Immune Evasion in Ebolavirus Infections

VP35

The VP35 is a part of the viral RNA polymerase complex, along with NP, VP30, and the L protein. VP35 is an important component for EBOV to prevent the activation of the innate immune response. The C-terminal region of VP35 contains an IFN inhibitory domain (IID) (32,40). The IID has three active regions. First, it contains a binding site for the NP. Second, it has a central basic patch that can bind to the phosphate backbone of dsRNA. Finally, the IID contains a region that binds and caps the ends of dsRNA. The last two regions prevent the activation of the RIG-I pathway and the protein kinase R (PKR) by effectively “hiding” the dsRNA from cellular sensor proteins. This mechanism has been confirmed by crystal structure analysis of VP35-dsRNA complexes (41). For the Marburg virus (MARV) VP35, these structures revealed that VP35 can bind both the caps of dsRNA molecules and fully coat the dsRNA backbone (3). Since the VP35-coated viral dsRNA cannot be detected, infected cells cannot detect the presence of the virus, and, consequently, cannot produce IFN-α or β, therefore preventing the establishment of an early antiviral state in the host.

VP35 was also shown to interact with the dsRNA binding protein 76 (DRBP76), a host protein that inhibits the replication of certain viruses. Shabman et al. (71) have shown that DRBP76 can bind VP35 and reduce EBOV replication. The interaction of VP35 and DRBP76 prevented the coimmunoprecipitation of VP35 with NP. Shabman et al. also showed that the presence of DRBP76 reduces the activity of the polymerase complex in a mini-genome assay. Interestingly, binding of DRBP76 to VP35 did not seem to inhibit VP35's anti-IFN activity, suggesting that the basic patch and the end-cap regions are unaffected by the presence of DRBP76. A compound that mimics DRBP76 and blocks the interaction of VP35 and NP may be a realistic target for the development of small-molecule inhibitors of EBOV.

Leung et al. (43) have proposed that VP35 is also capable of preventing IRF7 phosphorylation, and, consequently, dimerization, based on its ability to block the production of IFN in conventional DCs (cDCs) but not in plasmacytoid DCs (pDCs). By using an engineered Newcastle disease virus that expresses the EBOV VP35, they have shown that pDCs, but not cDCs, can relocate IRF7 to the nucleus and secrete IFN in large quantities. They proposed that in cDCs, VP35 blocks the TBK1/IKKɛ-mediated phosphorylation of IRF7, preventing RIG-I-dependent production of type I IFN. In contrast, pDCs sense RNA-virus infections by detecting endosomal ssRNA through TLR7, which leads to phosphorylation of IRF7 through IKKα/IRAK1, which are probably not inhibited by VP35.

Supporting this hypothesis, data published by Prins et al. (57) showed that VP35 interacts directly with TBK1/IKKɛ and is phosphorylated by these kinases. In addition, Prins et al. showed that in coimmunoprecipitation assays, the interaction of IRF3 and IRF7 with IKKɛ is greatly reduced in the presence of increasing amounts of VP35 (57). They demonstrated the effect of this inhibition by transfecting cells with a reporter plasmid expressing luciferase under the IFN-α4 promoter and vectors expressing IRF7 alone or IRF7 and VP35. The cells were subsequently infected with sendaï virus and showed a return to baseline in reporter activity in the presence of both IRF7 and VP35.

An additional role of VP35 is to facilitate the SUMOylation of IRF7 and IRF3 (11). As reported earlier, the SUMOylation of IRF3/7 on viral infection is carried out by TLR receptors and RIG-I pathways and seems to be a part of a negative feedback loop (36). In this case, VP35 would appear to accelerate the SUMOylation of IRF3/7 and reduce their transcriptional activity. Since it appears that PIAS1, a SUMO E3 ligase, is phosphorylated by IKKα, which can also phosphorylate IRF7, the SUMOylation of IRF7 and possibly IRF3 may be a way for cells to finely regulate the production of type I IFN and the associated genes. The EBOV VP35 seems to take advantage of this system to assist in inhibiting the production of type I IFN.

Studies comparing the structure and activity of VP35 from EBOV and RESTV have described functional and structural differences between the two proteins (41). The VP35 from RESTV has been shown to be more heat stable in its free form and to have reduced binding to dsRNA. These differences may help explain why RESTV was seen to have a reduced ability to inhibit the type I IFN response in infected cells compared with EBOV and MARV (31). These findings may explain why RESTV is less pathogenic than EBOV, but they apply mostly to species-independent roles of VP35 and, thus, are unlikely to explain why this difference in pathogenicity is more evident in humans than in NHPs (17).

The numerous adaptations of VP35 to prevent the production of type I IFN suggests that inhibition of these IFNs, and other associated gene products, is critical to the production of a pathogenic infection by EBOV.

VP24

The other protein from EBOV that interferes with IFN activity is VP24, which forms a part of the viral matrix, as the minor matrix protein, along with VP40, the major matrix protein. Inside the cell, VP24 appears to have two main roles: (i) It reduces the activity of the transcription/replication complex, and (ii) it interferes with IFN signaling.

It was suggested by Iwasa et al. (27) that VP24 reduces the activity of the EBOV transcription/replication complex to transition from the transcription/replication stage to the packaging stage of its cycle. By using a mini-genome assay in the 293 cell line, they have shown that the presence of VP24 greatly reduces the expression of the reporter gene. They suggested that VP24 acts at the level of the transcription/replication complex, because the levels of reporter mRNA and positive stranded genomes were also reduced. These data indicate that VP24, along with VP40, regulates the virus life cycle by stopping the activity of the transcription/replication complex.

VP24 also has important immune evasion functions through its ability to interfere with the JAK-STAT pathway. This signaling pathway is required for the cells to respond to type I and II IFNs. According to Shabman et al. (70), VP24 binds to a C-terminal site of karyopherin-α1 that overlaps the binding site of the heterogenous nuclear ribonuclear protein complex C1/C2 (hnRNP C1/C2). Karyopherin-α1 is a transporter protein; it binds proteins targeted for the nucleoplasm, such as phosphorylated STAT1, and brings them across the nuclear pore with the help of hnRNP C1/C2. Since VP24 interferes with the binding between karyopherin-α1 and hnRNP C1/C2, it prevents phosphorylated STAT1 from entering the nucleoplasm to activate gene transcription.

Shabman et al. have also shown that this interaction between VP24 and karyopherin-α1 redistributes hnRNP C1/C2 to the cytoplasm. They also suggest that hnRNP C1/C2 may have a role in EBOV replication, as it is able to prime cap-independent, IRES-dependent translation. hnRNP C1/C2 also stabilizes unprocessed mRNAs during splicing and nuclear export.

In addition, VP24 was recently shown to be able to interact directly with STAT1 (80,81). Although the exact effects of this interaction are still unknown, its importance is highlighted by the fact that STAT1 knockout mice can be lethally infected by wild-type filoviruses (63). This suggests that IFN signaling is likely important to the defense against EBOV and the other filoviruses.

Recent data suggest that, in HEK 293 cells, VP24 is also capable of inhibiting the phosphorylation of p38, thereby blocking IFN signaling through the MAPK pathway (23). Such an interaction has not been reported for other viruses, even in cases where the p38 MAPK pathway is critical for the establishment of the antiviral response, such as for hepatitis C virus. However, VP24 did not interfere with MAPK signaling in HeLa cells, which suggests that this interaction may be cell-type dependent.

Along with VP35, VP24 enables EBOV to block the early IFN response. The initial detection of dsRNA is made harder by VP35, while VP24 abrogates the response to IFN-α/β in infected cells. This means that even when RIG-I, or another dsRNA detection pathway, can detect the presence of EBOV, neighboring infected cells will not be able to respond to the IFN produced.

In light of these findings, it is unsurprising that VP24 has been coined as a molecular determinant of EBOV virulence in rodent models (13,48), requiring specific mutations for the virus to be completely lethal. What is more surprising is the fact that VP24 from EBOV, mouse-adapted EBOV, and RESTV show the same potential for blocking human and mouse phospho-STAT1 activity (64). In the case of 293T cells, RESTV VP24 seemed slightly better at blocking IFN signaling. This information suggests that even though VP24 adaptation is necessary to have a lethal virus, in the mouse model at least, this adaptation does not reduce VP24's activity in other species and also does not explain the differential pathogenicity of EBOV, mouse-adapted EBOV and RESTV.

The crystal structures of residues 1–233 of the SUDV and RESTV VP24 proteins were solved in 2012 and revealed that the protein folds into a pyramidal shape with conserved pockets at the bottom of faces 1 and 3 (81). In addition, the same study confirmed that STAT1 (amino acids 1-683) and VP24 interact directly in an ELISA and identified key residues that are important for this interaction by using deuterium exchange mass spectrometry (DXMS).

The MARV VP24 does not inhibit IFN signaling; in marburgviruses, this role is carried out by the major matrix protein VP40 (73). This suggests that the ability to interfere with IFN signaling evolved independently in each lineage after the divergence between the ebolaviruses and the marburgviruses about 10,000 years ago (9).

Infection of DCs

The ebolaviruses have evolved a number of mechanisms to prevent the effective initiation of the innate immune response. Given that EBOV replicates in certain innate immune cells (monocytes, macrophages, and DCs), it is not surprising that the virus is able to block many signaling pathways that would cause these cells to activate. As stated earlier, this blocking is the result of VP35 and VP24 interfering with the normal phosphorylation of IRF3/7 and SUMOylation of IRF7 as well as the nuclear transport of STAT1 and the phosphorylation of p38. In three recent publications, important sites for immune interference were identified in VP24 and VP35 based on structural studies (40,81) and viruses with mutations to abolish the activities of VP24 and VP35 were generated (44).

The regions responsible for immune interference were termed innate response antagonist domains (IRADs) and included one VP24 site and 5 sites within VP35. When viruses mutated to remove the activity of any one of these IRADs were used to infect human DCs, most of the inhibitory activity of the virus was removed. This suggests that neither VP24 nor VP35 alone is sufficient to block the activation of DCs.

Interestingly, EBOV is incapable of infecting pDCs (42). Results from Leung et al. (42) show that virus-like particles (VLPs) cannot even trigger endocytosis or any form of activation of pDCs. As a consequence, pDCs do not produce IFN-α in infection models. This selectivity in EBOV's infectivity may have evolved to prevent the activation of pDCs, which would induce a strong IFN response. As stated earlier, VP24 and VP35 are capable of preventing the phosphorylation of IRF7 by the RIG-I pathway but not by the TLR pathway, which would be activated in pDCs (via TLR7). By contrast, the infection of cDCs by EBOV leads to exponential virus growth and impaired activation of the cells (4).

Recently, the IFN-inducible transmembrane protein 1 (IFITM1) was shown to reduce the permissivity of various cell lines, which highlights the importance of an early IFN response (26).

sGP

The role of the sGP in the virus life cycle and pathogenesis in primates is still relatively poorly understood. Recent data support the importance of sGP in the context of animal infections (35,74). The 7U variant of the virus required five passages in Vero E6 cells before its editing site became predominantly 8U. In spite of this, the 8U variant required only a single passage in guinea pigs for its editing site to return to the wild-type sequence. This preference for a 7U variant was also observed in virus isolated from lethally infected macaques (35). These results suggest a very strong selective pressure in favor of the production of sGP. The authors hypothesized that the anti-tetherin activity of GP_1,2 (30,36) makes high levels of expression of this protein more favorable in vitro.

In the search for the role of sGP, it has been proposed that sGP might play a role in immune evasion by adsorbing antibodies directed against the GP₁ part of the full-length glycoprotein (51), which shares its first 295 amino acids with the sGP (69). However, the differences in the structure of the two proteins, such as the GP_1,2 being trimeric while the sGP forms a dimer (69), would probably reduce the number of epitopes they share. The lack of a crystal structure of EBOV sGP and of a complete structure for GP_1,2 makes comparing the available epitopes of these two proteins much harder. Nevertheless, recent empirical evidence suggests that sGP can subvert the antibody response of mice immunized against GP_1,2 when sGP is present during or after the GP_1,2 immunization (51). The proposed mechanism is that the large quantities of sGP produced during infection stimulate the production of antibodies that cross-react between sGP and GP_1,2, thereby reducing the amount of GP_1,2-specific antibodies. It is still unclear how important this immune subversion is to the pathogenesis of EBOV.

The EBOV sGP was found by Wahl-Jensen et al. (76) to restore the barrier function of human vascular endothelial cells exposed to the full-length GP_1,2 in the form of VLPs. Their results also showed that the presence of sGP partially reduced the increase in permeability induced by tumor necrosis factor (TNF)-α. They suggested that the virus might have evolved this characteristic to prevent white blood cells from accessing the site of infection.

More recently, it was proposed that the sGP may play a role in viral particle formation. Iwasa et al. (28) have published data suggesting that the sGP can replace the GP₁ subunit of the GP_1,2 to form infectious viral particles. Nonetheless, the reduced replication of VSV viruses pseudotyped with sGP-GP points to this role as a secondary and maybe purely stochastic one.

The amounts of sGP produced by the virus during infection and the massive bystander apoptosis observed in lymphocytes prompted Wolf et al. (78) to evaluate whether the protein could cause the death of these cells. The presence of purified sGP did not change the apoptosis rate of Jurkat cells, even in the presence of TNF-α or the death receptor ligands FASL and TRAIL.

It is important to note that the marburgviruses do not possess the editing site present in the EBOV genome, and, consequently, do not produce sGP (16). Despite this difference, the marburgviruses are at least as pathogenic as the ebolaviruses, and all known variants of Marburg have been pathogenic in humans. This suggests that sGP may not be a determinant of filovirus human pathogenicity, which is supported by the fact that RESTV also has 7U residues at positions 1021–1027 but is not pathogenic to humans.

Finally, one possible role that has not been explored for sGP would be to mostly be a “place-holder” protein. The full-length spike protein contains a very large and highly immunogenic mucin-like domain. The size of this domain eventually causes cellular detachment and death of the infected cell. The immunogenicity of the GP_1,2 protein could also lead to rapid clearance of infected cells by the innate immune system and the early adaptive responses. It is possible that the editing site evolved as a way to reduce the amount of GP_1,2 produced during infection to levels that enable efficient replication with the lowest impact on the survival of the infected cell.

Δ-peptide

The sGP is cleaved post-translationally to yield sGP and the Δ-peptide. Radoshitzky et al. (62) have shown that the presence of a Δ-peptide-Fc fusion protein can inhibit entry of EBOV into cells in a dose-dependent manner. However, this effect was only observed for the peptides from highly pathogenic viruses in humans. The Δ-peptide and other secreted forms of the glycoprotein (such as sGP, GP₁, and GP_1,2Δ) were shown to be unable to induce activation of human macrophages in vitro; only VLPs containing the GP_1,2 were able to activate the cells (75).

GP_1,2

The full-length GP forms the virion spike and mediates attachment to target cells as well as membrane fusion. It requires cleavage by cathepsins B & L in the phagosome in order to induce membrane fusion (10). This cleavage permits the interaction of the GP with another receptor inside the phagosome, the endosomal cholesterol trasnporter Niemann-Pick disease C1 (NPC1) (8). It is still unclear exactly with which surface protein(s) EBOV first makes contact, as NPC1 is probably only accessible after the virus has made contact with the target cell and has been taken up in a phagosome. It has been suggested that lectins are implicated because of the heavy glycosylation of the mucin-like domain of GP₁.

Aleksandrowicz et al. (1) have shown that EBOV enters through both macropinocytosis and by clathrin-mediated endocytosis, but the specific cell-surface receptor was not identified. It is possible that many such receptors exist for EBOV's target cell types, which makes identifying any specific receptor harder. The surface molecule DC-SIGN has been identified as promoting close contact between EBOV virions and the target cell without actually mediating the internalization of the attached virions (47).

Investigations into the cause of the apparent downregulation of some surface proteins, such as β1-integrin and the type I major histocompatibility complex (MHC-I) (72), revealed that the proteins were not downregulated, but rather hidden from detection by the mucin domain of the GP_1,2 protein (18). More recent reports also suggest that this steric shielding effect can block Fas-mediated apoptosis (52,53). It remains unclear whether the amount of GP_1,2 produced during a normal infection is sufficient to hide MHC-I, as this effect is seen in the context of cells transfected with a plasmid expressing GP_1,2 under a strong promoter such as the CMV promoter.

The EBOV GP has been shown to have anti-tetherin activity (30,36). Tetherin is a cell surface protein with two transmembrane domains joined by an extracellular loop. It has a role in preventing the release of enveloped viruses from the surface of the cell. It is thought to mediate this activity by having one transmembrane domain that remains inserted into the cell membrane while the other transmembrane domain is taken up in the viral envelope, effectively tethering the virus to the cell. Kühl et al. (36) have shown that the GP₂ subunit of EBOV is capable of relocalizing tetherin; although full-length GP_1,2 colocalizes with tetherin, it did not relocalize it.

This is consistent with a 2014 report suggesting that the transmembrane domain is required for anti-tetherin activity in filoviruses but that it is not sufficient to mediate the activity (21). This may explain the finding of soluble GP₁ subunits, in that having GP₂ alone prevents tetherin from being brought to the surface of the cell and the production of soluble GP₁ is only incidental. In addition, it is possible that the switch to 8U viruses after serial passages in vitro permits an increased production of GP₂, which is more useful in a system that is free of anti-EBOV GP antibodies than sGP.