AMP-Activated Protein Kinase Is Required for the Macropinocytic Internalization of Ebolavirus

Abstract

Identification of host factors that are needed for Zaire Ebolavirus (EBOV) entry provides insights into the mechanism(s) of filovirus uptake, and these factors may serve as potential antiviral targets. In order to identify novel host genes and pathways involved in EBOV entry, gene array findings in the National Cancer Institute's NCI-60 panel of human tumor cell lines were correlated with permissivity for EBOV glycoprotein (GP)-mediated entry. We found that the gene encoding the γ2 subunit of AMP-activated protein kinase (AMPK) strongly correlated with EBOV transduction in the tumor panel. The AMPK inhibitor compound C inhibited infectious EBOV replication in Vero cells and diminished EBOV GP-dependent, but not Lassa fever virus GPC-dependent, entry into a variety of cell lines in a dose-dependent manner. Compound C also prevented EBOV GP-mediated infection of primary human macrophages, a major target of filoviral replication in vivo. Consistent with a role for AMPK in filovirus entry, time-of-addition studies demonstrated that compound C abrogated infection when it was added at early time points but became progressively less effective when added later. Compound C prevented EBOV pseudovirion internalization at 37°C as cell-bound particles remained susceptible to trypsin digestion in the presence of the inhibitor but not in its absence. Mouse embryonic fibroblasts lacking the AMPKα1 and AMPKα2 catalytic subunits were significantly less permissive to EBOV GP-mediated infection than their wild-type counterparts, likely due to decreased macropinocytic uptake. In total, these findings implicate AMPK in macropinocytic events needed for EBOV GP-dependent entry and identify a novel cellular target for new filoviral antivirals.

INTRODUCTION

Filoviruses are enveloped viruses containing a negative sense, nonsegmented RNA genome (1). This virus family is composed of two genera: Ebolavirus and Marburgvirus. The viral glycoprotein (GP) is present on the surface of virions and is composed of a heterotrimeric complex consisting of GP1 that interacts with attachment factors and receptors and GP2 that mediates fusion between the host and viral membrane (2–12). Several cellular adherence factors/receptors have been identified that mediate filovirus internalization, including TIM-1 on epithelial cells and several C-type lectins on antigen-presenting cells (2, 8, 9, 13, 14). Upon internalization, Ebolavirus progresses through endocytic compartments with low-pH-dependent proteolytic processing of heavily glycosylated GP by cathepsins B and L (Cat B/L), generating a 19-kDa form (15, 16). This processed GP interacts with Niemann-Pick C1 (NPC1) in the late endosome or lysosomal compartment (5, 6, 17). Subsequent events that are poorly defined result in membrane fusion.

The route(s) of endocytic uptake of filoviruses remains controversial, with caveolin-mediated endocytosis, clathrin-mediated endocytosis, and macropinocytosis implicated in Zaire Ebolavirus (EBOV) entry (18–24). Strong evidence supports a role for macropinocytosis in a variety of cell types as virion-associated EBOV GP interactions with host cells stimulate actin polymerization events, leading to membrane blebbing, lamellipodium formation, and macropinocytosis resulting in virus uptake (20, 21, 23). In addition to actin polymerization, a variety of cell signaling molecules, including RhoC, phosphatidylinositol 3-kinase (PI3K), Rac1, protein kinase C (PKC), CDC42, PAK1, and Axl, are implicated in macropinocytic uptake of EBOV in various cells (20, 21, 22, 25–27).

AMP-activated protein kinase (AMPK) regulates metabolism by responding to levels of AMP present in the cell (for recent reviews, see references 28, 29, and 30). Not surprisingly, this critical sensor of metabolism has also been implicated in regulating replication of some viruses (31). Two regulatory subunits, β and γ, as well as the catalytic α subunit make up the heterotrimeric AMPK complex, which is conserved in all eukaryotic species (32). Anabolic processes in cells hydrolyze ATP to ADP or AMP. As these anabolic by-products accumulate, AMP binds to the γ subunit of AMPK and causes structural changes and subsequent phosphorylation of Thr172 in the catalytic subunit, which is required for AMPK activity (33). AMP and ADP binding also prevent dephosphorylation of the alpha subunit and subsequent inactivation of the AMPK complex by cellular phosphatases (34). Both Ca²⁺/calmodulin-dependent protein kinase kinase (CAMKK) and LKB1 phosphorylate and activate AMPK, but there are undoubtedly other upstream kinases (35). In general, AMPK activation causes a cellular switch from ATP-consuming to ATP-producing pathways such as fatty acid oxidation, glycolysis, and glucose uptake (28–30, 33).

In addition to a role in regulating metabolism, recent studies have implicated AMPK activity in actin polymerization and the induction of macropinocytosis and phagocytosis (36–40). For instance, macrophages from diabetic mice have reduced macropinocytosis, and leptin-enhanced AMPK activity increased macropinocytosis (38). Another recent study implicated Rac1 as a mediator of AMPK-dependent phagocytosis (36). Moser et al. extended this work, investigating virus internalization using an RNA interference (RNAi) screen of the Drosophila kinome, identifying the importance of AMPK activity for vaccinia virus (VV) uptake via macropinocytosis (39). However, pathways activated by AMPK that lead to actin-dependent macropinocytosis remain poorly defined. With the recent appreciation of the importance of macropinocytosis for filovirus infection, we sought to determine if AMPK is involved in EBOV entry. In this study, we demonstrate that AMPK is required for efficient EBOV entry. Recognition of the importance of this critical cellular kinase for EBOV infectivity opens a new avenue for the development of filovirus antivirals.

MATERIALS AND METHODS

Cells and cell lines.

HeLa and Vero cells were maintained in Dulbecco's modified Eagles medium (DMEM; Gibco-BRL) containing 10% fetal bovine serum (FBS) and penicillin-streptomycin. The NCI-60 panel of tumor lines was purchased from National Cancer Institute (NCI) Developmental Therapeutics Program (DTP). This panel of cells, which included 786-O and SNB-19 cells, was maintained in RPMI medium (Gibco-BRL) containing 5% FBS (HyClone) with penicillin-streptomycin. Wild-type (WT) and AMPK-deficient (AMPK^−/−) mouse embryonic fibroblasts (MEFs) were maintained in DMEM supplemented with 10% FBS, 1% penicillin-streptomycin, 1% HEPES, and 1% l-glutamate. All cells were maintained at 37°C and 5% CO₂ unless otherwise noted.

Peripheral blood mononuclear cells (PBMCs) were isolated from whole human blood using Ficoll-Hypaque as per the manufacturer's instruction (Sigma-Aldrich). Monocyte-derived macrophages (MDMs) were isolated by adherence on gelatin-coated flasks as previously described (41). Freshly isolated MDMs were plated at a density of 5 × 10⁵ cells/well in a 48-well format and allowed to differentiate for 5 days in RPMI medium containing 10% FBS, 10% autologous human serum, and granulocyte-macrophage colony-stimulating factor (10 ng/ml).

Evaluation of pharmacological inhibitors.

AMPK inhibitor compound C (also known as dorsomorphin) (6-[4-(2-piperidin-1-ylethoxy)phenyl]-3-pyridin-4-ylpyrazolo[1,5-a]pyrimidine), 5-(N-ethyl-N-isopropyl)amiloride (EIPA), and CA-074 were obtained from Sigma-Aldrich. Small-molecule inhibitors were incubated at concentrations noted in the figure legends with cells for 1 h prior to viral transduction or infection or at the time noted in the experiment. Cells were transduced with a range of EBOV GP-pseudotyped vesicular stomatitis virus-enhanced green fluorescent protein (VSV-EGFP) (multiplicity of infection [MOI] of 0.05 to 1) in the presence of the inhibitors at concentrations noted in each experiment. Four hours following transduction, the medium was removed and replaced with fresh medium that did not contain inhibitors.

Production of recombinant VSV.

Recombinant, replication-competent VSV that expresses mucin domain-deleted Zaire EBOV GP (EBOV GP/rVSV) was produced as previously described (42). Briefly, the EBOV glycoprotein that has a deletion of the GP1 mucin domain (EBOV GP) was cloned into the genome of a recombinant VSV (Indiana). The recombinant genome contained the EGFP gene in place of VSV-G. EBOV GP was cloned in front of EGFP. Infectious virus was generated as described previously (43) and titers were determined. The mucin domain-deleted EBOV GP was used extensively in this study as previous work has shown that this glycoprotein has a similar tropism to that of the full-length EBOV GP with higher pseudovirion titers (44, 45).

Production of infectious EBOV.

All work with replication-competent EBOV was performed with recombinant Zaire EBOV encoding GFP. This virus was originally supplied by John Towner, CDC, Atlanta, GA. Virus stocks were produced by infecting Vero cells at an MOI of 0.01 and cultivation for 5 days. At this time all cells were fluorescent, indicating complete spread of the virus. The culture supernatant was collected, and cell debris was pelleted by centrifugation at 2,000 × g at 4°C. The supernatant was then collected and overlaid onto a 20% sucrose cushion (20% sucrose, wt/vol, in 20 mM HEPES, pH 7.4), and virus was pelleted by ultracentrifugation at 100,000 × g for 4 h at 10°C. After the supernatant was removed, the pellet was rinsed with phosphate-buffered saline (PBS); the virus pellet was resuspended in PBS, aliquoted, and frozen at −80°C.

Production of VSV pseudovirions.

VSV pseudovirions were produced as previously described (11). Briefly, VSVΔG-EGFP virions bearing mucin domain-deleted EBOV GP, VSV G, or the Lassa fever virus (LFV) GPC were produced in HEK 293T cells by transfection of the glycoprotein-expressing plasmid, followed 24 h later with VSVΔG-EGFP virion transduction). Newly pseudotyped virions were collected for 48 to 72 h, and supernatant was filtered through a 0.45-μm-pore-size filter. Viral transduction was determined 24 h later by analyzing EGFP expression using flow cytometry.

VV production.

HeLa cells were infected with vaccinia virus (VV; Western Reserve strain) that expressed EGFP and incubated for 3 days. Infected cell populations were collected, pelleted, and subjected to freezing and thawing to release intracellular virions (46). Supernatants were collected and frozen at −80°C until use. To determine titers of the stock, serial dilutions of virus were added to Vero cells and quantified by EGFP expression at 24 h following infection.

EBOV transduction and infection studies. (i) CGA screen.

Details of our comparative genomics analysis (CGA) screen have been previously described (8). Briefly, cells of 54 of the 59 NCI-60 cell lines were transduced with equivalent quantities of VSVΔG-EGFP pseudotyped with either VSV-G or EBOV GP ΔO. Viral transduction efficiency was assessed 24 h posttransduction by flow cytometry-based analysis of EGFP expression. The average percentage of EGFP-expressing cells from three independent assays was used as a seed file for COMPARE analysis performed at http://dtp.nci.nih.gov/compare/. Statistical analysis of the correlation findings was performed using the SAS software package (SAS Institute). Cellular genes that correlated with both VSV-G- and EBOV GP-dependent transduction were eliminated from our analysis, focusing the studies on those cellular genes that correlated only with EBOV GP-dependent entry.

(ii) VSVΔG-EBOV pseudovirion transductions.

In studies in which compound C was present only during the first 4 h of transduction, equal numbers of cells were incubated with mucin domain-deleted EBOV GP- or LFV GPC-pseudotyped VSVΔG for 4 h. Most studies used an MOI of pseudovirions of 0.05, but a wide range of MOIs were found to be equivalently sensitive to compound C. Supernatant was then removed and replaced with fresh medium that did not contain inhibitors for a total of 24 h. Viral transduction was determined by analyzing EGFP expression via flow cytometry (BD FACSCalibur).

In studies in which compound C remained on the cells for the duration of the experiment, equal numbers of cells were incubated with VSVΔG-EBOV GP in the presence of compound C. Viral transduction was determined at 24 h by analyzing EGFP expression via flow cytometry.

(iii) EBOV GP/rVSV infections.

Equal numbers of cells were incubated with a dilution series of either WT VSV, vaccinia virus (VV), or VSV with the native glycoprotein replaced with the EBOV glycoprotein, all encoding an EGFP reporter gene. After 4 h, supernatant was removed and replaced with fresh medium. Cells were incubated for 24 h, lifted with trypsin, and fixed in 2% formaldehyde. Infection was analyzed by EGFP expression by either immunofluorescence (IF) microscopy or flow cytometry.

(iv) EBOV infections.

Vero-E6 cells were grown to 80% confluency and treated for 1 h with compound C in dimethyl sulfoxide (DMSO) or with DMSO alone, as indicated in Fig. 3B. Cells were then infected by addition of recombinant Zaire EBOV encoding EGFP at an MOI of 0.05. After 24 h, cells were imaged by an epifluorescence microscope. Fluorescent cells were counted from images using Cell Profiler software (Broad Institute, MA) using a customized algorithm (available upon request). All work with EBOV was performed at biosafety level 4 (BSL4) at Texas Biomedical Research Institute, San Antonio, TX.

Fig 3.

Compound C inhibits EBOV infection. Vero E6 cells were pretreated with compound C for 1 h, and replication-competent Zaire EBOV that expresses EGFP was added at an MOI of 0.05. After 24 h, cells were imaged and infected cells in micrographs were counted. Data were normalized to untreated cells and studies were performed in triplicate. (A) Bright-field and fluorescent images (EGFP) of a representative micrograph of EBOV infection in the presence of DMSO or 25 μM compound C. (B) Compound C dose-response curve for inhibition of EBOV infection. Shown are mean and standard error of the relative number of EBOV-infected cells normalized to untreated cells after treatment with compound C.

Confocal microscopy.

Vero cells (4.5 × 10⁴) were grown on poly-l-lysine-coated coverslips and incubated at 37°C for 2 h in the presence or absence of 10 μM compound C before incubation with VSVΔG-EBOV (MOI of ∼200) or equivalent concentrations of VSVΔG that does not contain a glycoprotein (no Env) for 30 min at 4°C. The supernatant was aspirated and replaced with prewarmed DMEM with and without compound C. Virus was allowed to internalize for 0 or 30 min at 37°C before coverslips were fixed in 2% paraformaldehyde as previously described (20). Cells were permeabilized with 0.2% Triton X-100 for 10 min at room temperature. Actin was visualized by staining with Alexa Fluor 647-phalloidin (Invitrogen) for 30 min at room temperature. Microscopy was performed on a Nikon C1 confocal microscope. Z-series (1.0-μm slices) encompassing the entire cellular field at ×600 under oil immersion were obtained. The total phalloidin signal (number of voxels) of the complete Z-series was determined using Imaris, version 7.6, software (Bitplane Scientific Software). The number of voxels detected was divided by the total number of cells within the analyzed Z-stack to generate the number of voxels/cell.

Dextran uptake studies.

Vero cells (6 × 10⁴) or wild-type and AMPK^−/− MEFs (4 × 10⁴) were seeded in a 48-well format. Vero cells were pretreated with compound C or EIPA for 1 h at 37°C, as indicated in the legend to Fig. 6. All cells were incubated with 70,000-kDa dextran (0.5 μg/ml) conjugated to fluorescein (Invitrogen) for 1 h at 37°C in the presence or absence of drug. Extracellular dextran was then removed with a 1× PBS wash. Dextran uptake was assayed for FL-1 intensity (mean fluorescence intensity [MFI]) via flow cytometry.

Fig 6.

Actin polymerization is stimulated by EBOV GP pseudovirions in an AMPK-dependent manner, stimulating macropinocytosis. (A) Time course of actin polymerization following EBOV GP pseudovirion incubation in the presence or absence of 10 μM compound C. Vero cells were preincubated with and without drug at 37°C for 2 h. Cells were then incubated with EBOV pseudovirions (MOI of 200) in the presence or absence of drug for 30 m at 4°C. Virus-containing medium was removed, and fresh medium with and without compound C was added before the temperature was shifted to 37°C for 30 m. Cells were fixed in 2% paraformaldehyde and permeabilized, and polymerized actin was imaged by confocal microscopy using Alexa Fluor 647-phalloidin. To quantify the polymerization activity, the total phalloidin staining of 6 to 12 images composed of 6 to 19 cells/image was quantified using the Imaris software program and averaged for the number of voxels/cell. (B) Compound C reduces dextran uptake into Vero cells. FITC-conjugated 70-kDa dextran was incubated with Vero cells treated with either 10 or 50 μM compound C or 5 or 25 μM EIPA, a well-established macropinocytosis inhibitor. One hour later, cells were lifted and assessed for fluorescent dextran uptake by flow cytometry. (C) AMPK^−/− MEFs have reduced dextran (Dex) uptake compared to WT MEFs. FITC-conjugated 70-kDa dextran was incubated with equivalent numbers of WT or AMPK^−/− MEFs for 1 h at 37°C. Cells were washed thoroughly and assessed for dextran uptake by flow cytometry. *, P < 0.05; **, P < 0.001.

Internalization assay.

Vero cells (1 × 10⁵) were plated in a 24-well format. Cells were preincubated with 25 μM EIPA or CA-074 or 50 μM compound C for 1 h before being cooled to 4°C. Cells were incubated with VSVΔG-EBOV pseudovirions for an additional hour at 4°C. For baseline values at time zero, medium was removed, and cells were incubated with two 5-min incubations in trypsin at 37°C to remove surface-attached virions. These cells were subsequently replated in fresh medium. To obtain rates of virus internalization inhibitor and unbound virus were removed at time zero, and medium containing the appropriate concentration of inhibitor was added at 37°C for the times noted in Fig. 7. Any surface-exposed virus was removed by two 5-min incubations with trypsin at 37°C. Twenty-four hours later, EGFP expression in cells was determined by flow cytometry to assess virus entry.

Fig 7.

AMPK is required for EBOV internalization from the cell surface. (A) Vero cells were incubated with a macropinocytosis inhibitor (EIPA; 25 μM), Cat B inhibitor (CA-074; 25 μM), AMPK inhibitor (compound C; 50 μM), or a DMSO control for 1 h. Treated cells were transduced with EBOV GP pseudovirions (MOI of 0.05) for an additional 4 h. Virus and inhibitor were then replaced with fresh DMEM. Viral transduction was determined 24 h later by EGFP expression via fluorescence-activated cell sorting. (B) Vero cells were incubated with the indicated inhibitor for 1 h before cells were chilled to 4°C. EBOV GP pseudovirions were then allowed to bind to the cells for an additional hour at 4°C. Surface-bound virus was either removed by trypsin treatment prior to shifting the temperature to 37°C (control, 0 min) or allowed to transduce for 1 h at 37°C in the presence of inhibitor or the DMSO control prior to trypsin treatment. After 1 h, cells were refreshed with medium not containing inhibitor. Viral transduction was determined 24 h later by EGFP expression via flow cytometry. **, P < 0.01; ***, P < 0.001.

Statistics.

Each experiment was performed in triplicate. Statistical significance was determined by a two-tailed Student's t test.

RESULTS

AMPKγ gene expression correlates with EBOV transduction.

We performed a large-scale bioinformatics screen called a comparative genomic analysis (CGA) screen to identify cellular genes important in EBOV GP-dependent transduction. Details of the screening approach have been previously reported (8, 22, 47, 48). This screen utilized the NCI-60 panel of human tumor lines, and this panel of cells was redundantly evaluated for mRNA expression in gene array assays, providing mRNA expression profiles for each cell line. These gene array findings are available in the public domain (http://dtp.nci.nih.gov). The permissivity of the cell lines to VSVΔG-EBOV or to VSVΔG pseudotyped with its native glycoprotein G (VSVΔG-G) was determined by incubating each cell line with equal numbers of pseudovirions. The resulting percentage of cells that were EGFP positive was correlated with the gene expression data of each line by the COMPARE algorithm. Genes identified to correlate with EBOV GP- or VSV-G-dependent transduction were ranked by the Pearson correlation coefficient (PCC). Those genes within the EBOV GP listing that correlated with both EBOV GP- and VSV-G-dependent transduction were subtracted, yielding a subset of genes that were likely to be specifically involved with EBOV GP-dependent entry rather than downstream VSV expression events. Importantly, several previously identified EBOV entry factors including Axl, the C-type lectins, and integrin β1 correlated strongly with EBOV transduction in our screen (2, 9, 14, 20, 49–55). In addition, several signaling molecules that previously were reported to be important for EBOV transduction/infection were also identified in this screen, including PI3K, phospholipase C (PLC), and PKC (20, 21, 23, 27). As our screen was successful in identifying these pathways, we mined the results to identify additional signaling molecules important in EBOV entry. Intriguingly, the γ2 subunit of AMPK correlated strongly with EBOV transduction (PCC of 0.489) (P = 0.0000875) (Fig. 1). As AMPK has recently been reported to be important for VV uptake via macropinocytosis (39) and as macropinocytosis is an important mechanism of EBOV entry (20, 21, 23, 26), we targeted this gene for further study.

Fig 1.

Findings from our comparative genome analysis screen. EBOV GP-mediated transduction and AMPK mRNA expression in the human NCI-60 tumor panel of cell lines. (A) Relative levels of AMPK mRNA in 54 of the 59 cell lines that compose the NCI-60 human tumor cell panel. mRNA levels were obtained from gene array studies that are available to the public at the NIH NCI DTP website (http://dtp.nci.nih.gov). (B) VSVΔG-EBOV transduction efficiency of the same NCI-60 cell lines. NSCLC, non-small-cell lung cancer. Leuk, leukemia; P, prostate; CNS, central nervous system.

A small-molecule inhibitor of AMPK inhibits EBOV infection.

In our initial studies, the AMPK requirement for EBOV entry was explored in several cell lines derived from disparate tissues using compound C (also known as dorsomorphin) that is a membrane-permeable, selective inhibitor of AMPK (56). SNB-19 cells, a neuroblastoma cell line, and the renal cell line 786-O are both in the NCI-60 cell panel and express abundant AMPKγ subunit mRNA (Fig. 1, central nervous system [CNS] 4 and renal 3). Vero, SNB-19, 786-O, and the cervical cancer line HeLa were incubated with compound C or vehicle control for 1 h. EBOV GP-pseudotyped VSV or LFV GPC-pseudotyped VSV that was used as a control virus throughout these studies was incubated with these cells for an additional 4 h. The drug and virus were removed, medium was refreshed, and viral transduction was analyzed by evaluating transduced cell EGFP expression via flow cytometry at 24 h. Compound C significantly reduced EBOV pseudovirion transduction in all cells tested although the ability of a 10 μM concentration to block EBOV transduction varied between the cell lines (Fig. 2A). Additionally, this inhibition was specific to EBOV GP-dependent transduction as compound C had no effect on LFV GPC-dependent entry. As the viral core of both pseudovirion stocks was VSV, our findings suggested that this inhibitor was specifically interfering with EBOV GP-dependent entry into these cells.

Fig 2.

Compound C inhibits EBOV transduction of cell lines. (A) Ability of compound C to block EBOV GP transduction in a number of cell lines. The indicated lines were treated with compound C (10 μM) or the DMSO control for 1 h at 37°C. Either EBOV GP or LFV GPC pseudovirions (MOI of 0.05) were added to the cultures. Virus was allowed to internalize for 4 h before virus and inhibitor were removed, and cells were refreshed with medium without inhibitor for an additional 24 h when transduction levels were determined by assessing EGFP expression in transduced cells by flow cytometry. Findings are shown as a percentage of EGFP expression in cells treated with compound C divided by EGFP expression in cells treated with DMSO. (B) Effect of compound C on EBOV GP pseudovirion transduction, LFV GPC pseudovirion transduction, and cellular cytotoxicity (viability) in Vero cells when cells were incubated with compound C or DMSO for 1 h at 37°C for the initial 4 h of the transduction. A range of MOIs from 0.1 to 1 was used for these studies with similar findings (n = 5). After this incubation, cells were refreshed with medium without inhibitor, and cells were assessed for EGFP expression at 24 h following transduction. (C) Effect of compound C on EBOV GP pseudovirion transduction, LFV GPC pseudovirion transduction, and cellular cytotoxicity (viability) in Vero cells incubated with compound C or DMSO for the duration of the 24-h experiment. (D) Time-of-addition experiments with compound C. Vero cells were precooled to 4°C and incubated with EBOV GP pseudovirions. Unbound virus was removed, and the cells were shifted to 37°C. Compound C (50 μM) or DMSO control was added at the indicated time following the temperature shift. At 6 h following initiation of transduction, virus and inhibitor were removed, and cells were refreshed with medium. Viral transduction was assayed 24 h later by flow cytometry. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

To more closely assess the effect of compound C on EBOV pseudovirion transductions, dose-response curves of the inhibitor were performed in Vero cells. In these studies, inhibitor was either removed after the first 4-h incubation of transduction (Fig. 2B) or maintained on the cells for the duration of the 24-h experiment (Fig. 2C). Compound C inhibited EBOV pseudovirion transduction with a 50% inhibitory concentration (IC₅₀) of ∼6 μM, regardless of whether the drug was maintained or removed from the culture after 4 h. In contrast, the tested concentrations of compound C had no effect on LFV GPC-dependent entry. Maintenance of the inhibitor on the cells for the duration of the experiment resulted in greater levels of cell cytotoxicity, and, consequently, in most of our studies that used compound C, the inhibitor was removed following incubation during the initial 4 h of transduction.

A time-of-addition study was performed to understand the time frame of compound C-dependent inhibition of EBOV transduction. Vero cells were incubated with equal amounts of EBOV pseudovirions for 1 h at 4°C. Unbound virus was removed, and the cells were shifted to 37°C, and compound C or DMSO was added to the cells at 0, 15, and 30 min following the temperature shift and every hour thereafter for 4 h. Virus and compound C were incubated with cells for a total of 6 h before being replaced with fresh medium. Addition of compound C at or close to the initiation of viral transduction significantly reduced EBOV pseudovirion entry (Fig. 2D). This effect was time sensitive, with compound C becoming ineffective when added after 4 h following initiation of transduction, implicating AMPK in early life cycle events.

Compound C was evaluated for inhibition of replication-competent EBOV that expresses EGFP during infection (Fig. 3). Infections were performed for 24 h in the presence of serial dilutions of compound C. The numbers of EGFP-expressing cells were assessed in these experiments. In a manner similar to our transduction studies, we found that the inhibitor was highly effective at blocking EBOV infection. Somewhat surprisingly, inhibition was more than 45-fold stronger than that observed in our transduction studies, with an IC₅₀ of 0.13 μM. This enhanced sensitivity to compound C of our infectious virus compared to our transducing virions is likely due to the fact that VSV replication is reduced in the presence of AMPK, as recently reported (57), and inhibition of AMPK activity may enhance replication of those VSV particles that are able to enter cells. This work provides strong evidence that AMPK activity is important for EBOV infection. Furthermore, as the only EBOV viral protein present in both the VSV transductions and EBOV infections was EBOV GP, these studies implicate AMPK activity in EBOV entry events.

AMPK is required for efficient EBOV GP-mediated infection of human macrophages.

Macrophages are critical targets of EBOV infection in vivo (58–63). Consequently, we next determined if EBOV GP-dependent infection of human macrophages requires AMPK. In these studies, we used a previously described, replication-competent, recombinant VSV with the EBOV GP and EGFP genes cloned into the VSV genome in place of the native G glycoprotein (EBOV GP/rVSV) (8). Purified, in vitro matured human monocyte-derived macrophages (MDMs) from two donors were incubated for 1 h with DMSO or compound C and infected with EBOV GP/rVSV (MOI of 1) for an additional 4 h. Virus and drug were removed, and the cells were washed and maintained for an additional 24 h. Viral infection was analyzed by EGFP expression using fluorescence microscopy. EBOV GP/rVSV readily infected MDMs, and compound C significantly reduced EBOV GP-mediated infection, suggesting that AMPK is important for EBOV GP-dependent infection of a primary, physiologically relevant cell type (Fig. 4A). The number of EGFP-expressing cells in the micrographs was quantitated by Volocity Image Analysis software (Perkin-Elmer), and an average of 452 cells were GFP positive in the control populations, whereas no EGFP-expressing cells were detected in the compound C-treated wells. Interestingly, the compound C treatment altered the morphology of MDMs, producing elongated cells; however, compound C cytotoxicity to MDMs was modest (Fig. 4B).

Fig 4.

AMPK inhibition significantly reduces EBOV GP-mediated infection of primary human MDMs. (A) MDMs were incubated with DMSO or compound C (50 μM) for 1 h. Treated cells were infected with EBOV GP/rVSV for an additional 4 h. Unbound virus and drug were removed, and viral infection was visualized for EGFP expression at 24 h postinfection by fluorescence microscopy. Experiments were performed in triplicate with two donors. Shown is a representative experiment. (B) Cytotoxicity of compound C in MDMs. Experiments were performed in quadruplicate with two donors, and the mean and standard deviation of the mean are shown.

Cells lacking functional AMPK are less susceptible to EBOV GP-mediated infection.

As our studies on AMPK had thus far involved a pharmacological inhibitor, we employed mouse embryonic fibroblasts (MEFs) lacking the α1/2 catalytic subunits of AMPK (AMPKα1/α2^−/−) to extend and verify our observations (39, 64–66). AMPK in these cells cannot become activated and is unable to phosphorylate downstream targets (39, 57, 64, 66). Wild-type (WT) and AMPK^−/− MEFs were infected with wild-type VSV (Indiana strain), vaccinia virus (VV), or EBOV GP/rVSV, each of which expressed EGFP as a reporter gene (MOI of ∼1). Virus expression was assayed at 24 h following infection by fixing the cells and analyzing EGFP expression using flow cytometry. AMPK^−/− MEFs displayed significantly lower permissivity to VV than WT MEFs, as previously described (Fig. 5) (39). Interestingly, EBOV GP/rVSV infection of AMPK^−/− MEFs was even more reduced than infection of WT control MEFs, with less than 10% of wild-type infection in the knockout MEFs. This reduction in EBOV GP-mediated infection was specific for the glycoprotein, as VSV-G-dependent infection was equivalent in both WT and AMPK^−/− MEFs, as previously reported (39). These findings provided additional evidence that AMPK activity in the host cell was required for optimal EBOV GP-dependent entry.

Fig 5.

AMPK^−/− MEFs support lower levels of EBOV infection than WT MEFs. Equal numbers of WT or AMPK^−/− MEFs were infected with VSV, VV, or EBOV GP/rVSV (MOI of 1). Viral infection was assayed by EGFP expression at 24 h postinfection via flow cytometry. ***, P < 0.001.

AMPK activity enhances actin polymerization by EBOV GP and dextran uptake.

Incubation of host cells with VV and EBOV causes large actin protrusions and increased macropinocytosis (21, 23, 39, 67, 68). To determine if AMPK was involved in actin polymerization in the presence of EBOV pseudovirions, we incubated Vero cells with VSV particles pseudotyped with EBOV GP (MOI of 200) or equivalent concentrations of VSV that did not contain a GP (Fig. 6A, No GP) in the presence and absence of 10 μM compound C. After a 30-min incubation in medium with or without drug at 4°C to allow virus binding and synchronized entry, medium was removed, and prewarmed medium with or without compound C was added to cells for 0 or 30 min before fixation and visualization. Polymerized actin stress fibers were detected by Alexa Fluor 647-phalloidin staining and confocal microscopy. Polymerized actin staining in 6 to 12 Z-series were captured, analyzed, and quantified by voxel numbers using Imeris imaging software (Fig. 6A). The amount of actin polymerization/cell appeared to be slightly higher at time zero in the presence of EBOV GP pseudovirions than with the no-Env pseudovirions although this difference was not statistically significant. However, as other investigators have previously found (21, 23), the addition of EBOV caused significant actin polymerization during entry in cells at 30 min. Polymerization was inhibited by compound C. As a reduction in actin polymerization would be anticipated to result in reduced macropinocytosis, we evaluated the ability of both Vero cells treated with compound C and WT and AMPK^−/− MEFs to take up high-molecular-weight dextran, a cargo commonly used to assess uptake by macropinocytosis. In our Vero cells, we also used EIPA, a specific inhibitor of the plasma membrane Na⁺/H⁺ antiporter that is known to block macropinocytosis and EBOV entry, as a control (20, 69). We found that both EIPA and compound C modestly, but significantly, inhibited fluorescein isothiocyanate (FITC)-conjugated 70-kDa dextran uptake (Fig. 6B). In a similar manner, dextran uptake by AMPK^−/− MEFs was significantly reduced compared to WT MEFs (Fig. 6C). In neither of these experiments was dextran uptake as profoundly inhibited by the absence of AMPK activity as EBOV infection, suggesting that either dextran uptake can occur in these cells independent of macropinocytosis or that other steps in EBOV infection in addition to macropinocytosis are regulated by AMPK.

Compound C inhibits EBOV GP pseudovirion internalization.

As dextran uptake was reduced in the absence of AMPK functionality, we next sought to determine if EBOV GP pseudovirion internalization was also reduced by the loss of AMPK activity. Additional inhibitors used to control this study included EIPA and CA-074. CA-074 is a cathepsin B inhibitor that allows virion internalization but blocks the proteolysis and subsequent fusion of GP in the late endosome/lysosome (15, 16, 70). In a first series of experiments, Vero cells were incubated with effective concentrations of EIPA, CA-074, compound C, or the DMSO control for 1 h before addition of EBOV GP pseudovirions. Transduction was performed in the presence of inhibitors for an additional 4 h before the virus and inhibitor were removed and replaced with fresh medium. Transduction was assessed 24 h later. As expected, all three compounds inhibited EBOV GP pseudovirion transduction of Vero cells (Fig. 7A). In a second set of experiments, these same concentrations of inhibitors were used in an internalization assay. Vero cells were incubated with EIPA, compound C, CA-074, or DMSO for 1 h before the cells were chilled to 4°C. Cells were then incubated with virus for an additional hour at 4°C to synchronize entry. The supernatant was replaced with prewarmed medium containing the inhibitor or DMSO for 1 h at 37°C. At times noted on the figure (0 min or 60 min), extracellular virus was removed from the cell surface by two 5-min incubations with trypsin. The treated cells were replated in fresh medium without inhibitors for 24 h. Cells that were treated with trypsin without the temperature shift displayed low levels of virus transduction compared to cells where virus was allowed to internalize for 1 h, demonstrating that the trypsin treatment removed surface-bound virions (Fig. 7B). After 1 h of internalization in the presence of EIPA, transduction was low when the cells were treated with trypsin, indicating that the virions remained on the surface of cells. Interestingly, a similar finding was observed in the presence of compound C, demonstrating that AMPK is also required for EBOV internalization from the cell surface. As expected, cathepsin B inhibitor CA-074 had no effect on internalization of the pseudovirions from the cell surface, and, thus, under this second set of conditions, subsequent transduction was not inhibited.

DISCUSSION

Uptake via cellular endocytic machinery allows incoming viruses to bypass the crowded cellular meshwork found in the cytoplasm near the plasma membrane, delivering the virus to a site suitable for replication (71, 72). Macropinocytosis is a well-established endocytic mechanism of uptake of large cargo that involves actin-dependent extension of lamellipodia that form vesicles (73). Macropinocytic uptake of eukaryotic viruses is relatively recently appreciated; however, already a number of genetically and morphologically disparate viruses have been reported to enter cells via this mechanism, including filoviruses, vaccinia virus, influenza A virus, and African swine fever virus (20, 21, 23, 68, 74, 75). Here, we found that loss of activity of a key metabolic regulator, AMPK, decreases actin polymerization and reduces both EBOV and dextran uptake.

AMPK was identified as potentially important for EBOV entry through a bioinformatics-based screen that correlated mRNA expression in a panel of tumor cell lines with permissivity of that cell panel for EBOV GP-mediated entry. EBOV GP pseudovirion transduction or EBOV infection was inhibited in the absence of AMPK activity, and this inhibition was specific for the viral glycoprotein as compound C had no effect on entry of VSV pseudovirions containing LFV GPC. This finding implicated AMPK-regulated pathways in EBOV GP-dependent entry events. As EBOV entry has been recently shown to be mediated by actin-dependent macropinocytosis (20, 21, 23), we compared actin polymerization in the presence or absence of compound C. EBOV pseudovirion-stimulated actin polymerization was inhibited in the absence of AMPK. Consistent with this, EBOV pseudovirions remained on the surface of cells with defective AMPK activity. Thus, we propose that AMPK activity enhances EBOV infection by regulating virus uptake into cells via macropinocytosis. It is possible that AMPK regulates additional steps in EBOV entry and expression since EBOV was much more effectively inhibited by the loss of AMPK activity than high-molecular-weight dextran uptake, a cargo frequently used to assess macropinocytosis.

AMPK activity has been shown to be important for macropinocytic uptake of VV (39). VV is reported to enter cells by stimulating actin polymerization, membrane blebbing, lamellipodium formation, and macropinocytosis (67, 76, 77), and Moser et al. extended these studies by demonstrating the importance of AMPK in VV entry (39). Our findings complement and broaden these studies by identifying the importance of AMPK activity for EBOV entry and infection as well.

Interestingly, activation of AMPK enhances entry and subsequent levels of infection of some enveloped viruses while it inhibits the replication of other enveloped viruses. Moser et al. recently showed that activation of this kinase robustly inhibits replication of the Bunyavirus Rift Valley fever virus through the reduction of lipid levels in infected cells (57). In this study, wild-type VSV was also modestly inhibited by AMPK activation. Thus, it is remarkable that we observed a profound inhibition of EBOV GP/rVSV infectivity by loss of AMPK activity since replication of our recombinant VSV would be expected to be enhanced by reduced activity of this kinase.

AMPK-dependent signaling pathways that result in macropinocytosis are not defined. Several recent studies hint at possible links between AMPK activity and the formation of actin-mediated lamellipodium formation and macropinocytosis (38, 39). Rac1, PI3K/Akt, protein kinase C, and myosin light chain kinase (MLCK), which have either been previously identified to be involved in EBOV entry and macropinocytosis and/or were identified in our CGA screen, are known to be phosphorylation targets downstream of AMPK activation (36, 78–82).

One possible means by which activation of AMPK may lead to altered macropinocytosis is by relocalization and altered function of components of the protruding lamellipodia. AMPK activation is known to phosphorylate the anticapping protein vasodilator-stimulated phosphoprotein (VASP) (37, 40, 83). Unphosphorylated VASP prevents actin capping and enhances actin filament elongation by binding to the barbed end of actin, whereas actin-capping proteins are critical in the formation of lamellipodia (84–87). Phosphorylation of VASP inhibits its anticapping activity, and sequestration or phosphorylation of VASP correlates with increased cell motility although how VASP is regulated at the tip of lamellipodia is not entirely clear (84, 88–91). Thus, a lack of AMPK-dependent VASP phosphorylation may explain the decreased motility observed in AMPK^−/− MEFs and reduced lamellipodium formation (39). Intriguingly, phosphorylated VASP has been observed at sites of EBOV entry (23).

We have previously shown that filovirus entry into some cells such as Vero cells require PI3K/Akt activity, whereas EBOV entry of other cells such as SNB-19 cells is independent of PI3K/Akt activity but requires PLC (20). Interestingly, here we show that EBOV transduction of both of these highly permissive lines required AMPK, suggesting that either the AMPK requirement is independent of both of these pathways, that AMPK can activate either pathway, or that AMPK is activated downstream from these pathways. Moser et al. speculate that one or more Rho GTPases may be activated downstream from AMPK activation (39). As we have shown that Rho B and C enhance EBOV GP-dependent uptake (22), our newest findings lend indirect support for this possibility.