Ebola Virus Uses Tunneling Nanotubes as an Alternate Route of Dissemination

Abstract

Ebola virus (EBOV) disease is marked by rapid virus replication and spread. EBOV enters the cell by macropinocytosis and replicates in the cytoplasm, and nascent virions egress from the cell surface to infect neighboring cells. Here, we show that EBOV uses an alternate route to disseminate: tunneling nanotubes (TNTs). TNTs, an actin-based long-range intercellular communication system, allows for direct exchange of cytosolic constituents between cells. Using live, scanning electron, and high-resolution quantitative 3-dimensional microscopy, we show that EBOV infection of primary human cells results in the enhanced formation of TNTs containing viral nucleocapsids. TNTs promote the intercellular transfer of nucleocapsids in the absence of live virus, and virus could replicate in cells devoid of entry factors after initial stall. Our studies suggest an alternate model of EBOV dissemination within the host, laying the groundwork for further investigations into the pathogenesis of filoviruses and, importantly, stimulating new areas of antiviral design.

Ebola virus (EBOV) disease is characterized by an exceptionally efficient spread of uncontrolled viral replication in host tissues, evasion of host immune defenses, and persistence in immune-privileged sites. Macrophages are among the initial cells targeted by EBOV [1]. EBOV propagates infection through the cell-free form, where virus particles enter the cell by macropinocytosis, replicate in the cytoplasm, and then egress from the cell surface to infect new cells [2–4].

EBOV is an enveloped virus with a negative-strand RNA genome. The glycoprotein (GP) mediates the entry of cell-free virions. Nucleoprotein (NP) forms complexes with VP35, VP30, and the RNA-dependent RNA polymerase L to facilitate virus genome replication and transcription. NP, VP35, and VP24 proteins are required for proper nucleocapsid formation. The matrix protein (VP40) is the main driving force for EBOV budding at the cell surface from virus-induced filopodia [1, 5, 6].

Viruses require host cell processes to maximize replication and spread. Several clinically significant viruses, including influenza virus, HIV-1, and SARS-CoV-2, were shown to exploit intercellular tunneling nanotubes (TNTs) to move their genome to naive cells and bypass host immunity [7–10]. TNTs are F-actin–based tubular structures several hundred microns long that connect 2 homo- or heterologous cells, providing continuity of plasma membrane and cytoplasm between the connected cells. This allows the cells to rapidly exchange components, including proteins, RNA, and organelles (eg, mitochondria and endosomes) [11, 12]. Because the transport speed through TNTs is rapid, it would be energetically more favorable for a virus to traffic the genome intercellularly to propagate infection, in contrast to undergoing the conventional replication cycle. Additionally, viral genome transfer by TNTs would broaden the host cell tropism, further contributing to disease development, and it would facilitate virus persistence in the host, effectively avoiding host immune defenses and countermeasures [7]. Understanding how viruses exploit TNTs during infection is therefore key to designing more effective treatments. Herein, we investigated the involvement of TNTs in EBOV intercellular dissemination.

METHODS

Cells and Viruses

Vero cells were cultured in Dulbecco's Modified Eagle Medium (Thermo Fisher) supplemented with 10% fetal bovine serum (Thermo Fisher). Pooled primary human umbilical vein endothelial cells (HUVECs; ATCC) were maintained in a vascular cell basal medium with added nutrients from an endothelial cell growth kit (ATCC). Peripheral blood was collected from healthy adult human donors after obtaining their consent according to the approved protocol (20180013HU; Institutional Review Board, University of Texas Health) to prepare monocyte-derived macrophages (MΦs) as described previously [13, 14]. MΦ cultures and HUVEC-MΦ cocultures were maintained in RPMI medium supplemented with 10% autologous serum. All cells were grown at 37 °C in a humidified 5% CO₂ incubator.

Experiments with replication-competent viruses were performed in the biosafety level 4 laboratory at the Texas Biomedical Research Institute (Texas Biomed) according to standard operating procedures and protocols approved by the institute's Biohazard & Safety and Recombinant DNA Committees. The NCBI accession numbers for filoviruses used in these studies were as follows: NC_002549, EBOV variant Mayinga; KF990213, recombinant EBOV variant Mayinga encoding green fluorescent protein (GFP); NC_001608, Marburg virus (MARV) strain Musoke; NC_006432, Sudan virus (SUDV) strain Gulu; and MT796851, Reston virus (RESTV) variant R08. EBOV, EBOV-GFP, and MARV were obtained from the Texas Biomed repository. SUDV and RESTV were provided by Ricardo Carrion (Texas Biomed). Recombinant Lassa virus (LASV) strain Josiah (NCBI accessions HQ_688672 and HQ_688674) encoding GFP was rescued as previously described [15]. All viruses were grown and concentrated as before [16], and viral titers were determined by incubating serial dilutions of the stocks on Vero cells. After 24 hours, EBOV-GFP– and LASV-GFP–infected cells were treated with Hoechst dye (Thermo Fisher) to stain nuclei, photographed by a Nikon automated system, and analyzed by CellProfiler software (Broad Institute) to quantify infected cells and nuclei (GFP positive). To determine titers of the wild type EBOV, MARV, SUDV, and RESTV, 10-fold serial dilutions of the viruses were added to Vero cells for 1 hour at 37 °C. After the virus was removed, the cells were overlaid with Dulbecco's Modified Eagle Medium containing 2% fetal bovine serum and 1.5% methyl cellulose (MilliporeSigma) for EBOV, MARV and RESTV or Eagle's Minimum Essential Medium supplemented with 2% fetal bovine serum and 1% agarose for SUDV. For EBOV, MARV, and RESTV, cells were fixed 10 days after infection and then stained with gentian violet dye (Ricca Chemical Company) to identify virus foci. For SUDV, at 7 days after infection, cells were overlaid with secondary overlay containing 8% neutral red (Thermo Fisher) for 24 hours to count virus foci.

Antibodies and Inhibitors

Antibodies to detect EBOV proteins were rabbit polyclonal antibody to NP, VP35, VP40; mouse monoclonal antibody to GP (4F3); and human monoclonal antibody to GP (KZ52; all obtained from Integrated BioTherapeutics). Human IgG isotype control antibody was from Thermo Fisher. A single-domain llama antibody (sdAb) differentially reactive to various ebolaviral NPs (sdAb ZE) was employed as a bivalent fusion protein to dimeric enhanced ascorbate peroxidase [17], while a sdAb specific for MARV NP (sdAb A) was used as a monomeric fusion [18]. The sdAb fusion proteins were detected with His-tag (D3I10) rabbit monoclonal antibody conjugated to Alexa Fluor 647 (Cell Signaling). A mouse monoclonal antibody to LASV NP (100LN) was obtained from Zalgen Labs. Antibodies to detect host proteins were rabbit polyclonal to Lamp-1 (MilliporeSigma) and COX IV (Cell Signaling), mouse monoclonal to TNFAIP2 and LC3B (Santa Cruz Biotechnologies), and α-tubulin conjugated to Alexa Fluor 555 (DM1A; Thermo Fisher).

Dimethylsufoxide (DMSO) was purchased from ATCC. EIPA (5-[N-ethyl-N-isopropyl] amiloride) and nocodazole were obtained from MilliporeSigma. Cytochalasin D (cytD) was obtained from MedChemExpress.

Plasmids

mPol-I/LASV-Lag, mPol-I/LASV-Sag/GFP-2A-NP, pCAGGS-LASV-NP, and pCAGGS-LASV-L have been described [15] and were provided by Juan Carlos de la Torre (Scripps Research). Rescue of recombinant LASV encoding GFP was conducted as detailed previously [15].

Previous work has also provided information on EBOV protein–expressing plasmids pCAGGS-NP, pCAGGS-VP35, pCAGGS-VP24, pCAGGS-VP30-GFP, pCAGGS-VP40, and pcDNA3-GP, as well as pCAGGS-EGFP [5, 16]. mRFP-N1 was a gift from Robert Campbell, Michael Davidson, and Roger Tsien (plasmid 54635; Addgene) [19]. HUVECs were used in all experiments requiring plasmid transfection due to their transfection efficiency.

Cell Viability

Cytotoxic properties of EBOV infection and compound/antibody treatments were tested in MΦs with CellTiter-Glo (Promega). In virus assays, cells plated into 96-well plates at 5 × 10⁴ cells/well were mock infected or infected with EBOV at a multiplicity of infection (MOI) of 1 or 5 for 24, 48, or 72 hours, with the inoculum remaining throughout the experiment. Different MOIs were used to determine whether the virus’s effect on cell viability is dose dependent. The effect of infection on cell viability was determined by an ordinary 1-way analysis of variance (ANOVA) Dunnett multiple-comparison test. In treatment studies, cells were left untreated; treated with six 2-fold serially diluted concentrations of EIPA, cytD, nocodazole, or GP KZ52 antibody; or treated with equal concentrations of the solvent, DMSO, or IgG isotype control for 72 hours in triplicate. The number of metabolically active cells was determined by measuring luciferase activity. In treatment studies, nonlinear regression analysis was performed with GraphPad software to select a nontoxic concentration range for each treatment, relative to the untreated samples, for studies of virus spread. Each experiment was performed 3 times.

Scanning Electron Microscopy

In homologous cultures, MΦs were plated into wells of 12-well plates at 2 × 10⁵ cells/well. In cocultures, MΦs and HUVECs were plated at a 1:1 ratio to 2 × 10⁵ cells/well. After 24 hours, cells were infected with EBOV at an MOI of 5 for 24 or 48 hours and then fixed. The virus inoculum remained throughout the experiment. The MOI of 5 was used to ensure that every cell was targeted by the virus. Twenty-four hours is sufficient for the virus to complete 1 cycle of replication, and the 48-hour time point was used to assess a time dependence of TNT formation. Samples were sputter coated with a layer of gold/palladium and then imaged by a JEOL scanning electron microscope. For each condition, the number of MΦs with TNTs in electron micrographs was counted manually in ≥100 cells and reported as a ratio to the total number of cells. TNT length and width were determined with ImageJ software in ≥50 TNTs per experimental condition. The results from 3 independent experiments were combined for analysis between mock- and EBOV-infected samples. The data were analyzed by a 2-way ANOVA Šídák multiple-comparison test.

Live Microscopy

To image TNT formation in MΦ cultures, cells cultured in 8-well slides (ibidi) at 10⁵ cells/well were infected with EBOV-GFP at a MOI of 1 or left uninfected for 24 hours. The virus inoculum remained throughout the experiment. The samples were photographed in fluorescence and brightfield modes via a 20× objective with air immersion.

To study nucleocapsid transfer between cells, 10⁵ HUVECs cotransfected with pCAGGS-NP/pCAGGS-VP35/pCAGGS-VP24/pCAGGS-VP30-GFP were plated into wells of a 12-well plate and imaged 48 hours later with the EVOS M7000 system (Thermo Fisher). The plate was placed into the onstage incubator at 37 °C with 5% CO₂. Images were acquired in fluorescence and brightfield modes every 10 seconds for 20 minutes via a 40× long working distance phase objective with air immersion. The images were analyzed and converted into a movie format with Celleste analysis software (Thermo Fisher). The speed of the nucleocapsid transfer was determined as follows: the distance that a nucleocapsid traveled / the time of the transfer.

Immunofluorescence Studies of Infected MΦs

All studies were performed in 8-well slides, with 10⁵ MΦs plated per well. The cells were infected with wild type EBOV, SUDV, RESTV, MARV, EBOV-GFP, or LASV-GFP at an MOI of 1 for 48 hours and then fixed. The virus inocula remained throughout the experiment. The samples were permeabilized with 0.1% Triton X-100 for 10 minutes and then blocked with 5% goat serum (GS; Thermo Fisher) for 1 hour. All primary antibody treatments were prepared in 5% GS and performed at 4 °C overnight, while Alexa Fluor–conjugated secondary antibody and phalloidin–Alexa Fluor 405 treatments (Thermo Fisher), as well as RedDot2 nuclear stain (Biotium), all diluted at 1:1000, were at room temperature for 1 hour. All primary antibodies detecting EBOV NP, VP35, VP40, and GP proteins and LASV NP proteins were diluted at 1:1000 in 5% GS. sdAb ZE antibody (SUDV and RESTV samples) and sdAb A antibody (MARV sample) were diluted to 100 ng/μL in 5% GS. Antibodies detecting host proteins were diluted in 5% GS at 1:200 (LC3B, Lamp-1), 1:500 (α-tubulin, COX IV), or 1:1000 (TNFAIP2). Z-stacks were acquired with a Leica STELLARIS 8 confocal microscope on ≥50 cells/sample. Three-dimensional reconstruction and analysis were done with Leica or Imaris software (Oxford Instruments). The ratio of virus-positive TNTs containing the tested host factor was determined in >100 TNTs/sample. The number of EBOV- and LASV-infected cells with TNTs was counted manually in >100 cells/sample and reported as a ratio to the total number of infected cells. The results were analyzed by a 2-way ANOVA Šídák multiple-comparison test (TNTs positive for virus and the host factor) or an ordinary 1-way ANOVA Dunnett multiple-comparison test (ratios of EBOV- and LASV-infected cells with TNTs).

RNAscope

MΦs plated at 10⁵ cells/well were incubated with EBOV at an MOI of 1 for 48 hours, with the virus inoculum present throughout the course of the experiment. EBOV RNA was detected with the RNAscope Fluorescent Multiplex Kit (version 2; Advanced Cell Diagnostics) according to the manufacturer's recommendations. Samples were treated with a probe-binding EBOV genomic NP open reading frame, and the signal was amplified by incubation with an amplifier conjugated to a fluorescent dye (Advanced Cell Diagnostics). Samples were subsequently stained with NP antibody and phalloidin–Alexa Fluor 405, imaged with the Leica STELLARIS 8 system, and analyzed as indicated previously.

Localization of Individual EBOV Proteins to TNTs

To determine whether viral proteins localize to TNTs, 2 × 10⁴ HUVECs were transfected with 1 µg of pCAGGS-EGFP or cotransfected with either pCAGGS-NP/pCAGGS-VP35/pCAGGS-VP24 or pCAGGS-VP40/pcDNA3-GP at equal concentrations, 0.5 µg of each plasmid, by using the NEON electroporation system (Thermo Fisher) according to the supplier's recommendations. Forty-eight hours later, cells were stained with rabbit NP antibody or costained with VP40/mouse GP antibodies. TNTs were counted and analyzed for the presence of viral proteins in ≥100 transfected cells per condition. The ratio of transfected cells with TNTs in each sample was determined as follows: number of cells with TNTs / total number of transfected cells. The results from 3 independent experiments were combined to be assessed for statistical significance by an ordinary 1-way ANOVA Dunnett multiple-comparison test.

EBOV Nucleocapsid Transfer

To determine whether nucleocapsids can move into naive cells, HUVECs were transfected with plasmids encoding either EBOV NP/VP35/VP24 or red fluorescent protein (RFP). After 24 hours, cells were detached, mixed at the 1:1 ratio, and plated into 8-well slides at 2 × 10⁴ cells/well. The cells were cocultured for an additional 48 hours to allow for TNT formation. The samples were subsequently fixed, stained with rabbit NP antibody, and imaged by a Leica STELLARIS 8 confocal microscope to obtain a section through the middle of the cell and TNTs.

Virus Replication Studies

Two identical sets of MΦs were plated into wells of 96-well plates at 5 × 10⁴ cells/well. Set 1 was incubated with EIPA at 25 µM, cytD at 20 µM, nocodazole at 20 µM, or an equal amount of DMSO or left untreated for 1 hour; after which, it was challenged with EBOV-GFP at an MOI of 0.01 in the presence of the treatments. In the neutralization assay, 500 infectious particles of EBOV-GFP were incubated with GP KZ52 antibody (20 µg/mL), human IgG isotype control (20 µg/mL), or medium for 30 minutes at 37°C and then overlaid onto cells. All treatments were performed in triplicate, and the samples were fixed 24 hours postinfection. Set 2 was incubated with EBOV-GFP at an MOI of 0.01 for 1 hour to allow for binding; afterward, it was washed and overlaid with a new medium containing EIPA, cytD, nocodazole, GP KZ52, GP KZ52/cytD, or GP KZ52/nocodazole treatment and appropriate controls, as just described and in triplicate. The treatments were repeated 24 and 48 hours postinfection. Cells were fixed 72 hours postinfection. All monolayers were treated with Hoechst dye (Thermo Fisher) to stain nuclei, photographed by a Nikon automated system, and analyzed by CellProfiler software to quantify infected cells and nuclei (GFP positive). Infection efficiency in each sample was determined as a ratio of infected cells and nuclei and reported relative to untreated samples. Statistical analysis was performed with a 2-way ANOVA Šídák multiple-comparison test.

Statistical Analysis

For each experiment, ≥3 independent tests were performed. All data were analyzed for statistical significance through ANOVA tests, with P ≤ .05 being considered significant.

RESULTS

EBOV Infection Triggers TNT Formation

MΦs are among the primary targets of EBOV infection and are believed to contribute to infection spread by an unclear mechanism [1]. To study morphologic features of primary human MΦs, we challenged them with EBOV-GFP. After 24 hours, the time sufficient for the virus to undergo 1 replication cycle [6, 20], we detected the formation of long, thin intercellular connections positive for infection without having a detrimental effect on cell viability (Figure 1A). To characterize these structures in detail, we stained infected cells with fluorescent phalloidin to visualize F-actin, a major component of TNTs, and with antibodies to detect viral NP and the host TNF-α–induced protein 2 (TNFAIP2), which localizes to TNTs and is key for their biogenesis [11, 12, 21]. As seen in Figure 1B, the connections floated free in the medium and contained TNFAIP2 and viral NP. Our data indicate that TNTs are indeed the structures triggered by EBOV infection.

Figure 1.

EBOV infection triggers TNT formation. A, MΦs were infected with EBOV-GFP at MOI = 1 for 24 hours and then imaged. Arrowheads point to connections between infected cells. In cell viability tests, MΦs were infected with EBOV at the indicated MOI and time. Cell viability was determined by CellTiter-Glo reagent, and the values for each time point were normalized to values of uninfected cells and are shown as means (±SDs) of 3 independent experiments. B, MΦs infected with EBOV at MOI = 1 for 48 hours were stained with phalloidin (pseudocolored gray, here and all subsequent phalloidin staining), RedDot2 dye (blue), and antibodies to TNFAIP2 and EBOV NP. Confocal images were acquired as Z-stacks and converted to 3-dimensional images. The maximum intensity projection of a Z-stack is shown in the larger image on the left. The arrowhead indicates an intercellular connection. The images on the right are a side view of the merged and individual channels of the sample. C, MΦ or MΦ-HUVEC cultures were infected with EBOV at MOI = 5 or left untreated for 24 or 48 hour and then imaged by scanning electron microscopy. Representative images of TNTs in infected cells are shown. HUVECs and MΦs were identified by cell morphology. MΦs are marked by stars in the coculture image. Examples of MΦ and HUVEC-MΦ connections are marked with yellow and blue-yellow arrowheads, respectively. D, The number of MΦs with TNTs at combined 24- and 48-hour time points in homologous cultures in electron micrographs was counted manually in ≥100 cells/sample and is reported as a ratio to the total number of cells (left panel). The length and width of TNTs in the micrographs were determined by ImageJ software in ≥50 TNTs/sample (middle and right graphs, respectively). ^*P < .05. ***P < .001. ****P < .0001. EBOV, Ebola virus; GFP, green fluorescent protein; hpi, hours postinfection; HUVEC, human umbilical vein endothelial cell; MΦ, monocyte-derived macrophage; MOI, multiplicity of infection; NP, nucleoprotein; ns, nonsignificant; TNT, tunneling nanotube.

Scanning electron micrographs showed that MΦs form TNTs in a homologous culture and in coculture with the HUVECs, which are commonly used to model EBOV infection in the endothelium (Figure 1C). These data show that MΦs have the potential to facilitate virus transfer to heterologous cell types with TNTs, thus expanding cell tropism. In homologous cultures, as compared with mock-infected samples, EBOV-infected MΦs formed significantly more TNTs as early as 24 hours postinfection. Cumulatively for the 24- and 48-hour time points, the TNTs were 8 to 204 µm long, which is consistent with published data [10], and 0.1 to 5 µm in diameter (Figure 1D). While the infection did not affect the overall length of the TNTs, TNTs induced by the virus at 24 hours were significantly smaller in diameter as compared with mock-infected cells (an average of 1.3 vs 1.8 µm), suggesting that EBOV may trigger different TNT subsets during its infectious cycle.

EBOV Proteins and Genome Localize to TNTs During Infection

Having observed that EBOV NP is associated with TNTs (Figure 1B), we next assessed whether additional viral proteins and/or viral genome localizes to TNTs. EBOV-challenged MΦs were coimmunostained with antibodies to VP40 and GP, which are required for virus particle formation at the cell surface [6], or with NP-specific antibody and a probe that binds the viral genome. No staining was observed in mock-infected samples. We detected VP40 and GP in nanotubes connecting infected cells, although they did not always appear to localize to the same TNTs simultaneously (Figure 2A). We also readily detected viral genomes associated with NP protein in the TNTs (Figure 2B), suggesting that viral nucleocapsids can move within virus-induced TNTs.

Figure 2.

EBOV proteins and RNA localize to TNTs during infection. MΦs infected with EBOV at MOI = 1 for 48 hours were stained with phalloidin, RedDot2 dye (blue), and either (A) antibodies to VP40 and GP or (B) antibody to NP and a probe binding NP RNA. Samples were analyzed as in Figure 1B. The solid arrowheads point to TNTs positive for all tested viral factors. The hollow arrowhead points to a TNT containing only VP40. EBOV, Ebola virus; GP, glycoprotein; MOI, multiplicity of infection; NP, nucleoprotein; TNT, tunneling nanotube.

EBOV Nucleocapsids Traffic Through TNTs in the Absence of Infection

EBOV nucleocapsids were shown to traffic into filopodia at the cell surface [5, 22]. Since EBOV RNA/NP complexes localize to TNTs in infected cells (Figure 2B), we tested whether viral nucleocapsids can translocate intercellularly and trigger TNT development under noninfectious conditions. We reconstituted nucleocapsids by coexpressing NP, VP35, and VP24 proteins in HUVECs [5] and, after 48 hours, stained the samples with NP-specific antibody. The nucleocapsids were found throughout the TNTs and appeared to be transported into connecting cells (Figure 3A). Interestingly, nucleocapsid reconstitution triggered a significant increase in the TNT numbers (Figure 3B), suggesting that these viral structures are sufficient to promote nanotube formation in infected specimens, potentially facilitating infection spread. Coexpression of VP40 and GP did not result in protein trafficking to TNTs (Figure 3A), in contrast to our observations in infected cells (Figure 2A), and did not affect the number of TNTs (Figure 3B). These findings suggest that an additional viral or host factor present during infection is required for the translocation of VP40 and GP in TNTs during infection.

Figure 3.

EBOV nucleocapsids traffic through TNTs in the absence of infection. A, HUVECs were transfected to express NP/VP35/VP24 proteins, VP40/GP proteins, or GFP or left untransfected. After 48 hours, cells were stained with phalloidin, RedDot2 (blue), and antibodies detecting NP or VP40/GP. The arrowhead points to nucleocapsids within the TNT. The maximum intensity projection of Z-stacks are shown. B, The ratio of transfected cells with TNTs in each sample was determined as follows: number of cells with TNTs / total number of transfected cells, and the data distribution are shown as a violin plot. C (left), HUVECs coexpressing NP/VP35/VP24/VP30-GFP for 48 hours were imaged by live-cell microscopy. Images were acquired in the brightfield and fluorescence modes for up to 120 frames. Each frame was photographed every 10 seconds for 20 minutes. Magnified images of the boxed region in the fluorescent mode are shown as selected frames in the right-side panels. The arrowhead points to a nucleocapsid trafficking through the TNT. C (right), Cells coexpressing NP/VP35/VP24/VP30-GFP for 48 hours were stained with phalloidin, sdAb ZE antibody, and VP35 antibody. The arrowhead points to a nucleocapsid within a TNT. D, HUVECs coexpressing either NP/VP35/VP24 or RFP for 24 hours were cocultured at the 1:1 ratio. After 48 hours, the samples were stained with phalloidin, RedDot2 (blue), and antibody detecting NP. A section through the middle of the cell and TNTs is shown. The dotted line marks the boundaries of an RFP-expressing cell. The blue and white arrowheads point to TNTs and nucleocapsids transferred to the RFP-expressing cell, respectively. ****P < .0001. EBOV, Ebola virus; GFP, green fluorescent protein; GP, glycoprotein; HUVEC, human umbilical vein endothelial cell; NP, nucleoprotein; ns, nonsignificant; RFP, red fluorescent protein; TNT, tunneling nanotube.

We next used live cell microscopy to study movement of fluorescently labeled nucleocapsids through the nanotubes. HUVECs coexpressing NP, VP35, VP24, and VP30-GFP to label reconstituted nucleocapsids [5] were examined for TNTs containing fluorescent nucleocapsids. Once identified, the areas were photographed in the brightfield and fluorescence modes every 10 seconds for up to 20 minutes. We observed that the nucleocapsids translocated through the nanotubes between cells at 8 to 27 nm/s (Figure 3C, Supplementary Movie 1), which is consistent with published data on nucleocapsid locomotion in filopodia [19]. To assess whether VP30-GFP was associated with nucleocapsids, we fixed and stained the samples with NP-specific sdAb-ZE and a VP35-specific antibody and assessed the colocalization of VP30-GFP with each signal. We determined that ∼80% of VP30-GFP colocalized with NP and 89% with VP35 (Figure 3C), confirming that VP30-GFP–positive structures are indeed viral nucleocapsids.

To determine whether nucleocapsids can move into naive cells, we transfected 2 separate HUVEC populations with plasmids encoding either EBOV NP, VP35, and VP24 or RFP. We then cocultured the 2 cell sets to allow for TNT formation and examined RFP-expressing cells for the presence of virus nucleocapsids. We observed nucleocapsid localization to TNTs and transfer into the naive cells (Figure 3D). Overall, our data show that EBOV nucleocapsids trigger the development of TNTs and translocate through them in a virus-free manner.

EBOV Localizes to Wide TNTs

MΦ TNTs are categorized as thin or wide, depending on structural constituents. TNTs with a narrower diameter contain only F-actin, whereas thicker TNTs, typically those >0.7 µm in diameter, contain F-actin and microtubules. Wide TNTs facilitate the transport of cellular organelles, including endosomes and mitochondria, by an adenosine triphosphate (ATP)– and tubulin-mediated mechanism [11, 23]. To determine which subset of nanotubes is exploited by EBOV, we assessed the relative ratios of virus-positive TNTs that contained α-tubulin and organelles. Staining EBOV-infected MΦs for endosomal proteins Lamp-1 and LC3B, α-tubulin, or mitochondrial protein COX IV revealed that the majority of EBOV NP signal in TNTs localized to nanotubes positive for the tested host factors (Figure 4), suggesting that EBOV traffics through wide TNTs.

Figure 4.

EBOV localizes to wide TNTs. MΦs infected with EBOV (MOI = 1) for 48 hours were stained with phalloidin and antibodies to (A) NP, Lamp-1, and LC3B or (B) NP and α-tubulin. Samples were imaged and analyzed as in Figure 1B. The solid arrowheads point to TNTs positive for a tested host factor. The hollow arrowheads point to TNTs containing F-actin only. C, The ratio of NP-positive TNTs containing the tested host factor was determined in >100 TNTs/sample and is shown as a mean (±SD) of 3 independent experiments. D, Infected MΦs were stained with phalloidin, RedDot2 dye (blue), and antibodies to NP and COX IV, and imaged as previously indicated. ****P < .0001. EBOV, Ebola virus; MΦ, monocyte-derived macrophage; MOI, multiplicity of infection; NP, nucleoprotein; TNT, tunneling nanotube.

EBOV Replicates in the Presence of Treatments Targeting Virus Entry

Our data show that EBOV nucleocapsids can migrate through nanotubes (Figure 3A and 3D) in a virus-free manner. Next, we tested whether the virus can replicate in the presence of treatments blocking virus entry. EIPA stops macropinocytosis used by the virus to enter cells [2]. cytD and nocodazole—potent inhibitors of actin and microtubule polymerization, respectively—were shown to block EBOV entry [24]. GP antibody clone KZ52 [25] neutralizes GP activity [26], blocking virus entry. To determine the nontoxic range for virus studies, we treated MΦs with 2-fold serial dilutions of each treatment for 72 hours and then determined cell viability (Figure 5A). Subsequent experiments used concentrations that reduced cell viability by <5%. Furthermore, treatment with 25µM EIPA, 5µM cytD, 10µM nocodazole, and 20-µg/mL GP KZ52 blocked infection by 98%, 96%, 23%, and 94%, respectively (Figure 5B), all observations consistent with published data [2, 24, 26].

Figure 5.

EBOV replicates in the presence of treatments targeting virus entry. A, MΦs were left untreated or were treated with EIPA, nocodazole, cytD, DMSO, GP KZ52 neutralizing antibody, or IgG isotype control at indicated concentrations for 72 hours. Cell viability was determined with CellTiter-Glo reagent. Nonlinear regression analysis was performed to select a nontoxic concentration (≥95% of cell viability) for each treatment relative to the untreated samples. B, MΦs were incubated with EIPA (25 µM), nocodazole (10 µM), cytD (5 µM), or an equal amount of DMSO or were left untreated for 1 hour. Cells were challenged with EBOV-GFP at MOI = 0.01 with treatments present. In the neutralization assay, 500 infectious particles of EBOV-GFP were incubated with GP KZ52 antibody or IgG isotype control (20 µg/mL) or medium for 30 minutes and then overlaid onto cells for 24 hours. Samples were stained with Hoechst dye, photographed, and analyzed by CellProfiler to quantify infected cells and nuclei (GFP positive). Infection efficiency in each sample was determined as number of infected cells / number of nuclei and reported relative to untreated samples. Mean infection efficiencies (±SDs) from 3 different experiments are shown. C, MΦs incubated with EBOV-GFP (MOI = 0.01) for 1 hour were overlaid with a medium containing EIPA, nocodazole, cytD, GP KZ52, GP KZ52/nocodazole, or GP KZ52/cytD treatments and appropriate controls, as previously indicated. The treatments were repeated 24 and 48 hours postinfection and analyzed as in panel B. D, Representative images of samples obtained in panel C. Samples were analyzed 72 hours postinfection as in panel B. ****P < .0001. cytD, cytochalasin D; DMSO, dimethylsufoxide; EBOV, Ebola virus; GFP, green fluorescent protein; GP, glycoprotein; MΦ, monocyte-derived macrophage; MOI, multiplicity of infection; TNT, tunneling nanotube.

We next incubated MΦs with EBOV-GFP for 1 hour to allow for virus entry, removed the inoculum, and added treatments as previously indicated. Under these conditions, the virus completes 1 round of replication but is expected to be progressively inhibited over time due to the reduced entry into additional cells. Surprisingly, we observed significant virus spread in MΦs treated with EIPA, nocodazole, and GP KZ52 after 72 hours. Relative to control-treated cells, infection efficiency reached 54% following EIPA treatment, 100% following nocodazole, and 63% following GP KZ52 (Figure 5C). We also observed efficient TNT formation in EIPA- and GP KZ52–treated samples, with visible nanotubes connecting infected and naive cells (Figure 5D). CytD reduced the number of EBOV-positive cells by >90% (Figure 5C), likely to inhibition of virus egress [20].

Because EBOV preferentially localizes to tubulin-positive TNTs (Figure 4B and 4C), we assessed virus replication in MΦs cotreated with GP KZ52 antibody and either nocodazole or cytD. Both cotreatments significantly reduced virus spread in the monolayers as compared with the GP KZ52 antibody treatment alone (Figure 5C and 5D). Our data show that EBOV can transfer intercellularly and replicate its genome under conditions when virus entry is blocked, thus bypassing the cell-free form.

LASV Does Not Trigger TNT Development

We next determined whether the ability to promote TNT development by EBOV is unique among viruses targeting primarily MΦs in the host. Similar to EBOV, LASV can efficiently replicate in MΦs but, in contrast to EBOV, does not activate them or facilitate proinflammatory cytokine production [27, 28]. MΦs challenged with either EBOV-GFP or LASV-GFP for 48 hours and stained for NP were examined by microscopy. As earlier (Figure 1), EBOV infection dramatically increased TNT development, whereas LASV did not (Figure 6A). We also did not observe efficient localization of LASV NP to existing TNTs (Figure 6B). Experiments with additional human pathogen filoviruses SUDV and MARV and the nonpathogenic RESTV showed that each of these MΦ-tropic viruses localized to nanotubes (Figure 6C), suggesting that filoviruses, not LASV, may widely exploit TNTs during their replication cycle.

Figure 6.

LASV does not trigger TNT development. A, MΦs infected with EBOV-GFP or LASV-GFP (MOI = 1) for 48 hours were stained with antibodies to NP. Maximum intensity projections of Z-stacks are shown in the left panel. The number of infected cells with TNTs was counted manually in >100 cells/sample and is reported as a ratio to the total number of infected cells (graph on the right). The bars are averages (±SD) of 3 independent experiments. B, LASV-GFP–infected MΦs were processed and imaged as in panel A. The arrowhead points to a TNT between infected cells. C, MΦs infected with SUDV, REST, or MARV (MOI = 1) for 48 hours were stained with phalloidin and either sdAb ZE antibody (SUDV and RESTV samples) or sdAb A antibody (MARV sample). *P < .05. EBOV, Ebola virus; GFP, green fluorescent protein; LASV, Lassa virus; MΦ, monocyte-derived macrophage; MARV, Marburg virus; MOI, multiplicity of infection; NP, nucleoprotein; RESTV, Reston virus; SUDV, Sudan virus; TNT, tunneling nanotube.

DISCUSSION

We demonstrated that infection of human MΦs with EBOV triggers formation of TNTs of various lengths and diameters, but only those containing tubulin and transport of cellular organelles also contained viral proteins and RNA. We further showed that EBOV nucleocapsids traffic between cells under noninfectious conditions. Importantly, we observed that EBOV infection could spread in cultures treated with virus entry inhibitors after the initial stall, suggesting that dissemination of virus replication can occur through a virus-free form.

The wide range of TNT length and width within the EBOV-infected MΦ population suggests an adaptability of MΦs to the environment through potentially diverse TNT functions. It is plausible that the thinner size of a subset of TNTs in EBOV-infected MΦs at the earlier time point (24 hours), as compared with mock-infected cells, represents a basal stress response to infection or is being used by the virus to remotely modulate cell responses, further contributing to disease development.

The significant increase in ATP production levels at the earlier time point of EBOV infection suggests that the virus may exploit or subvert mitochondrial functions for its own benefit. Electron micrographs of cell monolayers infected with EBOV clearly show localization of mitochondria in close proximity to viral inclusions containing nucleocapsids [29], and EBOV VP30 and VP40 proteins interact with components of mitochondrial complexes [30], suggesting that the organelle may fuel energetically consuming steps of the virus replication cycle, such as transcription and replication of virus genome and viral particle egress. Mitochondria continuously undergo a series of dynamic adaptations to maintain their distribution and morphology to facilitate essential cellular processes, including immunity. To short-circuit mitochondria-dependent immune activation, EBOV inhibits the critical mitochondrial antiviral signaling protein pathway, whose role is to trigger expression of type I interferons [31]. While VP35 protein inhibits the mitochondrial antiviral signaling protein pathway through disruption of binding of viral RNA to the RIG-I receptor [32], it is plausible that compromised mitochondrial dynamics can also contribute to this block.

EBOV VP40 and GP localize to the cell surface during infection [1, 5, 6] and therefore are expected to associate with TNTs, which share membranes with cells that they connect. However, we did not detect either protein on TNTs of transfected endothelial cells, possibly due to the lack of additional factors present during EBOV infection. The ability of viral nucleocapsids to traffic through and induce TNT formation in the absence of infection has wide-ranging implications for EBOV pathogenesis. Evidence suggests that EBOV RNA can persist at immune-privileged sites, such as ocular tissue, brain, and testes, for extended periods after apparent recovery and later reemerge as an infectious virus with a nearly identical sequence to the primary infection [1, 33–35]. It is unclear how EBOV genetic material moves within the tissues without detectable viremia, but it may occur through intercellular connections. In the brain, TNTs contribute to multilayered intercellular communication by connecting diverse cell types, thus facilitating proper brain physiology. TNTs can also promote the transfer of pathogenic protein aggregates, resulting in degenerative pathologies, including Parkinson and Alzheimer diseases, and cancer progression by controlling the microenvironment [36]. In ocular tissue, the prominent TNT network modulates eye physiology by connecting widely spaced cells. For example, MΦ TNT–mediated transfer of healthy lysosomes to diseased fibroblasts can reverse pathology associated with lysosomal storage disease in the cornea [7, 37]. Thus, it is plausible that EBOV can exploit the extensive network of TNTs to move to and persist in these tissues for an extended time. More studies are needed to determine whether EBOV can persist in tissues typically associated with active virus replication.

MΦ TNTs are structurally diverse, allowing for distinct functional properties. TNTs with wider diameters contain microtubules and transport cellular organelles by an ATP- and tubulin-mediated mechanism at a speed of about 1 µm/s [12, 23]. Although EBOV preferentially localizes to TNTs positive for tubulin, endosomes, and mitochondria, the intracellular movement of nucleocapsids is independent of tubulin and, according to our data and others [22], occurs at a slower speed of 11 to 27 nm/s. While not impossible, the evidence indicates that EBOV nucleocapsids traffic through TNTs by a mechanism that is distinct from those of organelles.

MΦ TNTs are formed by 2 processes: a temporary bridge forms between contacting cells after they touch and move away from each other, or filopodia present in one cell can extend toward another, converting to TNTs after contact [38, 39]. EBOV nucleocapsids are recruited into budding virions at cell surface filopodia [5], making it possible for these protrusions to fuse with neighboring cells and generate a hollow tunnel, allowing nucleocapsids to pass through. However, our findings that the coexpression of VP40 and GP, which drive virion budding, does not facilitate TNT formation make this model of TNT generation unlikely. Even though TNTs and filopodia share structural similarities, evidence is emerging that TNTs and filopodia are distinct cellular structures that form through separate mechanisms and may be regulated differently [40, 41]. We are currently investigating migration properties of EBOV-infected MΦs to address whether the affected cell movement may contribute to the increased TNT number, either through more efficient contact or TNT stabilization due to reduced cell movement. It is also possible that EBOV promotes TNT formation by a novel, virus-specific mechanism.

Currently available Food and Drug Administration–approved countermeasures efficiently target EBOV GP and include a successful preventative vaccine and monoclonal antibodies [42]. Our data showing EBOV replication in MΦ cultures treated with established inhibitors of viral GP activity suggest that these treatments may not be able to (1) eliminate virus whose spread occurs by a virus-free–mediated mechanism or (2) target virus persistence in immune-privileged sites and likely other tissues. More studies are needed to determine whether this phenomenon is observed for additional inhibitors of EBOV entry. Understanding how EBOV exploits TNTs will be essential to design an optimal treatment regimen for EBOV and other filovirus infections. Because TNTs share the cytoskeleton, membrane, and other host factors with the cells that they connect, targeting these major TNT components could be challenging due to the lack of specificity required for sufficient antiviral efficacy. The nucleoside analog cytarabine and the mTOR inhibitor everolimus inhibit TNT formation [43, 44] and therefore may be investigated as a part of combinatorial antiviral strategies, especially those targeting the viral genome replication step. We are currently utilizing laser microdissection technology, followed by proteomic and transcriptomic analyses, to identify host factors specifically involved in filovirus transfer between cells, which will aid with selection of targets for development of antivirals.

EBOV disease is characterized by efficient virus spread, uncontrolled replication, escape of immune defenses, and high levels of inflammation [1]. Upregulation of several proinflammatory cytokines correlates with increased mortality in infected individuals [1]. Inflammation exacerbates TNT formation, partly by a TNF-α–induced mechanism, where strongly induced TNFAIP2 triggers F-actin polymerization from the initiating cell surface [21, 45, 46]. We detected a strong association of TNFAIP2 with TNTs in EBOV-infected MΦs, suggesting that a virus-induced proinflammatory environment may direct TNT development. Our LASV results further support this hypothesis since LASV infection does not activate human MΦs [27, 28] and consequently does not promote TNT formation or viral NP localization to existing TNTs. Interestingly, viral NP translocated to nanotubes of MΦs challenged with SUDV, MARV, and RESTV. These pathogens upregulate proinflammatory cytokines [47–50], suggesting a mechanism that is shared by filoviruses, at least partially. TNTs may provide critical protection to virus spread due to shielding from circulating immune cells and bypassing virus-cell interactions that may trigger host defenses. Additional investigations are needed to determine which host environment is conducive to TNT development and why, as well as how TNT formation contributes to filovirus pathogenesis.

In summary, we demonstrate that EBOV infection of human MΦs triggers TNT formation, which supports the transport of viral nucleocapsids, thus identifying an alternate mode of virus dissemination with the established egress of nascent virions from the cell surface. This discovery advances our understanding of EBOV replication in the host and, importantly, will stimulate new areas for antiviral design.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank numerous donors for blood donations. We are incredibly grateful to Juan Carlos de la Torre for providing plasmids for rescue of recombinant LASV-GFP, Ricardo Carrion for providing stocks of SUDV and RESTV, Luis Branco for providing NP antibody to detect LASV infection, and Barbara Hunter for assistance with the electron microscope imaging. We also thank the Texas Biomed select agent program and biocontainment team for outstanding technical support.

Author contributions. O. S. conceived and designed the study. M. A. D., C. G. L., V. K., L. M.-S., R. T., and O. S. performed experiments. E. A., L. M.-S., L. J. S., A. H., and L. S. S. contributed reagents. M. A. D., C. G. L., and O. S. analyzed the data. M. A. D. and O. S. wrote the manuscript. C. G. L., E. A., L. M.-S., A. H., and L. S. S. edited the manuscript.

Supplement sponsorship. This article appears as part of the supplement “10^th International Symposium on Filoviruses.”

Financial support. This work was supported by the National Institute of Allergy and Infectious Diseases (grants R21AI154336 and R21AI151717); and Texas Biomed Forum (grants 2017 and 2020).