The adaptor complex AP-2 regulates post-endocytic trafficking through the non-clathrin Arf6-dependent endocytic pathway Free

Endocytosis is the process by which cells internalize extracellular contents, fluids, and cell-surface proteins. Multiple forms of endocytosis exist and can be classified based on their dependence on the coat protein clathrin (Conner and Schmid, 2003; Nichols and Lippincott-Schwartz, 2001). Clathrin-mediated endocytosis (CME) is far better characterized than clathrin-independent endocytosis, and much of the molecular machinery required for the former has been identified. Cargoes that enter the clathrin pathway, such as the transferrin receptor (TfnR), contain within their cytoplasmic domains internalization signals that are recognized by proteins known as clathrin adaptors (Bonifacino and Traub, 2003; Ohno, 2006; Traub, 2003; Wendland, 2002). Adaptors have diverse roles in CME, including stimulation of clathrin-coat assembly, binding and clustering of cargo at sites of invagination, and recruitment of accessory proteins that facilitate coat formation, cargo recruitment and vesicle formation (Bonifacino and Traub, 2003; Robinson and Bonifacino, 2001; Traub, 2003). The large GTPase dynamin functions at the latest stages of CME to promote vesicle scission (Sever et al., 2000a; Sever et al., 2000b; Song and Schmid, 2003). Shortly following internalization into vesicles, endocytosed proteins are found in an early endosome marked by Rab5 and EEA1 (Christoforidis et al., 1999; Hiroyama and Exton, 2005; Simonsen et al., 1998; Zerial and McBride, 2001). From this compartment, potential fates of cargo include sorting to the lysosome for degradation, or recycling to the plasma membrane.

The best-characterized clathrin adaptor is the adaptor protein 2 (AP-2) complex. The importance of this heterotetrameric complex is highlighted by the finding that small interfering RNA (siRNA)-mediated depletion of its various subunits completely inhibits endocytosis of certain clathrin-pathway cargoes such as the TfnR (Hinrichsen et al., 2003; Keyel et al., 2006; Motley et al., 2003). AP-2 is generally believed to function exclusively at the plasma membrane, where it is highly concentrated at clathrin-coated pits (Robinson, 1987; Robinson, 1989; Robinson and Pearse, 1986). Phosphatidylinositol (4,5)-bisphosphate (PtdIns(4,5)P₂) has an essential role in recruiting AP-2 to the plasma membrane by binding to basic motifs in the α- and μ2-subunits of AP-2 (Gaidarov et al., 1996; Gaidarov and Keen, 1999; Rohde et al., 2002).

By comparison, our understanding of endocytic trafficking through clathrin-independent pathways is less complete. Recently, much interest has focused on the role of the ADP-ribosylation factor 6 (Arf6) GTPase in endocytosis (Donaldson, 2003; Donaldson and Radhakrishna, 2001). The literature on Arf6 is complex, with Arf6 functioning in CME in some cell types (such as CHO and MDCK cells) (Altschuler et al., 1999; D'Souza-Schorey et al., 1995; Palacios et al., 2002), and non-clathrin endocytic trafficking in others (such as macrophages, neuroendocrine and HeLa cells) (Arnaoutova et al., 2003; Brown et al., 2001; Donaldson and Radhakrishna, 2001; Naslavsky et al., 2003; Niedergang et al., 2003; Radhakrishna and Donaldson, 1997; Zhang et al., 1999; Zhang et al., 1998). In macrophages, Arf6 activity is required for phagocytosis, a clathrin-independent form of endocytosis (Niedergang et al., 2003; Zhang et al., 1999; Zhang et al., 1998). The role of Arf6 in clathrin-independent trafficking has been most extensively studied in HeLa cells, and is supported by multiple lines of evidence. First, cargoes of the Arf6 pathway [such as major histocompatibility complex class I (MHCI), β1 integrin, and glycosylphosphatidyl inositol (GPI)-anchored proteins] lack the sorting signals recognized by AP-2. Second, endosomes harboring Arf6-pathway cargo exclude classical clathrin cargo such as transferrin (Tfn) (Brown et al., 2001; Naslavsky et al., 2003; Radhakrishna and Donaldson, 1997). Third, expression of a mutant form of the adaptor protein AP180/CALM that sequesters clathrin and blocks CME does not affect internalization of Arf6-pathway cargo (Naslavsky et al., 2004). Fourth, internalization of Arf6 cargo does not require dynamin, which is essential for CME (Blagoveshchenskaya et al., 2002; Le Gall et al., 2000). Together, these observations demonstrate that Arf6 regulates endocytosis through a route distinct from CME in these cells.

Recent studies have begun to shed light on the mechanisms of internalization and recycling through the Arf6 pathway. Endocytosis through this pathway requires cholesterol, as suggested by its sensitivity to the sterol-binding agent filipin (Naslavsky et al., 2004). Once internalized, endocytic cargo may be delivered to a distinct recycling compartment termed the tubular endosome, and recycled to the plasma membrane in an Arf6-, Rab22- and Rab11-dependent manner (Powelka et al., 2004; Weigert et al., 2004). Alternatively, Arf6 endosomes may fuse with endosomes derived from the clathrin-dependent pathway, with subsequent delivery of cargo to lysosomes (Naslavsky et al., 2003). Fusion of these distinct endosomal populations requires the production of phosphatidylinositol (3,4,5)-trisphosphate (PtdIns(3,4,5)P₃) (Naslavsky et al., 2003).

Despite these advances, many questions remain regarding the selection of cargo into the Arf6 pathway and the mechanisms of sorting to the recycling vs degradative routes. Recent work showed that Arf6 directly associates with AP-2 in a GTP-dependent manner, and that Arf6 functions cooperatively with PtdIns(4,5)P₂ to recruit AP-2 to membranes (Paleotti et al., 2005). This led the authors to conclude that the function of Arf6 in the context of clathrin-dependent endocytosis is to promote recruitment of AP-2 to the plasma membrane. However, we considered the possibility that the physical association of AP-2 and Arf6 reflects a role for AP-2 in clathrin-independent trafficking through the Arf6 pathway. In this study, we examine the effects of AP-2 depletion on trafficking through this endocytic route. Our results suggest a role for AP-2 in post-endocytic sorting of Arf6 cargo, specifically in promoting recycling over lysosomal targeting.

Arf6 is believed to function in clathrin-independent endocytosis in HeLa cells based on multiple lines of evidence. However, previous studies have not utilized siRNA to examine whether endocytic trafficking of Arf6 cargo proceeds normally in cells depleted of clathrin. Furthermore, effects of clathrin or AP-2 depletion on post-endocytic trafficking have not been directly examined. To address these issues, we first knocked down clathrin expression by using siRNA oligonucleotides directed against clathrin heavy chain (CHC). HeLa cells were subjected to two rounds of transfection with siRNA oligonucleotides. Immunoblotting confirmed a substantial reduction in levels of CHC protein (Fig. 1A). To determine whether this degree of depletion was functionally significant, we examined internalization of the TfnR, a classic cargo of the clathrin pathway, by monitoring the uptake of fluorescently labeled Tfn. Live cells were incubated with Tfn–Alexa-Fluor-546 for 30 minutes at 37°C. Cells were then fixed, labeled with anti-clathrin antibody and analyzed by confocal microscopy. Cells transfected with control siRNA (directed against firefly luciferase), hereafter referred to as control siRNA cells, exhibited uniform Tfn uptake (Fig. 1B). As expected, Tfn uptake was significantly inhibited in cells where clathrin expression was knocked down (Fig. 1C). In addition, some clathrin knockdown cells displayed an altered localization pattern of Tfn, in which low levels of ligand aggregated in a perinuclear region (Fig. 1C, arrows). While such a phenotype has not previously been described in clathrin-depleted cells, it has been reported in cells overexpressing the clathrin hub region (Bennett et al., 2001), which functions in a dominant-negative manner. Similar Tfn localization has also been reported in cells expressing a dynamin mutant that blocks CME (van Dam and Stoorvogel, 2002), a phenotype that was also observed in our system (Fig. 1D, arrow).

Because this altered pattern of Tfn staining had not previously been reported for clathrin-depleted cells, we wished to confirm that it did not arise from some irregularity in our live uptake procedure. We therefore examined localization of the TfnR itself by immunofluorescence microscopy. Clathrin status was monitored by Tfn–Alexa-Fluor-488 uptake because the clathrin and TfnR antibodies were of the same species and IgG subtype, precluding their simultaneous use. As seen in Fig. 1E, TfnR accumulated at the plasma membrane in clathrin knockdown cells. In addition, low levels of TfnR localized to perinuclear clusters in some cells (Fig. 1E, arrows), as seen with Tfn. In adjacent cells, where clathrin expression was intact, the TfnR was found on vesicles throughout the cell (Fig. 1E). The nature of the perinuclear aggregate that forms in cells where clathrin is inactivated has not been well characterized (Bennett et al., 2001; van Dam and Stoorvogel, 2002), but it is reasonable to surmise that it arises from secondary effects of prolonged clathrin and/or dynamin inactivation on post-Golgi and/or endosomal trafficking. Nevertheless, our data indicate that the extent of clathrin reduction in our system is substantial. Therefore, we hereafter refer to these cells as clathrin-depleted for simplicity, recognizing that knockdown of expression may be incomplete.

We next explored whether trafficking through the Arf6 pathway was affected by clathrin depletion by monitoring internalization of two representative cargoes, MHCI and β1 integrin (Donaldson and Radhakrishna, 2001; Radhakrishna and Donaldson, 1997). Antibodies against the extracellular domain of MHCI or β1 integrin were incubated with live cells, and allowed to internalize for 4 hours. Shorter periods of uptake were also examined, but gave less robust labeling (data not shown). Cells were then fixed and analyzed by confocal microscopy. In contrast to TfnR, neither MHCI (Fig. 2A) nor β1 integrin (Fig. 2B) appeared to accumulate at the plasma membrane in clathrin knockdown cells compared with wild-type cells in the same field. However, the intracellular distribution of these cargoes was altered. In wild-type cells, intracellular vesicles that contain MHCI and β1 integrin were more uniformly dispersed. In clathrin-depleted cells, both proteins localized to vesicles that were clustered in the perinuclear area, reminiscent of what we observed with Tfn. However, Tfn and β1 integrin localized to distinct vesicles (Fig. 2B), consistent with previous reports, which showed that these cargoes segregate into distinct endosomal compartments (Radhakrishna and Donaldson, 1997; Naslavsky et al., 2003). Thus, unexpectedly, depletion of clathrin appears to affect Arf6-dependent trafficking, perturbing not the internalization of Arf6 cargo but their intracellular distribution.

Given these surprising results, we wished to investigate whether AP-2 depletion similarly perturbed trafficking through the Arf6 pathway. HeLa cells were subjected to two rounds of transfection with siRNA oligonucleotides directed against the α-adaptin or μ2-subunit of AP-2. Immunoblotting confirmed significant reduction in levels of α-adaptin and μ2-protein in cells expressing either siRNA (Fig. 3A), consistent with previous reports (Janvier and Bonifacino, 2005; Motley et al., 2003). In control siRNA cells, anti-α-adaptin labeled punctate structures that represent clathrin-coated pits at the plasma membrane (Fig. 3B). In addition, this antibody labeled the nucleus non-specifically. In HeLa cells transfected with α-adaptin siRNA, the majority of cells exhibited a significant reduction or complete loss of the anti-α-adaptin signal at coated pits, while all cells gave a uniform non-specific nuclear signal (Fig. 3C, left panel). Uptake of Tfn–Alexa-Fluor-546 was strongly inhibited in the α-adaptin knockdown cells relative to control cells (Fig. 3C, middle panel). Furthermore, anti-TfnR staining confirmed substantial plasma membrane accumulation of the receptor in cells depleted of α-adaptin (Fig. 3C, right panel). In some α-adaptin-depleted cells, perinuclear aggregation of the TfnR was also observed (Fig. 3C, right panel), similar to what was seen in the clathrin-depleted cells. Depletion of the μ2 subunit of AP-2 also inhibited Tfn uptake (Fig. 3D).

Having confirmed the functional inactivation of AP-2, we proceeded to monitor the effects of AP-2 depletion on the trafficking of Arf6-pathway cargo (Donaldson and Radhakrishna, 2001; Radhakrishna and Donaldson, 1997). Live cells were incubated with antibody against MHCI or β1 integrin, then fixed and analyzed by confocal microscopy. Neither of these cargoes accumulated at the plasma membrane in cells depleted of AP-2 when compared to control cells, indicating that AP-2 is dispensable for their internalization (Fig. 4). However, similar to what was observed in clathrin-depleted cells, both MHCI and β1 integrin decorated vesicles that clustered in the perinuclear region in cells depleted of α-adaptin or μ2 (Fig. 4). This altered distribution did not arise from perturbations in Arf6 localization, which was indistinguishable in AP-2-depleted, clathrin-depleted and control cells (supplementary material Fig. S1).

The immunofluorescence microscopy analyses above suggested that depletion of clathrin or AP-2 did not block internalization of Arf6 cargoes but, instead, induced their intracellular clustering. However, because of this altered distribution, it was difficult to definitively affirm that their endocytosis was unaffected. To explore this quantitatively, we utilized flow cytometry-based assays. We first measured whether depletion of clathrin or AP-2 increased relative levels of MHCI at the cell surface. As a positive control, we confirmed that surface levels of TfnR were significantly increased in clathrin- and α-adaptin-depleted cells compared with control siRNA cells (Fig. 5A). By contrast, the relative proportion of MHCI at the cell surface was unaltered by depletion of clathrin or α-adaptin (Fig. 5A), supporting our conclusions from the immunofluorescence microscopy analysis.

We next measured the initial rate of endocytosis for MHCI. Surface MHCI was labeled by incubating cells with anti-MHCI antibody on ice. Cells were then transferred to 37°C for various times to allow internalization. Antibody remaining at the cell surface was detected using fluorescently conjugated secondary antibody, and quantified by flow cytometry. Depletion of clathrin had no effect on the initial rate of MHCI internalization relative to control cells (Fig. 5B). The internalization rate was only modestly reduced in the α-adaptin-depleted cells (Fig. 5B). Although statistically significant, this difference had no measurable effect on the steady-state ratio of levels of surface MHCI: intracellular MHCI (Fig. 5A), perhaps due to a compensatory increase in biosynthesis. Interestingly, analysis of longer times points in the endocytosis assay revealed a striking difference in the trafficking of MHCI specifically in α-adaptin-depleted cells. After the initial decline in surface MHCI, levels reached a plateau between 30-120 minutes in both control and clathrin-depleted cells (Fig. 5C). This plateauing presumably reflects recycling of internalized MHCI back to the plasma membrane. However, surface levels of MHCI continued to decline throughout the 120-minute period in α-adaptin-depleted cells (Fig. 5C), suggesting that recycling of surface-derived MHCI is perturbed.

We surmised that the apparent reduction in MHCI recycling in the α-adaptin-depleted cells might result from enhanced lysosomal targeting and degradation of surface-derived MHCI. Consistent with this possibility, we noticed that, at all time points in the endocytosis assays above, the absolute MHCI signal was lower in the α-adaptin-depleted cells compared with the control and clathrin-depleted cells (data not shown). This was confirmed by measuring internal pools of MHCI. Live cells were incubated with Alexa-Fluor-488-conjugated anti-MHCI for 2 hours. The pool at the cell surface was quenched with anti-Alexa-Fluor-488 antibody, and the remaining internal pool was quantified by flow cytometry. As shown in Fig. 6A, intracellular levels of MHCI in α-adaptin-depleted cells were ∼80% of control cells (n=4, P=0.0003), whereas there was no significant difference between clathrin-depleted cells and control cells.

As an additional quantitative approach, levels of endogenous MHCI protein were determined by immunoblotting. MHCI levels in cells depleted of α-adaptin were 67.7±9.3% of those found in control cells (n=7, P=0.01). To determine whether the reduction in MHCI levels arose from enhanced lysosomal targeting, we examined the effects of bafilomycin A1 (BafA1), an inhibitor of the vacuolar ATPase that prevents lysosome-mediated degradation. Treatment of α-adaptin-depleted cells with BafA1 restored MHCI levels to 175.0±14.1% (n=4, P=0.01) of those in untreated α-adaptin-depleted cells (Fig. 6B). These results suggest that loss of AP-2 leads to enhanced targeting of Arf6-pathway cargo to the lysosome, and to its subsequent degradation.

We next examined whether surface MHCI was degraded more rapidly in AP-2-depleted cells by using cell-surface biotinylation assays. Cell-surface proteins were labeled by incubation with biotin on ice. Excess biotin was removed, and cells were returned to 37°C for 4 hours to allow resumption of trafficking. Cell extracts were prepared, subjected to streptavidin-agarose pulldown assays to isolate biotin-conjugated proteins, then immunoblotted using various antibodies. To ensure efficient functional inactivation of clathrin and AP-2, we first examined the levels of biotinylated TfnR at the plasma membrane. As expected, the level of biotinylated TfnR at the cell surface (t=0 time point of the assay) was greatly increased in clathrin- and α-adaptin-depleted cells when compared to control cells, indicative of a block in internalization (Fig. 6C). By contrast, surface levels of MHCI at t=0 were not elevated in cells depleted of α-adaptin (Fig. 6D) or clathrin (Fig. 6E) when compared with control cells, confirming our flow cytometry assays. After 4 hours of internalization, the amount of MHCI remaining was considerably lower in α-adaptin-depleted cells compared with control cells (Fig. 6D,F). This effect was specific for cells lacking AP-2, as no significant difference in MHCI degradation was observed between control and clathrin-depleted cells (Fig. 6E,F). In sum, these experiments suggest that AP-2 has a unique role in post-endocytic sorting of Arf6-pathway cargo, specifically in diverting them from lysosomal degradation.

Whereas our biochemical and flow-cytometry-based assays indicated a specific requirement for AP-2 in trafficking of Arf6-pathway cargo (Figs 5 and 6), our initial immunofluorescence microscopy studies suggested that the distribution of MHCI and β1 integrin vesicles is perturbed in both clathrin- and AP-2-depleted cells (Figs 2 and 4). To gain insights into this apparent discrepancy, we further examined the vesicular aggregates observed in the clathrin- and AP-2-depleted cells, surmising that perturbations in the dynamics and/or function of endosomes and/or lysosomes cause the enhanced degradation of MHCI in AP-2-depleted cells. Consistent with this idea, Bonifacino and colleagues previously showed that depletion of clathrin and, to a lesser extent, μ2 altered the subcellular distribution of several lysosomal markers, including CD63, Lamp1 and Lamp2 (Janvier and Bonifacino, 2005). Vesicles harboring these markers had a swollen, clumped appearance, leading the authors to conclude that lysosome morphology and distribution were altered in cells that lack functional clathrin or AP-2. We observed a similar clustering of CD63-positive vesicles in cells depleted of α-adaptin (Fig. 7A), μ2-subunit (Fig. 7B) or clathrin (Fig. 7C). By comparison, CD63 vesicles were more uniformly dispersed in adjacent wild-type cells (Fig. 7A-C). Co-labeling with Tfn in the clathrin-depleted cells revealed that the CD63 clusters were in close proximity to but distinct from the Tfn clusters (Fig. 7C).

We predicted that the altered morphology and distribution of CD63 in AP-2-depleted cells might be accompanied by perturbations in endosomal and/or lysosomal trafficking, and that the enhanced degradation of surface-derived MHCI might result from increased delivery to a degradative compartment. If this were the case, one might expect to see increased localization of MHCI with late endosome and lysosomal markers. To test this, we examined whether the clusters of β1 integrin were positive for CD63 in AP-2- and clathrin-depleted cells. Some degree of β1 integrin colocalization with CD63 was observed in control cells (Fig. 7A-C; Fig. 7E shows a magnified view of the control cell labeled with asterisk in the merge panel of Fig. 7B). Strikingly, the level of β1 integrin that colocalized with CD63 was increased in both α-adaptin- and μ2-subunit-depleted cells (Fig. 7A,B). By contrast, whereas clathrin depletion also caused clustering of β1 integrin vesicles, it did not enhance its colocalization with CD63. Rather, it led to a decrease in colocalization with CD63 (Fig. 7C). Quantification of β1 integrin colocalization with CD63 is presented in Fig. 7D. Together, these results indicate that, whereas inactivation of either clathrin or AP-2 alters the distribution of late endosomes and lysosomes within the cell, loss of AP-2 but not clathrin enhances lysosomal targeting of Arf6-pathway cargo.

To further characterize the compartment where β1 integrin was clustered in AP-2-depleted cells, we tested the presence of additional endosomal and lysosomal markers. We first examined the late endosome and lysosome marker Lamp1. Because antibodies against Lamp1 and β1 integrin had been generated in the same species and were of the same IgG subtype, they could not be used simultaneously. As an alternative we co-stained for CD63 and Lamp1, and found that they completely colocalized at the clustered vesicles in AP-2-depleted cells (Fig. 8A). It can thus be inferred that the β1 integrin clusters are positive for Lamp1. By contrast, the early endosome marker EEA1, although also clustered in the perinuclear region, was largely excluded from the β1 integrin vesicles (Fig. 8B). Furthermore, β1 integrin and TfnR segregated to distinct aggregates in AP-2-depleted cells (Fig. 8C). Further contrasting these cargoes, we found that TfnR and CD63 did not significantly colocalize in AP-2-depleted cells. Whereas both clustered in the perinuclear region, CD63 decorated distinct vesicles and TfnR exhibited a hazier staining pattern (Fig. 8D). Thus, depletion of AP-2 appears to selectively enhance localization of Arf6 cargo but not clathrin cargo to late endosomal and lysosomal compartments.

Our work reveals that depletion of AP-2 has more widespread effects on endocytic trafficking than previously appreciated, affecting endocytic trafficking through the non-clathrin, Arf6-regulated pathway. Depletion of AP-2 has a slight effect on the initial rate of internalization, and also causes enhanced lysosomal targeting and degradation of Arf6-pathway cargo. Our biochemical studies indicate that the half-life of surface-derived MHCI molecules is reduced in cells that lack AP-2. Accordingly, total levels of MHCI are modestly but consistently reduced in AP-2-depleted cells, and can be restored with bafilomycin A1. Furthermore, colocalization of Arf6 cargo with late endosomal and lysosomal markers is enhanced in cells depleted of AP-2.

The decrease in the initial rate of MHCI internalization in AP-2-depleted cells suggests that AP-2 has a function independent to that of clathrin. Mechanistically, one could envision that AP-2 is recruited to regions of the plasma membrane enriched in PtdIns(4,5)P₂ generated through activation of phosphatidylinositol (4)-phosphate 5-kinase (PtdIns(4)P-5-kinase), an effector of Arf6 (Honda et al., 1999). AP-2 could function in this context to promote vesicle formation by recruiting proteins that induce membrane curvature, such as epsin (Chen et al., 1998; Ford et al., 2002), or BAR-domain-containing proteins, such as amphiphysin (David et al., 1996; Farsad et al., 2001; Peter et al., 2004; Slepnev et al., 2000). Consistent with this notion, previous studies have shown that Arf6 can bind to AP-2, and cooperate with phosphoinositides to promote AP-2 recruitment to liposomes (Paleotti et al., 2005). Alternatively, the reduced rate of MHCI internalization in AP-2-depleted cells could be indirect. Loss of AP-2 might generally perturb plasma membrane dynamics, for example by affecting PtdIns(4,5)P₂ availability and/or distribution.

Our observations extend recent reports indicating that depletion of AP-2 and clathrin has more pervasive effects on the endosomal system than initially recognized. Janvier and Bonifacino showed that depletion of clathrin and, to a lesser extent, AP-2 induced clustering of Lamp-containing vesicles and altered the morphology of lysosomes (Janvier and Bonifacino, 2005). However, their study did not examine whether the vesicle clusters were functionally distinct in the two depleted populations, or whether lysosome-mediated degradation was perturbed. Our studies indicate that this is in fact the case. Robinson and colleagues recently showed that in cells that express the human immunodeficiency virus (HIV) Nef protein, depletion of AP-2 enhanced delivery of MHCI to a pre-lysosomal compartment (Lubben et al., 2007). Morphometric analysis of transmission electron microscopic images indicated that the abundance of vacuoles and multivesicular bodies (MVBs) was not elevated in AP-2-depleted cells, suggesting that this enhancement instead arose from an alteration in cargo sorting. Notably, they observed enhanced MHCI degradation only upon simultaneous coexpression of Nef and depletion of AP-2; no increase in MHCI degradation was observed in cells depleted of AP-2 alone. This contrasts our results, which show that AP-2 depletion by itself perturbs degradative trafficking of MHCI. This discrepancy might be explained by the modest effects of AP-2 depletion on steady-state levels of MHCI. Also, different methods were utilized to monitor the half-life of MHCI: Lubben et al. used metabolic labeling with ³⁵S, whereas we specifically monitored the plasma-membrane-derived pool using surface biotinylation (Lubben et al., 2007). The molecular mechanism by which the depletion of Nef and AP-2 cooperatively induced MHCI degradation was not addressed, but the authors speculated that altered membrane composition and dynamics of endosomes caused increased delivery of MHCI to a pre-lysosomal compartment.

Several models of how AP-2 might regulate post-endocytic and degradative trafficking through the Arf6 endocytic pathway may be considered. It is most likely that AP-2 exerts its effects from the plasma membrane, because extensive studies identify this as its sole or at least predominant residence (Robinson, 1987; Robinson and Pearse, 1986; Traub et al., 1996). One possible mechanism by which AP-2 might function in the Arf6 pathway is suggested by recent work on its role in clathrin-dependent endocytosis (Lakadamyali et al., 2006). This study posited that initial sorting of clathrin-dependent cargoes occurs at the plasma membrane, prior to their delivery to early endosomes. It was suggested that early endosomes are not uniform, but comprise two distinct populations: one that is dynamic and matures rapidly to late endosomes and lysosomes, and another larger static population that matures much more slowly and is preferentially recycled. Knockdown of AP-2 dramatically reduced the size of this latter population, without significantly affecting the abundance of the rapidly maturing endosomes (Lakadamyali et al., 2006). It is possible that distinct early endosome populations similarly exist for the Arf6 pathway, with AP-2 selectively promoting the formation of a pool that is preferentially recycled. However, because markers of endosome maturation have not been defined for the Arf6 pathway as they have for the clathrin pathway, it is currently not technically feasible to test this hypothesis.

Another possible mechanism is that AP-2 functions from an intracellular location. Whereas the majority of studies identify the plasma membrane as the primary residence of AP-2, ultrastructural analysis indicated that a population may also exist at the lysosome (Traub et al., 1996). These authors suggested that AP-2 promotes retrograde trafficking from the lysosome. We thus considered the possibility that AP-2 localizes transiently at this site, but that detection may be obscured by its comparatively overwhelming concentration at clathrin-coated pits. In an attempt to unmask such a potential localization, we examined α-adaptin localization in cells depleted of clathrin or treated with chlorpromazine, a compound that induces mis-assembly of clathrin-coated pits and redistribution of AP-2 to intracellular vesicles (Subtil et al., 1994; Wang et al., 1993). However, we did not detect a convincing localization of α-adaptin at endomembranes under either of these conditions (data not shown).

Yet another possibility is that effects of AP-2 knockdown on Arf6 cargo are indirect, arising secondarily through effects on the clathrin pathway. As discussed in the Introduction, a subset of endosomes derived from the Arf6 pathway normally fuse with those from the clathrin pathway (Naslavsky et al., 2003). Depletion of the larger static pool of clathrin-derived endosomes in α-adaptin-depleted cells (as described in Lakadamyali et al., 2006) would leave the rapidly maturing pool of endosomes as the major one for Arf6 endocytic vesicles to fuse with, resulting in enhanced lysosomal degradation of Arf6 cargo. However, one conundrum with this model is that clathrin depletion would also be predicted to reduce the static endosome population, but MHCI degradation was not enhanced in cells lacking clathrin. This might be because clathrin is required for subsequent sorting from early endosomes to late endosomes, which would be inhibited in clathrin-depleted cells. Consistent with this possibility, we found that colocalization of MHCI with CD63 was reduced in clathrin-depleted cells (Fig. 7).

Regardless of the mechanism underlying enhanced lysosomal degradation of Arf6 cargo, our study indicates that depletion of AP-2 has much broader effects on trafficking than simply inhibiting internalization of a subset of clathrin cargo. Interestingly, it has recently been reported that another adaptor classically associated with CME, epsin, can also function in non-clathrin endocytic trafficking (Chen and De Camilli, 2005; Sigismund et al., 2005). These studies suggested that epsin can promote the internalization of cargo through caveolae (Sigismund et al., 2005), and that this function might be inhibited by its interaction with clathrin (Chen and De Camilli, 2005). Thus, endocytic regulators that have so far been viewed as being specific for the clathrin pathway might function more broadly in non-clathrin routes as well.

HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, penicillin, streptomycin, fungizone and GlutaMax (Gibco BRL). Cultures were maintained at 37°C in 5% CO₂. Cells were transfected using FuGENE6 (Roche), Lipofectamine 2000 or Oligofectamine (Invitrogen) as detailed below. HA-dynamin PH^*/pUHD was provided by Mark Lemmon (University of Pennsylvania), HA-Arf6WT/pLNCX was generously provided by Julie Donaldson (NIH).

Anti-MHCI (W6/32) was used for immunofluorescence and FACS analysis. Anti-β1-integrin (TS2/16) was used for immunofluorescence. Anti-α-adaptin (M-300), anti-HA (sc-805), and anti-EEA1 (N-19) were from Santa Cruz Biotechnologies, Inc. Anti-clathrin and anti-TfnR (CD71) antibodies were purchased from BD Biosciences. Anti-Lamp1 (H4A3) was from Developmental Studies Hybridoma Bank. Cy3-conjugated donkey anti-mouse IgG and fluorescein isothiocyanate (FITC)-conjugated donkey anti-rabbit IgG were purchased from Jackson ImmunoResearch. Alexa-Fluor-633 goat anti-rabbit IgG, Alexa-Fluor-488 goat anti-mouse IgG, Alexa-Fluor-546–Tfn, and Alexa-Fluor-488–Tfn were purchased from Molecular Probes.

siRNA oligonucleotides were purchased from Dharmacon, Inc. The α-adaptin-subunit target sequence was 5′-GAGCAUGUGCACGCUGGCCA-3′. The μ2-subunit target sequence was 5′-AAGUGGAUGCCUUUCGGGUCA-3′. The clathrin-heavy-chain target sequence was 5′-CCUGCGGUCUGGAGUCAAC-3′. The control siRNA used was an oligonucleotide targeted against firefly luciferase and had the sequence 5′-AACGUUACCGCGGAAUACUUCGA-3′. HeLa cells were subjected to two rounds of transfection with siRNA oligonucleotides by using Oligofectamine (Invitrogen) as to the manufacturer's instructions. After the second transfection, cells were reseeded and allowed to grow one additional day before their use in FACs analysis, immunofluorescence or immunoblotting.

Immunofluorescence was performed as previously described, except that saponin (0.1%) was used in lieu of Triton X-100 (Masuda-Robens et al., 2003). Briefly, cells were fixed with formaldehyde, permeabilized, incubated with primary antibodies for 2-3 hours, then washed. Samples were labeled with fluorescently labeled secondary antibodies for 1 hour, washed, then mounted with SloFade (Molecular Probes). For live antibody-uptake experiments, anti-MHCI or anti-β1-integrin (TS2/16) were added to the cell medium of living cells at 37°C for various times as detailed in the figure legends. Where indicated, cells were incubated with Alexa-Fluor-488- or Alexa-Fluor-546-conjugated Tfn for 1 hour on ice, then transferred to 37°C for 30 minutes to allow internalization. Cells were then fixed and processed as above. Samples were viewed on a Zeiss 510 laser scanning confocal microscope with a C-Apochromat 63×1.2Wcorr objective at excitation wavelengths of 488 nm (FITC), 546 nm (Cy3) or 633 nm (Cy5).

To measure rates of endocytosis, knockdown or control HeLa cells were seeded to confluence. Cells were released with PBS, 10 mM EDTA, resuspended in cold growth medium, and pelleted in a microfuge at 3000 rpm for 5 minutes at 4°C. Surface MHCI was labeled by incubating the resuspended pellet in anti-MHCI for 45 minutes on ice. Cells were washed three times in cold medium and then resuspended in cold medium. Aliquots were dispensed into Eppendorf tubes, then incubated at 37°C to allow internalization. At the indicated times, samples were transferred to an ice bath to halt trafficking. To detect anti-MHCI remaining at the cell surface, FITC-conjugated anti-mouse IgG was added, and samples were incubated for an additional 45 minutes on ice. Samples were then washed twice and fixed in PBS containing 1% formaldehyde. Cells were washed twice in FACs buffer, then twice in sheath fluid (Fisher). Data were processed using CellQuest Pro software (BD Biosciences).

To measure steady-state levels of MHCI and TfnR, cells were collected as above. To measure their cell-surface levels, samples were resuspended in FITC-conjugated antibody diluted in growth medium and incubated for 45 minutes at 4°C. Anti-MHCI was from Biodesign International; anti-TfnR was from BD Pharmingen. Cells were washed three times in FACs buffer, then fixed, washed and stored as above. Total levels of MHCI and TfnR were measured as follows: samples were immediately fixed in 1% formaldehyde/PBS, then washed twice in growth medium containing 2% saponin before resuspension in medium containing FITC-conjugated antibody. After 45 minutes incubation at 4°C, cells were washed twice in FACs buffer containing 2% saponin, then twice in sheath fluid.

To monitor internal MHCI pools, anti-MHCI was labeled with Alexa-Fluor-488 (Molecular Probes) per manufacturer instructions. Cells were incubated with anti-MHCI conjugated to Alexa-Fluor-488 for 10 minutes on ice, followed by incubation for 2 hours at 37°C. Samples were washed twice in cold medium, and anti-MHCI conjugated to Alexa-Fluor-488 that had remained at the cell surface was quenched by incubating cells for 30 minutes on ice in medium containing anti-Alexa-Fluor-488 (Molecular Probes) or in anti-cyclin-E antibody as a negative control. Internal anti-MHCI conjugated to Alexa-Fluor-488 was then quantified by flow cytometry.

To quantify MHCI levels, cells were lysed in 50 mM Tris pH 7.5, 100 mM NaCl, 2 mM MgCl₂, 0.1% sodium dodecyl sulfate (SDS), 0.5% sodium deoxycholate, 1% Triton X-100, 10% glycerol containing leupeptin, pepstatin and phenylmethysulfonyl fluoride. Lysates were incubated on ice for 10 minutes, then microcentrifuged at 14,000 rpm for 10 minutes at 4°C. The clarified supernatant was immunoblotted with antibody against MHCI (3B10.7) and either antibody against p70S6k, MAPK or actin as loading controls, then analysed using enhanced chemiluminescence (Amersham). Protein levels were determined by measuring band intensity using ImageJ software (NIH), and MHCI values were normalized against the loading controls for each individual experiment (n=7). MHCI levels in control siRNA cells were designated as 100%, and levels in α-adaptin knockdown cells were calculated relative to this value. Bafilomycin-A1 (Sigma) was used at 250 nM overnight.

siRNA cells were prepared as described above and seeded at 2.5×10⁵ cells per 35-mm well. Cells were allowed to grow for one additional day before the assay was performed. Cells were washed three times with PBS-CM (PBS, 0.1 mM CaCl₂, 2 mM MgCl₂), incubated on ice for 20 minutes in PBS-CM containing 0.5 mg/ml NHS-SS-biotin (Pierce), washed twice with ice-cold PBS-CM and residual biotin was quenched by incubating on ice with freshly made PBS-CM, 100 mM glycine for 10 minutes. After quenching, cells were washed twice with ice-cold PBS-CM, then once with pre-warmed growth medium. Samples were lysed immediately (t=0), or following incubation at 37° C for 4 hours (t=4 hours). Samples were lysed in 500 μl of RIPA buffer (20 mM Tris pH 7.5, 150 mM NaCl, 2 mM EDTA, 5 mM β-glycerophosphate, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitors. Biotinylated proteins were immobilized using NeutraAvidin Protein (Pierce) for 1.5 hours at 4°C. The beads were washed twice in RIPA buffer. Proteins were eluted by boiling in sample buffer, fractionated by SDS-PAGE, then immunoblotted using the indicated antibodies. MHCI levels were quantified by measuring band intensities using ImageJ software. The amount of MHCI that had remained after 4 hours was normalized against the amount at t=0 for each respective siRNA cell population. Data represent the average of at least three independent experiments, each performed in duplicate.

siRNA cells were allowed to take up anti-β1-integrin antibody for 4 hours. For the clathrin siRNA cells, Alexa-Fluor-546-conjugated Tfn was added during the final 30 minutes of incubation to identify the clathrin-depleted cells. Cells were fixed and subjected to indirect immunofluorescence using anti-CD63 (hybridoma 439) antibody. At least two fields were taken from two independent experiments, and used for the quantification of each siRNA cell line. From each of these fields, at least nine knockdown cells were chosen for analysis. Using the Metamorph `Count Nuclei' function, β1-integrin- and CD63-positive spots were determined based on predetermined threshold intensity as well as diameter. A binary mask was then made for each spot, and unwanted pixels were masked. On average, 25 β1 integrin spots and 25 CD63 spots were found per knockdown cell after unwanted pixels were masked (thus, ∼200-250 positive spots for each field containing 9-11 cells were analyzed). Using the Metamorph `Measure Colocalization' function, the area of β1 integrin spots that overlapped with CD63 spots was then determined in percent. Owing to variability in staining from experiment to experiment, the effects of clathrin or AP-2 depletion on colocalization of β1 integrin with CD63 was determined by comparing knockdown cells and WT cells within the same field. The average area of β1 integrin overlap with CD63 in wild-type cells was defined as 100%, and the degree of overlap in knockdown cells was calculated relative to this value.

This work was supported by the National Institutes of Health, National Cancer Institute (Grant CA81415 to M.M.C.). We thank Michael S. Marks and Christopher G. Burd for valuable scientific discussions, and the critical evaluation of this manuscript. We are also grateful to Subba Rao Gangi Setty and Kate Herman for excellent technical advice.