Flat clathrin coats on endosomes mediate degradative protein sorting by scaffolding Hrs in dynamic microdomains

Endocytosis of membrane proteins is important for nutritional uptake, immune functions and modulation of signalling responses (Conner and Schmid, 2003). Most endocytosed membrane proteins are either recycled to the plasma membrane or transported to lysosomes for degradation. Ubiquitylation of membrane proteins serves as sorting signal for entry into intraluminal vesicles of multivesicular endosomes (MVEs) (Hicke and Dunn, 2003; Dupre et al., 2001; Reggiori and Pelham, 2001). When the MVEs fuse with late endosomes or lysosomes, the intraluminal vesicles and their content are degraded (Katzmann et al., 2002; Raiborg et al., 2003; Gruenberg and Stenmark, 2004). The ubiquitin-binding protein Hrs (hepatocyte-growth-factor-regulated tyrosine kinase substrate), which is recruited to endosomes through interaction with the endosomal lipid phosphatidylinositol (3)-phosphate [PtdIns(3)P] (Gaullier et al., 1998; Raiborg et al., 2001c), mediates the efficient sorting of ubiquitylated membrane proteins into MVEs (Raiborg et al., 2002; Katzmann et al., 2002; Raiborg et al., 2003). Besides interacting with cargo and PtdIns(3)P, Hrs is responsible for the endosomal recruitment of several cytosolic proteins that cooperate in cargo binding and MVE biogenesis. The Hrs-mediated recruitment of endosomal sorting complex required for transport (ESCRT)-I to endosomes is essential for proper degradation of ubiquitylated membrane proteins (Bache et al., 2003a; Pornillos et al., 2003; Katzmann et al., 2003; Bilodeau et al., 2003; Lu et al., 2003). Likewise, the signal-transducing adaptor molecule STAM, which is in a tight complex with Hrs, binds ubiquitin and is important for receptor degradation (Takata et al., 2000; Bilodeau et al., 2002; Bilodeau et al., 2003). Hrs can also interact with the terminal domain (TD) of clathrin heavy chain (Raiborg et al., 2001a), but the functional significance of this interaction has not been investigated.

Clathrin is a coat protein with well-known functions in the secretory and endocytic pathways (Brodsky et al., 2001; Kirchhausen, 2000), mediating the sorting of membrane proteins into coated vesicles that bud from the trans-Golgi network, the plasma membrane and endosomal tubules. In addition to the canonical role of clathrin in the biogenesis of clathrin-coated transport vesicles, several reports have described the presence of a flat clathrin coat of uncharacterised function on the vacuolar parts of endosomes (Raposo et al., 2001; Sachse et al., 2002; Raiborg et al., 2002). Since this clathrin coat contains Hrs and ubiquitylated cargo, it has been hypothesised that it is involved in Hrs-mediated degradative protein sorting (Sachse et al., 2002; Raiborg et al., 2002; Clague, 2002). Alternatively, the Hrs-containing clathrin coat could be involved in Hrs-mediated recycling of non-ubiquitylated cargo to the plasma membrane (Hanyaloglu et al., 2005; Yan et al., 2005), or it could serve to inhibit fusion with other endosomes (Sun et al., 2003). In either of these models, the exact role of clathrin is not clear. Clathrin could either function as a scaffolding protein that restricts Hrs and, thereby, ubiquitylated cargo in microdomains on the endosome membranes (Sachse et al., 2002; Raiborg et al., 2002; Clague, 2002), serve as a mechanical barrier to inhibit membrane fusion (Sun et al., 2003), act as an allosteric regulator of Hrs or other proteins involved in the sorting process (Jackson et al., 2003), or mediate budding of vesicles from vacuolar endosomes (although such buds have not been detected by electron microscopy). It thus seems pertinent to design experiments that can address the function of clathrin on vacuolar endosomes.

Here, we have tested the hypothesis that clathrin is required for the function of Hrs in degradative protein sorting. We show that clathrin recruitment to endosomes by Hrs is needed for efficient degradation of endocytosed epidermal growth factor (EGF), whose receptor is a prototypic example of a ubiquitylated membrane protein destined for lysosomes (Haglund et al., 2003; Mosesson et al., 2003). Further, we find that the function of clathrin in degradative sorting depends on its ability to scaffold Hrs in dynamic microdomains on the endosome membrane, to which ESCRT-I is recruited. These findings are consistent with the idea that clathrin functions to concentrate Hrs in restricted microdomains on the endosomal membrane, thus facilitating sorting of ubiquitylated cargo.

We have shown previously that the C-terminus of Hrs can bind specifically to clathrin via its C-terminal clathrin-box motif (Raiborg et al., 2001a) (Fig. 1A). Likewise, we found that full-length wild-type Hrs (Hrs_wt) but not a mutant that lacks the 5-residue clathrin-box motif (Hrs_ΔCB) can bind to immobilised clathrin TD (Fig. 1B), and that this does not require tyrosine phosphorylation of Hrs (Fig. 1C). To examine whether Hrs is required for efficient recruitment of clathrin to vacuolar endosomes, we studied the presence of endosomal clathrin in cells whose Hrs had been knocked down by Hrs-specific small interfering RNA (siRNA) oligonucleotides. In siRNA-treated cells, a significant reduction in Hrs levels was observed, as expected (Bache et al., 2003b) (Fig. 1D). Whereas clathrin could be easily detected on EEA1-positive endosomes in Hrs-containing control cells (Fig. 2A, inset), it was almost absent from endosomes in Hrs-depleted cells (Fig. 2B,C, Hrs-depleted cells are indicated by arrows, insets with dashed lines), although a fine punctate clathrin staining, corresponding to clathrin-coated pits and vesicles, could be detected both in the Hrs-depleted cells and in control cells. This provides direct evidence that Hrs is required for efficient recruitment of clathrin to endosomal vacuoles. To assess the importance of the Hrs clathrin-box motif in clathrin recruitment to endosomes, we included rescue transfections with myc-Hrs_wt and myc-Hrs_ΔCB. In Hrs-depleted cells transfected with siRNA-resistant myc-Hrs_wt at close to endogenous level (Fig. 2B, arrowheads), clathrin could be detected on endosomes (Fig. 2B, inset with solid lines). By contrast, when Hrs-depleted cells were transfected with myc-Hrs_ΔCB (Fig. 2C, arrowhead), the endosomes were still devoid of clathrin (Fig. 2C, inset with solid lines). This indicates that the clathrin-box motif is responsible for the ability of Hrs to recruit clathrin to endosomes. Since Hrs associates directly with endosomal membranes (Raiborg et al., 2001c) and is necessary (this study) and sufficient (Raiborg et al., 2001a) for efficient clathrin recruitment to vacuolar endosomes, it may be regarded as a clathrin adaptor on such endosomes.

Previous studies have revealed that Hrs is accumulated in restricted microdomains on the endosome membrane, strongly colocalising with clathrin (Raiborg et al., 2002; Raiborg et al., 2001a; Sachse et al., 2002). By contrast, EEA1-containing regions of the endosome membrane, which are involved in endosome fusion (Zerial and McBride, 2001), are largely devoid of clathrin and Hrs (Raiborg et al., 2001a; Sachse et al., 2002). To investigate whether clathrin is essential for the localisation of Hrs in restricted microdomains, we set out to study the relative localisations of Hrs and EEA1 in cells whose endogenous clathrin had been depleted by plasmid-based siRNA. To study endosomal microdomains, the clathrin-depleted cells were transfected with a constitutively active form of the early-endosomal GTPase Rab5 (Rab5^Q79L), to increase the fusion rate of endosomes (Stenmark et al., 1994). This gives rise to enlarged early endosomes on which microdomains can be easily distinguished by confocal immunofluorescence microscopy (Raiborg et al., 2002; Raiborg et al., 2001a; Pelkmans et al., 2004). In agreement with previous studies (Gillooly et al., 2003; Raiborg et al., 2001a; Sachse et al., 2002), there was little colocalisation of Hrs and EEA1 on control Rab5^Q79L-enlarged endosomes, whereas Hrs colocalised strongly with clathrin (Fig. 3A,C). In cells that were virtually devoid of clathrin - as judged by immunostaining with anti-clathrin antibodies - similarly enlarged endosomes as in control cells could be observed (Fig. 3B), indicating that endosome biogenesis does not depend on clathrin-mediated endocytosis. In such clathrin-depleted cells, both Hrs and EEA1 could still be detected on endosomes. Interestingly, however, clathrin-depleted endosomes showed a strongly increased colocalisation of Hrs and EEA1 (Fig. 3B,C). This indicates that depletion of endosomal clathrin by siRNA leads to a mixing of Hrs and EEA1-positive regions, possibly due to lateral diffusion of Hrs in the endosome membrane.

The finding that clathrin is required for the restricted localisation of Hrs on endosomes raised the question whether the ability of Hrs to bind clathrin is essential for its distribution into microdomains. To address this, Hrs-depleted and Rab5^Q79L-transfected cells were co-transfected with either myc-Hrs_wt or myc-Hrs_ΔCB and studied by confocal microscopy with antibodies against EEA1 and the myc-tagged construct. Strikingly, in Rab5^Q79L-transfected cells, the clathrin-binding-deficient myc-Hrs_ΔCB mutant showed a strong colocalisation with EEA1 on endosomes (Fig. 3E,F) in contrast to myc-Hrs_wt, which showed as little colocalisation with EEA1 as endogenous Hrs (Fig. 3D,F). Taken together, these data indicate that endosomal clathrin, by binding to the clathrin-box of Hrs, serves a scaffolding function on the endosomal membrane, restricting Hrs in microdomains devoid of EEA1.

Since clathrin seems to function as a scaffold to restrict Hrs to distinct microdomains, we asked whether clathrin binding is sufficient to redirect an otherwise non-colocalizing endosomal protein into these microdomains. To address this question, we transfected HeLa cells with a chimeric protein consisting of a tandem FYVE-domain construct (2xFYVE), which has been employed extensively as a probe for PtdIns(3)P (Gillooly et al., 2000), recombinantly fused to the C-terminal part of Hrs that includes the clathrin-box motif (2xFYVE-Hrs CT) (Fig. 4A). At low expression levels, this chimera localised to endosomes through binding of endosomal PtdIns(3)P and led to a recruitment of clathrin to endosomes (not shown). Previous studies have shown that the C-terminal part of Hrs alone is cytosolic (Raiborg et al., 2001c). PtdIns(3)P is enriched in EEA1-positive regions of the endosome membrane, and the 2xFYVE probe has been found to colocalise strongly with EEA1 (Gillooly et al., 2003). By contrast, the clathrin-binding 2xFYVE-Hrs CT probe did not colocalise with EEA1 (Fig. 4B, quantitated in Fig. 4F), but was instead found within yellow fluorescence protein (YFP)-Hrs-positive microdomains on Rab5^Q79L-enlarged endosomes (Fig. 4C, quantitated in Fig. 4F). A control construct consisting of the 2xFYVE probe fused to the C-terminus of Hrs in which the clathrin-box motif had been deleted (2xFYVE-Hrs CT_ΔCB) (Fig. 4A), behaved as a regular 2xFYVE probe, strongly colocalising with EEA1 (Fig. 4D, quantitated in Fig. 4F), but not with YFP-Hrs (Fig. 4E, quantitated in Fig. 4F). Taken together, these findings suggest that clathrin binding is sufficient to sequester endosomal proteins in restricted microdomains.

If the Hrs/clathrin microdomains were functional sorting platforms, one would expect the sorting components to be dynamic in order to relay the ubiquitylated cargo to downstream components of the sorting machinery, such as ESCRT complexes. Fluorescence recovery after photobleaching (FRAP)-experiments in live cells have shown previously that endosomal GFP-clathrin light chain (LC) exchanges with a cytosolic pool (Pelkmans et al., 2004). To study whether endosome associated Hrs shows similar dynamics, we performed live cell confocal microscopy of HeLa cells that had been co-transfected with Rab5^Q79L and either GFP-clathrin LC or YFP-Hrs. Both clathrin LC and Hrs were found in microdomains on the endosome membranes in live cells. A 3D-reconstruction of the enlarged endosomes showed that the Hrs microdomains were spread throughout the perimeter of the spherically shaped endosomes (not shown). There seemed to be some lateral movement of the microdomains in the membrane, although fluorescence loss in photobleaching (FLIP)-analysis indicated that there was very little, if any lateral diffusion of Hrs within the endosome membrane (not shown). This indicates that the microdomains are separate entities. When whole endosomes were photobleached, FRAP-analysis showed that both Hrs and clathrin were rapidly recruited from the cytosol onto a nearly identical microdomain pattern (Fig. 5A), indicating that each microdomain exchanges its sorting components with a cytosolic pool. The fluorescence recovered uniformly along the whole length of each microdomain, rather than from the edges. This indicates that the microdomains do not grow from the edges, but rather the whole microdomain represents a dynamic pool of Hrs and clathrin that undergoes signficant exchange with cytosolic pools. The recovery curves were corrected for bleaching caused by continued imaging, and the data were fit into a diffusion model (Pelkmans et al., 2004; Yguerabide et al., 1982) to determine the kinetics. The half-time of the recovery after photobleaching was approximately 17-19 seconds for both proteins (Fig. 5B,C) and the mobile fraction of both Hrs and clathrin in the microdomains were calculated to be between 80 and 90 percent (see Materials and Methods). The finding that Hrs and clathrin showed similar dynamics is consistent with the idea that Hrs recruits clathrin to these structures, and suggests that the kinetics of clathrin exchange are mainly determined by the exchange rate of Hrs.

The sorting complex ESCRT-I has been proposed to operate downstream of Hrs in the endosomal sorting pathway. Previous reports have shown that Hrs facilitates the recruitment of the ESCRT-I-component Tsg101 to endosomes, although its precise localisation has not been determined (Bache et al., 2003a; Pornillos et al., 2003; Katzmann et al., 2003; Bilodeau et al., 2003; Lu et al., 2003). A dynamic exchange of endosomal Hrs and clathrin with a cytosolic pool could facilitate the recruitment of Tsg101 into the coat. However, whereas Hrs localises to early endosomes, the majority of Tsg101 is found on late endosomes (Bache et al., 2003a). We therefore wanted to determine whether Tsg101 can be found in the Hrs microdomains and performed immunofluorescence confocal microscopy on Rab5^Q79L-transfected HEp-2 cells that were labelled for endogenous Tsg101 and Hrs or EEA1. Compared with EEA1 and Hrs, which are enriched on endosome membranes, Tsg101 was less abundant on the limiting membrane of the enlarged endosomes and was frequently found in restricted microdomains that were devoid of both Hrs and EEA1 (Fig. 6A). We could, however, detect localisation of Tsg101 to some (albeit not all) Hrs microdomains. Because Tsg101 tends to aggregate when overexpressed (Bache et al., 2003a), we were not able to perform FRAP analysis using GFP-Tsg101. Quantifications of single Rab5^Q79L-endosomes showed that Tsg101 was enriched in the Hrs microdomains compared with EEA1-positive regions (Fig. 6A,B, notice that these quantifications underestimate the colocalisation with Hrs compared with EEA1, because the Hrs microdomains are smaller). The minor pool of Tsg101 that colocalises with EEA1 could represent Tsg101 that remains associated with endosome membranes by binding to ubiquitylated EGFRs.

Having established a role for clathrin in controlling the localisation of Hrs on the endosome membrane, we next asked whether clathrin recruitment is needed to mediate the function of Hrs in degradative protein sorting. As a marker for endosomal sorting and lysosomal degradation, we used fluorescently labelled EGF, which is endocytosed and efficiently degraded together with its receptor in lysosomes. In agreement with previous studies of the EGF receptor (Bache et al., 2003b), the degradation of endocytosed EGF was significantly inhibited by siRNA-mediated depletion of Hrs, and increased amounts of the EGF receptor could be detected at the plasma membrane (not shown). The Hrs knockdown was efficient in virtually all cells as judged by immunofluorescence confocal microscopy (not shown), and by western blotting (Fig. 1D shows a representative siRNA-experiment). Since high-level overexpression of Hrs inhibits EGFR trafficking (Raiborg et al., 2001a; Petiot et al., 2003), we employed conditions of low-level Hrs expression for rescue experiments (see Materials and Methods). When Hrs-depleted cells were transfected with siRNA-resistant myc-Hrs_wt at close to endogenous level, the degradation was rescued (Fig. 7A, arrowheads), with the same amount of internalised EGF remaining after the chase period as in cells not depleted of Hrs (Fig. 7C). By contrast, transfection of siRNA-treated cells with siRNA-resistant myc-Hrs_ΔCB at a similar level did not rescue the EGF degradation (Fig. 7B, arrowheads; Fig. 7C). Instead, EGF was detected in Hrs_ΔCB-positive endosomes (Fig. 7B, arrows), suggesting that transport through early endosomes was delayed. This demonstrates that the clathrin-box motif is important for the function of Hrs in degradative protein sorting. Control experiments (not shown) demonstrated that similar amounts of EGF were found in Hrs-depleted cells and in cells rescued with Hrs_wt or Hrs_ΔCB after a 15-minute internalisation, indicating that Hrs and its clathrin-box are not essential for endocytosis of EGF. Taken together, these results indicate that clathrin recruitment and thereby correct localisation of Hrs within microdomains on the endosome membrane is required for efficient sorting of EGF towards lysosomes for degradation. It is worth noticing that we did not observe any reduction in the amount of Hrs labelling on endosomes in clathrin-depleted cells, and that myc-Hrs_ΔCB was as strongly expressed on endosomes as myc-Hrs_wt (data not shown). This indicates that loss of endosomal clathrin does not cause instability of the endosomal pool of Hrs, and that the observed requirement for clathrin reflects its direct involvement in the function of Hrs.

In this report we show that a reciprocal relationship exists between Hrs and clathrin on endosomal membranes, which ensures efficient degradation of internalised EGF. On the one hand, Hrs is required for efficient recruitment of clathrin to vacuolar endosomes. On the other hand, clathrin is required for clustering of endosomal Hrs into restricted microdomains that facilitate cargo sorting. The fact that Hrs homologues in yeast, nematodes and flies also contain putative clathrin-box motifs argues for a conserved role for clathrin in Hrs-mediated protein sorting. Deletion of a putative clathrin-box motif in Vps27, the yeast homologue of Hrs, did not prevent the intravacuolar accumulation of an ubiquitylated reporter protein at steady-state, but it cannot be excluded that transport kinetics are affected in this mutant (Katzmann et al., 2003). Our results are consistent with the hypothesis that clathrin facilitates degradative protein sorting by concentrating Hrs in restricted microdomains on the endosome membrane (Sachse et al., 2002; Raiborg et al., 2002). This could enable efficient sorting of ubiquitylated membrane proteins through low-affinity interactions (Katzmann et al., 2002; Raiborg et al., 2003).

Although we have shown here that Hrs is important for the recruitment of the coat protein clathrin to early endosomes, additional proteins also contribute to endosomal clathrin recruitment. For instance, the AP-1 complex is thought to be involved in the recruitment of clathrin to buds of tubular recycling endosomes (Stoorvogel et al., 1996; Robinson, 2004; Deneka et al., 2003). In addition, recent investigations suggest that there are additional endosomal proteins with similar functions to Hrs. The GGA3 protein is located to the trans-Golgi network and early endosomes, and interacts with clathrin and ubiquitin as well as with ESCRT-I (Puertollano and Bonifacino, 2004). Likewise, the endocytic proteins Tollip and Tom1 form a complex that is recruited to endosomes by the FYVE-domain protein Endofin, and interact with ubiquitin and clathrin (Yamakami et al., 2003; Seet et al., 2004; Katoh et al., 2004). The functional relationship between Hrs and these proteins has not been investigated, but because depletion of Hrs by siRNA led to a strong depletion of clathrin from early endosomes (Fig. 2), Hrs probably is a major contributor to the early-endosomal recruitment of clathrin. We noticed that GGA3 levels were unaffected in Hrs-depleted cells (not shown), indicating that the observed effects were not indirect through destabilisation of GGA3. However, STAM, which localises to endosomes through its tight association with Hrs (Bache et al., 2003b), was recently found to bind clathrin (McCullough et al., 2006). It is therefore possible that both subunits of the Hrs-STAM complex contribute to the recruitment of clathrin to vacuolar endosomes. The finding that Hrs_ΔCB, which is able to bind STAM but unable to bind clathrin, did not recruit clathrin to endosomes, argues that Hrs is the most important of these two subunits with respect to clathrin recruitment.

Because of the paucity of antibodies that yield sufficiently strong labelling by electron microscopy, we have studied the relative localisations of Hrs, EEA1 and Tsg101 by confocal microscopy in cells expressing Rab5^Q79L, which expands the size of EEA1-containing endosomes. Even though this Rab5 mutant has been reported to delay endosomal maturation (Rink et al., 2005), the relative localisations of the sorting machineries are probably preserved on the Rab5^Q79L-expressing endosomes (Pelkmans et al., 2004). Indeed, the localisation of Hrs and EEA1 to distinct microdomains was originally established in Rab5^Q79L-expressing endosomes (Raiborg et al., 2001a) and was later confirmed by electron microscopy of untransfected cells (Sachse et al., 2002). It is worth noticing though, that there was an enrichment of Tsg101 on Rab5^Q79L-expressing endosomes compared with EEA1-containing endosomes from untransfected cells (not shown). This is consistent with the possibility that ESCRT-I recruitment and dissociation are associated with endosomal maturation (Katzmann et al., 2002; Gruenberg and Stenmark, 2004). The detected localisation of Tsg101 to some of the Hrs-containing microdomains is thus unlikely to be an artifact of the Rab5^Q79L expression and is in agreement with the fact that untransfected cells contain a pool of endosomes that contain both Hrs and Tsg101 (Bache et al., 2003a).

Electron micrographs of Hrs/clathrin-containing microdomains indicate that the endosomal clathrin coat is differently organised than the clathrin-coated pits at the plasma membrane. Compared with clathrin-coated pits, which are curved and contain a one-layered, well-organised coat, the Hrs-containing clathrin coat is flat and bi-layered, with an electron-dense inner layer and a fuzzier outer layer (Sachse et al., 2002; Raiborg et al., 2002). It will be interesting in the future to study the structure of this coat in more detail. During their lifetime, the clathrin-coated pits exchange their clathrin with a cytosolic pool (Wu et al., 2001). Our FRAP experiments in live cells show similar kinetics for endosomal clathrin to those reported for the plasma-membrane-associated clathrin, and are in agreement with the kinetics analysis of endosomal clathrin recovery measured by others (Pelkmans et al., 2004). Importantly, our live cell imaging and FRAP analyses show that, similar to clathrin, endosome-associated Hrs also dynamically exchanges with a cytosolic pool. Hrs and clathrin show very similar kinetics, supporting the idea of Hrs as an endosomal clathrin adaptor.

It has been suggested that clathrin exchange at the plasma membrane is involved in the transformation of hexagons to pentagons, which will introduce curvature (Wu et al., 2001). Alternatively, such exchange could be involved in error correction to ensure the correct insertion of hexagons and pentagons into the curved lattice (Wu et al., 2001). Since the endosomal Hrs/clathrin coat always appears flat, it is more likely that the dynamic behaviour is important for its function in cargo sorting, possibly to facilitate transfer of cargo to downstream sorting components such as the ESCRT proteins. One possible scenario is that dissociation of Hrs and clathrin generate space for the recruitment of ESCRT-I proteins within the Hrs/clathrin coat itself, rather than at the edges. This would ensure efficient transfer of cargo from Hrs to Tsg101 as the endosomes mature, and prior to invagination of the endosomal membrane, which is likely to happen at the edges of the coat. Consistent with this model, we could detect Tsg101 within a subset of the Hrs microdomains.

The present results indicate that clathrin and the clathrin-binding ability of Hrs are important to restrict Hrs in microdomains that are negative for EEA1. The reason why EEA1 is not evenly distributed along the endosome membrane is not clear, although interaction with the GTP-bound form of Rab5 could serve to limit its localisation (Simonsen et al., 1998). In clathrin-depleted cells, Hrs was found within EEA1-positive regions rather than being evenly distributed on the surface of the endosome membrane. We cannot rule out that the remaining Hrs microdomains were maintained through incomplete knockdown of clathrin. However, because both EEA1 and Hrs contain a FYVE domain that binds specifically to the endosomal lipid PtdIns(3)P and is required for their recruitment to endosomes (Gillooly et al., 2001), it is reasonable to assume that the localisation of PtdIns(3)P becomes decisive for the localisation of Hrs under conditions where the binding of Hrs to clathrin is impaired. Indeed, EEA1-positive regions of early-endosomal membranes are enriched in PtdIns(3)P (Gillooly et al., 2003) and may thus explain the increased colocalisation of Hrs and EEA1 in clathrin-depleted cells. The finding that a PtdIns(3)P-binding 2xFYVE probe was redirected from EEA1-containing regions towards the Hrs/clathrin microdomains when fused to a clathrin-binding sequence indicates that clathrin-binding proteins can cluster together on the endosome membrane. This enforces our conclusion that clathrin operates as an endosomal scaffold, restricting the localisation of Hrs and its binding partners. The restricted localisation of Hrs, caused by clathrin binding, could be important for keeping the sorting machinery separate from the EEA1-dependent fusion machinery, thereby enabling efficient sorting of ubiquitylated cargo.

Consistent with this model, we found that clathrin recruitment by Hrs is important for the Hrs-mediated degradation of internalised EGF. In myc-Hrs_ΔCB expressing cells, the remaining EGF accumulated mainly in early endosomal compartments. Since EGF is known to be associated with its receptor along the degradative endocytic pathway, this could be due to the retention of ubiquitylated EGF-receptor and its ligand within a Hrs-containing sorting machinery capable of ubiquitin binding but incapable of binding clathrin. Deleting the clathrin-box motif, which is only five amino acid residues in the extreme C-terminus of Hrs, does not change the ability of Hrs to bind any of its known interactors other than clathrin, such as STAM, Eps 15 or Tsg101 (Raiborg et al., 2001b; Bache et al., 2003a), and is thus a specific means of studying the effect of clathrin binding without interfering with other interactions.

Even though we did not detect any difference in clathrin binding in serum-starved and EGF-stimulated cells, we cannot rule out that phosphorylation of Hrs by activated growth factor receptors can regulate the affinity for clathrin further. However, the observation that Hrs does not need to be phosphorylated to bind clathrin (Fig. 1C) fits with the idea of Hrs as a key sorting component that not only recognises activated growth factor receptors but also cell adhesion molecules (Palacios et al., 2005), misfolded membrane proteins (Sharma et al., 2004) and other types of receptors that do not induce Hrs phosphorylation (Marchese et al., 2003). Thus, clathrin is probably important for the Hrs-mediated degradative sorting of different types of cargo. It will be interesting to study in future experiments whether clathrin is also required for Hrs-mediated recycling.

Based on the above findings and the current literature, we propose the following model for the function of Hrs and clathrin in endosomal protein sorting (Fig. 8): Ubiquitylated cargo is initially retained in restricted Hrs/clathrin coats on endosomes and prevented from being recycled. The dynamic nature of the coat, exchanging with a cytosolic pool, allows recruitment of ESCRT-I through transient openings within the coat. Clathrin generates a scaffold that concentrates ubiquitylated cargo and sorting machinery and facilitates efficient transfer of cargo from Hrs to ESCRT-I within the coat. As the endosome matures, Hrs and clathrin dissociate (Sachse et al., 2002; Raiborg et al., 2001a), and ESCRT-I remains on the late endosome membrane associated with ubiquitylated cargo. Subsequently, other ESCRTs are recruited and mediate invagination of the endosome membrane and further sorting of cargo into MVEs. Although further experiments will be required to prove or disprove this model, the present work provides, to our knowledge, the first experimental evidence that clathrin is important for a trafficking pathway that does not involve formation of coated carriers.

Anti-myc antibodies were from 9E10 hybridoma cells. Human anti-EEA1 serum (Mu et al., 1995) was a gift from Ban-Hock Toh, Melbourne, Australia. Affinity-purified antibodies against recombinant Hrs have been described before (Raiborg et al., 2001a). Mouse anti-clathrin heavy chain (HC) was from Acris Antibodies GmbH (Hiddenhausen, Germany). Rhodamine-, Cy2-, Cy3- Cy5- and horseradish peroxidase-labelled secondary antibodies were from Jackson Immunoresearch Inc. (West Grove, PE). The anti-phosphorylated-tyrosine antibody was from BD Transduction Laboratories, San Jose, CA. The mouse anti-Tsg101 antibody was from GeneTex (San Antonio, TX)

The pGEX-2T-clathrin-TD and GFP-clathrin LC constructs were a gift from James Keen (Kimmel Cancer Institute, Philadelphia, PA). The pcDNA3-myc-Hrs constructs indicated were generated by PCR with mouse Hrs as the template. The pcDNA3-myc-2xFYVE-Hrs CT constructs were generated by PCR with the FYVE domain of Hrs as a template. All PCR-amplified DNAs were verified by sequencing. The YFP-Hrs construct was generated by subcloning mouse Hrs into pEYFP-C1 (Clontech, Mountain View, CA). The pcDNA3-Rab5^Q79L construct has been described previously (Stenmark et al., 1994). The siRNA specific for human Hrs (synthesised by Dharmacon Research, Inc, Lafayette, CO) has been described previously (Bache et al., 2003b). A scrambled sequence of the Hrs-specific RNA duplex was used as a negative control. In rescue experiments, constructs based on siRNA-resistant (mouse-derived) Hrs were used in transfections. The vector for plasmid-based siRNA against clathrin has been described previously (Grimmer et al., 2005).

HeLa or HEp-2 cells were cultured as recommended by ATCC. pcDNA3 constructs and the plasmid-based clathrin-siRNA construct were expressed in cells using FuGENE 6™ (Roche Diagnostics, Basel, Switzerland) or Effectene™ (Qiagen, Inc. Valencia, CA) transfection reagents according to the instructions from the manufacturers. To obtain enlarged endosomes, the pcDNA3-Rab5^Q79L construct was expressed for 48 hours. For the knock-down of clathrin, the cells were co-transfected with the plasmid-based siRNA and pcDNA3-Rab5^Q79L for 72 hours. To reintroduce murine Hrs at close to endogenous level in Hrs-depleted cells, the pcDNA3myc-Hrs constructs were expressed for 17 hours using 20 ng plasmid DNA per well (24-well plate) using Effectene as the transfection reagent. Transfections with siRNA oligonucleotides were performed as previously described (Bache et al., 2003b).

GST-clathrin-TD (a GST-fusion construct of residues 1-579 of clathrin heavy chain) was produced in Escherichia coli BL21 (DE3) cells as described previously (Raiborg et al., 2001a). The recombinant protein was purified on glutathione-Sepharose 4B (Amersham Pharmacia Biotech AB, Uppsala, Sweden) after lysis of the bacteria in B-PER™ reagent (Pierce, Rockford, IL) according to the instructions from the manufacturers.

Aliquots (25 μl) of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech, Uppsala, Sweden) were washed three times with assay buffer (25 mM Hepes pH 7.2, 125 mM potassium acetate, 2.5 mM magnesium acetate, 5 mM EGTA and 1 mM dithiothreitol) before incubation with 0.2 μmol of either GST-clathrin-TD or GST in 200 μl assay buffer for 30 minutes at room temperature. The beads were washed twice in assay buffer before incubation for 1 hour at 4°C, with cell lysate from HeLa cells that had been transiently transfected for 24 hours with the myc-Hrs constructs indicated. The cells had been serum-starved for 4 hours and stimulated or not with 50 ng/ml EGF for 15 minutes prior to lysis. The cells were lysed in assay buffer containing 0.1% NP-40 and protease and phosphatase inhibitor cocktails (Sigma). Finally, the beads were washed four times in assay buffer supplemented with protease and phosphatase inhibitors and resuspended in SDS-PAGE sample buffer. Myc-Hrs associated with the beads was detected by SDS-PAGE followed by immunoblotting with the mouse anti-myc antibody. The lysates were also subjected to immunoprecipitation with the anti-myc antibody. The immunoprecipitated myc-Hrs was deteced by immunoblotting with an anti-phosphorylated-tyrosine antibody.

HeLa or HEp-2 cells grown on coverslips were permeabilised with 0.05% saponin, then fixed with 3% (w/v) paraformaldehyde and stained for fluorescence microscopy as described (Raiborg et al., 2001a). Coverslips were examined with a Zeiss LSM 510 META confocal microscope equipped with Plan-Apochromat ×63/1.4 and Neo-Fluar ×100/1.45 oil-immersion objectives.

HEp-2 cells were first depleted of endogenous human Hrs by siRNA for 72 hours. The siRNA-treated cells were then seeded onto coverslips in a 24-well tissue culture dish and the following day transfected with siRNA-resistant pcDNA3-myc-Hrs constructs for 17 hours. Transfected cells were incubated with 100 ng/ml Rhodamine-labelled EGF (Molecular Probes, Eugene OR) for 15 minutes, washed twice in Dulbecco's modified Eagle's medium (DMEM) and incubated for 3 hours in DMEM supplemented with 10% foetal calf serum (FCS) and 10 μg/ml cycloheximide to allow EGF degradation. The cells were then processed for confocal microscopy. For the quantification of cell-associated Rhodamine-labelled EGF, confocal images of randomly selected transfectants were recorded at fixed intensity settings below pixel-value saturation and analysed by post-image processing. All pixel values above background level (defined as mean values obtained in unlabelled cells) were integrated for each cell by using the histogram function in the Zeiss LSM Software, version 3.2. Between 20 and 30 cells from each transfection and time point were quantified from four independent experiments. Since high-level overexpression of Hrs inhibits EGF trafficking (Raiborg et al., 2001a), cells expressing the myc-tagged Hrs constructs at more than five times of the endogenous Hrs levels (as judged by measuring the total intensity of the Hrs labelling in untransfected control cells) were excluded from the analysis.

For quantification of colocalisation of Hrs and EEA1 on enlarged endosomes, confocal images of single endosomes were analysed. Using the colour range and histogram tools of Adobe Photoshop 7.0, the amount of endosomal Hrs pixels that colocalised with EEA1 was quantified and presented as the percentage of the total amount of Hrs pixels per endosome. All pixels with an intensity level over 140 were counted. Colocalisation was detected as yellow (green-red) or cyan (blue-green) pixel values in the merged images. For quantification of colocalisation of the 2xFYVE-Hrs CT constructs and EEA1 or YFP-Hrs, the amount of overlapping pixels on endosome membranes was counted by using the histogram function in the Zeiss LSM Software, version 3.2. The Zeiss LSM Software was also used to determine the amount of endosome-associated Tsg101 that colocalised with EEA1 or Hrs.

We thank Sébastien Wälchli for providing the plasmid for siRNA-mediated depletion of clathrin, James Keen for providing the GFP-clathrin LC plasmid, and Scott Emr for helpful suggestions. J.W. is a senior scientist and C.R. is a postdoctoral fellow of the Norwegian Cancer Society. L.M. is a postdoctoral fellow of the FUGE programme. This work was also supported by the Top Research Programme, the Research Council of Norway, and the Novo-Nordisk Foundation.