studies in Drosophila indicate that PCP proteins localize to the cell junctions just apical to the adherence junction (15). Interestingly, PCP proteins are initially uniformly distributed around subapical plasma membranes and their asymmetrical localizations precede all other signs of PCP during wing development in Drosophila (16). Corresponding to this, the alignment of noncentrosomal microtubules are reorganized so that the majority of noncentrosomal microtubules are aligned along the proximal-distal axis with an excess of the plus ends oriented distally prior to the onset of the PCP signaling events (13,14,17). This microtubule dynamics are controlled by Dachsous and Fat implicating that the Dachsous/Fat/Four-jointed pathway may provide long-range directional information to reorganize the microtubule cytoskeleton for polarized delivery of PCP proteins (17).
Newly synthesized integral PCP proteins are delivered along the secretory transport pathway to the plasma membrane where they perform their physiological functions. Packaging of Vangl2 into vesicles at the ER, the first step of the secretory transport pathway, depends on one of the COPII subunits, Sec24B (18). The selectivity of this sorting is illustrated by the behavior of the Vangl2 cytosolic, C-terminal looptail mutant protein, which is unable to be packaged into COPII vesicles and thus cannot be exported out of the ER (18). Correspondingly, the looptail mutation of Vangl2 or a mutation in Sec24B causes severe defects in neural tube closure during mouse embryonic development (18).
The trans-Golgi network (TGN) is another important station in the secretory transport pathway. At the TGN, elaborate cargo sorting machineries are employed to accurately package the correct cargo molecules into the corresponding transport carriers that are targeted to specific destinations (19). Defects in protein sorting at the TGN cause mistargeting of a variety of cargo molecules and induce various physiological defects (19). Understanding how PCP proteins are sorted into transport vesicles at the TGN will not only provide important information to explore how they are delivered to the plasma membrane but also reveal novel insights into how their asymmetric localizations are established.
Several key participants in TGN sorting have been identified, and these include small GTPases of the Arf family and cargo adaptors. Upon GTP binding, the Arf proteins undergo conformational changes in which the N-terminal lipid-binding motif is exposed to bind TGN membranes and the switch domains change the orientation to recruit various cytosolic cargo adaptors (20,21). Once recruited on the TGN membranes, cargo adaptors bind cargo molecules thereby enriching them into transport vesicles (19). Sorting of a subset of soluble cargo proteins at the TGN is regulated by the actin-severing proteins ADF/cofilin, the lumenal Golgi-resident protein (Cab45), and the secretory pathway Ca 2ϩ -ATPase isoform 1 (SPCA1) (22)(23)(24)(25).
We have used mammalian cells as a model system to investigate the molecular mechanisms of PCP sorting. This system allowed us to perform not only microscopy and biochemical assays, but also the in vitro assay combined with biochemical manipulations, and thus to define the sorting signals and binding sites. Our previous analysis indicates that sorting of Vangl2 at the TGN depends on one of the Arf family proteins, Arfrp1, and the clathrin-associated adaptor protein complex-1 (AP-1) (26). Further analysis indicates that Arfrp1-AP-1 interaction opens the cargo-binding pocket of AP-1 to allow AP-1 to directly interact with the tyrosine sorting motif on the Vangl2 cytosolic domain (26). Interestingly, unlike Vangl2, TGN export of Fzd6 is independent of Arfrp1 and AP-1 suggesting that TGN export of Vangl2 and Fzd6 are mediated by different cargo sorting machineries (26). In this study, we sought to utilize mammalian cells to analyze whether Vangl2 and Fzd6 are packaged into different vesicles at the TGN and to investigate the molecular mechanism that mediates TGN export of Fzd6.

Vangl2 and Fzd6 are packaged into separate vesicles in an assay that reconstitutes vesicle budding from the TGN in vitro
We previously demonstrated that TGN export of Vangl2 and Fzd6 depends on distinct cargo sorting machineries. One possible consequence of this behavior is that Vangl2 and Fzd6 may be sorted into separate vesicles. To test this, we reconstituted release of Vangl2 and Fzd6 into vesicles from the TGN in vitro and then tested whether the two proteins were packaged together or separately. A TGN vesicle-budding reaction using digitonin-permeablized cells has been reported (27). We performed the TGN vesicle-budding assay using COS7 cells transfected with HA-Fzd6 or HA-Vangl2 (Fig. 1A). After a 20°C incubation, HA-Fzd6 showed a strong TGN-accumulated pattern with no detectable surface-localized pattern in over 80% of the expressing cells (Fig. S1, A-C, and quantification in G). When cells were shifted to 32°C, the percentage of cells showing TGN-accumulated Fzd6 dropped to below 5% and the majority of cells showed a clear surface-localized pattern of Fzd6 (Fig. S1, D-F, and quantification in G). To reconstitute packaging of Fzd6 into vesicles at the TGN, COS7 cells were incubated at 20°C to accumulate newly synthesized HA-Fzd6 at the TGN, then permeablized by digitonin, and incubated at 30°C with rat liver cytosol, GTP, and an ATP regeneration system in the presence of a GTPase-defective form for Sar1A, Sar1A(H79G), to inhibit vesicle budding from the ER. After incubation, the released vesicles were separated from the donor membranes by centrifugation and analyzed by immunoblotting with antibodies against TGN46, Sec22, and the HA tag. TGN46 is an integral membrane protein that is transported from the TGN to the cell surface (28) and Sec22 is a soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) that directly binds COPII and regulates ER-to-Golgi trafficking (29). Thus TGN46 and Sec22 can be used to monitor TGN export and ER export respectively.
Using this assay, we reproducibly detected release of TGN46, Sec22, and HA-Fzd6 into vesicles when permeabilized cells were incubated in the absence of Sar1A(H79G) (Fig. 1B, lane 5). Addition of Sar1A(H79G) to the reaction inhibited the vesicular release of Sec22 but not TGN46 and HA-Fzd6, which were packaged dependent on the concentration of cytosol (Fig. 1, B, lanes 2-4, and C). Export of Fzd6 depended on GTP and ATP ( Fig. 1, D, lane 1, and E). A nonhydrolyzable analog of GTP, GMPPNP, significantly reduced packaging of Fzd6 into vesicles TGN sorting of a planar cell polarity protein Frizzled6 ( Fig. 1, D, lane 4, and E). HA-Fzd6 (WT) in the vesicle fraction in the presence of Sar1A(H79G) was resistant to endoglycosidase H (endo H) digestion (Fig. 1F, lanes 1 and 2) but sensitive to peptide:N-glycosidase F (PNG-F) digestion (Fig. 1F, lanes 3  and 4, asterisk), whereas TGN46, which is highly O-glycosylated, was resistant to both endo H digestion and PNG-F diges-TGN sorting of a planar cell polarity protein Frizzled6 tion (Fig. 1F), suggesting that HA-Fzd6 in the vesicle fraction contains glycans consistent with modifications in the Golgi. Based on these data, we propose that the discharge of WT Fzd6 in vesicles originates primarily at the TGN.
Using the same assay, we detected cytosol-dependent release of WT HA-Vangl2 ( Fig. 1G) but not the tyrosine mutant (HA-Vangl2 Y279A,Y280A ) into the vesicle fraction ( Fig. 1, K-L). Export of Vangl2 and TGN46 into vesicles was significantly reduced by GMPPNP (Fig. 1G, lane 6). The GTPase-defective form for Arfrp1, Arfrp1(Q79L), inhibited Vangl2 release into vesicles in a concentration-dependent manner ( Fig. 1, G, lanes 4 and 5, and H), whereas packaging of TGN46 into vesicles was only slightly inhibited by Arfrp1(Q79L) (Fig. 1G) indicating that the sorting of Vangl2 in vesicles originates primarily at the TGN. We then analyzed the effects of the Arfrp1(Q79L) mutant on Fzd6 release into vesicles. The result indicates that, at the lower concentration, Arfrp1(Q79L) did not affect Fzd6 release into vesicles (Fig. 1, I, lane 3, and J). At the higher concentra-tion, Arfrp1(Q79L) partially inhibited release of Fzd6 in the vesicle fraction ( Fig. 1, I, lane 4, and J). This analysis suggests that Arfrp1(Q79L) has a stronger inhibiting effect on vesicular release of Vangl2 than Fzd6. We then analyzed the effects of Arfrp1(Q79L) on release of Fzd6 and Vangl2 into vesicles in cells co-expressing these two cargo proteins. The result indicates that the Arfrp1(Q79L) mutant inhibited Vangl2 release into vesicles (Fig. 1M). In contrast, Arfrp1(Q79L) had a weaker inhibiting effect on Fzd6 release into vesicles than Vangl2 release into vesicles (Fig. 1M).
Vangl2, Fzd6, and TGN46 were all packaged into buoyant membranes (Fig. 1N). We used negative stain EM to visualize the morphology of the buoyant membrane structures produced in the in vitro budding reaction. The digitonin-treated COS7 cells were visualized by negative stain EM ( Fig. 2A). We detected numerous donut-shaped structures in the buoyant membrane fraction (Fig. 2, D-DЈ, and E). The average diameter and standard deviation of all of the structures seen on the grids

TGN sorting of a planar cell polarity protein Frizzled6
is 47.15 Ϯ 22.29 nm. The donut shape probably represented vesicles collapsed by the negative stain procedure. When we performed the vesicle-budding reaction in the absence of GTP and an ATP regeneration system or in the presence of the nonhydrolysable GTP analog, GMPPNP, the number of donutshaped structures was significantly reduced (Fig. 2, B and E, C and E). These analyses indicate that the nonsedimenting membranes in the budding reaction are transport vesicles not fragments of the Golgi or reformed ministacks.
To test whether Vangl2 and Fzd6 were packaged into vesicles together or separately, we performed the TGN vesicle-budding assay using COS7 cells co-transfected with FLAG-Vangl2 and HA-Fzd6 in which ϳ80% of the transfected cells co-expressed both constructs. We then immunoisolated the vesicles containing FLAG-Vangl2 using an antibody against the FLAG tag. The immunoisolation approach robustly captured Vangl2-enriched vesicles (Fig. 3, A and B). Strikingly, the immunoisolated Vangl2-enriched vesicles did not contain Fzd6 or TGN46 (Fig.  3, A and C). Likewise, TGN46 was not detected in vesicles isolated using a Fzd6 construct containing a FLAG tag in the C-terminal cytosolic domain (Fig. 3, D-F). As a control for our vesicle isolation experiment, we performed the vesicle-formation assay using cells co-transfected with FLAG-Vangl2 and HA-Vangl2 and then immunoisolated the vesicles containing FLAG-Vangl2. We found a fraction of the vesicles enriched with HA-Vangl2 was co-isolated with FLAG-Vangl2 (Fig. 3, G-I), suggesting that the separation of Vangl2 and Fzd6 was not due to disruption of vesicles. The fraction of vesicles enriched with HA-Vangl2 that were not co-isolated with FLAG-Vangl2 may be derived from cells expressing HA-Vangl2 but not FLAG-Vangl2. We analyzed the morphology of the immunoisolated vesicles by negative stain EM. The structures in the immunoisolated vesicle fraction containing Fzd6-FLAG or FLAG-Vangl2 showed donut shapes (Fig. 3, J-K), which are similar to the morphology we observed in the buoyant membrane fraction from the vesicle-formation assay (Fig. 2D). We did not detect these donut-shaped structures in the control experimental group in which the budding assay was performed without cytosol. There were no clear vesicle coat structures surrounding the immunoisolated vesicles possibly due to the reason that vesicle coat proteins were released from the vesicles after a 2-day period to immunoisolate the vesicles. Another possibility is that the immunoisolation works only for uncoated vesicles as the coat may block the antibody-antigen interaction. These results indicate that Vangl2, Fzd6, and TGN46 are packaged into separate vesicles.

Clathrin and epsinR regulate sorting of Fzd6 at the TGN
An HA-Fzd6 mutant construct lacking the C-terminal cytosolic domain (HA-Fzd6-⌬cyto) accumulated at the juxtanuclear area and colocalized with a TGN marker (Golgin97) (Fig. 4, A-F), suggesting that this domain of Fzd6 is important for export of Fzd6 from the TGN. HA-Fzd6-⌬cyto mutant was incorporated into vesicles in a budding reaction conducted in the absence of Sar1A(H79G) (Fig. 4G, lane 2). However, in the presence of Sar1A(H79G), packaging of the HA-Fzd6-⌬cyto mutant into vesicles was nearly completely blocked ( Fig. 4, G, lane 3, and H), suggesting that HA-Fzd6-⌬cyto folded well enough to be packaged into COPII vesicles at the ER but lacked information or a structure necessary for sorting into vesicles packaged at the Golgi membrane. Moreover, the N-terminal one-third region of the Fzd6 cytosolic domain was sufficient to restore Sar1A(H79G)-insensitive packaging of Fzd6 into vesicles ( Fig. 4

, I, lane 3, and J).
To identify the molecular machinery that regulates sorting of Fzd6 at the TGN, we analyzed whether clathrin is an important regulator that interacts with Fzd6. We detected a weak interaction between clathrin heavy chain (CHC) and Fzd6 through an immunoprecipitation experiment (Fig. S2A). The weak interaction was possibly due to the reason that clathrin interacts with Fzd6 indirectly through clathrin adaptors. As a second approach to test the interaction between Fzd6 and clathrin, we performed GST-pulldown experiments by incubating purified GST-tagged Fzd6 cytosolic domain fragments with cytosol prepared from COS7 cells. We found that clathrin from cell lysates was pulled down by the full-length Fzd6 C-terminal cytosolic domain (aa 497-709), Fzd6 C-terminal cytosolic domain depleted of the conserved KTXXXW motif that binds the PDZ domain of Dishevelled (aa 504 -709) and the first one-third region of the Fzd6 C-terminal cytosolic domain (aa 497-567) (Fig. S2B, lanes 1-3). In contrast, the last two-thirds region of the Fzd6 C-terminal cytosolic domain (aa 568 -709) did not pulldown clathrin (Fig. S2B, lane 4). We then performed siRNA knockdown analysis to test whether CHC mediates TGN export of Fzd6. We used HeLa cells for our knockdown analysis because most of the commercially available pre-designed siRNAs are against human proteins. The expression of CHC was significantly reduced after siRNA treatment (Fig. 5G). Around 85% of siRNA-treated cells showed strong accumulations of HA-Fzd6 at the juxtanuclear area, colocalized with the TGN marker, Golgin97, whereas only 13% of mock-treated cells showed TGN-accumulated Fzd6 (Fig. 5, A-F and H). This suggests that CHC is important for TGN export of Fzd6. Knockdown of CHC also caused accumulations of Vangl2 at the TGN indicating that both Vangl2 and Fzd6 depend on CHC for traffic from the TGN (Fig. 5H). In contrast, knockdown of CHC did not affect TGN export of an apical cargo, p75-GFP ( Fig. 5H) (30).
Clathrin interacts with cargo molecules through the clathrin-adaptor proteins. We analyzed the localizations of Fzd6 in HeLa cells treated with siRNAs against some of the TGN-and endosome-localized clathrin adaptor or adaptor subunits including the ␥ subunit of AP-1 (␥1), the ␦ subunit of AP-3 (␦3), or epsinR. Cells treated with each of these siRNAs did not show detectable defects in HA-Fzd6 localization at steady state, possibly due to the presence of functional redundancy among the clathrin adaptors. To test whether knockdown of a specific clathrin adaptor can cause a kinetic delay of TGN export of Fzd6, we incubated HeLa cells at 20°C in the presence of cycloheximide to synchronize a pool of newly-synthesized HA-Fzd6 in the TGN. As the HA tag is localized on the extracellular domain of Fzd6, we performed a surface-labeling experiment to label the surface-localized HA-Fzd6. After a 20°C incubation, HA-Fzd6 was accumulated at the juxtanuclear area, colocalized with TGN46 and the majority of cells showed no detectable surface-localized Fzd6 (Fig. 6, A-H). When cells were shifted to 32°C, Fzd6 in most control siRNA-treated cells showed a TGN sorting of a planar cell polarity protein Frizzled6 detectable surface-localized pattern but also a certain degree of TGN-and ER-localized pattern (Fig. 6, I-L). The remaining Fzd6 at the TGN and the ER area after the temperature shift condition may be caused by overexpression of Fzd6. Another possible reason is that cycloheximide may not completely inhibit the synthesis of the exogenously overexpressed Fzd6. It is also possible that some of the TGN-accumulated Fzd6 was retrieved back to the ER after temperature shift. We then quantified the percentage of cells showing a detectable surface pattern of Fzd6 at the same exposure time. After the temperature was released to 32°C, around 70% of cells treated with control siRNA showed a detectable surface pattern (Fig. 6V). Through this method, we found that cells treated with siRNA against one of the TGN-localized cargo adaptor, epsinR, showed a significant reduction of the percentage of cells showing surface-localized Fzd6 (Fig. 6, M-P, T, and V), whereas knockdown of ␥1 or ␦3 did not affect surface delivery of Fzd6 (Fig. S3, A-C). The majority of cells treated with siRNA against ␥1 showed strong accumulations of HA-Vangl2 at the juxtanuclear area at the steady state (Fig. S3, D-I), consistent with our previous analysis (26). In contrast, knockdown of epsinR did not cause a defect in the TGN export of HA-Vangl2 (Fig. 6W). These analyses indicate that TGN export of Vangl2 and Fzd6 depends on distinct cargo sorting machineries.
We tested whether the phenotype of epsinR knockdown can be rescued by expressing an siRNA-resistant epsinR construct by mutating the siRNA-targeting site, and fused this construct to a FLAG tag and an IgG-binding ZZ domain (epsinR RE -FLAG). Western blot analysis indicates that expression of the mutant form but not the endogenous form of epsinR was resistant to siRNA treatment (Fig. 6U). Expression of epsinR RE -FLAG restored the percentage of cells with detectable surface HA-Fzd6 to the control level (Fig. 6, Q-S and V), suggesting that expressing epsinR rescues the knockdown phenotype. We analyzed the localization of Fzd6 in cells that were transfected with another siRNA that targets a different region in epsinR. HeLa cells transfected with this siRNA also showed a reduction of the expression level of epsinR (Fig. S4Q) and a significant reduction of the percentage of cells showing surface-localized Fzd6 (Fig. S4, I-P and R).
To test whether epsinR interacts with Fzd6, we performed GST-pulldown experiments using GST-tagged Fzd6 cytosolic domain fragments (aa 497-467 and 568 -709) and lysates from COS7 cells expressing epsinR fused to FLAG and an IgG-binding ZZ domain (used in combination to increase the specificity for affinity purification) at its C terminus (hereafter referred to as epsinR-FLAG). A GST fusion containing the N-terminal one-third region of the Fzd6 cytosolic domain bound epsinR-FLAG in COS7 cell lysates, whereas the GST fusion containing the other two-thirds of the Fzd6 cytosolic domain did not (Fig.   7A). Epsins are composed of an ENTH domain at the N terminus followed by a long, unfolded domain. The unfolded domain but not the ENTH domain of epsinR interacted with the GSTtagged Fzd6 cytosolic domain (Fig. 7B).
We then purified epsinR by immunoprecipitation from cell lysates of COS7 cells expressing epsinR-FLAG. FLAG-tagged epsinR co-purified with CHC (Fig. 7C, lanes 1 and 2). The immunoprecipitated epsinR-CHC complex bound the N-terminal one-third region of the Fzd6 cytosolic domain but not the other part of Fzd6 cytosolic domain (Fig. 7D).
The results from the GST-pulldown assay are consistent with our vesicular release assay (Fig. 4, I and J) suggesting that the N-terminal one-third region of the Fzd6 cytosolic domain contains a sorting motif that is important to bind the epsinR-CHC complex for incorporation of Fzd6 into vesicles. Sequence alignment of this region in Fzd6 and another PCP protein, Friz-zled3, across species revealed several conserved motifs including a KTXXXW motif (Fig. 8A, green box) and a polybasic motif (Fig. 8A, red box). We found that deleting the polybasic motif in Fzd6 resulted in the accumulation of mutant Fzd6 at the juxtanuclear area, colocalized with Golgin97 in the majority of the cells (Fig. 8, B-G). In contrast, deleting the KTXXXW motif or other conserved regions in the Fzd6 cytosolic domain had no effect on Fzd6 localization.
To test whether the polybasic motif was critical for packaging of Fzd6 into transport vesicles at the TGN, we performed the TGN vesicle budding reaction using HA-tagged Fzd6 WT or the Fzd6 polybasic mutant construct. The efficiency of packaging of mutant Fzd6 into transport vesicles was significantly reduced (Fig. 8, H and I). To test whether the polybasic motif was important to bind epsinR, we compared the efficiency of GST pulldown using wild type and mutant versions of HAtagged Fzd6. Fzd6 lacking the polybasic region displayed reduced binding to epsinR-FLAG in COS7 cell lysates (Fig. 8, J and K) and reduced binding to the immunoprecipitated epsinR-CHC complex (Fig. 8, L and M). These analyses suggest that epsinR forms a stable complex with clathrin and this complex binds to the polybasic motif on the Fzd6 cytosolic domain to mediate sorting of Fzd6 into transport vesicles at the TGN.

Binding of Fzd6 to epsinR causes disassociation of epsinR from AP-1
EpsinR and AP-1 are both clathrin adaptors that interact with each other. To test whether binding of AP-1 to epsinR interferes with the interaction between epsinR and Fzd6, we mutated the AP-1-binding motif on epsinR. As previously reported (31), mutating the two key amino acids on epsinR, Asp-349 and Asp-371, blocked the interaction between epsinR and AP-1 (Fig. 9A). We then performed a GST-pulldown experiment using purified GST-tagged Fzd6 cytosolic fragment (aa . In each replicate of the experiment, the intensity of the protein of interest in each reaction condition was normalized to the sum of the intensities of that protein from all reaction conditions performed in that replicate. J-K, the TGN vesicle-formation assay was performed using COS7 cells expressing FLAG-Vangl2 or FLAG-Fzd6. The medium speed supernatant fractions were incubated with anti-FLAG M2 affinity gel. The bound fraction was eluted using FLAG peptides and the eluted fraction was analyzed by negative stain EM. Scale bar, 100 nm.

TGN sorting of a planar cell polarity protein Frizzled6
497-567) and cell lysates from HEK293T cells transfected with plasmids encoding a WT FLAG-tagged epsinR C terminus (epsinR ⌬ENTH-FLAG) or transfected with plasmids encoding an AP-1-binding deficient mutant version of the FLAG-tagged epsinR C terminus (epsinR ⌬ENTH D349R,D371R -FLAG). Interestingly, mutating the AP-1-binding sites on epsinR enhanced the interaction between epsinR and Fzd6 (Fig. 9B), suggesting that binding of AP-1 to epsinR interferes with the interaction between Fzd6 and epsinR. As both Fzd6 and AP-1 bind to the C-terminal unfolded region of epsinR, we hypothesize that binding of Fzd6 to epsinR causes dissociations of AP-1 from epsinR. To test this, we immunoprecipiated the FLAG-tagged epsinR C terminus and incubated the immunoprecipiated proteins (epsinR ⌬ENTH IP) with purified GST-Fzd6 cytosolic domain fragments (aa 497-567 or 568 -709). As expected, ␥1-adaptin co-immunoprecipitated with epsinR ⌬ENTH-FLAG (Fig. 9C). The GST-Fzd6 cytosolic domain fragment (aa 497-567) but not the GST-Fzd6 cytosolic domain fragment (aa 568 -709) bound to epsinR ⌬ENTH IP (Fig. 9C, lanes 2 and 3). Remarkably, we found that the affinity between epsinR and ␥1-adaptin was significantly reduced in the presence of purified a GST-Fzd6 cytosolic domain fragment (aa 497-567) (Fig. 9C, lane 2). In contrast, a purified GST-Fzd6 cytosolic fragment (aa 568 -709) did not interfere with the interaction between epsinR and ␥1-adaptin (Fig. 9C,  lane 3). We then performed the experiment using three different concentrations of Fzd6 (497-567), we found that coimmunoprecipitation of AP-1 was progressively reduced by increasing concentrations of Fzd6 (497-567) (Fig. 9, D, lanes 1-3, and E). In contrast, the Fzd6 cytosolic fragment depleted of the KR motif did not cause a release of AP-1 from epsinR ( Fig. 9, D, lane 4, and E). These results indicate that binding of Fzd6 to epsinR causes dissociation of epsinR from AP-1 and this may allow these two cargo adaptors

TGN sorting of a planar cell polarity protein Frizzled6
to be separated from each other to perform distinct cargo sorting functions.

Discussion
The TGN is an essential transport hub in the secretory transport pathway. Although significant progress has been achieved in understanding biosynthetic cargo protein sorting at the TGN, several important aspects remain to be further investigated. How specific and robust is the cargo sorting process that takes place at the TGN? Are the biochemical approaches that proved crucial to understanding other coat protein-mediated vesicle budding events applicable to study sorting at the TGN? Moreover, the molecular mechanisms that regulate TGN sorting of signaling receptors that mediate important physiological and pathological processes remain to be defined.
Using a cell-free reconstitution approach that captures packaging of cargo proteins into vesicles from the TGN, we showed that the three cargo proteins, Vangl2, Fzd6, and TGN46, are packaged into separate vesicles. These analyses support a view of differential sorting of Vangl2 and Fzd6 at the TGN and suggest that the packaging of cargo proteins into transport vesicles at the TGN is highly specific with perhaps only a minor contribution by nonspecific packaging such as bulk flow. We probed the molecular mechanism that regulates packaging of Fzd6 into transport vesicles and found that epsinR forms a stable complex with clathrin that interacts with a polybasic motif on the Fzd6cytosolic domain to facilitate its sorting into transport vesicles.
The cargo sorting process is a critical step to ensure the fidelity of protein transport along the complicated TGN exit routes. Defects in TGN sorting will cause mistargeting of proteins thereby inducing physiological defects (19). Consistent with our finding, knockdown of AP-1 subunits in the Drosophila wing disrupts the polarized pattern of PCP proteins and induces defects in PCP signaling processes such as multiple cellular hair defects (32). The Drosophila homolog of epsinR, lqfR, is important for various developmental processes including wing hair patterning suggesting its involvement in planar cell polarity (33,34). However, the majority of the defects in lqfR mutants are rescued by exon 6 of the lqfRa gene encoding a domain that is homologous to Tel2, a protein that regulates telomere maintenance and DNA repair (35). In zebrafish, an insertion within an intron of the gene encoding epsinR causes defects in epidermal homeostasis (36). This phenotype is similar to that observed in zebrafish mutants bearing mutations in lethal giant larvae 2 (ldl2), an important regulator of cell polarity (36,37). EpsinR and Ldl2 function synergistically in regulating epidermal homeostasis (36). The ENTH domains of two epsins in yeast, Ent1 and Ent2, bind guanine nucleotide triphosphatase-activating proteins for Cdc42 and disrupting this interaction leads to cell polarity defects (38). In mammalian cells, epsinR is important for retrograde transport of Shiga toxin, TGN46, and mannose 6-phosphate receptor from the early endosomes to the TGN (39).
EpsinR contains an N-terminal ENTH domain and a C-terminal unfolded region. The epsinR ENTH domain interacts directly with various SNAREs (40 -44). We mapped the Fzd6binding site on epsinR to the C-terminal unfolded region but not the ENTH domain, suggesting that epsinR could possibly interact simultaneously with both SNAREs and Fzd6 to package them together into transport vesicles. The C-terminal unfolded region of epsinR also binds clathrin (19). Here we showed that epsinR forms a stable complex with clathrin. By analogy to other vesicle coat proteins, it seems likely that clathrin plays an indirect role through epsinR in the sorting of Fzd6 into transport vesicles.
Currently, several sorting motifs have been identified that are recognized by clathrin adaptors. These motifs include tyrosinebased motifs that are recognized by adaptor complexes, dileucine-based motifs that are recognized by adaptor complexes and GGAs, and NPXY motifs that are recognized by clathrin adaptor proteins, Dab2 and ARH (45). The C-terminal unfolded region of epsins 1-3 contains ubiquitin-interacting motifs that are implicated to interact with ubiquitinated cargo proteins (46 -48). A clathrin adaptor protein, Hrb, recognizes the longin domain of the SNARE Vamp7 (49). Another group of clathrin adaptors ␤-arrestins recognize the phosphorylated C-terminal tail and the cytosolic surface of the transmembrane bundle of the G protein-coupled receptors (50). We found that TGN export of Fzd6 depends on a conserved polybasic motif on its C-terminal cytosolic domain. The polybasic motif also regulates TGN export of the reptilian reovirus p14 protein (51,52). This motif mediates the interaction between activated Rab11 and p14 as detected in cell lysates (53). Thus the polybasic motif may interact with multiple binding partners. Another PCP protein, Frizzled3, also contains a polybasic motif in its C-terminal cytosolic domain indicating that the epsinR-clathrin complex may also regulate sorting of Frizzled3 at the TGN.
The C-terminal unfolded region of epsinR also directly interacts with the appendage domain of AP-1 (45). Evidence suggests that these two clathrin adaptors may be functionally related. In yeast, the epsin-related proteins, Ent3 and Ent5, function redundantly with AP-1 to regulate intracellular retention of Chs3p (54). In mammalian cells, epsinR and AP-1 are shown to depend on each other for their maximum incorporation into clathrin-coated vesicles (42). Based on proteomic analysis of purified clathrin-coated vesicles (CCVs) in control cells and in cells where epsinR is depleted by sequestering in a "knocksideways" approach, the Robinson laboratory showed that: 1) epsinR is a major, near-stoichiometric component of

TGN sorting of a planar cell polarity protein Frizzled6
CCVs and 2) that epsinR knocksideways blocks the production of the entire population of intracellular CCVs (55). This result would appear to conflict with our observation of a cargo-selec-tive effect of epsinR on the traffic of Fzd6 versus Vangl2. The knocksideways approach, although depleting a protein more quickly than the knockdown approach, has the disadvantage of  A and B, purified GST-tagged Fzd6 cytosolic domain fragments were incubated with lysates from COS7 cells transfected with full-length FLAG-tagged epsinR or FLAG-tagged epsinR ENTH domain or FLAG-tagged epsinR C-terminal unfolded region. After incubation, the bound fraction was analyzed by immunoblot. C, FLAG-tagged epsinR from cell lysates of transiently transfected COS7 cells were immunoprecipitated and the proteins were eluted using FLAG peptide (lanes 1 and 2), proteins remained on beads after elution (lane 3) and proteins in the flow-through fraction (lane 4) were resolved by SDS-PAGE and analyzed by silver staining. D, GST-tagged Fzd6 cytosolic domain fragments immobilized to GSH beads were incubated with the immunoprecipitated epsinR-CHC complex. After incubation, the bound fraction was analyzed by immunoblot.

TGN sorting of a planar cell polarity protein Frizzled6
sequestering all other proteins to which the target is tightly and quantitatively bound. Thus the knocksideways approach may cause a more general defect in the traffic of clathrin-mediated cargo, including processes that are not epsinR or AP-1-dependent. This is reflected in their previous analysis (42,56) as mentioned under "Discussion": "However, it is important to remember that knocking down epsinR or AP-1 produces a much weaker phenotype than knocking them sideways" (55).
In the current study, we did not detect any defects in Fzd6 localization in cells depleted of AP-1 or defects in Vangl2 localization in cells depleted of epsinR. The in vitro vesicle-formation assay suggests that Vangl2 and Fzd6 are packaged in separate vesicles and their vesicular release depends on the AP-1-binding motif on Vangl2 and the epsinR-binding motif on Fzd6, respectively. This evidence suggests that epsinR and AP-1 can function independently of each other to regulate sorting of Vangl2 and Fzd6 at the TGN. Moreover, our results indicate that binding of Fzd6 to epsinR reduces the affinity between epsinR and AP-1. Thus the presence of a specific cargo protein, Fzd6, can cause dissociation of epsinR from AP-1 to allow them to perform distinct cargo sorting functions.
We suggest that an initial step in the establishment of epithelial cell-lateral polarity is established by the sorting of membrane-signaling proteins such as Vangl2 and Fzd6 by different Golgi vesicle-budding coats and that the final disposition of these receptors on opposing lateral surfaces of an epithelial cell are reinforced by association with PCP proteins at the proximal and distal plasma membrane domains. In the Drosophila wing, Frizzled and Dishevelled are preferentially delivered to the distal site of cell boundaries (13,14), consistent with polarized trafficking of PCP proteins. Noncentrosomal microtubules are aligned along the proximal-distal axis with an excess of the plus ends oriented distally prior to the onset of the PCP signaling events in the Drosophila wing (13,14,17). This suggests a mechanism whereby distinct vesicles may be trafficked in opposing orientation along polarized microtubules. An analysis of the protein composition of the Vangl2-enriched and Fzd6enriched vesicles could provide important clues to the means by which these vesicles are directed to opposing cell-surface domains.

Constructs, reagents, cell culture, immunofluorescence, and transfection
Small interference siRNAs against clathrin heavy chain or against epsinR and other clathrin adaptor subunits were pur-chased from Qiagen (Valencia, CA). The target sequence against clathrin heavy chain was TAATCCAATTCGAAGAC-CAAT. The target sequences against epsinR were CAGGCTT-CGTGAAGAGCGAAA and GGTCACGAATGTTAAAAGA. The target sequence against ␥1-adaptin was TAGCACAGGT-TGCCACTAA and the target sequence against ␦3-adaptin was CGCTGAAAATTCCTATGTT (26). HeLa cells and HEK293T cell lines were kindly provided by the University of California-Berkeley Cell Culture Facility and were confirmed by short tandem repeat profiling. COS-7 cells were obtained from ATCC (catalog number ATCC CRL-1651, RRID: CVCL_0224). All cell lines were tested negative for mycoplasma contamination. HeLa, HEK293T, and COS7 cells were maintained in GIBCO Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (FBS), 10 milliunits/ml of penicillin, and 0.1 mg/ml of streptomycin. Transfection of siRNA or DNA constructs into HeLa cells or COS7 cells and immunofluorescence were performed as described (26). A temperature shift experiment was performed as described (26). The cycloheximide was used at the concentration of 100 g/ml. To label surface-localized Fzd6 containing an HA tag in its extracellular N-terminal cytosolic domain, HeLa cells expressing HA-Fzd6 were incubated with the mouse anti-HA antibody (1:200) in PBS containing FBS (250 l in 10 ml) for 40 min on ice. After several washes with PBS, cells were fixed for 15 min with 4% paraformaldehyde in PBS and then a normal immunofluorescence procedure was performed. Images were acquired with a Zeiss Axioobserver Z1 microscope system or Leica STED TCS SP5 II Confocal Laser Scanning Microscope.
The plasmid encoding p75-GFP was kindly provided by the Rodriguez-Boulan lab (Weill Cornell Medical College, New

TGN sorting of a planar cell polarity protein Frizzled6
York). HA-tagged mouse Vangl2 was cloned in pCS2 (18). The plasmid encoding HA-Frizzled6 was generated by cloning fulllength mouse Frizzled6 amplified from Frizzled6-Myc into pcDNA4/TO and an HA tag was inserted by PCR between Phe-22 and Thr-23. The plasmid encoding HA-Frizzled6-FLAG was generated by inserting a FLAG tag between Ser-677 and Pro-678 in HA-Frizzled6 in pcDNA4/TO. The plasmids encoding GST fusions of the Frizzled6 cytosolic fragment (aa 497-567) and the Frizzle6 cytosolic fragment (aa 568 -709) were generated by cloning fragments amplified from HA-Friz-zled6 into pGEX-2T. The plasmids encoding full-length epsinR, epsinR-ENTH (aa 1-166), and epsinR ⌬ENTH (aa 167-625) fused to FLAG and an IgG-binding ZZ domain were generated by cloning fragments amplified from human epsinR cDNA into the modified pcDNA4/TO vector encoding 3ϫ FLAG tag and ZZ domain (kindly provided by the Zhong lab, University of Texas Southwestern, Dallas, TX).

Immunoprecipitation, protein purification, and binding assay
Immunoprecipitation of FLAG-tagged epsinR was performed by incubating 50 l of compact anti-FLAG M2-agarose affinity beads with 10 ml of 0.5 mg/ml of cell lysates from COS7 cells transfected with epsinR-FLAG in HKT buffer (100 mM KCl, 20 mM Hepes, pH 7.2, 0.5% Triton X-100) with mixing at 4°C overnight. After incubation, the beads were washed 4 times with 10 ml of HK buffer (100 mM KCl, 20 mM Hepes, pH 7.2) and the bound material was eluted with 1 ml of 0.2 mg/ml of FLAG peptides in HK buffer (100 mM KCl, 20 mM Hepes, pH 7.2) at 4°C for 4 h. The eluted fraction was then dialyzed against HK buffer.
Purification of GST-tagged protein was performed as described (26). Binding assays were carried out with 4 l of compact GSH beads bearing around 1 g of GST-tagged Fzd6 C-terminal cytosolic domain fragments. The beads were incubated with 0.4 g of purified epsinR-CHC complex in 400 l of binding buffer (100 mM KCl, 20 mM Hepes, pH 7.2, 5 mM MgCl 2 , 0.5% Triton X-100) containing 0.1 mg/ml of BSA, or incubated with 150 l of 0.2 mg/ml of cell lysates from COS7 cells transfected with FLAG-tagged epsinR WT or truncated constructs in binding buffer at 4°C for 90 min. After incubation, the beads were washed four times with 500 l of binding buffer and the bound material was analyzed by immunoblot.

Vesicular release assay from the TGN
The TGN vesicular release assay was performed using COS7 cells transfected with HA-Fzd6 or HA-Vangl2. The approach was a modified version of the vesicle-budding reaction designed previously to detect COPII vesicle formation at the ER (18). Day 1 after transfection, cells grown in one 10-cm dish at around 70% confluence were incubated at 20°C for 2 h in Opti-MEM I reduced-serum medium supplemented with 10% fetal bovine serum to accumulate newly-synthesized Fzd6 or Vangl2 at the TGN. Cells were then permeablized in 3 ml of ice-cold KOAc buffer (110 mM potassium acetate, 20 mM Hepes, pH 7.2, 2 mM magnesium acetate) containing 40 g/ml of digitonin on ice for 5 min, and the semi-intact cells were then sedimented by centrifugation at 300 ϫ g for 3 min at 4°C. The cell pellets were resuspended in 1 ml of high salt KOAc buffer (1 M potassium acetate, 20 mM Hepes, pH 7.2, 2 mM magnesium acetate) and incubated on ice for 5 min. After incubation, permeabilized cells were washed twice with 1 ml of KOAc buffer and resuspended in 100 l of KOAc buffer. The budding assay was performed by incubating semi-intact cells (around 0.02 OD/reaction) with 2 mg/ml of rat liver cytosol in a 100-l reaction mixture containing 200 M GTP and an ATP regeneration system (40 mM creatine phosphate, 0.2 mg/ml of creatine phosphokinase, and 1 mM ATP) in the presence or absence of a 0.5 g of Sar1A(H79G) mutant protein. After incubation at 30°C for 1 h, the reaction mixture was centrifuged at 14,000 ϫ g to remove cell debris and large membranes. The medium speed supernatant was then centrifuged at 100,000 ϫ g to sediment small vesicles. The pellet fraction was then analyzed by immunoblot. For density gradient flotation assay, the high speed pellet was resuspended in 100 l of 35% OptiPrep and overlaid with 700 l of 30% OptiPrep and 30 l of KOAc buffer. The samples were centrifuged at 55,000 rpm in a TLS55 rotor in a Beckman ultracentrifuge for 2 h at 4°C. After centrifugation, fractions were collected from the top to the bottom of the tube and aliquots were analyzed by SDS-PAGE and immunoblot.
To immunoisolate vesicles enriched with Vangl2 or Fzd6, the volume of the reaction mixture was scaled up to 750 l. The medium speed supernatant was then incubated with 30 l of compact anti-FLAG M2-agarose affinity beads at 4°C overnight. After incubation, the beads were washed 4 times with 1 ml of KOAc buffer and eluted with 50 l of KOAc buffer containing 0.1% Triton X-100 and 0.2 mg/ml of FLAG peptides. The eluted fraction was then analyzed by immunoblot. For immunoisolation of vesicles for the negative stain analysis, the elution buffer did not contain Triton X-100. Figure 9. Binding of Fzd6 to epsinR causes disassociation of epsinR from AP-1. A, HEK293T cells were transfected with plasmids encoding a WT FLAGtagged epsinR C terminus (epsinR ⌬ENTH-FLAG) or transfected with plasmids encoding an AP-1-binding deficient mutant version of FLAG-tagged epsinR C terminus (epsinR ⌬ENTH D349R,D371R ). Day 1 after transfection, FLAG-tagged epsinR constructs were immunoprecipitated and eluted by FLAG peptides and analyzed by immunoblot using the indicated antibodies. B, a purified GST-tagged Fzd6 cytosolic domain fragment (GST-Fzd6 497-567) was incubated with lysates from HEK293T cells transfected with plasmids encoding epsinR ⌬ENTH-FLAG or epsinR ⌬ENTH D349R,D371R . After incubation the bound proteins were analyzed by immunoblot. C, cell lysates from HEK293T cells transfected with plasmids encoding epsinR ⌬ENTH-FLAG without the IgG-binding ZZ domain were incubated with M2-agarose conjugated with anti-FLAG antibodies. Subsequently, the immunoprecipitated proteins were incubated with purified GST-tagged Fzd6 cytosolic domain fragments (aa 497-567 or 568 -709). After incubation, the proteins bound to M2-agarose were eluted by FLAG peptides and analyzed by immunoblot using the indicated antibodies. Data shown are a representative example of three biological repeats. D, cell lysates from HEK293T cells transfected with plasmids encoding epsinR without the IgG-binding ZZ domain were incubated with M2-agarose conjugated with anti-FLAG antibodies. Subsequently, the immunoprecipitated proteins were incubated with the indicated amount of a purified GST-tagged Fzd6 cytosolic domain fragment (aa 497-567, WT) or the indicated amount of a purified GST-tagged Fzd6 cytosolic domain fragment (aa 497-567, ⌬KR). After incubation, the proteins bound to M2-agarose were eluted by FLAG peptides and analyzed by immunoblot using the indicated antibodies. E, quantification of the levels of the bound ␥1-adaptin based on three independent replicates (mean Ϯ S.D.). In each replicate of the experiment, the intensity of the bound ␥1-adaptin was normalized to the sum of the intensities of the bound ␥1-adaptin from all reaction conditions performed in that replicate.

TGN sorting of a planar cell polarity protein Frizzled6
Negative stain electron microscopy analysis Carbon-coated 400-mesh copper grids were glow discharged using PELCO easiGlow TM Glow Discharge Cleaning System (Ted Pella, Inc., Redding, CA). Ten-microliters of purified and concentrated vesicle sample aliquots were air dried onto the copper grid and negatively stained for EM observation. To obtain better contrast, sample grids were post-stained with 5% uranyl acetate for 2 min and further dried before being examined using a Hitachi H-7650 transmission electron microscope with a CCD camera operating at 80 kV (Hitachi High-Technologies Corporation, Japan).