Clathrin adaptor AP1B controls adenovirus infectivity of epithelial cells

Abstract

Adenoviruses invading the organism via normal digestive or respiratory routes require the Coxsackie-adenovirus receptor (CAR) to infect the epithelial barrier cells. Because CAR is a component of tight junctions and the basolateral membrane and is normally excluded from the apical membrane, most epithelia are resistant to adenoviruses. However, we discovered that a specialized epithelium, the retinal pigment epithelium (RPE), anomalously expressed CAR at the apical surface and was highly susceptible to adenovirus infection. These properties of RPE cells correlated with the absence of the epithelial-specific clathrin adaptor AP1B. Furthermore, knockdown of this basolateral sorting adaptor in adenovirus-resistant MDCK cells promoted apical localization of CAR and increased dramatically Adenovirus infectivity. Targeting assays showed that AP1B is required for accurate basolateral recycling of CAR after internalization. AP1B knock down MDCK cells missorted CAR from recycling endosomes to the apical surface. In summary, we have characterized the cellular machinery responsible for normal sorting of an adenovirus receptor and illustrated how tissue-specific variations in such machinery result in drastic changes in tissue-susceptibility to adenoviruses.

To gain access to a target organism, pathogens must develop adaptive mechanisms to overcome the formidable barrier presented by polarized epithelia lining the digestive or respiratory tracts [reviewed in (1–3)]. Coxsackie and adenoviruses invading the organism via normal intestinal or respiratory routes use Coxsackie-adenovirus receptor (CAR) as a primary receptor (4). Interestingly, CAR is a normal transmembrane protein component of tight junctions (TJ) and basolateral membranes, but it is excluded from the apical surfaces of most epithelial cells (5). Cocksackie viruses gain access to CAR by triggering specific signaling events that mediate the opening of TJ and facilitate penetration (3, 6) but adenoviruses have no similar mechanism available and require mechanical disruption of target epithelial tissues for penetration (7) (2, 8). However, the retinal pigment epithelium (RPE) in situ is highly susceptible to adenoviruses introduced by subretinal injection (9, 10), and confluent RPE monolayers in culture can be efficiently inoculated with adenoviruses added to the apical medium (11). What is the mechanism involved in this facile adenovirus infection of some native epithelia?

Polarized epithelial cells are characterized by an asymmetric distribution of plasma membrane proteins into apical and basolateral domains, separated by the tight junction (TJ) fence (12–14). Their polarized surface organization results from intracellular sorting of their apical and basolateral plasma membrane (PM) proteins at 2 major intracellular compartments, trans Golgi network (TGN), and recycling endosomes (RE). Sorting is carried out by cellular machinery that recognizes sorting signals present in apical and basolateral PM proteins that mediates their incorporation into transport vesicles for delivery to their respective residence domains at the cell surface (12, 13, 15). Recently, it has become apparent that most basolateral proteins require clathrin for accurate sorting (16) and previous work has shown that the clathrin adaptor AP1B facilitates the sorting of proteins with tyrosine-based basolateral sorting motifs (17, 18). Additional work has shown that AP1B localizes to RE and sorts basolateral proteins not only in their recycling route (19) but also in the biosynthetic route (20, 21), supporting the concept that newly synthesized PM proteins may transit from the Golgi complex to the endosomal compartment before reaching the cell surface [reviewed by (12, 22)].

Here, we systematically analyzed the mechanisms that facilitate adenovirus infection of some native epithelia. Our results show that lack of expression of the epithelial-specific clathrin adaptor AP1B in RPE correlates with high infectivity by adenoviruses and demonstrate that knockdown of AP1B by siRNA dramatically enhances adenovirus infectivity of a non-susceptible epithelium. They also illustrate the molecular mechanism responsible for such variation in infectivity. In epithelial cells that contain AP1B, CAR is routed directly from the TGN to the basolateral surface and is incorporated slowly to the TJ after several rounds of internalization and basolateral recycling from the RE. Epithelial cells that lack AP1B also route newly synthesized CAR to the basolateral surface but fail to recycle endocytosed CAR accurately to the basolateral domain. Instead, they transcytose CAR to the apical surface, resulting in increased apical expression of this receptor and higher adenovirus infectivity.

Results

RPE Displays Anomalously High Adenovirus Infectivity.

To study adenovirus infection of polarized epithelial cell lines, we used replication defective adenovirus serotype 5, in which sequences encoding transcription factors E1A/E1B (required for transcription of all viral proteins) are replaced by sequences encoding GFP. Infective viruses were generated by complementation of E1A/E1B in HEK-293 cells. In cells other than HEK-293 cells, inoculation with these viruses reproduces all aspects of adenovirus infection with the exception of the production of capsid proteins and virus assembly. Inoculation of several different polarized epithelial cell lines grown on Transwell filters from the apical side, the most likely route used by the virus in vivo (7), resulted in dramatic differences in infection levels with adenovirus serotype-5. Confluent monolayers of Caco-2 (human intestinal epithelial cell line), Calu-3 (human respiratory epithelial cell line), and MDCK (dog kidney epithelial cell line) were poorly infected. On the other hand, confluent monolayers of human fetal RPE (primary cultures) and ARPE-19 (human RPE cell line) were infected at much higher levels (5–20 fold). Importantly, brief pretreatment of the monolayers with 2.5 mM EDTA before inoculation resulted in 5–20 fold higher infection of Caco-2, Calu-3, and MDCK while only marginally increased infectivity of RPE cells (Fig. 1). These experiments are consistent with the known localization of CAR to TJ and the basolateral membrane and its absence from the apical membrane in many epithelia (5, 23). They also suggest the hypothesis that the higher infectivity of some native epithelia, that is, RPE, results from the presence of accessible adenovirus receptors at the apical membrane.

Fig. 1.

Adenovirus infection of polarized epithelial monolayers. Confluent monolayers of epithelial cells, grown on Transwell^R chambers, were inoculated with Adenovirus-5 (Ad5CiG), encoding GFP, added to the apical medium. The infection levels were quantified by measuring GFP fluorescence (see Experimental Procedures). Caco-2 (human intestinal), Calu-3 (human respiratory), and MDCK (dog kidney) were poorly infected whereas human fetal RPE (primary cultures) and ARPE-19 (human RPE cell line) were efficiently infected. Pretreatment with EDTA before inoculation dramatically increased infection levels of Caco-2, Calu-3, and MDCK but only marginally increased infection levels of RPE cells. The extent of viral transduction is cell-type dependent. (Scale bar, 20 μm.) Data are represented as mean ± SD, n = 3.

CAR Is Present at the Apical Surface of RPE.

To test this hypothesis, we studied the localization of CAR in the various epithelial cell lines used in Fig. 1. Exposure of both sides of intact monolayers grown on filters to a monoclonal antibody against the ectodomain of human CAR demonstrated the presence of endogenous CAR in the basolateral membrane of Caco-2 and Calu-3 cells, but no staining of TJ or of the apical surface (Fig. 2). In striking contrast, the human CAR antibody labeled both apical and basolateral surfaces of intact human RPE cells. Upon fixation and permeabilization, CAR was also detected at the TJ of all epithelial cell types (Fig. S1). CAR expression in all these cell lines was also confirmed by western blot (Fig. S2). These experiments supported our working hypothesis, that is, that the higher apical infectivity of RPE by adenoviruses might be explained by immunoaccessible CAR at the apical surface.

Fig. 2.

CAR is expressed apically in polarized RPE cells. Polarized monolayers of human epithelial cell lines Caco-2 and Calu-3, grown on Transwell^R chambers, were fixed with paraformaldehyde and incubated in the presence of anti-human CAR antibody (RmcB) added to both apical and basal media, overnight at 4 °C, washed, fixed, and exposed to fluorescent second antibody added to both sides. Note that Caco-2 and Calu-3 cells display CAR only on the basolateral side whereas human fetal RPE and ARPE-19 cells display CAR in both apical and basolateral surfaces. (Scale bar, 10 μm.)

Cells Expressing Apical CAR Lack the Clathrin-Adaptor AP1B.

CAR displays a typical tyrosine-based basolateral signal of the YxxΦ type (where Y is tyrosine, x is any amino acid, and Φ is any hydrophobic amino acid) (5). As this class of basolateral sorting signals is typically recognized by AP1B (see Introduction), we therefore tested the hypothesis: Might the apical localization of CAR in RPE cells be explained by the absence of AP1B in these cells?

The epithelial-specific adaptor AP1B shares with the ubiquitous AP1A 3 of its 4 subunits (γ, β1, and σ), but it differs from AP1A in the possession of a different medium (μ) subunit, μ1B instead of μ1A. Analysis of different epithelial cell lines by RT-PCR (Fig. 3A) and western blot (Fig. 3B) demonstrated the presence of μ1A in all of them; by contrast, only RPE cells lacked μ1B.

Fig. 3.

RPE cells lack clathrin adaptor AP1B. (A) RT-PCR and (B) western blot of μ1A and μ1B in various epithelial cell lines. Note absence of μ1B in RPE cells. Calu-3 samples were run at a later time.

Knockdown of AP1B in MDCK Cells Promotes Apical Infectivity with Adenoviruses.

If lack of AP1B were indeed responsible for the apical localization of CAR in RPE cells, knockdown of AP1B in MDCK cells would result in increased expression of CAR at the apical surface. Staining of intact monolayers of MDCK cells knocked down for AP1B (20) and overexpressing human CAR demonstrated the presence of this protein at both apical and basolateral surfaces, whereas in control cells CAR was only detected at the basolateral surface (Fig. 4A). In agreement with these findings, AP1B KD MDCK cells were approximately 15 times more susceptible to adenoviruses from the apical surface than wild-type MDCK cells (Fig. 4 B and C). The same phenotype was observed in a μ1B knockdown clonal cell line derived from the Fischer rat thyroid (FRT) epithelial cell line (21). Permanent expression of human μ1B (resistant to the shRNA targeting the canine sequence) into μ1B KD MDCK cells (Fig. S3) restored the basolateral expression of CAR (Fig. 4A) and the low apical infectivity by adenoviruses (Fig. 4 B and C).

Fig. 4.

Apical CAR localization and high adenovirus infectivity of AP1B-KD MDCK cells. (A) Wild-type (WT) MDCK cells, AP1B knockdown (μ1B KD) MDCK cells, and μ1B-KD MDCK cells expressing human μ1B-HA were fixed and analyzed for surface expression of overexpressed human CAR by immunofluorescence, exactly as in Fig. 2. Note that CAR is basolateral in WT MDCK cells, non-polar in μ1B KD MDCK cells, and basolateral in μ1B KD/human μ1B-HA MDCK cells. (B and C) AP1B knock-down MDCK cells show higher susceptibility to apical adenovirus infection than WT MDCK cells and μ1B KD/μ1B-HA MDCK cells. (Scale bar, 10 μm.) Data are represented as mean ± SD, n = 6.

Although to date we have not succeeded in reconstituting AP1B into RPE cells due to technical difficulties in permanently transfecting these cells, our experiments with MDCK cells support our hypothesis that lack of AP1B in RPE cells promotes apical expression of CAR and increased apical infectivity with adenoviruses.

As an additional control for these experiments, we studied the infectivity of wild-type and AP1B-KD MDCK cells by Ad5F7CiG, a serotype-5 adenovirus bearing a subgroup B serotype-7 fiber protein (24), modified to express chloramphenicol acetyl transferase and GFP. Viruses from this subgroup have been reported to use CD46 as a primary receptor instead of CAR (25). CD46 localizes to the basolateral surface of MDCK cells even after mutation of a putative tyrosine signal (26, 27), suggesting that this protein would not be affected by the absence of AP1B. Indeed, Ad5F7CiG did inefficiently infect control or μ1B-knockdown MDCK cells from the apical side (Fig. S4). Only after TJ disruption by EDTA treatment was this virus able to transduce MDCK monolayers, in agreement with the basolateral localization of CD46. These data support our hypothesis that in the absence of AP1B, the depolarization of CAR, and not a non-specific event, indeed promotes the higher infectivity of Ad5 from the apical side in AP1B-knockdown MDCK cells.

Delivery of Newly Synthesized CAR to the Basolateral Membrane Is Independent of AP1B.

Clathrin adaptors have characteristic subcellular distributions that determine the sorting processes in which they participate (28). They are found in 2 varieties, tetrameric (AP1, AP2, AP3, and AP4) and monomeric (GGAs). AP2 localizes at the plasma membrane, where it cooperates with clathrin and other accessory proteins in the endocytosis of a variety of receptors, whereas AP1, AP3, AP4, and GGAs localize to intracellular organelles, the example, the Golgi complex and recycling endosomes (RE), where they participate in inter-organellar trafficking. AP1B localizes to RE and participates in the sorting of basolateral proteins such as transferrin receptor (TfR) and VSV-G protein at both biosynthetic and recycling routes of recently polarized (<1 day confluent) MDCK cells. However, in MDCK monolayers polarized for over 4 days, biosynthetic delivery of newly synthesized TfR to the basolateral membrane was independent of AP1B, suggesting that an adaptor other than AP1B was involved in this route and that AP1B controlled the polarity of TfR at the level of the recycling route (19–21).

How does AP1B control the basolateral sorting of CAR? Does it control its biosynthetic delivery to the basolateral membrane or its accurate basolateral recycling after internalization? To investigate the first possibility, we carried out a pulse-chase analysis of CAR-GFP in fully polarized cells using optical microscopy (29) and biochemical surface-capture assays that measure the time course of delivery of PM proteins in the biosynthetic route (12, 13, 15). Using these assays, we observed that both control MDCK cells and AP1B-knockdown MDCK cells sorted newly synthesized CAR-GFP with normal kinetics to the basolateral membrane, suggesting that CAR reaches the PM via an AP1B-independent route. However, after several additional hours of chase, AP1B-knockdown cells missorted CAR to the apical surface (Fig. 5 A and B), suggesting that AP1B sorts CAR post-endocytically, in the recycling route of this receptor to the basolateral PM.

Fig. 5.

Newly synthesized CAR is delivered preferentially to basolateral PM. (A) CAR-GFP delivery to PM. Confluent monolayers of WT and μ1B-KD MDCK cells were microinjected with cDNA encoding human CAR-GFP, incubated at 37 °C for 1 h to express the protein, then at 20 °C to allow accumulation of CAR-GFP at the TGN, and chased at 37 °C for different times. Fixed monolayers were imaged with a laser scanning confocal microscope (LSCM). Note basolateral delivery of CAR-GFP in both WT and μ1B-KD MDCK cells in the first hours of trafficking. At late chase time points, some loss of polarity of CAR-GFP is observed in μ1B-KD MDCK cells. (Scale bar, 10 μm.) (B) Domain selective biotinylation and surface capture of CAR. Confluent monolayers of WT and μ1B-KD MDCK cells overexpressing human CAR-GFP were pulsed with ³⁵S methionine/cysteine, chased for different times, and the percentage of CAR arriving to apical or basolateral domains retrieved by domain selective biotinylation and streptavidin precipitation. Note that CAR is delivered preferentially to the basolateral membrane in both WT and μ1B-KD MDCK cells, but it progressively loses polarity at late chase points in μ1B-KD MDCK cells. Data are represented as mean ± SD, n = 3.

CAR Undergoes Several Rounds of Endocytosis Before Incorporation to TJ and Transcytoses to the Apical Surface of AP1B-Deficient Epithelia.

If CAR requires AP1B to accurately recycle to the basolateral surface, knocking-down of AP1B should result in aberrant transcytosis of endocytosed CAR to the apical surface. To investigate this hypothesis, we developed antibody-binding assays for CAR endocytosis and transcytosis (see Experimental Procedures).

The endocytosis assay (Fig. 6A) indicated that MDCK cells internalized CAR efficiently with an approximate half-life of 20 min and that lack of AP1B did not interfere with this mechanism. By contrast, we observed a dramatic transcytosis of CAR in AP1B-knockdown MDCK cells, at the rate of 10% of the amount present at the basolateral surface per hour, which was not observed in wild type MDCK cells (Fig. 6B). These experiments indicate that AP1B is involved in normal basolateral recycling of CAR from RE to the basolateral membrane and that its absence promotes incorporation of CAR at RE into a transcytotic route to the apical membrane.

Fig. 6.

AP1B-KD MDCK cells transcytose CAR to the apical PM. (A) CAR endocytosis. An antibody-based assay was used to quantify CAR endocytosis in subconfluent WT and AP1B-KD MDCK cells permanently expressing human CAR. Cells were exposed to CAR monoclonal antibody for 2 h at 0 °C, chased for 30, 60, or 120 min at 37 °C, fixed with 4% paraformaldehyde and stained with Cy5-donkey anti-mouse secondary antibody. After a second fixation with 4% paraformaldehyde, the cells were permeabilized with 0.1% Triton X-100 and stained with Cy3-donkey anti-mouse antibodies. Green Cy5 fluorescence represents surface CAR whereas red Cy3 fluorescence represents internalized CAR. Results were quantified using a Nikon widefield microscope and plotted in bottom graph. The x-y fluorescence images shown below were obtained with a Leica SP2 scanning confocal microscope. Note that CAR endocytosis proceeds at the same rate in control and AP1B KD MDCK cells. (Scale bar, 10 μm.) (B) CAR transcytosis. Fully polarized (4 days confluent) monolayers of WT and AP1B-KD MDCK cells expressing human CAR and grown on polycarbonate filters were exposed to MαCAR antibody on the basal side (2 h, 0 °C), chased at 37 °C with rabbit anti mouse (RαM) antibodies on the apical side for different times, fixed, stained with IRDye800-GαR antibody and infra-red scanned with Odyssey^R. After a wash, second fix, permeabilization with 0.1% saponin, RαM staining and IRDye800-GαR, cells were again infrared scanned. Results are plotted in the bottom graph. Note that CAR is transcytosed at the rate of 10% per hour by AP1B KD MDCK but not by WT MDCK cells. The fluorescence image below is a confocal image of parallel samples decorated with GαΜ-Alexa488 (green) and GαR-Alexa555 (red) using a similar protocol to observe transcytosis at the microscope level. Green fluorescence represents basolateral CAR whereas red fluorescence represents CAR transcytosed to the apical membrane. (Scale bar, 10 μm.) (C) AP2 depletion in fully polarized MDCK-WT by siRNA decreases CAR internalization from the basolateral membrane. In all panels, data are represented as mean ± SD, n = 3.

Additional AP2 knockdown experiments demonstrated the involvement of AP2 in basolateral internalization of CAR (Fig. 6C).

Discussion

Here, we focused on elucidating the mechanism underlying the very facile infection of certain epithelia by adenoviruses. Screening of various epithelial cell types demonstrated that RPE was much more susceptible to infection by adenoviruses serotype 5 than respiratory, kidney, or intestinal epithelia. Analysis of the localization of CAR by immunofluorescence showed that RPE anomalously localized CAR to the apical surface. Additional studies demonstrated that RPE cells do not express the clathrin adaptor AP1B and that MDCK cells knocked-down of APIB by siRNA (20) also localized CAR to the apical surface. These experiments conclusively established that lack of AP1B confers epithelial cells with high infectivity to adenoviruses due to anomalous apical localization of CAR.

To further understand how AP1B contributes to the more frequently observed basolateral and TJ localization of CAR, we carried out a careful analysis of its biosynthetic and recycling routes. Surprisingly, since our previous work has shown that AP1B participates in the biosynthetic route of VSV-G protein, which like CAR exhibits a tyrosine-based (YxxΦ) basolateral signal (20), AP1B knockdown did not affect the basolateral delivery of newly synthesized CAR from the TGN but, instead, interfered with the accurate post-endocytic recycling of CAR to the basolateral membrane and promoted missorting of CAR to the apical membrane. AP2 knockdown strongly interfered with this step. Taken together, our results demonstrated that CAR is normally targeted in an AP1B-independent fashion from the TGN to the basolateral membrane, where it undergoes several rounds of AP2-dependent endocytosis and accurate basolateral recycling mediated by AP1B (Fig. 7). Future work must identify the adaptor involved in basolateral sorting of newly synthesized CAR.

Fig. 7.

CAR trafficking in polarized epithelial cells. (A) Epithelial cells that express AP1B efficiently target newly synthesized CAR to the basolateral membrane via an unknown adaptor (1), internalize CAR via AP2 and likely clathrin (white dots) into recycling endosomes (RE) (2) and recycle CAR accurately to the basolateral surface via AP1B/clathrin-mediated sorting (3). CAR is excluded from the apical membrane of these cells and therefore adenoviruses or Cocksackie viruses require opening of TJ for efficient infection. (B) Epithelial cells lacking AP1B target newly synthesized CAR to the basolateral membrane (1), internalize CAR via AP2 into RE (2), but fail to accurately recycle CAR to the basolateral surface (3) due to lack of AP1B. CAR is transcytosed to the apical PM by an unknown mechanism (4) where it is available for adenovirus infection.

The paradigm we introduce here to describe how lack of a clathrin adaptor confers epithelial susceptibility to adenoviruses may help understand how other pathogens normally enter epithelial cells. CD155, the receptor for human poliovirus (30) and for animal α herpes viruses (31) is known to interact with AP1B in epithelial cells (32). Reoviruses, which causes gastrointestinal and respiratory illnesses, initiate infection by binding the junction adhesion molecule (JAM) via their capsid protein σ1 (33). Of note, RPE cells localize large pools of JAM to their apical surface; furthermore, this localization is phenocopied by MDCK cells depleted of AP1B by siRNA. Lastly, RPE displays relatively high apical susceptibility to herpes simplex virus type 1 infection compared with MDCK cells, which might be attributed to the anomalous apical localization of 1 of several possible receptors used by this virus (34, 35).

In summary, our experiments provide a molecular explanation for the resistance of most epithelial tissues and the anomalous susceptibility of RPE to adenoviruses. They describe the complete route followed by a TJ protein and the involvement of clathrin adaptors in this route. Further studies should explore in detail whether other transmembrane components of the TJ (e.g., claudins, occludin,) use a similar route and, importantly, to what extent AP1B controls epithelial invasion by other human pathogens. Additionally, it is likely that other epithelia found to lack AP1B might show similar enhanced viral infectivity as RPE.

Experimental Procedures

AP1B Knockdown and Reconstitution.

Generation of stable clones of MDCK μ1B knockdown was reported elsewhere (20). Briefly, a pair of custom-synthesized sense and antisense oligonucleotides encoding the 19-mer μ1B-targeted sequence (1B-M16) were annealed and cloned into a pSuper. retro.puro vector (Oligoengine) following manufacturer's instructions. The resulting vector, pSuper-1B-M16, was transfected into MDCK cells by electroporation (Amaxa). Puromycin-resistant clones were picked and further characterized by RT-PCR and western blot analysis. AP1B-KD cells were transfected with pCB6μ1B-HA, encoding human μ1B tagged with an HA epitope, and selected for 14 days with 1 mg/mL G418 in the presence of 2.5 μg/mL puromycin. Several clones resistant to both puromycin and G418 were tested for expression of HA, and basolateral expression of human transferrin receptor (overexpressed from adenoviruses), as criteria for recovery of AP1B expression and function (20). Among these, clone 9 was selected for further studies on CAR localization and apical infectivity for adenoviruses.

For more details see SI Text.

Viral Infection.

Inoculation of epithelial monolayers was carried out at a multiplicity of infection of 5,000 viral particles (in 400 μL) per cell (particles to pfu ratio ≈100) in opti-MEM or in Ca/Mg-free Hank's Balanced Salt Solution (HBSS) containing 2.5 mM EDTA at 37 °C, added only to the top chamber of the filters. After viral transduction, cells were dissociated by trypsinization and collected into Eppendorf tubes with 500 μL complete medium at 4 °C. Cells were washed with cold HBSS and lysed with 250 μL RIPA (150 mM NaCl, 1% SDS, 1 mM EDTA, 025% sodium deoxycholate, 50 mM Tris-HCl, pH 7.4) buffer containing protease inhibitors on ice for 1 h in the dark. Samples were transferred to an opaque 96-well plate and read in a Gemini SoftMax plate reader with filters set at 485/509 nm Ex/Em for GFP using uninfected cells as blank. Propidium iodide (PI) was then added, and filters were set at 535/617 nm. GFP readings were normalized to PI values.

Plasmids and Transfections.

pCDNA3.1 encoding the human isoform of CAR was kindly provided by Jeffrey M. Bergelson. A plasmid encoding a GFP-tagged form of CAR was generated as follows: human CAR was amplified with the following primers (5′ATCTGGTACCATGGCGCTCCTGC3′/5′GATACGCGTTACTATAGACCCATCCTTGC), changing the stop codon by the first 3 nucleotides of the MluI restriction site. The PCR product was purified, digested with KpnI and MluI and cloned into pCB7. The sequence encoding the spacer-GFP was amplified with the following primers (5′ATGTAGCCGTCTCCCCGCCGAACAG3′/5′TAGTTCTAGATTACTTGTACAGCTCGTC3′) from the plasmid VSVG-sp-GFP (36). This PCR product was cloned between the MluI and XbaI sites of the pCB7CAR, thus tagging CAR with eGFP on the cytoplasmic side after a 23-amino acid spacer.

All transfections for transient expression of both full-length hCAR and hCAR-GFP were performed by electroporation of fully polarized monolayers, as previously described (37).

Targeting Assays.

Optical Microscopy Assay.

Both wild-type and AP1B-depleted MDCK cells were plated at confluency on glass-bottom dishes (Mattek P35G-1.0–14-C). After 3 days the cells were subjected to nuclear microninjection with 5 μg/mL CAR-GFP in microinjection buffer. Cells were allowed to express the protein for 1.5 h at 37 °C. Dishes were transferred to a 20 °C block to allow protein accumulation at the TGN for 1 h. Cells were transferred to 32 °C in the presence of 100 μg/mL cycloheximide for the chase. At each time point, cells were fixed in PFA and processed for confocal microscopy.

Biochemical Targeting Assay.

MDCK cells stably expressing CAR-GFP, confluent for 4.5 days on filters, were incubated with 5 mM butyrate (Sigma) for 12 h before experiments to boost CAR-GFP expression. Cells were then pulse-labeled for 20 min with 0.5 mCi/filter of ³⁵SMet/Cys (NEG-072, Perkin-Elmer) and immediately chased at 37 °C in complete DMEM supplemented with 3 mM cold methionine and cysteine. At each time point, cells were processed for surface delivery as described elsewhere (20). For more information, see SI Text.

Endocytosis Assay.

The endocytosis assay used in AP1B-knockdown MDCK cells is a modification of a previously published protocol (38). For more information, see SI Text.

AP2 was knocked-down in MDCK cells according to previous protocols (16). Monolayers of AP2-knockdown MDCK cells, fully polarized on filters for 3 days, were incubated with anti-CAR antibody in 1% BSA OptiMEM in the basal chamber for 2 h at 0 °C, washed, and chased at 37 °C. At each time point, cells were chilled on ice and fixed in 4% PFA at 0 °C for 10 min. Cy5-conjugated secondary antibodies were used to label the bound anti-CAR antibodies at the surface and PI in 0.1% TX-100 was used as a counterstain. Samples were scanned in a Typhoon Trio with the appropriate filters for each fluorophore. Internalization was calculated as surface signal over PI.

Trancytosis Assay.

A fluorescence based capture assay of secondary antibody on the apical chamber was used to measure trancytosis of anti-CAR antibody bound to basal membrane of fully polarized WT or AP1B-knockdown MDCK cells expressing hCAR. For more information, see SI Text.