Vacuolar Protein Sorting Pathway Contributes to the Release of Marburg Virus

VP40, the major matrix protein of Marburg virus, is the main driving force for viral budding. Additionally, cellular factors are likely to play an important role in the release of progeny virus. In the present study, we characterized the influence of the vacuolar protein sorting (VPS) pathway on the release of virus-like particles (VLPs), which are induced by Marburg virus VP40. In the supernatants of HEK 293 cells expressing VP40, different populations of VLPs with either a vesicular or a filamentous morphology were detected. While the filaments were almost completely composed of VP40, the vesicular particles additionally contained considerable amounts of cellular proteins. In contrast to that in the vesicles, the VP40 in the filaments was regularly organized, probably inducing the elimination of cellular proteins from the released VLPs. Vesicular particles were observed in the supernatants of cells even in the absence of VP40. Mutation of the late-domain motif in VP40 resulted in reduced release of filamentous particles, and likewise, inhibition of the VPS pathway by expression of a dominant-negative (DN) form of VPS4 inhibited the release of filamentous particles. In contrast, the release of vesicular particles did not respond significantly to the expression of DN VPS4. Like the budding of VLPs, the budding of Marburg virus particles was partially inhibited by the expression of DN VPS4. While the release of VLPs from VP40-expressing cells is a valuable tool with which to investigate the budding of Marburg virus particles, it is important to separate filamentous VLPs from vesicular particles, which contain many cellular proteins and use a different budding mechanism.In recent years, virus-like particles (VLPs), which are formed upon recombinant expression of the viral matrix and/or surface glycoproteins, have been recognized as representing powerful tools for developing novel vaccines and investigating certain aspects of the viral replication cycle (24, 44, 59, 63). Matrix proteins from many enveloped RNA viruses, including retroviruses, rhabdoviruses, filoviruses, paramyxoviruses, orthomyxoviruses, and arenaviruses, are able to induce VLPs (10, 14, 18, 28-30, 48, 49, 52). Increasing evidence also indicates that budding activity, and thus the release of VLPs, is often influenced by a complex interplay with components of the endosomal sorting complexes required for transport (ESCRTs), which mainly constitute the vacuolar protein sorting (VPS) pathway (16, 38, 42, 54). ESCRTs trigger the formation and budding of vesicles into the lumina of multivesicular bodies (MVBs), and the constituents of the ESCRTs are recycled by the activity of VPS4, an AAA-type ATPase. Expression of dominant-negative (DN) VPS4 mutants, which lack the ability to bind or hydrolyze ATP, blocks recycling of the ESCRTs and induces the formation of enlarged endosomes lacking internal vesicle accumulation (2, 3, 7). The inward budding of vesicles into the MVBs is topologically similar to the budding of viruses, since the vesicles bud away from the cytosol and into the lumen (reviewed in references 1, 20, and 26). Therefore, it is not entirely surprising that viruses use the cellular ESCRT machinery to organize the budding of viral progeny. Interactions between viral matrix proteins and ESCRTs occur through tetrapeptide motifs, known as late domains, which were first identified in retroviruses. Known late domains consist of the amino acid sequence P(T/S)AP, PPxY, or YxxL, where “x” represents any amino acid (19, 25, 62). The P(T/S)AP motif, for example, mediates interaction with tumor susceptibility gene 101 (Tsg101) (16, 36, 57); the PPxY motif mediates binding to WW domains of Nedd4-like ubiquitin ligases (9, 22); and the YxxL motif mediates interaction with AIP1/Alix (35, 47, 58). Recently, a novel late-domain motif, FPIV, has been identified in paramyxoviruses (46), and it is thought that additional late-domain motifs remain to be discovered (for a review, see reference 5).Inhibition of the VPS pathway has been shown to inhibit the budding of various viruses that are released with the help of ESCRTs. However, the budding of viruses and VLPs depends on the activity of ESCRTs to different degrees. Downregulation of Tsg101, a member of the ESCRT-I complex, inhibited the release of VLPs mediated by lymphocytic choriomeningitis virus Z protein and Marburg virus (MARV) VP40 (42, 54) but did not substantially inhibit the release of Gag-induced VLPs of Moloney murine leukemia virus and Rous sarcoma virus or that of matrix protein-induced VLPs of rabies virus (16, 27, 38). Expression of DN VPS4 inhibited the release of VLPs induced by the Gag proteins of Rous sarcoma virus and Moloney murine leukemia virus (16, 38) as well as that of VLPs induced by Lassa virus Z protein (55) but had no effect on the budding of rabies virus and cytomegalovirus (13, 27). These data indicate that in spite of the presence of late-domain motifs, a block in the VPS pathway may not always be critical for the budding of VLPs. In addition, the lack of known late domains in many enveloped viruses raises the question of whether they use other entry points into the VPS pathway or whether they exploit entirely different mechanisms of budding (60). To date, knowledge of how viral matrix proteins engage cellular machineries, such as the VPS pathway, to induce viral budding at the plasma membrane is very limited (8).The matrix protein VP40 of MARV contains only one known late-domain motif, PPPY, and a recent study showed that mutation of this late domain inhibited the release of VP40-induced VLPs. In addition, depletion of Tsg101 reduced the release of VP40-induced VLPs, suggesting that ESCRT-I is involved in this process (54). Whether a functional VPS pathway is important for the release of MARV VP40-induced VLPs or MARV particles remains unknown.VLPs induced by many viral matrix proteins have a morphology similar to that of cellular vesicles, which makes it difficult to separate the spherical VLPs from released cellular vesicles (4, 17, 53). In contrast, VLPs induced by the filovirus matrix protein VP40 are elongated and similar in morphology to viral particles (30, 49). Nevertheless, we observed that the supernatants of cells expressing VP40 contained various populations of particles with different morphologies. This raised the questions of whether the different particles are released by the same mechanism, whether they are all induced by VP40, and whether they are dependent on the same cellular pathways.The aim of the present study was to analyze the populations of particles released from cells expressing the MARV matrix protein VP40 and to gain further insights into the interaction between MARV and the cellular machinery involved in the budding of VLPs and MARV particles.We found that cells expressing VP40 released vesicular and filamentous particles, which could be separated by gradient centrifugation. Fractions with mainly vesicular particles represented a mixture of vesicles containing exclusively cellular proteins and vesicles also containing VP40 and few short filamentous particles. Longer filamentous particles, whose morphology resembled that of MARV particles but which displayed a much higher variability in length (400 nm to 5 μm), were found in denser gradient fractions. Filamentous VP40-induced VLPs were able to sort out cellular proteins efficiently. Release of VP40-induced filamentous VLPs was supported by the late-domain motif present in VP40, and inhibition of the cellular ESCRT machinery reduced the amount of these VLPs in the supernatant. Interestingly, the release of VLPs induced by a mutant of VP40 lacking the late domain was also reduced by inhibition of the cellular ESCRT machinery. Expression of a DN mutant of VPS4 diminished the budding of infectious MARV particles by 50%, a finding consistent with the idea that the activity of the ESCRT machinery supports viral budding but is not essential.