Epsin 1 is a cargo-specific adaptor for the clathrin-mediated endocytosis of the influenza virus

Imaging Epsin 1 in Live Cells.

To detect the intracellular distribution of epsin 1, we imaged endogenous epsin 1 and clathrin in BS-C-1 cells by using immunofluorescence. Epsin 1 appeared as punctate structures and colocalized with clathrin extensively (Fig. 1A): 86 ± 3% of the epsin 1-stained structures showed a clathrin signal, and 75 ± 4% of the clathrin-stained structures displayed an epsin 1 signal, consistent with the observed colocalization between epsin 1 and clathrin-coated structures (13, 14). To facilitate live-cell imaging of epsin 1, we constructed a fusion protein containing epsin 1 and Venus, a variant of the enhanced yellow fluorescent protein (EYFP) (23), and transfected BS-C-1 cells with epsin 1-Venus. Epsin 1-Venus also displayed discrete structures and colocalized with endogenous clathrin to a similar extent (Fig. 1B): 86 ± 5% of the epsin 1-Venus structures colocalized with immunostained clathrin structures, and 83 ± 4% of the clathrin structures exhibited an epsin 1-Venus signal. These observations suggest that the fusion construct of epsin 1-Venus was targeted to the correct cellular locations and revealed nearly all epsin 1-containing structures.

Influenza Virus Colocalizes with Epsin 1 during Clathrin-Mediated Endocytosis.

The influenza virus enters cells through multiple endocytic pathways (8–10). About 60% of the virus particles enter by clathrin-mediated endocytosis, and the remaining 40% enter through a clathrin- and caveolin-independent pathway (10). Upon internalization, the virus releases its genome into the cytoplasm via pH-dependent membrane fusion of the viral envelope with the acidic endosome (7, 24). To probe the involvement of epsin 1 in influenza viral entry, we tracked individual virus particles in live cells expressing epsin 1-Venus. For virus imaging, the virus was labeled with a red lipophilic dye, DiD, which did not perturb influenza infectivity (25) and was added to cells in situ at 37 °C. Upon binding and internalization, the virus particles exhibited rapid and directed movements in a microtubule-dependent manner [supporting information (SI) Movie S1]. Pretreatment of cells with nocodazole, a microtubule-depolymerizing drug, abolished these movements. Eventually the virus particles fused with the endosomal membrane, as indicated by a strong increase in the DiD signal because of the spreading of DiD molecules into the larger endosomal membrane and the resulting dequenching of DiD fluorescence (Movie S1) (25). As a control, dequenching of DiD fluorescence was not observed in cells treated with ammonium chloride, a lipophilic base known to raise endosomal pH and inhibit pH-dependent viral fusion (25).

In the following experiments, we used rapid microtubule-dependent movement and viral fusion as the criteria to indicate that viral entry occurred successfully. Viral fusion is a critical step in the influenza infection pathway, allowing the genetic material of the virus to be released into the cell. It has been shown that influenza infection is inhibited when endosomal pH is raised to prevent fusion (8, 24), and that isolated genetic materials of influenza are capable of inducing infection when microinjected into the cytosol (26, 27). Thus, a criterion that involves viral fusion likely represents infectious entry. When virus particles displaying rapid microtubule-dependent movement and viral fusion were identified, we then traced the virus particles back to analyze the colocalization of the virus particles with epsin 1-Venus before their entry. Quantitative analysis showed that 56 ± 6% of the internalized virus particles colocalized with epsin 1-Venus before entry, whereas the remaining 44 ± 6% entered cells without overlapping with epsin 1-Venus (n = 71; Fig. S1). The ratio is similar to the determined partition of influenza virus particles between the clathrin-mediated and clathrin-independent endocytic pathways (10), leading to a hypothesis that the virus particles internalized through clathrin-mediated endocytosis also colocalized with epsin 1.

To test this hypothesis, we performed three-color imaging experiments in which the virus particles were tracked in cells that coexpressed epsin 1-Venus and clathrin ECFP, the fusion protein of clathrin light chain and the enhanced cyan fluorescent protein (ECFP). It has been shown that the clathrin fusion protein functions as normally as its endogenous counterpart (10, 28), and that the fluorescent clathrin signal reveals >96% of the clathrin-coated structures (10). In cells coexpressing epsin 1-Venus and clathrin-ECFP, extensive colocalization was observed between the two fluorescent proteins (Fig. S2A). Two classes of influenza entry behaviors were observed in these cells. Among the virus particles that successfully entered cells (n = 73), 52 ± 6% appeared to recruit both epsin 1 and clathrin to the virus-binding sites (Fig. 2 A and C and Movie S2), whereas the remaining 48 ± 6% entered cells without colocalizing with either clathrin or epsin 1 (Fig. 2 B and D and Movie S3). Interestingly, the recruitment of epsin 1 and clathrin always appeared to occur simultaneously within the time resolution (≈1 s) of our experiments (Fig. 2C). The fluorescence intensities of epsin 1 and clathrin then increased in a synchronized manner before coat disassembly occurred, suggesting a potential cooperative recruitment of the two proteins. These observations suggest the participation of epsin 1 in the clathrin-mediated endocytosis of the influenza virus.

Epsin 1 Is Required for the Clathrin-Mediated Endocytosis of Influenza.

Epsin 1 has a multidomain structure that interacts with several components of CCPs (11). To test whether epsin 1 plays a critical role in the clathrin-mediated endocytosis of the influenza virus or simply appears in the CCPs enclosing virus particles by a “piggyback” mechanism, we knocked down epsin 1 in cells by using siRNA. The knockdown efficiency was assessed by immunofluorescence and Western blot. In the immunofluorescence image of the siRNA knockdown cells, the number of epsin 1-stained spots and the total epsin 1 fluorescence were reduced by 84% and 87%, respectively, compared with those in the nontreated or nontargeting siRNA treated cells (Fig. 3A). A dramatic reduction in the expression of epsin 1 was also observed by Western blot analyses (Fig. S3A).

We then tested the entry mechanism of the influenza virus by tracking individual virus particles in the epsin 1 knockdown cells expressing clathrin-EYFP. Among the virus particles that successfully entered cells (n = 108), only 18 ± 4% were internalized through CCPs, whereas the great majority (82 ± 4%) of the virus particles were internalized without colocalization with any clathrin-coated structures. This partition ratio was in contrast to those observed in nontreated cells and nontargeting siRNA treated cells, where the majority of the internalized virus particles entered through CCPs (Fig. 3B), suggesting that epsin 1 was important for the clathrin-mediated endocytosis of the influenza virus.

The total percentage of the virus particles that entered and fused per cell, however, did not change, and the infectivity of the virus was also uninhibited by epsin 1 knockdown (Fig. S4), suggesting that the virus did not immediately commit to one of the two pathways upon binding. Thus, when the clathrin-dependent, entry pathway was inhibited in the absence of epsin 1, most of the virus particles were routed to the clathrin-independent entry pathway. This property would have made it difficult to dissect the role of specific proteins in each individual entry pathway in a conventional biochemical assay demonstrating the power of the single-particle tracking approach.

Interestingly, despite the structural similarity to epsin 1 (29), epsin 2 knockdown did not have a similar effect on influenza viral entry. A nearly identical fraction of virus particles entered through the clathrin pathway in epsin 2 knockdown cells as in control cells (Fig. 3B), despite efficient epsin 2 knockdown (Fig. S3B). Consistent with this result, the double knockdown of both epsin 1 and epsin 2 showed a similar effect as epsin 1 single knockdown (Fig. 3B and Fig. S3C).

The UIM Motifs of Epsin 1 Are Critical for the Clathrin-Mediated Endocytosis of Influenza.

Epsin 1 has several UIMs that interact with polyubiquitin chains and ubiquitinated receptors (Fig. 4A), potentially allowing epsin 1 to capture endocytic cargo receptors by using this mechanism (17, 18, 20, 30). To test whether the UIMs play a role in the clathrin-mediated entry of influenza, we constructed an epsin 1 mutant that lacked the UIM domains (epsin1ΔUIM), fused the mutant construct to either Venus or ECFP, and transfected the cells with the fluorescent-protein-tagged epsin1ΔUIM (Fig. 4A). At low expression levels, epsin1ΔUIM-Venus appeared as discrete spots that colocalized extensively with endogenous clathrin (Fig. S2B), indicating that epsin1ΔUIM was still able to interact with CCPs efficiently. However, when we tracked influenza virus particles that successfully entered cells expressing epsin1ΔUIM-Venus, only 8 ± 3% of the virus particles colocalized with epsin1ΔUIM before their entry, whereas the remaining 92 ± 3% did not show any colocalization with epsin1ΔUIM before internalization (n = 78; Fig. 4B). This is in contrast to the major fraction of virus particles (56 ± 6%) observed to colocalize with epsin 1 before their entry in epsin 1-Venus expressing cells, suggesting that the removal of the UIMs significantly reduced the interaction between epsin 1 and influenza virus receptors.

Next, we overexpressed epsin1ΔUIM-ECFP in clathrin-EYFP-expressing cells. Overexpression of epsin1ΔUIM-ECFP was indicated by a strong and diffuse ECFP signal. The density of clathrin-coated structures in epsin1ΔUIM-ECFP overexpressing cells was similar to that observed in cells not expressing epsin1ΔUIM-ECFP (Fig. S5A). Moreover, when transferrin was added to these cells, 88% of the transferrin colocalized with clathrin-EYFP-labeled structures, and 70% of the clathrin-EYFP structures were labeled with transferrin (Fig. S5B). These results suggest that the overexpression of epsin1ΔUIM-ECFP did not perturb the overall distribution of CCPs, and that nearly all CCPs were labeled with clathrin-EYFP. However, when we tracked influenza virus particles in these cells, only 6 ± 4% of the successfully internalized virus particles entered through CCPs, whereas the overwhelming majority (94 ± 4%) took the nonclathrin pathway (n = 33). This is in stark contrast to the cells not expressing epsin1ΔUIM, where 57% of the internalized virus particles (n = 44) entered through clathrin-mediated endocytosis (Fig. 4C). These results indicate that overexpression of epsin1ΔUIM severely inhibit the clathrin-mediated entry of the influenza virus.

We also tested whether the tandem UIM motifs themselves can serve as a dominant-negative mutant by constructing a mutant protein containing only the three UIMs fused with ECFP (UIM-ECFP) (Fig. 4A). In cells overexpressing UIM-ECFP, a significant reduction was also observed in the fraction of the virus particles taking the clathrin pathway, although the reduction was not as strong as that induced by the epsin1ΔUIM dominant-negative mutant (Fig. 4C). A plausible explanation for the difference is that epsin1ΔUIM, with all of the CCP-interacting motifs of epsin 1 retained, can efficiently compete with the endogenous epsin 1 for clathrin recruitment but does not interact with influenza receptors, thus potently depleting clathrin from being available for the influenza receptors. However, the UIM motifs likely do not bind influenza receptors as well as full-length epsin 1, considering that the lipid-, clathrin-, AP-2-, and Eps15-interacting domains of epsin 1 may effectively increase the avidity between epsin 1 and influenza receptors and thus have a weaker inhibitory effect.

The above results indicate a critical role for the UIM motifs of epsin 1 in clathrin-mediated influenza viral entry. To test whether this effect was specific to epsin 1, we sought to probe the role of other UIM-containing CCP components, such as Eps15 and Eps15R (31), in influenza uptake. Efficient double knockdown of both Eps15 and Eps15R was achieved with siRNAs (Fig. S6A). In contrast to epsin 1 knockdown, the fraction of the virus particles entering through the clathrin pathway was not significantly affected by the Eps15 and Eps15R double knockdown (Fig. S6B).

Epsin 1 Is Not Required for the Clathrin-Mediated Endocytosis of Transferrin, EGF, and LDL.

We noticed that the overexpression of epsin1ΔUIM did not perturb the cellular distribution of clathrin (Fig. S5 A and B). Moreover, in cells overexpressing epsin1ΔUIM-ECFP, transferrin was internalized to the same extent as in control cells (Fig. S5 C and D), suggesting that the dominant-negative effect of epsin1ΔUIM is cargo-specific. One possible explanation of the different effect of epsin1ΔUIM on the uptake of transferrin versus influenza is that the clathrin-binding affinity of epsin1ΔUIM, although similar to that of the endogenous epsin 1, may be lower than that of other CCP adaptors, such as AP-2. As a result, overexpression of epsin1ΔUIM does not efficiently deplete clathrin for transferrin endocytosis, which uses AP-2 as the adaptor.

To test whether epsin 1 is indeed a cargo-specific adaptor, we measured the effect of epsin 1 knockdown on the internalization of transferrin, EGF, and LDL. Highly efficient epsin 1 knockdown was achieved by using siRNA (Fig. 5A and Fig. S3A). The knockdown did not affect clathrin distribution; a similar density of fluorescent clathrin spots was observed by immunofluorescence in both the nontreated and the epsin 1 knockdown cells (Fig. 5A). The uptake of transferrin was not affected by epsin 1 knockdown (Figs. 5 A and B), and extensive colocalization (approximately 70%) was observed between transferrin and clathrin before the internalization of transferrin (Fig. S7), indicating that transferrin entered through the clathrin pathway. In contrast, the knockdown of AP-2 by μ2-adaptin siRNA efficiently blocked the internalization of transferrin (Figs. 5 A and B). Similar to transferrin, EGF and LDL also entered the epsin 1 knockdown cells as efficiently as in the nontreated cells (Fig. S8 and Fig. 5B). Because EGF may explore alternative clathrin-independent endocytic pathways (22, 32, 33), we tested whether EGF still entered cells through CCPs by adding EGF to the epsin 1 knockdown cells that expressed clathrin-EYFP. In these cells, EGF quickly clustered, and over 90% of the EGF clusters colocalized with clathrin-EYFP before entry (Fig. S9 A and C). Similarly, approximately 90% of the LDL particles were internalized after colocalizing with clathrin-EYFP (Fig. S9 B and C), consistent with a previous report on normal LDL entry in epsin 1 knockdown cells (34). In addition, all three ligands showed essentially the same entry kinetics in epsin 1 knockdown cells as in control cells (Fig. S10). These results indicate that the clathrin-mediated endocytosis of transferrin, EGF, and LDL did not require epsin 1.

Cells, Viruses, Antibodies, Plasmids, and siRNAs.

Procedures for cell culture, virus labeling, immunofluorescence, transfection, siRNA knockdown, and ligand uptake are described in SI Text.

Live-Cell Fluorescence Imaging.

Tracking of the influenza virus, EGF, and LDL in live cells was performed at 37°C on an inverted Olympus IX-71 microscope. ECFP was excited by a 457-nm Ar ion laser (Melles-Griot), EYFP or Venus was excited by a 532 nm Nd:YAG laser (Crystalaser), and DiD, Alexa633, or Alexa647 was excited by a 633-nm He-Ne laser (Melles-Griot). Multicolor imaging was performed as described (7). All emissions were collected by a 1.4 N.A. objective (Olympus). The short- and long-wavelength emissions were separated by 650-nm long-pass dichroic mirrors (Chroma) and imaged by a CCD camera (CoolSNAP HQ, Roper Scientific). The emission filters used are a 665- nm long-pass filter (Chroma) for DiD, Alexa633, or Alexa647; a 570/40- or 585/70-nm band-pass filter (Chroma) for EYFP or Venus; and a 480/40-nm band-pass filter (Chroma) for ECFP. The image acquisition rates were 0.5–2 Hz. Image analysis and single-particle tracking were carried out by using custom-written software as described (7, 10).