Marburg virus exploits the Rab11-mediated endocytic pathway in viral-particle production

ABSTRACT Filoviruses produce viral particles with characteristic filamentous morphology. The major viral matrix protein, VP40, is trafficked to the plasma membrane and promotes viral particle formation and subsequent viral egress. In the present study, we assessed the role of the small GTPase Rab11-mediated endocytic pathway in Marburg virus (MARV) particle formation and budding. Although Rab11 was predominantly localized in the perinuclear region, it exhibited a more diffuse distribution in the cytoplasm of cells transiently expressing MARV VP40. Rab11 was incorporated into MARV-like particles. Expression of the dominant-negative form of Rab11 and knockdown of Rab11 decreased the amount of VP40 fractions in the cell periphery. Moreover, downregulation of Rab11 moderately reduced the release of MARV-like particles and authentic MARV. We further demonstrated that VP40 induces the distribution of the microtubule network toward the cell periphery, which was partly associated with Rab11. Depolymerization of microtubules reduced the accumulation of VP40 in the cell periphery along with viral particle formation. VP40 physically interacted with α-tubulin, a major component of microtubules, but not with Rab11. Taken together, these results suggested that VP40 partly interacts with microtubules and facilitates their distribution toward the cell periphery, leading to the trafficking of transiently tethering Rab11-positive vesicles toward the cell surface. As we previously demonstrated the role of Rab11 in the formation of Ebola virus particles, the results here suggest that filoviruses in general exploit the vesicle-trafficking machinery for proper virus-particle formation and subsequent egress. These pathways may be a potential target for the development of pan-filovirus therapeutics. IMPORTANCE Filoviruses, including Marburg and Ebola viruses, produce distinct filamentous viral particles. Although it is well known that the major viral matrix protein of these viruses, VP40, is trafficked to the cell surface and promotes viral particle production, details regarding the associated molecular mechanisms remain unclear. To address this knowledge gap, we investigated the role of the small GTPase Rab11-mediated endocytic pathway in this process. Our findings revealed that Marburg virus exploits the Rab11-mediated vesicle-trafficking pathway for the release of virus-like particles and authentic virions in a microtubule network-dependent manner. Previous findings demonstrated that Rab11 is also involved in Ebola virus-particle production. Taken together, these data suggest that filoviruses, in general, may hijack the microtubule-dependent vesicle-trafficking machinery for productive replication. Therefore, this pathway presents as a potential target for the development of pan-filovirus therapeutics.

MVD is a severe and often fatal disease with a mortality rate of up to 88% in humans (1).Currently, neither specific vaccines nor therapeutics are approved for the treatment and prevention of MVD (2,3), although the potential effectiveness of vaccines and therapeutics have been demonstrated in animal models (4).
MARV and Ebola virus (EBOV) encode seven structural genes in their single-stranded negative-sense RNA genomes that assemble to yield filamentous viral particles (5).The virus-encoded major matrix protein VP40 self-assembles to form virus-like particles (VLPs), which resemble the morphology of authentic viruses and are released from the cell surface even when VP40 is expressed by itself (6)(7)(8)(9).While EBOV VP40 predominantly targets the plasma membrane (PM) as a site for VLP formation and release, MARV VP40 accumulates in multivesicular bodies in the cytoplasm in addition to the PM.VLP-positive multivesicular bodies are enriched at the PM for subsequent budding of MARV particles and VLPs (10)(11)(12).
Several host proteins have also been identified as key factors in VP40 trafficking to the PM.For example, a Ras GAP-related actin-binding scaffolding protein, IQGAP1 (25), a component of the host COPII vesicular transport system, Sec24C (26), and a microtubule network (27) appear to be responsible for the intracellular transport of VP40 to the budding site.However, a comprehensive understanding of the underlying molecular mechanisms is lacking.
Growing evidence suggests that various viruses exploit the Rab11-mediated endocytic pathway during their assembly process by trafficking viral components to their egress sites for incorporation into virions (28).Rab11 is one of the most well-charac terized small GTPases and, as a member of the Rab family, predominantly distributed in recycling endosomes and post-Golgi vesicles (29)(30)(31)(32).Rab11 is indispensable for the regulation of cargo recycling through recycling and secretory vesicles (29,33) and other cellular processes, including cell migration (34), ciliogenesis (35), and phagocytosis (36).
Only a few studies have demonstrated the role of the Rab11-mediated pathway in the filovirus life cycle.Liquid chromatography-linked tandem mass spectroscopy revealed that the Rab11b isoform is incorporated into authentic filovirus virions (47,48).Our group has clarified the critical role of the Rab11-dependent vesicle traffic pathway in the VP40-mediated release of Ebola VLPs (49).
This study assessed the role of the Rab11-mediated pathway in the MARV repli cation cycle.We found that the transient expression of MARV VP40 promoted the diffuse cytoplasmic distribution of Rab11, which is normally localized in the perinuclear region.We also observed that Rab11 was incorporated into Marburg VLPs.The expres sion of a dominant-negative form and knockdown of Rab11 decreased VP40 distribu tion to the cell periphery.Moreover, we demonstrated that the release of Marburg VLPs and progeny infectious MARV was moderately reduced upon downregulation of Rab11.We further demonstrated that VP40 modulates the distribution of microtubules toward the cell periphery, which was often associated with Rab11.Depolymerization of microtubules reduced the accumulation of VP40 in the cell periphery along with viral-particle formation.VP40 physically interacted with α-tubulin, a major component of microtubules, but not with Rab11.Our findings suggest that MARV VP40 interacts with microtubules and facilitates their distribution toward the cell periphery, leading to the trafficking of transiently tethering Rab11-positive vesicles toward the cell surface and the subsequent release of virus particles to establish efficient viral egress.

MARV VP40 induces a dispersed distribution of Rab11
First, we investigated the effect of transiently expressed filovirus VP40 on the subcellu lar localization of Rab11.In plasmid-transfected and mCherry-expressing cells, Rab11 was predominantly distributed in the perinuclear region (Fig. 1A and B), suggesting its involvement in recycling endosomes and post-Golgi vesicles (30).Then, we eluci dated the intracellular distribution of transiently expressed filovirus VP40.MARV and EBOV VP40 were similarly distributed in multiple subcellular compartments, such as the cytoplasm, nucleus, and PM, along with filamentous structures from the cell surface (50) (Fig. 1C and D; insets, arrows).In accordance with previous findings (11,12), we observed that MARV VP40 showed large aggregates distributed in the cytoplasm (Fig. 1D, insets, arrowhead), which was not observed for EBOV VP40 (Fig. 1C).Co-immunofluorescence staining revealed that the expression of both VP40s induced a diffuse distribution of Rab11 throughout the cytoplasm (Fig. 1C and D; insets).However, no efficient co-locali zation of filovirus VP40 with Rab11 was observed.
We measured the distance between Rab11-positive vesicles and the closest nucleus and the ratio of the total volume of each Rab11-positive signal observed in the indicated regions outward from the nucleus.We confirmed that the overexpression of mCherry and filovirus VP40 did not affect the surface area of cell bodies and nucleus (Fig. 1E).Ratio of Rab11 that was distributed closer to the periphery of cells increased by expression of EBOV or MARV VP40 (Fig. 1F and G).For example, while 0.9% or 1.3% of the total volume of the Rab11-positive signal was distributed in the region more than 20 µm away from the nucleus of the control or mCherry-expressing cells, respectively, 7.7% or 7.5% of the Rab11-positive signal fraction was distributed in the same region of the cells expressing EBOV or MARV VP40, respectively (Fig. 1F and G).Modulation of Rab11 distribution was not observed in cells transiently expressing the MARV glycoprotein (GP; Fig. 2A) or nucleoprotein (NP; Fig. 2B), which are also major components of viral particles.Moreover, neither GP nor NP was shown to be co-localized with Rab11 (Fig. 2).We also confirmed the effect of the MARV GP and NP on the distribution of Rab11-positive vesicles by measuring the distance between Rab11 signals and the closest nucleus (Fig. 2C through E).

Rab11 is incorporated in MARV VLPs
We examined the incorporation of Rab11 into MARV particles using two independent analyses.We generated Marburg VLPs by co-expressing VP40 along with the viral GP and NP in Expi293F cells, purified them by ultracentrifugation, and confirmed their morphol ogy by negative-staining electron microscopy (Fig. 3A).Membranous objects were not observed in the control fractions, which were obtained from the culture supernatants of the backbone plasmid-transfected cells (Fig. 3A).Western blotting analysis revealed that VP40 and Rab11 were present in the purified VLP fractions (Fig. 3B), indicating that Rab11 is incorporated into MARV particles.We further confirmed the incorporation of Rab11 into Marburg VLPs using a protease protection assay (49,51).Purified VLPs were treated with or without Triton X-100 at room temperature for 10 min and further incubated at the same condition in the presence or absence of trypsin, followed by western blotting analysis using antibodies against VP40 and Rab11.While both VP40 and Rab11 were similarly detected in Marburg VLPs treated with trypsin or Triton X-100 alone, trypsinization of Triton X-100-treated VLPs abolished the signals of VP40 and Rab11 (Fig. 3C), further confirming the incorporation of Rab11 in VLPs.

Effect of the dominant-negative form of Rab11 on VP40 distribution and subsequent Marburg VLP formation
Because MARV VP40 modulated the intracellular distribution of Rab11 (Fig. 1) and Rab11 was incorporated into Marburg VLPs (Fig. 3), we further assessed the role of Rab11 in VP40-mediated MARV particle formation.We transiently expressed the GFP-fused wild-type (GFP-wtRab11) or the dominant-negative form of Rab11 (GFP-dnRab11) and MARV VP40 in Vero-E6 cells.The GFP-dnRab11 has an amino-acid substitution (S25N) and preferentially keeps the GDP-bound status, inhibiting the recycling of cargo from the intracellular recycling compartments to the cell surface (30,52,53).The effect of the Rab11 derivatives on the distribution of VP40 was examined using immunofluorescence staining.When GFP-wtRab11 was co-expressed with mCherry, it exhibited a predominant distribution in the perinuclear region (Fig. 4A, left) similar to endogenous Rab11 (Fig. 1A).Consistent with the results for endogenous Rab11 (Fig. 1D), efficient co-localization of GFP-wtRab11 with VP40 was not observed.
In contrast, GFP-wtRab11 was distributed diffusely throughout the cytoplasm of MARV VP40-expressing cells.Similar to that of the cells expressing GFP (Fig. 4B, inset and line scan), a fraction of VP40 was distributed to the periphery of the cells with intense clusters (Fig. 4C, inset and line scan).In contrast, GFP-dnRab11 was diffusely distributed in the cytoplasm and partially formed perinuclear clusters in both VP40-positive (Fig. 4D) and -negative (Fig. 4A, right) cells.Upon GFP-dnRab11 expression, VP40 formed large aggregates in the cytoplasm without an intense distribution at the PM (Fig. 4D).

Rab11 contributes to the distribution of VP40 in the PM and subsequent Marburg VLP formation
Next, we assessed the role of Rab11 in the trafficking of VP40 to the PM by knocking down endogenous Rab11 using siRNAs targeting two Rab11 isoforms: Rab11a (54) and Rab11b (55).These two isoforms share 89% amino acid sequence homology and contribute to vesicle-mediated cargo recycling toward the cell surface.Rab11a and Rab11b show ubiquitous and tissue-specific expression, respectively.We treated Vero-E6 cells with siRNAs against the two Rab11 isoforms and confirmed Rab11 downregulation using western blotting (Fig. 5A and B).We then assessed the effect of Rab11 knockdown on MARV VP40 distribution.In both untreated and control siRNA-treated cells, VP40 similarly formed clusters in the cytoplasm and cell periphery and thread-like structures in the PM (Fig. 5C and D, insets, line scans).In contrast, cells in which Rab11 isoforms were downregulated showed the reduction of the formation of VP40-positive clusters at the PM and filamentous structures from the cell periphery (Fig. 5E, inset, line scan).We further determined the effect of Rab11 downregulation on VP40-mediated Marburg VLP formation.HEK293 cells were transfected with siRNAs targeting Rab11 isoforms.At 72 h post-transfection (h.p.t.), while a minimal effect of the knockdown of Rab11 isoforms on the expression of the three MARV structural proteins was confirmed (Fig. 6A and  B), expression plasmids for VP40, NP, and GP were transfected to produce VLPs.Rab11 downregulation resulted in a 30% reduction in the formation of Marburg VLPs (Fig. 6C  and D).

Effect of Rab11 downregulation on viral particle production in MARV-infec ted cells
We assessed the effects of Rab11 knockdown on viral particle production in MARVinfected cells.HEK293 and Vero-E6 cells were treated with siRNAs for Rab11, and the knockdown efficiency was confirmed using western blotting (Fig. 7A and B).The siRNA-treated cells were infected with MARV at a multiplicity of infection (MOI) of 0.01.
Two days post-infection, the effect of Rab11 downregulation on MARV infection was determined using titration.We observed a minimal effect of Rab11 knockdown on MARV infection in both cell types (Fig. 7C).To elucidate the effect of Rab11 downregulation on the production of progeny MARV, we infected siRNA-treated cells with MARV at an MOI of 0.01, harvested supernatants at various times during incubation, and determined the MARV titer.We found that the production of infectious viral particles was indeed moderately reduced in both cell types (Fig. 7D and E), indicating that Rab11 is involved in MARV particle formation and subsequent viral egress.

VP40 exploits the microtubule network for self and tethering Rab11-positive vesicle trafficking toward the cell surface for efficient VLP formation
Rab11 links the vesicles to the cytoskeleton by interacting with specific motor pro teins and mediating active and directed vesicle transport along microtubules or actin filaments (56).We examined the distribution of microtubules in the cells expressing VP40 using immunofluorescence staining.In control cells, the microtubule network was visualized to be extended from the microtubule organizing center in the perinuclear region (Fig. 8A).In contrast, VP40-positive cells exhibited the decrease of polarized distribution in the perinuclear region and distribution of microtubule network toward the cell periphery (Fig. 8A and B, insets, line scans).A fraction of VP40 appears to be co-localized with tubulin (Fig. 8B, inset, line scan).Treatment with nocodazole, a microtubule depolymerizer, resulted in the decrease of cluster formation of MARV VP40 at the cell periphery (Fig. 8C, inset, line scan).Nocodazole treatment also negatively affected the production of VLPs (Fig. 8D and E).The data suggest that VP40-induced microtubule network distribution toward the PM is likely involved in the transport of VP40 itself at the cell periphery and in VLP formation.
Because VP40 was not efficiently co-localized with diffusely distributed Rab11 (Fig. 1D), we next elucidated whether VP40-induced microtubule distribution toward the cell periphery is responsible for the dispersed distribution of tethering Rab11-positive vesicles.A previous study demonstrated that the fusion of the red-fluorescent protein to the N-terminus of MARV VP40 does not disrupt its functionality (57).Thus, Vero-E6 cells were transfected with expression plasmids for mCherry fused to the N-terminus of VP40 and VP40 in a 1:5 ratio.Expression of mCherry-VP40 also induced an increase in microtubule distribution toward the cell periphery (Fig. 8F and G, inset, line scans).The dispersed Rab11-positive vesicles were often detected adjacent to microtubules in mCherry-VP40-positive cells (Fig. 8G, inset).The intensity-based co-localization analysis indicated that approximately 40% of Rab11 signals were co-localized with α-tubulin in multiple cells.
Finally, we investigated the interaction of MARV VP40 and α-tubulin by immunopreci pitation.Transiently expressed VP40 was co-immunoprecipitated with α-tubulin (Fig. 9A) and its interaction was not disrupted in the presence of nocodazole (Fig. 9B), suggesting that VP40 associates with the depolymerized form of tubulin.Their physical interaction was also confirmed by the detection of α-tubulin in the precipitants using the anti-VP40 antibody (Fig. 9A and B).In contrast, no interaction was observed between co-expressed VP40 and GFP-fused Rab11 derivatives (Fig. 9C), which is consistent with the data of lack of efficient co-localization of VP40 with endogenous Rab11 (Fig. 1D).Taken together, these data suggest that VP40 associates with the microtubule network for trafficking of itself and tethering Rab11-positive vesicles toward the cell periphery, contributing to MARV particle formation.

DISCUSSION
In the present study, we demonstrated that the MARV VP40 exploits the microtubuledependent trafficking of Rab11-positive vesicles to the PM for the release of MARV particles.Endogenous Rab11 was distributed more diffusely in the cytoplasm upon MARV VP40 expression (Fig. 1 and 2) and was subsequently incorporated into VLPs (Fig. 3).We also observed that blocking Rab11 function by either the expression of a dominant-negative form of Rab11 or knockdown of Rab11 by siRNA decreased the amount of VP40 distributed to the periphery of the cells (Fig. 4 and 5).We further demonstrated that Rab11 downregulation interfered with the release of both VLP (Fig. 6) and infectious viral particles (Fig. 7).
The diverse functions mediated by Rab11 are regulated by interactions with Rab11 family interacting proteins (Rab11-FIPs) (58).Various FIP members are involved in the egress process to release RSV (42) and filamentous IAV virions (43) from the PM, suggesting that Rab11-associated effector proteins synergistically upregulate the scission process of filoviruses in addition to the ESCRT machinery.Negative effects of nocodazole treatment on the VLP formation may be derived from the reduction of traffic of Rab11-positive vesicles at the budding sites and subsequent scission process.Rab11 forms complexes with specific motor proteins via interactions with distinct adaptor proteins and promotes the bidirectional transport of recycling endosomes, which is mediated by the microtubule network and actin-filament-dependent transport (56).Rab11 has been suggested to form a ternary complex with the myosin Vb motor protein and FIP1 to tether Rab11-positive vesicles to actin filaments at the microtubule-actin junction and contribute to delivery to the PM (59).Rab11-FIP3 is responsible for the formation of a ternary complex with the cytoplasmic dynein motor protein and mediates vesicle transport from peripheral sorting endosomes to the centrally located endosomal recycling compartment through the microtubule minus-end transport system (60).In addition, Rab11-interacting protrudin facilitates the interaction between Rab11 and kinesin family member (KIF) 5a and subsequent microtubule plus-end directed vesicle transport (61).Another plus-end-directed kinesin motor, KIF3, interacts directly with the FIP5 adaptor protein and regulates endocytic protein recycling at the perinuclear recycling endosome (62).were transfected with MARV VP40.At 4 h.p.t., the cells were treated with 1 µg/mL nocodazole and further incubated for 44 h.As a control, the cells were treated with DMSO.The cells and culture medium were harvested.TCLs were subjected to western blotting with the antibodies against VP40 or α-tubulin (D).The intensity of the band's correspondence to VP40 was quantified and normalized with that of α-tubulin (E, gray).VLPs in the culture medium were purified by ultracentrifugation and subjected to western blotting with an antibody for VP40 (D).The intensity of the bands was quantified corresponding to VP40 (E, pink).The experiment was performed three times independently, and representative blots and the mean ± SD are shown.**, P < 0.01 versus respective control (Student's t test).(F and G) VP40-mediated dispersed distribution of Rab11 is dependent on the microtubule network.Vero-E6 cells were transfected with the expression plasmids for mCherry-fused MARV VP40 and VP40 at a ratio of 1:5.At 24 h.p.t., the cells were harvested, and the subcellular distribution of VP40, Rab11, and α-tubulin was analyzed using immunofluorescence staining.We further observed that VP40 partly associated with microtubules (Fig. 8B and 9A) and promoted the microtubule network distribution toward the cell periphery (Fig. 8B  and G).Depolymerization of microtubules under treatment with nocodazole abrogated the cluster formation of VP40 at the cell periphery and VLP production (Fig. 8C through  E).The interaction of VP40 with α-tubulin remained even after treatment with noco dazole (Fig. 9B).These data indicate that the intracellular transport of VP40 and VLP formation require microtubule network dynamics.Moreover, dispersed Rab11 signals partly co-localized with microtubules in cells expressing VP40 (Fig. 8G).These data suggest that VP40 interacts with microtubules and modulates their dynamics, leading to the trafficking of microtubule-associated Rab11-positive vesicles toward the PM.
A previous study demonstrated that EBOV VP40 directly associates with the microtubule network via its partially homologous sequence with the tubulin-binding motif of the host microtubule-associated protein 2 and induces tubulin polymerization (27).Ten amino acids in EBOV VP40 have been shown to be identical to the tubulin-bind ing motif of MAP2 consisting of 31 amino acids.MARV VP40 possesses three amino acids that are homologous to the motif in MAP2, suggesting that unknown mechanisms, independent of this motif, may be responsible for its interaction with microtubules.
In our previous study, two VP40 mutants, which were functionally defective in VLP formation due to failure of dimerization and further oligomerization, did not affect the intracellular distribution of Rab11 (49), suggesting that dimer formation of VP40 is important for Rab11-mediated EBOV particle production.MARV and EBOV VP40 share a 49% amino acid sequence homology.Crystal structure studies have shown that the N-terminal domains of both VP40s are similar in structure and contribute to dimer formation, which further assembles into a flexible filamentous matrix (63,64).In contrast, the C-terminal domain of MARV VP40 is more loosely folded than that of EBOV VP40 and exhibits an extended, highly basic patch covering one side.Thus, further stoichiometric analyses are required to understand the mechanism by which MARV VP40 interacts with microtubules.Moreover, multiple host cytoskeletal factors, including actin (65)(66)(67) are responsible for the intracellular transport of VP40.Thus, these complexes of Rab11 effector proteins and various motor proteins likely mediate the traffic of VP40 that associates with cytoskeletal networks to the PM, which is responsible for virion formation and egress.Further characterization of the Rab11-associated effector proteins underlying this process is required to better understand the detailed molecular mechanisms by which Rab11 contributes to the intracellular trafficking of VP40 and the formation of filovirus particles.
Data in our previous and present studies showed that neither EBOV VP40 (49) nor MARV VP40 efficiently co-localize with endogenous Rab11 (Fig. 1) and GFP-wtRab11 (Fig. 4).A possible explanation for their inefficient co-localization is that VP40 is associated indirectly via the microtubule network with Rab11, which is expressed in the vesicles.Our study also demonstrated a lack of direct association between VP40 and GFP-Rab11 derivatives (Fig. 9C).This is because only a few Rab11-positive vesicles are associated with precipitated microtubule and/or the interaction may be disrupted by detergent treatment.Further investigation by use of live-cell imaging and/or the native immuno precipitation in the context of an authentic MARV infection system can be applied to confirm their indirect association under the physiological condition.
The effect of the treatment of siRNA against Rab11 on the release of VLPs and viral particles was moderate (Fig. 6 and 7), which may be partly due to the incom plete knockdown of Rab11 molecules and/or the existence of a Rab11-independent mechanism responsible for viral particle formation.In addition, the lack of statistical significance for the reduction of progeny MARV by knockdown of Rab11 at certain time points after infection is also likely due to the variability derived from the cell-based titration assay.Nevertheless, the reduction of virus titer exhibited statistical significance in both cell types with the treatment of siRNA against Rab11 at 72 h.p.i., wherein MARV replication reaches a plateau in the same cell type (68,69)., the cells were treated with 1 µg/mL nocodazole for 44 h.As a control, the cells were treated with DMSO.The cells were harvested and lysed with the tubulin-binding buffer.Lysate was incubated with polyclonal antibodies against α-tubulin or MARV VP40 and Protein G Sepharose (+Ab).As a control, lysate was incubated with Protein G Sepharose alone (−Ab).Immunoprecipitants were subjected to western blotting using anti-VP40 antibody.The intensity of the band's correspondence to co-precipitated VP40 or α-tubulin was quantified and normalized with that of α-tubulin or VP40, respectively.The experiment was performed three times independently, and representative blots (A) and the mean ± SD (B) are shown.n.s., not significant versus respective control (Student's t test).(C) Interaction of VP40 with Rab11.HEK293T cells were transfected with the expression plasmids for VP40 along with GFP, GFP-wtRab11, or GFP-dnRab11.At 48 h.p.t., the cells were harvested and lysed with RIPA buffer.Lysate was incubated with anti-GFP antibody-conjugated magnetic beads.Immunoprecipitants were subjected to western blotting using antibodies against VP40, GFP, or α-tubulin.The experiment was performed three times independently, and representative blots are shown.
Taken together, our study demonstrates that MARV exploits the microtubuledependent Rab11-mediated PM-directed vesicle-trafficking pathway for the release of viral particles.This indicates that two distinct genera of the family Filoviridae commonly exploit the Rab11-dependent endocytic pathway for efficient viral particle formation, which may offer new potential targets for the development of pan-filovirus therapeutics.

Image analysis
For quantification of the intracellular distribution of Rab11-positive vesicles, images for more than 10 cells in three to five fields were acquired with a z-stack of approximately 20 slices at 0.2 µm intervals.Z-stack images were reconstructed with imaging software (IMARIS; OXFORD Instruments, Oxfordshire, UK), and the surface area (µm 2 ) of the cell and the nucleus of the examined cells were analyzed using the ImarisCell module.Distance (µm) between the Rab11-positive signals and the closest nucleus and the volumes in single cells were analyzed using the ImarisCell module.The cellSens (Evident Scientific) was used to calculate the maximal fluorescence intensity of the channels to determine the distribution of the individual fluorescence signals.To determine the localization of Rab11 and microtubules, z-stack images were further processed by deconvolution.Images are shown by maximal intensity projection.For intensity-based co-localization analysis of Rab11 and α-tubulin, images for more than 10 cells in three to five fields were acquired with a z-stack of approximately 20 slices at 0.2 µm intervals.Z-stack images were reconstructed and analyzed using the ImarisCo-loc module.

Characterization of Rab11 incorporation in MARV VLPs
Marburg VLPs were prepared following the procedures for preparing Ebola VLPs (51,70).Equal amounts of the pCAGGS expression plasmids for MARV VP40, GP, and NP were transfected into Expi293F cells using ExpiFectamine (Thermo Fisher Scientific) according to the manufacturer's instructions.At 48 h.p.t., the culture supernatants were harvested and centrifuged at 440 × g for 5 min and then at 2,380 × g for 15 min to remove detached cells and cell debris, respectively.The VLPs were precipitated through a 30% sucrose cushion by ultracentrifugation at 14,860 × g for 1 h at 4°C with an SW32Ti rotor (Beckman, Fullerton, CA, USA).The precipitated VLPs were resuspended in TNE buffer [10 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 1 mM EDTA] and fractionated using a 30%-60% sucrose gradient in TNE buffer at 89,527 × g for 2.5 h at 4°C with the SW32Ti rotor.The expression of MARV VP40 in each fraction was determined by western blotting using a mouse anti-MARV VP40 monoclonal antibody (clone 1-17-1; 1:1,000 dilution).The fractions corresponding to VLPs were collected, and VLPs were concentrated by ultracentrifugation at 14,860 × g for 1 h at 4°C with the SW32Ti rotor.The amount of protein in the VLP suspension was determined using the Bradford protein assay kit (Bio-Rad Laboratories, CA, USA).The morphology of the purified VLPs was confirmed using transmission electron microscopy.The incorporation of Rab11 into Marburg VLPs was characterized by a western blotting-based protease protection assay (49).About 2 µg VLPs was treated with or without 0.05% Triton X-100 for 10 min, then in the presence or absence of 0.1 mg/mL trypsin (Sigma-Aldrich) for another 10 min at room temperature.VLPs were then incubated in Laemmli sample buffer for 5 min at 95°C, followed by western blotting using the mouse monoclonal antibodies against MARV VP40 (clone 1-17-1; 1:1,000 dilution) and Rab11 (BD Biosciences, Franklin Lakes, NJ, USA; 1:1,000 dilution).Band intensity was quantified using CSAnalyzer4 software Ver. 4 (ATTO Corporation, Tokyo, Japan).

Negative-staining electron microscopy
VLPs were fixed with 2.5% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in 0.1 M cacodylate buffer (pH 7.4) (Nacalai Tesque Inc., Kyoto, Japan) overnight at 4°C.Each sample was loaded onto a 200-mesh copper grid with a carbon-coated plastic film (Nisshin EM, Tokyo, Japan) immediately following glow discharge and negatively stained with uranyl acetate solution (1%, wt/vol) for 15 s.The morphology of each sample was observed using a JEM-1400Flash microscope (JEOL, Tokyo, Japan) with an acceleration voltage of 80 kV.

Effect of downregulation of Rab11 or microtubule modulators on Marburg VLPs production
To elucidate the effect of the downregulation of Rab11, siRNAs targeting human Rab11a and Rab11b (Thermo Fisher Scientific) were transfected into HEK293 cells using TransIT-X2.As a control, a siRNA encoding a sequence that does not target any known gene (Thermo Fisher Scientific) was transfected.At 72 h.p.t, Rab11 downregulation was confirmed by western blotting using the mouse monoclonal antibody for Rab11 (BD Biosciences; 1:1,000 dilution) and a rabbit polyclonal antibody for α-tubulin (Medical & Biological Laboratories; 1:1,000 dilution).The siRNA-treated cells were transfected with expression plasmids encoding MARV VP40, NP, and GP using TransIT-X2.At 48 h.p.t., the culture supernatant was harvested and centrifuged at 440 × g for 5 min and then at 2,380 × g for 15 min.VLPs were precipitated through a 20% sucrose cushion by ultracentrifu gation at 14,860 × g for 1 h at 4°C with an SW32Ti rotor.The precipitated VLPs were suspended in the TNE buffer.Expression of MARV proteins in the total cell lysates (TCLs) and the purified VLPs was determined by western blotting using mouse monoclonal antibodies against MARV VP40 (clone 1-17-1; 1:1,000 dilution), NP (clone FS0609; 1:1,000 dilution, gifted by Dr. Ayato Takada), GP (clone 127-8; 1:1,000 dilution, gifted by Dr. Ayato Takada), and a rabbit polyclonal antibody for α-tubulin (Medical & Biological Laborato ries; 1:1,000 dilution).Band intensity was quantified using the CSAnalyzer4 software.To examine the effect of microtubule modulators on VLP formation, HEK293T cells were transfected with the expression plasmids for MARV VP40 using TransIT-X2 (Mirus).At 4 h.p.t., the cells were treated with or without 1 µg/mL nocodazole (Sigma-Aldrich) for 44 h.As a control, the cells were treated with dimethyl sulfoxide (DMSO) (Nacalai Tesque Inc.).Expression of VP40 in the TCLs and production of harvested Marburg VLPs in the supernatant were determined as described above.

Effect of Rab11 downregulation on the release of authentic MARV
siRNAs targeting human Rab11a and Rab11b (Thermo Fisher Scientific) or control siRNAs were transfected into HEK293 or Vero-E6 cells using Transit-TKO (Mirus).At 24 h.p.t., Rab11 downregulation was confirmed by western blotting using a mouse monoclonal antibody for Rab11 (BD Biosciences; 1:1,000 dilution) and a mouse monoclonal antibody for β-actin (Sigma-Aldrich, clone AC-15; 1:1,000 dilution).To determine the effect of Rab11 knockdown on infection with MARV-Angola (GenBank accession: KY047763), we infected siRNA-transfected cells with MARV at an MOI of 0.01 and incubated them for 1 h at 37°C.The cells were washed three times with DMEM and covered with Eagle's minimal essential medium containing 2% FBS (Thermo Fisher Scientific) and 1.2% carboxymethyl cellulose (Sigma-Aldrich).After incubation for 2 days, the cells were fixed with 10% phosphate-buffered formalin overnight at 4°C, and the effect of Rab11 knockdown on viral infection was determined using a focus-forming assay (FFA).MARV-infected cells were observed by immunofluorescence staining with a mouse anti-MARV VP40 antibody (clone 1-17-1; 1:400 dilution).The MARV titers were quantified by measuring the number of fluorescent foci using a ZOE fluorescent cell imager (Bio-Rad).To analyze the effect of Rab11 downregulation on the release of progeny MARV, we inoculated siRNA-transfected cells with MARV at an MOI of 0.01 24 h after siRNA treatment and further incubated them for 1 h at 37°C.The cells were washed three times with DMEM and cultured in DMEM containing 2% FBS.Supernatants were collected at 0, 24, 48, and 72 h post-infection, and the titer of MARV in the supernatant was measured using a median tissue culture infectious dose (TCID 50 ) assay on Vero-E6 cells.
All infectious work with MARV was performed in the biosafety level-4 laboratory at the Integrated Research Facility at the Rocky Mountain Laboratories, Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Hamilton, Montana, USA.All experiments followed standard operating procedures which were approved by the Institutional Biosafety Committee.

Immunoprecipitation
To determine the association of MARV VP40 with α-tubulin, HEK293T cells were transfected with the expression plasmid for MARV VP40 using TransIT-X2.The backbone plasmid was transfected as a control.At 48 h.p.t., the cells were lysed with the tubulinbinding buffer [80 mM HEPES (pH 7.6), 10% Triton X-100, 1 mM MgCl 2 , and 1 mM EGTA] containing 1 mM Phenylmethylsulfonyl fluoride (PMSF), 2 µg/µL Leupeptin, 2 µg/µL Pepstatin A, and 2 µg/µL Aprotinin for 1 h at room temperature.After ultracentrifugation at 12,000 × g for 10 min, the supernatants were incubated with 1 µg/mL rabbit polyclonal antibody against α-tubulin (Medical & Biological Laboratories) or with a 1 µg/mL rabbit polyclonal antibody for VP40 (Integrated BioTherapeutics, Rockville, MD, USA) for 1 h at room temperature.Protein G Sepharose (Cytiva, Marlborough, MA, USA) was added to the supernatants and incubated for 1 h at room temperature.As a control, cell lysates were incubated with Protein G Sepharose alone.Sepharose was washed three times in the same buffer and immunoprecipitants were subjected to western blotting using a mouse monoclonal antibody against MARV VP40 (clone 1-17-1; 1:1,000 dilution) and α-tubulin (Abcam; 1:1,000 dilution).

Statistical analysis
For image analysis, images of more than 10 cells in 3-10 fields were acquired per sample, and the analyzed data were subjected to Student's t test for statistical analysis.For western blot analysis, FFA, and TCID 50 assay, experiments were independently performed more than three times and subjected to Student's t test or one-way ANOVA for statistical analysis.

FIG 1
FIG 1 Filovirus VP40 induced a dispersed distribution of Rab11.(A-D) Effect of filovirus VP40 expression on the intracellular distribution of endogenous Rab11.Vero-E6 cells were transfected with the expression plasmids for (B) mCherry, (C) Ebola virus (EBOV) VP40, and (D) Marburg virus (MARV) VP40.(A) A backbone plasmid was transfected as a control.At 24 h post-transfection, the cells were harvested, and the subcellular distribution of VP40 and Rab11 was analyzed using immunofluorescence staining.Ten cells in three to five fields were analyzed; the representative images are shown.The nuclei were counterstained with Hoechst 33342, and the dotted lines indicate their outlines.Insets show the box area; cell peripheries were determined by phase contrast images and shown in lines; arrows show filamentous (Continued on next page)

FIG 1 ( 4 FIG 2
FIG 1 (Continued) structures on the cell surface; arrowheads indicate the intracellular patched structures; scale bars: 10 µm.(E-G) Quantification of the intracellular distribution of Rab11 upon expression of filovirus VP40.(E) The surface area (µm 2 ) of the cell and the nucleus were measured using IMARIS imaging software.(F) The distance (µm) between Rab11-positive vesicles and the closest nucleus and their volumes in single cells were measured using the IMARIS imaging software.The ratio of the total volume of Rab11-positive signals distributed in the indicated regions in single cells was measured.Ten cells were analyzed; the results are shown as the mean ± SD. (G) Summary of the % total volume of Rab11-positive vesicles detected at the indicated distance from the closest nucleus of the analyzed cells.n.s.; not significance, *, P < 0.05; **, P < 0.01 versus respective control (Student's t test).

FIG 2 (FIG 3 6 FIG 4
FIG 2 (Continued) shown in pale blue bars.A.U.; arbitrary unit.Scale bars: 10 µm.(C-E) Quantification of the intracellular distribution of Rab11 upon expression of MARV GP and NP.(C) The surface area (µm 2 ) of the cell and the nucleus were measured using IMARIS imaging software.(D) The distance (µm) between Rab11-positive vesicles and the closest nucleus and their volumes in single cells were measured using IMARIS imaging software.The ratio of the total volume of Rab11-positive signals distributed in the indicated regions in single cells was measured.Ten cells were analyzed; the results are shown as the mean ± SD. (E) Summary of the % total volume of Rab11-positive vesicles detected at the indicated distance from the closest nucleus of the analyzed cells.n.s.; not significance, *, P < 0.05; versus respective control (Student's t test).

FIG 5 (
FIG 5 (Continued) or (E) siRNAs for Rab11, Vero-E6 cells were further transfected with the expression plasmid of VP40.At 48 h.p.t., the subcellular distribution of VP40 and Rab11 was analyzed using immunofluorescence staining.Ten cells in 5-10 fields per sample were analyzed; the representative images are shown.The nuclei were counterstained with Hoechst 33342, and the dotted lines indicate their outlines.The insets show the boxed areas.The plots indicate the individual fluorescence intensity along each of the periphery of individual cells was determined by the phase contrast images and shown in corresponding dotted lines.The location of the PM is shown in pale blue bars.A.U.; arbitrary unit.Scale bars in the large panels: 10 µm.

FIG 6
FIG 6 Effect of Rab11 downregulation on Marburg VLPs formation.(A) Downregulation of Rab11.HEK293 cells were transfected with siRNAs against Rab11a and Rab11b.At 72 h.p.t., the cells were transfected with the expression plasmids for MARV VP40, NP, and GP.At 48 h.p.t., the cells and culture medium were harvested.TCLs were subjected to western blotting with the antibodies against VP40, NP, GP, Rab11, or α-tubulin.(B) The intensity of the band's correspondence to VP40, NP, and GP was quantified and normalized with that of α-tubulin.(C) VLPs in the culture medium were purified by ultracentrifugation, and VLPs were subjected to western blotting with the antibodies against VP40, NP, or GP.(D) The intensity of the band's correspondence to each protein was quantified.The experiment was performed three times independently, and representative blots (A and C) and the mean ± SD are shown (B and D).*, P < 0.05; **, P < 0.01; ***, P < 0.001 versus respective control (one-way ANOVA).

FIG 7
FIG 7 Effect of Rab11 downregulation on MARV particle production.(A) Downregulation of Rab11.HEK293 or Vero-E6 cells were transfected with siRNA against Rab11a and Rab11b.At 24 h.p.t., the expression of Rab11 in siRNA-treated cells was assessed by western blotting.The experiment was performed three times independently, and the representative blot is shown.(B) The intensity of corresponding bands to Rab11 was normalized with that of the β-actin band.The mean and SD are shown.(C) The effect of Rab11 knockdown on infection with MARV.siRNA-treated cells were inoculated with MARV at an MOI of 0.01 and subjected to the focus-forming assay.(D and E) The effect of Rab11 knockdown on the release of MARV production.Virus titers in the culture media collected at 0, 24, 48, and 72 h post-infection were analyzed by the TCID 50 assay using Vero-E6 cells.The experiment was performed three times independently, and the geometric mean ± SD are shown.n.s., not significant, *P < 0.05; ***P < 0.001; ****P < 0.0001 versus respective control (Student's t test).

FIG 8
FIG 8 VP40 exploits the microtubule network for self and Rab11-positive vesicle trafficking for efficient VLP formation.(A-C) Effect of VP40 on the distribution of the microtubule network.Vero-E6 cells were transfected with the expression plasmids for MARV VP40.At 4 h.p.t., the cells were treated with DMSO (B) 1 µg/mL nocodazole (C) and further incubated for 20 h.As a control, the cells were transfected with a backbone plasmid (A).The subcellular distribution of VP40 and α-tubulin was analyzed using immunofluorescence staining.Ten cells in three fields were analyzed; the representative images are shown.The nuclei were counterstained with Hoechst 33342, and the dotted lines indicate their outlines.Cell peripheries were determined by phase contrast images and shown in lines.Insets show the box area.The plots indicate individual fluorescence intensities along each line.The location of the PM is shown in pale blue bars.A.U.; arbitrary unit.Arrows represent co-localized signals.Scale bars: 10 µm.(D and E) Effect of depolymerization of microtubule on VLP formation.HEK293T cells Ten cells in three fields were analyzed; the representative images are shown.The nuclei were counterstained with Hoechst 33342, and the dotted lines indicate their outlines.Cell peripheries were determined by phase contrast images and shown in dotted lines.Insets show the box area.Proportion of Rab11 co-localized with α-tubulin was measured using the IMARIS imaging software and shown in the image.Scale bars: 10 µm.

FIG 9
FIG 9 VP40 interacts with α-tubulin but not with Rab11.(A and B) Interaction of VP40 with α-tubulin.HEK293T cells were transfected with the expression plasmids for VP40.At 4 h.p.t., the cells were treated with 1 µg/mL nocodazole for 44 h.As a control, the cells were treated with DMSO.The cells were