Vaccinia Virus Infection Requires Maturation of Macropinosomes

The prototypic poxvirus, vaccinia virus (VACV), occurs in two infectious forms, mature virions (MVs) and extracellular virions (EVs). Both enter HeLa cells by inducing macropinocytic uptake. Using confocal microscopy, live‐cell imaging, targeted RNAi screening and perturbants of endosome maturation, we analyzed the properties and maturation pathway of the macropinocytic vacuoles containing VACV MVs in HeLa cells. The vacuoles first acquired markers of early endosomes [Rab5, early endosome antigen 1 and phosphatidylinositol(3)P]. Prior to release of virus cores into the cytoplasm, they contained markers of late endosomes and lysosomes (Rab7a, lysosome‐associated membrane protein 1 and sorting nexin 3). RNAi screening of endocytic cell factors emphasized the importance of late compartments for VACV infection. Follow‐up perturbation analysis showed that infection required Rab7a and PIKfyve, confirming that VACV is a late‐penetrating virus dependent on macropinosome maturation. VACV EV infection was inhibited by depletion of many of the same factors, indicating that both infectious particle forms share the need for late vacuolar conditions for penetration.

vacuoles formed by membrane fission when the protrusions collapse back onto the plasma membrane (16).
To date, over 20 different viruses from diverse families have been shown to use macropinocytosis for infectious entry (reviewed in 15,31). Amongst these is vaccinia virus (VACV), a large, enveloped, dsDNA virus characterized by its structural complexity and cytoplasmic life cycle. VACV, the prototypic member of the poxvirus family, was used as the vaccine for the eradication of smallpox (32). During infection, VACV produces two types of infectious particles: mature virions (MVs) with one envelope membrane and extracellular virions (EVs) with two membranes. For internalization and infection, both trigger their own macropinocytosis (33)(34)(35)(36)(37)(38).
The cellular factors required for macropinocytic uptake of VACV MVs and EVs include epidermal growth factor receptor (EGFR) and receptor tyrosine kinase (RTK) and RTK signaling, actin and myosin dynamics, small Rho GTPases and protein kinase C (PKC) as well as PI(3)kinase activity (33)(34)(35)(36)38,39). However, the macropinocytic trafficking requirements for VACV infection remain undefined. To this end, we have investigated the pathway taken by VACV MVs and EVs after macropinocytosis in HeLa cells. The maturation of VACV-induced macropinosomes was analyzed using microscopy in fixed and live cells. A targeted siRNA screen was used to identify cellular factors involved in VACV endocytic trafficking and a variety of perturbations allowed analysis of the critical cellular processes involved.

VACV MV-containing macropinosomes undergo maturation
To investigate the association of VACV with early and late endocytic vacuoles, we allowed MV particles to bind to HeLa cells in the cold and used confocal microscopy to follow the distribution of the virus after warming to 37 ∘ C for 0-4 h. The MVs used in these experiments contained a mCherry-tagged version of the core protein A4 [West-ern Reserve (WR) mCherry-A4] (40), which allowed us to visualize individual particles. To distinguish between internalized and surface-bound virions, we stained those accessible on the plasma membrane of non-permeabilized cells with the monoclonal antibody (MAb) 7D11, which recognizes the viral membrane protein L1. The HeLa cells expressed EGFP-tagged versions of EE or LE/LY proteins, Rab5, Rab7 or LAMP1. To localize an additional EE marker, EEA1, cells were permeabilized in some of the experiments and subjected to immunofluorescence staining directed against this factor.
When the images were subjected to colocalization analyses using automated particle detection ( Figure S1), 40.0% of virions were found to internalize by 30 min ( Figure 1A, light blue line), whereafter the number of intracellular virions identifiable as spots began to decline ( Figure 1A, light blue line). Penetration and disassembly of the viral cores in the cytosol are known to begin between 20 and 30 min after endocytosis (39,41). Thus, the subsequent decline in detectable viral spots observed between 60 and 120 min was consistent with the known kinetics of uncoating of the MV genome and core degradation (39,41,42).
Incoming viruses were found in Rab5a-positive, and to a lesser extent in EEA1-positive, vacuoles within 5 min after warming, indicating that the first maturation step occurred quickly after macropinocytosis ( Figure 1A, dark blue and green lines). Colocalization of internalized viruses with these early endocytic markers peaked at 18.0% for

816
Traffic 2015; 16: 814-831 Rab5a and 8.0% for EEA1 by 15 min ( Figure 1B,C), after which it started to decrease ( Figure 1A, Rab5: dark blue line, and EEA1: green line). These results suggested that attachment of these proteins to macropinosomes containing the virus occurred for only a short period after vacuole formation. We noted that the colocalization of virus with EEA1 was lower than Rab5a. This could be due to the fact that EEA1 is only found on a subset of early macropinosomes, as is the case with classical EEs (43). As EEA1 is required for docking and homotypic fusion of early macropinosomes (20), another possibility is that fusion of virus-containing vacuoles is a rapid event and thus the association of EEA1 with VACV-containing macropinosomes is brief.
Coincident with the decrease in Rab5a, MV-containing macropinosomes acquired late macropinosome markers Rab7a and LAMP1. While virions were not detected in Rab7a-positive macropinosomes at 5 min, colocalization with Rab7a increased to 8.0% by 15 min p.i., peaking at 30 min with 26.0% of MV-containing macropinosomes ( Figure 1A, red line, and 1D). The appearance of LAMP1 was delayed relative to Rab7a, with only 19.0% of virion-containing vacuoles positive for LAMP1 at 30 min ( Figure 1A, purple line). By 60 min, 28.0% of the MV-positive macropinosomes had acquired LAMP1 ( Figure 1E), which then declined over the next 2 h in parallel with the loss of Rab7 ( Figure 1A, purple line). In confirmation of these results, a similar percentage of VACV MVs were found to colocalize with endogenous Rab5a (16.7%), Rab7a (26.4%) and LAMP1 (30.6%) at the peak time points of colocalization determined in Figure 1A (15, 30 and 60 min, respectively) ( Figure S2).
When HeLa cells transiently expressing EGFP-tagged Rab5a, Rab7a or LAMP1 were infected with WR mCherry-A4 MVs and imaged live, coordinated movement of virions within vacuoles positive for these markers was observed ( Figure 1F and Movies S1-S6). Live-cell imaging revealed that infected cells often contained exceptionally large Rab5a-positive compartments that acquired virus particles through fusion with smaller vesicles (Movies S1 and S2). At later time points, VACV was found within EGFP-Rab7a and LAMP1-EGFP vesicles (Movies S3-S6), both of which were capable of homotypic fusion events (Movies S4 and S6). In agreement with the low pH requirement of VACV penetration by fusion out of vacuoles (33,(44)(45)(46)(47), we occasionally observed virus cores escaping from LAMP1-positive late macropinosomes ( Figure 1F and Movie S5).
The results indicated, in accordance with previous reports (39,46), that VACV MV endocytosis in HeLa cells is rapid and relatively efficient. Vacuoles that contained the internalized virus recruited Rab5a and, in some cases, EEA1 immediately after formation. In the time span of 30-60 min after warming, vacuoles underwent Rab5a to Rab7a conversion followed by the acquisition of LAMP1, an integral membrane protein. Following the acquisition of LAMP1, the total number of internalized cores started to decrease. Consistent with this, when release of cores into the cytosol was observed, it occurred from LAMP1-containing vacuoles. We have previously shown that virus-containing vacuoles positive for the endosomal content marker dextran do not contain transferrin at any time (34,35) confirming that they are distinct from classical endosomes.

Endocytosis-directed RNAi screen suggests that macropinosome maturation is critical for VACV infection
To establish which cellular trafficking factors VACV MVs require for infection, we performed a high-throughput RNAi screen in human tissue culture cells (HeLa) as outlined in Figure S3A. Using a recombinant VACV that expresses EGFP under an early viral promoter (WR E EGFP), we could readily distinguish infected from non-infected cells based on EGFP expression 6 h post-infection (h p.i.). Automated fluorescence microscopy and image analysis were employed to score for cell factors required for early stages of infection up to and including translation of early viral genes. The siRNA library contained three siRNAs against 162 human genes known to be involved in endocytosis and endocytic membrane traffic (Table S1) (48).
The screen was performed in triplicate and hits defined as proteins that, when depleted, caused at least a 40% decrease in infection with two or three out of the three siRNAs. Thirty-three cell factors fulfilled the criteria ( Figure 2A and Table S2). When the hits were clustered using our NETWORK VISUALIZATION ALGORITHM (42), which takes in account physical interactions defined in STRING (49) and cellular functions annotated in DAVID (50,51), we found that the hits fell into six clusters: membrane trafficking, Rab conversion, endosomal positioning and acidification, PI metabolism and the regulation of actin ( Figure 2A and Table S2).
Depletion of Rab5c, Rab7a, Rab34, RABGEF1 (Rabex5), SNX3, PIKfyve and PIK3R4 caused significant inhibition in VACV early gene expression ( Figure 2B). To follow up these factors, the siRNA with the greatest effect was rescreened against VACV MVs in parallel with Semliki Forest virus (SFV), an alphavirus that enters cells by clathrin-mediated endocytosis and penetrates from EEs ( Figure 2C). Depletion of clathrin light chain (CLTC), required for SFV, and the small GTPase Rac1 which is required for VACV infection served as controls ( Figure 2C; inset). While VACV infection was inhibited by at least 50% in all cases, knockdown of the genes had no effect on SFV infection. That the cells could still support SFV infection indicated that the inhibition of VACV was not due to a general defect in the endosome system or cell fitness. Collectively, the list of validated factors indicated that for infection by VACV MVs, macropinosomes must undergo both early and late stages of maturation.

VACV infection requires Rab7a function
As a classical LE marker, Rab7a is important for LE maturation and lysosome fusion. To determine if Rab7a function was required during VACV trafficking and infection, cells were transfected with wild type (WT), constitutively active (C/A), or dominant negative (D/N) versions of EGFP-Rab7a. Confocal microscopy of the transfected cell populations showed that the EGFP-Rab7a variants localized as expected, with WT and C/A Rab7a residing on a perinuclear vesicle population, and D/N Rab7a displaying a diffuse staining throughout the cell ( Figure 3A).
The cells were infected with a VACV recombinant expressing mRFP from an early/late promoter (WR E/L mRFP). At 6 h p.i., cells were collected and analyzed by flow cytometry for transfected cells that were also infected. Cells expressing C/A or D/N Rab7a showed a statistically significant (p < 0.05) average decrease in infection of 41.56 and 30.42%, respectively, when compared to cells expressing the WT protein ( Figure 3B).
Microscopy-based localization of EGFP-SNX3 during VACV infection indicated that SNX3 was recruited to MV-containing vacuoles with similar kinetics as Rab7a ( Figure S4). Collectively, the localization, RNAi screening and infection data indicate that recruitment of functional Rab7a to MV-containing macropinosomes is critical for VACV infection.

VACV entry depends on PIKfyve
PIs are essential for the progressive maturation of endosomes and macropinosomes (52,53). When cells expressing the PtdIns(3)P probe EGFP-FYVE EEA1 were infected with WR mCherry-A4 MVs and analyzed 15 min after warming by confocal microscopy, about 16.0% of the viruses were located in PtdIns(3)P-positive vacuoles ( Figure 4A,B). In addition, we observed recruitment of the PtdIns(3)P-binding probe to VACV MV-containing macropinosomes by live-cell imaging (Movie S7). Consistent with its role in maturation, the colocalization between VACV MVs and EGFP-FYVE EEA1 dropped steadily after 30 min with <8.0% of virions localizing with PtdIns(3)P-positive compartments at 60 min ( Figure 4A,B). These findings suggested that the VACV-containing macropinosomes acquired PtdIns(3)P at early stages of virion trafficking which is then lost as the macropinosomes mature.
Our siRNA screen revealed that PIK3R4, the regulatory subunit of the PtdIns(3)P-producing PIK3C3, as well as PIKfyve were required for VACV infection ( Figure 2). PIKfyve catalyzes the conversion of PtdIns(3)P to PtdIns(3,5)P 2 , a process of importance for the maturation of endosomes and macropinosomes (3,30,54). To determine if PIKfyve-mediated conversion of PtdIns(3)P to PtdIns(3,5)P 2 was important for VACV infection, cells were infected with WR E EGFP MVs in the presence of increasing concentrations of YM201636, an inhibitor of PIKfyve. When analyzed by flow cytometry 6 h p.i., a dose-dependent inhibition of VACV MV infection, up to 50%, was observed ( Figure 4C).
To ensure that YM201636 prevented infection at the level of virus entry, virus binding and macropinocytosis bypass assays were performed. WR EGFP-A4 MVs were first allowed to bind to cells on ice in the absence or presence of YM201636. When cell-associated fluorescence was quantified by flow cytometry as described (55), no defect in virus binding to cells was observed ( Figure 4D).
Taking advantage of the acid-mediated fusion activity of VACV MVs (46), virus penetration can be forced at the plasma membrane by low pH, thus bypassing endocytosis.
We have previously employed this strategy to test the specificity of entry inhibitors (34,55). To determine whether YM201636 inhibited MV entry, WR E EGFP MVs were bound to cells in the presence or absence of YM201636 and subsequently treated with pH 5.0 media ( Figure 4E). The epidermal growth factor inhibitor Iressa, previously shown to inhibit VACV infection at the level of entry (33), was used as a positive control for bypass. For both, exposure to pH 7.4 media served as a bypass control. In cells treated with YM201636 and low pH, VACV infection could be partially restored ( Figure 4E). The nearly twofold rescue was significant (p = 0.0016) and comparable to that seen in control cells treated with Iressa and low pH ( Figure 4E).
With the presence of PtdIns(3)P on MV-containing macropinosomes and the requirement of PIKfyve activity for VACV entry, we concluded that PIKfyve mediate PI exchange and, by extension, PtdIns(3,5)P 2 synthesis was

VACV EV infection relies on macropinosome maturation factors
In addition to MVs, a second infectious form of virus is produced during the VACV life cycle known as EVs. EVs consist of an MV-like particle surrounded by an additional Golgi-derived membrane containing cellular and at least six additional viral proteins (56). We have previously demonstrated that similar to MVs, EVs enter HeLa cells by inducing their own macropinocytic uptake (55). To overcome the topological constraint of having two membranes, the outer EV membrane is shed in an acid-dependent fashion within macropinosomes to free the underlying MV for fusion (36).
To determine if EVs share similar trafficking requirements as MVs, a subset of the LE factors required for MV trafficking were screened for their effect on EV infection (Table  S2). We have previously shown that during preparation a large fraction of EVs are disrupted (36). To circumvent any consequences of disrupted EVs or contaminating MVs on investigation of EV entry, preparations of EVs were treated with the MAb 7D11. This antibody is directed against L1R, a component of the VACV fusion machinery, which is only present in the MV membrane. Thus, MVs and disrupted EVs are rendered fusion incompetent by 7D11 neutralization, while intact EVs remain unaffected (illustrated in Figure 5A).
For the screen, WR E EGFP EVs were prepared and subjected to 7D11 neutralization as previously described (36). Neutralization of EV preparations resulted in a threefold reduction (from 57 to 18%) in infection relative to untreated EV preparations ( Figure 5B). The screen was performed in triplicate as described for Figure 2. Of the 10 factors assessed, depletion of 7 caused at least 40% reduction in infection with 2 or 3 of the 3 siRNAs and were defined as hits (Table S3). These include Rab5c, Rab7a, Rab34, SNX3, PIKP5K1B, RILP and FYCO ( Figure 5C). Depletion of clathrin light chain (CLTC), required for SFV, or Rac1, required for VACV infection, again indicated that the inhibition of EV infection was not due to a general defect in the endosome system or cell fitness ( Figure 5D). Collectively, these screening results indicated that for infectious entry both VACV MVs and EVs must traffic through macropinosomes that undergo maturation.

Discussion
VACV MVs and EVs initiate entry into host cells by triggering their own macropinocytic uptake (34,35,37,38). About 30 min later, the viral core is delivered into the cytosol after fusion of the viral envelope with the limiting membrane of endocytic vacuoles (39,46). In this study, we followed the viruses during the time of passage within the endocytic network, and analyzed the maturation of the virus-containing macropinosomes.
The first conversion that macropinosomes undergo after fission from the plasma membrane is the recruitment of EE factors such as Rab5 and its effector EEA1 as well as the accumulation of PtdIns(3)P (17,18,20). We observed MVs in Rab5a-positive, and to a lesser extent in EEA1-positive, vacuoles already 5 min after warming. The number within Rab5a-positive vacuoles reached a maximum of 18% of internalized virions by 15 min. The association of macropinosomes with EEA1 was less prominent than with Rab5 reaching a maximum of 8.0%. That this Rab5 effector and tethering factor is only found on a subset of EEs (43,57) may explain the low level of colocalization we observed. Perhaps, early macropinosomes, such as classical endosomes, exist as several distinct subsets.
Another possibility is that due to the asynchronous nature of VACV infection, combined with the transient role of EEA1 in promoting homotypic fusion of early macropinosomes, we only observe a fraction of actual virus EEA1 colocalization events (20,22,28). In fact, we could observe multiple rapid fusion events between virus-containing, Rab5a-positive early macropinosomes by live-cell imaging (Movies S1 and S2). It is possible that the fusion events also involved one of the hits in our RNAi screen, vesicle-associated membrane protein 7 (VAMP7). It has been shown to mediate pinosome fusion events in Dictyostelium discoideum (58).
Most of the early vacuoles were positive for PtdIns(3)P. We could follow the generation of this lipid in virus-containing vacuoles by live-cell imaging (Movie S7). We and others have previously shown that the vacuoles containing MVs also contain fluid markers such as dextran, and that they are devoid of transferrin, a cellular cargo molecule present in early and recycling endosomes (34,35,37,38). This is consistent with reports indicating that classical endosomes and macropinosomes give rise to distinct transport vesicle populations (28).
Having identified SNX3 as a hit in the RNAi screen, we investigated its localization in cells that had internalized VACV MVs. SNX3 was predominantly found in MV-containing LMs ( Figure S4). Interestingly, SNX3 has been shown to localize to SCVs where it facilitates SCV maturation to late compartments (25).
Maturation of virus-containing vacuoles continued about 15 min after warming with Rab5a being gradually replaced by Rab7a. LAMP1, a membrane glycoprotein found on LE and LYs, arrived shortly after this conversion. Vacuoles positive for both Rab7a and LAMP1 are known to serve as terminal organelles during trafficking of the fluid phase marker dextran (60). The change in Rabs and other markers thus suggested that VACV-containing macropinosomes underwent a full program of maturation, in line with the relatively late timing of VACV MV penetration. The VACV strain used here, WR, has been shown to require low pH (pH = 4.5-5.0) for envelope fusion (46,61,62). However, inhibition of WR MV infection by lysosomotropic agents varies within the range of 40-60%, and other VACV strains, such as IHD-J, fuse from macropinosomes in a pH-independent fashion (33,44). This suggests that other macropinosome factors, in addition to low pH, may be required for VACV fusion. It is also possible that the acidification of macropinosomes or their sensitivity to lysosomotropic agents is not identical to that of canonical endosomes.
Rab7a depletion and the expression of either D/N or C/A mutants inhibited MV infection. The Rab7a effectors identified as hits in the RNAi screen, RILP, FYCO1 and OSBPL1A, are all involved in endosome movement. FYCO1 interacts with kinesins, and OSBPL1A promotes binding of RILP to dynein to mediate the movement of Rab7-positive vesicles along microtubules (MTs) (63,64). Another hit, Rab34, confirmed by D/N and C/A expression ( Figure S5) has been implicated in regulation of lysosomal positioning through interaction with RILP (65).
The involvement of these factors implied that MV-containing macropinosomes interact with MTmediated motors for transport of the vacuoles toward the perinuclear region of the cell where the majority of LYs are located (3). However, as previously reported (39), neither depolymerization of MTs with nocodazole or colchicine, nor stabilization with taxol had any effect on VACV early infection events ( Figure S6). Although in agreement with the co-existence of MT-dependent and -independent macropinocytic pathways (66), in light of these findings we suspect that MT-based transport of VACV-containing macropinosomes is not necessary for VACV infection, but rather serves to increase the likelihood of macropinosome fusion with LEs or LYs, which could serve to speed entry kinetics.
PI kinases also play a central role in regulating the maturation of endosomes, phagosomes and macropinosomes (67). Our RNAi screen indicated that PI(3)K, PI(4)K, PI(5)K and PIKfyve were needed for MV infection (Figure 2). Previous studies have demonstrated a role for PI(3)K during multiple stages of VACV infection (68,69) including entry (34,39,70). Consistent with this, PtdIns(3)P was detected on early MV-containing macropinosomes until 30 min suggesting that these compartments undergo a lipid switch during their maturation ( Figure 4A,B).
The identification of PIKfyve as a hit in the RNAi screen and the inhibitory effect of a PIKfyve inhibitor, YM201636, indicated that conversion of PtdIns(3)P to PtdIns(3,5)P 2 is critical for VACV MV infection. PIKfyve facilitates Rab7a-independent fusion of LEs with each other and with LYs (30). In addition, the PIKfyve-activating kinase, Akt, previously implicated in VACV entry (70,71), was a hit in our screen. Interestingly, Akt has been reported to phosphorylate PIKfyve to facilitate lysosomal trafficking and degradation of EGFR (72).
VACV EVs also enter cells by macropinocytosis but the process of penetration is more complicated due to the additional envelope membrane (35,37). That the hits were largely overlapping with VACV MVs strongly suggested that EVs also require macropinosome maturation for productive infection. As EVs shed their outermost membrane within macropinosomes (35), and rely on the underlying low pH-dependent MV fusion machinery (46,61,62,73), it is not too surprising that they are subject to the same requirements as MVs.
Viruses provide many advantages as tools to characterize the complex conversions that take place before cargo-carrying vacuoles fuse with LYs. Using VACV as ligand, we found that the process of macropinosome maturation shares many features with the maturation of endosomes (3), phagosomes (74) and autophagosomes (75). Our findings indicated that maturation was essential for the infectivity of VACV MVs and EVs ( Figure 6). It was evident that VACV penetration occurred from macropinosomes quite late in the maturation program. This was illustrated by the finding that PIKfyve and Akt were needed, that delivery to Rab7a-positive compartments was necessary but not sufficient for infection and is consistent with the pH optimum of VACV fusion which is as low as 4.5 (46,61). It is even conceivable that fusion with LYs is required for productive infection. Collectively, our findings support the classification of VACV as a late-penetrating virus (14), and provide new cellular targets for the development of antiviral agents against viruses that use macropinocytosis as an entry route.

Cells and viruses
HeLa ATCC cells were propagated in DMEM including 10% fetal calf serum (FCS), non-essential amino acids and Glutamax (Life technologies). Recombinant VACVs used in the study were generated using the VACV WR strain as previously described (34). These include fluorescent expressors WR E EGFP and WR E/L EGFP as well as viruses containing fluorescent proteins fused to the N-terminus of the structural core protein A4 (WR EGFP-A4 and WR mCherry-A4).

Antibodies and inhibitors
Antibodies directed against EEA1 and mRFP were purchased from Cell Signaling and Clontech, respectively. The mouse MAb 7D11 (α-L1) was kindly provided by Bernard Moss with permission of Alan Schmaljohn (University of Maryland). Secondary antibodies were obtained from Invitrogen. Stocks of YM201636 (Calbiochem), IRESSA (LC Laboratories), Taxol, Colchicine and Nocodazole (Sigma) were prepared in dimethyl sulfoxide and stored at −20 ∘ C.

Colocalization assays
Cells were transfected with plasmids on coverslips in 24-well plates and infected with VACV mCherry-A4 (multiplicity of infection, MOI = 20) in FCS-free medium. Virus was bound to cells for 30 min in a cold water bath. Cells were then washed and incubated in a 37 ∘ C water bath for indicated time points (up to 4 h). Bound non-internalized particles were stained with the MAb 7D11 on ice without permeabilization. Fixed cells were analyzed with a Zeiss LSM 510 Meta confocal microscope using a 63× objective and seven to ten Z-stacks were acquired per image. All images represent maximum projects when not indicated otherwise. Cells analyzed for endogenous markers were first subjected to 7D11 staining as above, fixed and permeabilized for 2 min with 0.1% Triton-X-100 prior to staining with antibodies directed against each marker, followed by Alexa Fluor 488 secondary antibody. Antibodies used were Rab5 [Cell Signaling;

Quantification of colocalization of internalized VACV with endosomal markers
IMARIS software was used to automatically quantify the fraction of internalized virions that colocalize with the various endocytic markers (Rab5, EEA1, Rab7, LAMP1 and FYVE-EGFP) ( Figure S1). For each three-dimensional confocal stack, virus particles and endocytic vesicles were automatically detected using the IMARIS spot colocalization tool. For each image, the thresholding value was set to 200 nm to ensure maximum accuracy of particle detection. Internalized viruses were defined as mCherry-positive virions that did not colocalize with VACV particles positive for 7D11 immunostaining (i.e. non-internalized viruses). For this, mCherry viruses were considered as colocalizing if their median fluorescence was equal to or above the threshold used to identify

Inhibitor studies
HeLa cells were seeded in 24-well plates and pre-treated with different concentrations of YM201636 for 1 h. Cells were then infected with WR E EGFP (MOI = 1) in the presence of the inhibitor. At 6 h p.i., samples were harvested, prepared for flow cytometry and the percentage of infected cells was determined on the basis of EGFP fluorescence.

Rab mutants
HeLa cells were seeded in six-well plates and transfected with plasmids expressing EGFP-tagged WT Rab7a and the respective C/A and D/N variants. After 18 h, cells were infected with WR E/L mRFP (MOI = 1). Cells were harvested 6 h p.i. and scored for both EGFP and mRFP expression by flow cytometry. For microscopy, cells were transfected and prepared as described above for colocalization assays.
Virus was allowed to bind to cells at 4 ∘ C. Cells were then washed several times and prepared for fluorescence-activated cell sorter (FACS) analysis in order to quantify virus binding based on cell-associated EGFP fluorescence intensity. formaldehyde. For EVs, BSC40 cells were infected 24 h prior to screening and α-L1R-neutralized EVs were prepared as previously described (35). Neutralized EV preparations were then used for infection.

Acknowledgments
This work was funded by the Swiss National Foundation (Ambizione and Sinergia grants), by an ERC advanced grant and core funding to the MRC Laboratory for Molecular Cell Biology, University College London.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1: Automated analysis of VACV internalization and endosome colocalization. A) Detection of endocytic vesicles (endosomes) and viral particles using the IMARIS software. Digital images of endocytic vesicles, total virus particles (VACV tot) and external virus particles (VACV ext) generated using the IMARIS spot detection are displayed. B) Automatic detection of internalized virions (VACV int) with IMARIS. For this, internalized mCherry-positive virus particles were defined as those that have no associated VACV ext fluorescence. C) The IMARIS 'spot colocalization' tool was used to automatically detect the colocalization between internalized virions (VACV int) and endocytic vesicles (endosomes). A virion and an endocytic vesicle were considered to colocalize when their distance (from the center or each spot) was ≤200 nm. White arrowheads indicate internalized virus particles that colocalize with endosomal vesicles. Non-internalized virions (VACV ext) were used as negative controls (light-blue arrowhead). Figure S2: Colocalization of VACV with endogenous Rab5, Rab7 and LAMP1. A-C) HeLa cells were bound with VACV WR mCherry-A4 MVs at an MOI of 2 at 4 ∘ C. Cells were washed and shifted to 37 ∘ C for the indicated time points. Non-permeabilized cells were then subjected to immunostaining with α-L1R to distinguish external (blue) versus internalized (red) virions. To visualize endogenous Rab5 (A), Rab7 (B) or LAMP1 (C), cells were permeabilized and immunostained using antibodies directed against these various markers. Insets display colocalization events in the xy, yz and xz planes. White arrows represent colocalization events. D) The percent colocalization between internalized virions and the various endocytic markers was determine using IMARIS automated colocalization analysis as described in Figure S1. At least 30 total cells from three independent experiments were analyzed for each marker. Results displayed as the average ± SD.

Figure S3: RNAi screen workflow, image analysis and cell number correction.
A) The 'usual suspects' siRNA library consists of two 384-well plates. Three copies of the library (six plates in total) were used in each experiment. The siRNAs were introduced into HeLa ATCC cells by reverse transfection. At 72 h post-transfection, cells were infected with WR E EGFP MVs. At 6 h p.i., cells were fixed, nuclei stained with DAPI and the EGFP signal was enhanced by immunofluorescence staining using an α-EGFP antibody. Assay plates were then imaged using an IMAGEXPRESS Microscreening system. The screen was repeated three independent times and the results were shown as the mean of the triplicates. B) Image analysis was performed using an in-house MATLAB-based software that allowed for automatic digital detection and scoring of nuclei and EGFP-positive infected cells (scale bars, 50 μm). C) To correct for the effect on infection index due to deleterious effects of RNAi transfection on cell number variability, an infection index checkerboard was used for correction. Infection of a gradient of cells treated with control siRNA (AllStarNegative) was used to determine the correlation between the number of cells and the corresponding infection index. This was then used to create a normalization curve that was applied to the screening data to eliminate any cell number bias on infection. Any siRNA target wells displaying <200 cells were discarded from the analysis.       Table S2) whose depletion resulted in a 40% or greater decrease in EV infection.