Dual modes of rabies P-protein association with microtubules: a novel strategy to suppress the antiviral response

Nucleocytoplasmic trafficking of transcription factors is a central mechanism in cellular processes such as development and innate immunity (Gardner and Montminy, 2005; Jans et al., 2000; Pemberton and Paschal, 2005; Poon and Jans, 2005; Randall and Goodbourn, 2008; Reich and Liu, 2006; Ziegler and Ghosh, 2005). Conventionally, protein nuclear trafficking involves the interaction of nuclear localisation sequences (NLSs) or export sequences (NESs), located within the cargo protein, with import receptor proteins (importins) or export receptor proteins (exportins), which mediate translocation of the cargo across the nuclear envelope (Jans et al., 2000; Pemberton and Paschal, 2005; Poon and Jans, 2005). To activate transcription, transcription factors must be within the nuclear compartment; thus, modulation of transcription factor nuclear import and/or export is a potent regulatory mechanism. Numerous mechanisms are reported to regulate nuclear transport of transcription factors and other nucleocytoplasmic proteins by affecting the interaction of importins and exportins with NLSs or NESs, respectively (Gardner and Montminy, 2005; Jans et al., 2000; Pemberton and Paschal, 2005; Poon and Jans, 2005). For example, the transcription factors signal transducers and activators of transcription (STATs) STAT1 and STAT2 exist in a latent monomeric state until activation by phosphorylation and dimerisation, accompanied by activation of NLS function, results in accumulation of STAT dimers in the nucleus (Randall and Goodbourn, 2008).

Recent data, however, indicate that nuclear import-export can be subject to regulation additional to simple signal recognition by importins or exportins. Specifically, components of the cytoskeleton, particularly the microtubule (MT) network, can affect nuclear trafficking of specific proteins (Campbell and Hope, 2003; Giannakakou et al., 2000; Lam et al., 2002; Mikenberg et al., 2007; Moseley et al., 2007b; Rathinasamy and Panda, 2008; Roth et al., 2007). Intriguingly, there exist two opposing modes of MT-regulated nuclear import. In the first (MT-facilitated import), the nuclear localisation of proteins such as p53, parathyroid hormone-related protein, retinoblastoma protein (Rb), and the rabies virus P-protein (RPP) can be facilitated by MTs (Campbell and Hope, 2003; Giannakakou et al., 2000; Lam et al., 2002; Mikenberg et al., 2007; Moseley et al., 2007b; Rathinasamy and Panda, 2008; Roth et al., 2007). This generally involves the MT motor dynein, which mediates movement on MTs from the cell periphery to the perinuclear region (Dohner et al., 2005), and therefore, MTs and dynein are thought to shuttle protein cargo toward the nucleus, resulting in enhanced nuclear import kinetics (Lam et al., 2002; Moseley et al., 2007b; Roth et al., 2007). In the second mode (MT-inhibited import), protein association with MTs, involving sequestration to the cytoplasm, results in inhibition of nuclear import (Campbell and Hope, 2003; Haller et al., 2004; Malki et al., 2005; Yamasaki et al., 2005; Ziegler and Ghosh, 2005).

Importantly, it has been shown that MT facilitation or MT inhibition of import is a property of a select group of nuclear-trafficking proteins and that conventional importin-mediated nuclear import of other proteins is independent of the MT cytoskeleton, including import of interferon (IFN)-activated STAT1 (Lillemeier et al., 2001; Roth et al., 2007). Thus, MTs do not have an integral role in the nuclear import of all proteins, but interact directly or indirectly with certain proteins via specific sequences therein (MT-association sequences, MTASs) to affect subcellular localisation. This is supported by the fact that NLS-containing MT-facilitated-type proteins can revert to conventional MT-independent import as a result of deleting the MTAS, and that a particular form of MTASs (dynein light chain association sequences, DLCASs) found in a number of viral or cellular proteins including RPP, can confer a MT-facilitated phenotype on MT-independent NLS-containing proteins (Moseley et al., 2007b).

Nuclear trafficking of transcription factors is central to the innate immune response to viral infection, which is mediated by the IFN system (Chelbi-Alix et al., 2006; Haller et al., 2006; Randall and Goodbourn, 2008). Viral infection results in the activation of transcription factors such as NF-κB and IRF3, which translocate to the nucleus to activate IFN gene transcription (Randall and Goodbourn, 2008). The IFN gene products, such as IFNβ, act in an autocrine and paracrine fashion, binding to IFN receptors on the cell surface to initiate Jak-STAT-signalling pathways and activate STAT1 and STAT2 (see above). Nuclear-localised STAT transcription factor complexes then activate specific sequences [IFN-sensitive response elements (ISREs) for IFNα/β] in the promoter regions of IFN-sensitive genes, culminating in the expression of hundreds of IFN-responsive proteins that establish the cellular antiviral state (Chelbi-Alix et al., 2006; Haller et al., 2006; Randall and Goodbourn, 2008). Viral evasion of the IFN response is therefore an essential component of viral infection, and is achieved by the expression of virus-encoded proteins (IFN-antagonists), including the rabies virus protein RPP, which inhibit the IFN system via diverse mechanisms (Brzozka et al., 2005; Brzozka et al., 2006; Chelbi-Alix et al., 2006; Randall and Goodbourn, 2008; Shimizu et al., 2006; Vidy et al., 2005). The capacity of a virus to antagonise IFN is also a determinant of its ability to infect multiple species (Randall and Goodbourn, 2008), the significance of which is highlighted by recent outbreaks of emerging zoonotic viruses in human populations (e.g. SARS and Nipah virus) and the persistence of others, such as rabies virus (the cause of >50,000 human fatalities per year) and Ebola virus.

Here, we show for the first time that the P3 protein (a truncated version of RPP, produced by leaky scanning in rabies-virus-infected cells) (Fig. 1) (Chenik et al., 1995), forms MT-inhibited-type interactions with MTs. These interactions are dependent on the oligomeric state of RPP, suggesting that it may be able to switch dynamically between MT-inhibited and MT-facilitated modes. This represents the first evidence that a protein can undergo dual modes of MT association to regulate nuclear trafficking. We further show that P3 protein can cause the MT-independent-type protein STAT1 to form MT-inhibitory associations, resulting in sequestration of STAT1 away from the nucleus and inhibition of IFN signalling that is dependent on MTs and on the oligomeric state of RPP. This is the first report of MT function in viral IFN antagonism and identifies a novel mechanism to regulate STAT transcriptional activity, whereby the IFN-antagonist protein tethers STATs to the cytoskeleton, preventing signal transduction vital to the establishment of an antiviral state.

We previously showed that the DLC-AS region within residues 139-172 of RPP is able to confer MT-facilitated nuclear import on MT-independent-type protein constructs containing the RPP NLS (residues 174-297) or heterologous NLSs (Moseley et al., 2007b). The main rabies-virus-expressed form of RPP able to localise to the nucleus is the P3 protein (residues 54-297); this is due to the absence of residues 1-53, which inactivates the strong NES (NES1) that causes nuclear exclusion of full-length RPP, P1 (residues 1-297) (Fig. 1).

To analyse the role of MTs in nuclear import of P3, we transfected Vero cells to express GFP-P3 or GFP-RPP_139-297 before treatment with or without the MT-disrupting drug nocodazole (NCZ) and analysis of subcellular localisation by quantitative confocal laser scanning microscopy (CLSM) (Fig. 2) to determine the ratio of nuclear to cytoplasmic fluorescence (Fn/c) (Moseley et al., 2007a; Moseley et al., 2007b; Roth et al., 2007). The results confirmed previous observations that GFP-RPP_139-297 uses a MT-facilitated import mechanism, as indicated by a reproducible and significant decrease in nuclear accumulation of GFP-RPP_139-297 in NCZ-treated cells compared with untreated cells (Fig. 2B) (Moseley et al., 2007b). Conversely, GFP-P3 showed a significant increase in nuclear localisation following NCZ treatment, which is indicative of a MT-inhibited mechanism (Fig. 2B). The nuclear import of MT-independent-type proteins such as GFP and GFP-fused NLSs (e.g. the SV40 large-T antigen NLS, ppUL44 human cytomegalovirus NLS) was unaffected by NCZ (data not shown) (see Moseley et al., 2007b). Thus, the MT-facilitated mechanism enhancing nuclear import of the short RPP fragment RPP_139-297 was switched to a MT-inhibited mechanism in P3.

Although NES1 is inactive in P3, it contains a second NES (NES2) in the C-terminal domain (CTD; residues 174-297) (Fig. 1) (Moseley et al., 2007a; Pasdeloup et al., 2005). We therefore examined the effect of the nuclear export inhibitor leptomycin B (LMB, which specifically affects transport mediated by the exportin CRM1) on MT-dependent nucleocytoplasmic localisation of P3. As previously reported, LMB treatment significantly increased nuclear localisation of GFP-P3 (Fig. 2B) (Moseley et al., 2007a). Significantly, this localisation was further increased by combined treatment with LMB and NCZ (Fig. 2B). Thus, it appears that P3 nuclear localisation is negatively regulated by the combined action of nuclear export and MT-inhibitory mechanisms.

Although the DLC-AS of RPP is able to confer a MT-facilitated phenotype on proteins (Moseley et al., 2007b), it does not appear to result in stable association with MTs in transfected cells, i.e. DLC-AS-containing proteins do not decorate MTs in a fashion detectable by CLSM (Moseley et al., 2007b), which is consistent with dynamic, transient association of the RPP-DLC-AS with MTs for MT-facilitated nuclear import. By contrast, a stable, MT-inhibited mode of interaction would be expected to be revealed by association with MTs when viewed by microscopy. Routine CLSM indicated that GFP-P3 does associate with cytoplasmic structures (Fig. 2A). To characterise these structures, we used high-resolution CLSM to analyse Vero cells transfected with GFP-P3 and detected a clear association with filamentous structures in the cytoplasm (Fig. 3A upper panel and not shown). We observed a similar association of transfected GFP-P3 with filamentous structures in HeLa, BSR, U-373-MG and Cos-7 cells (not shown). To examine the contribution of different domains of RPP to MT association, we expressed in Vero cells several forms of RPP fused to GFP: P1 (residues 1-297, which contains the functional NES1); RPP_139-297 (which contains the DLC-AS and the CTD); RPP_174-297 (the CTD alone, which contains the NLS and NES2); and the regions RPP_1-172 and RPP_54-172 (which contain the RPP self-association region, with or without, respectively, an intact NES1) (see Fig. 1). As can be seen in Fig. 3A (lower panel), and in contrast to GFP-P3, none of GFP-P1, GFP-RPP_139-297, GFP-RPP_174-297, GFP-RPP_1-172 or GFP-RPP_54-172 showed any detectable association with cytoplasmic filaments. Treatment of cells with the MT-stabilising drug taxol (TAX) resulted in a more extensive interaction of GFP-P3 with filaments, indicating that the structures represent MTs, but no cytoplasmic filament association of the other proteins was detected (Fig. 3A). Thus, it appears that none of the modules tested are sufficient alone to mediate stable MT-association of GFP, and that this mode of association requires several domains within P3. Intriguingly, it also appears that although the sequences required for stable MT association are found within P3 (residues 54-297), this property is inhibited by the presence of the additional residues 1-53 in P1.

As expected, GFP-P1 and RPP_1-172, which contain the functional NES1, were excluded from the nucleus, whereas GFP-P3, GFP-RPP_139-297, GFP-RPP_174-297, and GFP-RPP_54-172 were distributed between the nucleus and cytoplasm to different extents (Moseley et al., 2007a; Pasdeloup et al., 2005). Interestingly, although GFP-RPP_174-297 and GFP-RPP_139-297 showed moderate nuclear accumulation (Fn/c for GFP-RPP_139-297 of 2.12±0.05; mean ± s.e.m. for n>40) (Fig. 2B, Fig. 3A), GFP-RPP_54-172 accumulated to a much greater extent (Fn/c= 3.4±0.12; Fig. 3A and not shown), to a level comparable with that of GFP-P3 in the presence of LMB and NCZ (Fn/c=3.59±0.21; Fig. 2B). Thus, RPP_54-172 appears to possess nuclear-localising activity that might contribute to P3 nuclear accumulation.

To confirm that the filaments observed in P3-transfected cells were indeed MTs, we treated GFP-P3-expressing cells with or without TAX or NCZ before fixation and immunostaining for the MT component β-tubulin. This revealed a clear colocalisation of filamentous GFP-P3 with tubulin in the form of MTs, confirming that GFP-P3 associates with the MT cytoskeleton (Fig. 3B and supplementary material Fig. S1A). Treatment of cells with TAX resulted in more extensive association or bundling of the GFP-P3-MT structures whereas NCZ treatment clearly disassembled the structures, resulting in a diffuse appearance of GFP-P3 and tubulin (Fig. 3B). We also confirmed that P3 was associated with MTs in intact, live cells by transfecting cells to coexpress GFP-P3 with α-tubulin fused to mCherry (tubulin-mCherry) (Shaner et al., 2004) or mCherry-P3 with α-tubulin fused to GFP (tubulin-GFP), finding that filamentous P3 clearly colocalises with α-tubulin-containing filaments (supplementary material Fig. S1B,C).

To confirm the physical association of GFP-P3 with MTs, we purified MTs and MT-associated proteins from cell lysates as previously described (Roth et al., 2007). Cells were transfected to express GFP-P3 or, as negative controls, GFP or GFP fused to the NLS from SV40 large-T antigen (GFP-T-ag), neither of which associate with MTs (Roth et al., 2007). As shown in Fig. 3C, GFP-P3, GFP alone and GFP-T-ag were present in the soluble (S) fraction, but only P3 was present in the MT fraction (pellet, P). Thus, the biochemical data (Fig. 3C) indicate that P3 can physically associate with MTs isolated from lysed cells, and this supports the key observations from CLSM that GFP-P3 associates with MTs in the context of live, intact cells (Fig. 3A,B and supplementary material Fig. S1).

In vitro, RPP exists in an equilibrium between monomeric and oligomeric forms (Gigant et al., 2000). Since RPP_54-139 contains the RPP self-association domain (Mavrakis et al., 2004), RPP_139-297 would be predicted to exist as a monomer, whereas P3, which can interact stably with MTs (see above), can exist in both monomeric and oligomeric states. Thus, interaction of RPP with MTs might depend strongly on oligomerisation. To test this possibility, we replaced the region 54-139 of P3 with a heterologous dimerisation module, using the Ariad inducible homodimerisation kit (see Materials and Methods). The module, Fv1, is a variant of the FKBP protein that can be induced to dimerise by the addition of a cell-permeable dimerisation reagent (DR), AP20187. To monitor the subcellular localisation of Fv1-containing protein, the Fv1 was fused in-frame C-terminal to GFP in pEGFP-C1 with RPP fragments subsequently cloned in-frame C-terminal to the GFP-Fv1 open reading frame. As shown in Fig. 4A, no filamentous structures were detected in cells expressing GFP-Fv1 alone or GFP-Fv1 fused to various forms of RPP. However, following addition of DR to cells, a clear association of GFP-fusion protein with MTs was detected in cells expressing GFP-Fv1-RPP_139-297, which became obvious following TAX treatment, indicating that RPP_139-297 contains the MTAS sequence and is necessary and sufficient to mediate stable RPP-MT association, dependent on dimerisation. The specificity of this effect was demonstrated by the fact that neither GFP-Fv1 nor GFP-RPP_139-297 could be induced to associate with filamentous structures following treatment with DR with or without TAX (Fig. 4A). Treatment with TAX alone did not induce association of Fv1-containing GFP-fusion proteins with MTs (see supplementary material Fig. S2A).

We were also able to detect some MT association of GFP-Fv1-RPP_174-297 in DR-treated cells, which was more obvious in cells treated with both DR and TAX, but in both cases, MT association was considerably reduced compared with that observed for GFP-Fv1-RPP_139-297 in equivalently treated cells (Fig. 4A). This indicated that efficient dimerisation-dependent MT association of the region RPP_139-297 requires both the DLC-AS (RPP_139-172) and CTD (RPP_174-297). However, no clear association of DR-dimerised GFP-Fv1-RPP_139-172 with MTs was observed in cells treated with or without TAX (Fig. 4A). Thus RPP_139-172 is not sufficient to mediate MT association, but facilitates MT association conferred by RPP_174-297. To confirm that the filamentous structures with which dimerised GFP-Fv1-RPP_174-297 and GFP-Fv1-RPP_139-297 associate are MTs, transfected cells treated with DR and/or TAX were fixed and immunostained for tubulin revealing that the filament-associated GFP-fused proteins in cells treated with DR or DR and TAX were colocalised with MTs (Figs 4B and supplementary material Fig. S2B), whereas no filamentous MT association of GFP-Fv1-RPP_174-297 and GFP-Fv1-RPP_139-297 was observed in untreated cells or cells treated with TAX alone (supplementary material Fig. S2B).

Previously, we showed that MT facilitation of nuclear import of RPP_139-297 is dependent on the DLC-AS (within residues 139-172), and can be inhibited by mutation (D₁₄₃/Q₁₄₇-A) to prevent association with the dynein light chain LC8 (Moseley et al., 2007b; Poisson et al., 2001; Raux et al., 2000). Thus, in common with other MT-facilitated proteins (e.g. Rb and p53) (Giannakakou et al., 2000; Roth et al., 2007), MT-facilitated-nuclear import of RPP_139-297 appears to involve dynein components. MT-inhibitory-type interactions of proteins, however, do not appear to depend on dynein.

We therefore examined the MT association of P3 and dimerised RPP_139-297 harbouring the D₁₄₃/Q₁₄₇-A mutation by transfection and CLSM analysis. Both P3(D₁₄₃/Q₁₄₇-A) and dimerised Fv1-RPP_139-297(D₁₄₃/Q₁₄₇-A) retained the capacity to decorate the cytoplasmic MT filaments of transfected cells (Fig. 5A), implying that stable MT association of these proteins is independent of LC8 association via the DLC-AS. As expected, Fv1-RPP_139-297(D₁₄₃/Q₁₄₇-A) does not associate with filamentous structures in cells not treated with DR (supplementary material Fig. S3A).

Cargo interactions with dynein and MTs frequently involve the dynein-associated dynactin complex (Dohner et al., 2005). The complex can be disassembled by overexpression of the dynactin component dynamitin, which is used as a standard method to examine dependency of processes, including MT-facilitated nuclear import, on dynein and dynactin function (Giannakakou et al., 2000; Roth et al., 2007). We therefore tested the capacity of P3 and dimerised RPP_139-297 to associate stably with MTs in cells overexpressing dynamitin fused to the monomeric fluorescent mCherry protein, which has previously been used to confirm the role of dynein and dynactin in nuclear import of p53 and Rb (Roth et al., 2007). As shown in Fig. 5B, overexpression of dynamitin-mCherry did not prevent stable MT association of GFP-P3 or dimerised GFP-Fv1-RPP_139-297. In cells expressing dynamitin-mCherry that were not treated with DR, GFP-Fv1-RPP_139-297 showed no association with filaments, as expected (supplementary material Fig. S3B). As a positive control for dynamitin overexpression, we confirmed our previous observations (Roth et al., 2007) that the nuclear localisation of the MT-facilitated-type protein GFP-Rb is significantly reduced in cells coexpressing dynamitin-mCherry compared with cells not expressing dynamitin-mCherry. Expression of dynamitin-mCherry significantly (P>0.0006) reduced nuclear localisation of GFP-Rb (Fn/c=12.9±3.2 and 23.8±2.9, n>38 for GFP-Rb in cells expressing or not expressing dynamitin-mCherry, respectively). Thus, it appears that stable P3-MT association does not involve the commonly used dynein-dynactin complex and, in contrast to the MT-facilitated-type interaction of RPP_139-297, is not dependent on association with dynein LC8 so that the MT-facilitated and MT-inhibited modes of RPP-MT interaction appear to be physically distinct.

RPP is multifunctional, with roles as the rabies virus polymerase (L-protein) cofactor and as the IFN antagonist (Brzozka et al., 2005; Brzozka et al., 2006; Chelbi-Alix et al., 2006; Randall and Goodbourn, 2008; Shimizu et al., 2006; Vidy et al., 2005). P3 appears to be specialised for rabies-virus-mediated IFN antagonism, because it lacks the N-terminal interaction site for the L-protein (Chelbi-Alix et al., 2006) (Fig. 1), and is implicated in rabies virus effects both on STAT signalling and on the IFN-induced promyelocytic leukaemia nuclear bodies (PML-NBs) (Blondel et al., 2002; Chelbi-Alix et al., 2006; Vidy et al., 2007). RPP has been reported to interact with STATs and to inhibit their nuclear localisation (Brzozka et al., 2006; Vidy et al., 2005), but the mechanisms underlying this are not fully understood. We hypothesised that P3, via its stable interaction with MTs, might be able to draw STAT1 to the MT network causing its sequestration in the cytoplasm and thereby effecting an MT-inhibitory-type mechanism. We therefore examined whether P3 and STAT1 form complexes at MTs. As can be seen in Fig. 6A (upper panel), STAT1-dsRed showed a diffuse localisation in Vero cells, consistent with the known distribution of non-activated STAT1. We observed no localisation of STAT1 with cytoskeletal filaments, even if the MTs were stabilised by TAX treatment (Fig. 6A, upper panel). However, when coexpressed with GFP-P3, STAT1-dsRed became associated with MT-like filaments, where it colocalised with GFP-P3. The filaments were rendered more easily detectable following TAX treatment and were completely disrupted following NCZ treatment (Fig. 6A, lower panel).

We next tested the capacity of RPP_139-297 to cause association of STAT1 with MTs, dependent on dimerisation. As can be seen in Fig. 6B and supplementary material Fig. S3C, no localisation of GFP or dsRed with MTs was observed in cells expressing non-dimerised GFP-Fv1-RPP_139-297 with or without treatment with TAX or NCZ. However, following treatment with DR, both proteins were observed to associate with MT-like structures, which were eliminated by NCZ treatment and enhanced by TAX treatment (Fig. 6B). Thus, the association of P3 with MTs appears to be able to cause translocation of STAT1 from a diffuse localisation in the cytosol to become strongly associated with MTs.

The capacity of P3 to cause stable association of STAT1 with MTs indicated that P3 might use a novel MT-dependent mechanism to inhibit nuclear import of STAT1 and thereby disable the innate immune response; i.e. P3 might be able to switch STAT1 nuclear import from a MT-independent to a MT-inhibited-type mechanism. We therefore examined quantitatively the nuclear accumulation of STAT1-dsRed in cells expressing or not expressing GFP-P3, with or without treatment with IFNα or IFNγ and with or without a disrupted MT network (Fig. 7A,B).

Quantitative analysis confirmed that, in the absence of GFP-P3 expression, STAT1-dsRed was diffusely distributed in Vero cells (Fig. 7A,B; Fn/c ∼1.5). This was unaffected by NCZ treatment, indicating that the distribution of latent STAT1 does not involve any significant MT-dependent component (Fig. 7A,B). Treatment of cells with IFNα (Fig. 7A,B) or IFNγ (not shown) resulted in a significant translocation of STAT1-dsRed to the nucleus (Fig. 7A,B; Fn/c ∼5.5 for IFNα), which was unaffected by NCZ treatment (Fig. 7B), indicating a MT-independent mechanism of IFN-induced STAT1 nuclear import, which is consistent with previous findings (Lillemeier et al., 2001). In the absence of P3 expression, the cytoplasmic fraction of STAT1-dsRed showed no association with cytoplasmic filaments whether cells were treated or not with IFN (Fig. 7A).

Coexpression of GFP-P3 with STAT1-dsRed resulted in nuclear exclusion of STAT1 (Fig. 7A,B; Fn/c ∼0.5) with no significant increase in nuclear accumulation upon treatment with NCZ or IFN alone (Fig. 7B). Thus, in cells expressing P3, MT association is not solely responsible for the nuclear exclusion of STAT1-dsRed, and treatment with IFNα (Fig. 7A,B) or IFNγ (not shown) alone is not sufficient to effect nuclear accumulation of STAT1-dsRed. However, when cells were treated with both IFNα and NCZ (Fig. 7A,B) or IFNγ and NCZ (not shown), a significant enhancement of nuclear accumulation was observed (Fig. 7A,B; Fn/c ∼1.5 for IFNα) within 2 hours of IFN treatment, which is consistent with a significant relocalisation of STAT1-dsRed protein to the nucleus in IFN-activated cells in the absence of an intact MT network. Expression of GFP without the P3 fusion partner did not block nuclear accumulation of STAT1-dsRed (data not shown). Thus, P3 expression inhibited IFN-activated nuclear accumulation of STAT1-dsRed and this effect was dependent on MTs.

We also examined the effect of IFN and NCZ on the subcellular localisation of endogenous STAT1 in cells expressing or not expressing GFP-P3. Vero cells were transfected and treated as above before fixation and immunostaining using anti-STAT1 for CLSM and image analysis (Fig. 7C,D). Nuclear localisation of endogenous STAT1 in nontransfected cells was significantly increased in IFN-activated cells and was unaffected by NCZ (Fig. 7C,D). This is consistent with the observations made for STAT1-dsRed and confirmed that IFN-dependent STAT1 nuclear accumulation is MT independent.

In contrast to STAT1-dsRed in live intact cells, the nuclear localisation of endogenous STAT1 in fixed untreated cells was not significantly affected by GFP-P3 expression (Fig. 7C,D). This is probably due to the fact that STAT1 is overexpressed in the case of STAT1-dsRed. However, IFN-induced nuclear accumulation of endogenous STAT1 was inhibited in cells expressing GFP-P3, unless the cells were treated with NCZ (Fig. 7C,D); this is consistent with the effects observed for STAT1-dsRed expressed in live cells. Thus, it appears that GFP-P3 can switch endogenous STAT1 and STAT1-dsRed from MT-independent to MT-inhibited nuclear import as a mechanism to inhibit STAT1 nuclear accumulation in response to IFN.

Following IFNα binding to the IFN receptor, STAT1 becomes phosphorylated at Y701 by JAK1 tyrosine kinase, which precedes STAT1 accumulation in the nucleus (Randall and Goodbourn, 2008). Thus, the above data indicate that GFP-P3 impacts on the nuclear localisation of phosphorylated STAT1. To confirm this effect we examined the nuclear localisation of Y701-phosphorylated STAT1 (STAT1-P) specifically by immunostaining fixed cells with a STAT1-P-specific antibody and analysing its nuclear localisation. In cells not expressing GFP-P3, STAT1-P nuclear localisation was independent of MT disruption by NCZ (supplementary material Fig. S4), but in GFP-P3-positive cells, nuclear localisation of STAT1-P was significantly enhanced by MT disruption, indicating that STAT1-P nuclear import is inhibited by P3, with dependence on MTs. This is consistent with results observed for total endogenous STAT1 or transfected STAT1-dsRed.

To establish the functional impact of the MT-dependent effect of P3 on STAT1 nuclear import, we examined the capacity of IFNα to stimulate STAT1-dependent activation of transcription using a luciferase reporter gene assay with luciferase expression under the control of an ISRE promoter (Vidy et al., 2005; Vidy et al., 2007). Cells were transfected with pISREluc and pRL-TK with or without cotransfection with GFP-P3 or GFP lacking a fusion partner, and the expression of luciferase was assessed following IFNα treatment of cells with or without NCZ pretreatment.

In cells expressing pISREluc and pRL-TK only, IFNα treatment significantly induced expression of luciferase compared with untreated cells (∼tenfold increase, Fig. 7E). As expected from the lack of an effect of NCZ on STAT1 nuclear localisation (Fig. 7A-D), NCZ treatment did not significantly affect expression of luciferase in cells treated without or with IFNα (Fig. 7E). Identical results were obtained in cells cotransfected with pIRESluc, pRL-TK and GFP (data not shown).

In GFP-P3-expressing cells, we observed a significant reduction (to about 20%) in luciferase induction in response to IFNα treatment compared with cells not expressing GFP-P3 (Fig. 7E). In contrast to cells transfected with pISREluc and pRL-TK vector only, NCZ treatment of cells cotransfected with GFP-P3 produced a significant increase (>twofold) in induction of luciferase by IFNα, which is consistent with a substantial role for a MT-inhibitory-type mechanism in P3-mediated negative regulation of IFNα signalling.

To examine this effect in the context of rabies virus infection, we infected pISREluc/pRL-TK transfected cells with CVS strain rabies virus and tested the responsiveness to IFNα, with or without NCZ treatment, as above. As can be seen in Fig. 7E, IFNα-stimulated luciferase expression was significantly reduced (to about 40%) in infected cells compared with uninfected cells. In similar fashion to GFP-P3-transfected cells, this decrease was dependent on the MT network, because NCZ treatment significantly enhanced (∼threefold) luciferase expression.

As our data implicated dimerisation in MT association of P3 and P3-associated STAT1 (see above), we next tested the effect of dimerisation of RPP on IFN-induced transcriptional activation. Cells were transfected with pISREluc/pRL-TK and vectors for expression of GFP-Fv1-RPP_139-297 or GFP-Fv1, before treatment with DR and/or NCZ followed by IFNα treatment. IFN treatment of cells expressing the luciferase vectors and GFP-Fv1 or GFP-Fv1-RPP_139-297 resulted in a significant increase in luciferase expression (Fig. 8A). The level of induction of luciferase in GFP-Fv1-transfected cells was equivalent to that in cells expressing pISREluc/pRL-TK only (not shown) confirming that GFP-Fv1 did not affect IFN-dependent signalling. DR treatment had no effect on IFN induction of luciferase expression in cells transfected with pISREluc/pRL-TK vectors only (not shown) or cotransfected with GFP-Fv1 (Fig. 8A), but caused a significant inhibition of luciferase induction in cells expressing GFP-Fv1-RPP_139-297, implying that dimerisation of the C-terminal region of RPP activates an IFN-antagonist mechanism.

To examine the role of MTs in this dimerisation-dependent mechanism, we treated pISREluc/pRL-TK/GFP-Fv1-RPP_139-297-transfected cells with DR and/or NCZ before addition of IFNα. Treatment of DR-treated cells with NCZ entirely negated the inhibitory effect of DR on luciferase induction, such that IFN-induced luciferase expression in cells treated with DR and NCZ was not different to that in untreated cells or cells treated with NCZ alone (Fig. 8B). This indicates that the dimerisation-dependent inhibition of IFN signalling by RPP is dependent on the integrity of the MT cytoskeleton.

Thus, STAT1 nuclear import and associated transcriptional activation of ISRE-dependent genes would appear to be independent of MTs unless the cells coexpress P3 protein. In the latter case, STAT1 appears to become tethered to MTs by P3, switching STAT1 nuclear import from a MT-independent to a MT-inhibited-type mechanism, with concomitant effects on cellular IFN signalling. The results strongly support the hypothesis that this mechanism is a crucial component in viral antagonism of IFN signalling.

This study establishes a role for MTs in viral antagonism of IFN responses for the first time. Specifically, we have shown that the rabies virus IFN antagonist P3 is able to cause interaction of STAT1 with MTs in order to prevent STAT1 nuclear import, thereby disabling the IFN-signalling pathway that is vital to the innate antiviral immune response.

The mechanisms evolved by viruses to evade the innate immune response are extraordinarily varied, with IFN antagonists frequently able to perform a myriad of functions, both in the basic viral life cycle and in evasion of the antiviral response (Randall and Goodbourn, 2008). This might be particularly true of viruses with limited coding capacity such as rabies virus, which, in common with a number of other RNA viruses [e.g. Nipah, measles and Sendai (Randall and Goodbourn, 2008; Shaw et al., 2004)], uses products of the P gene for IFN-antagonist functions. RPP displays many archetypal features of IFN antagonists: it is multifunctional (with housekeeping roles in viral replication as well as roles in IFN antagonism) (Chelbi-Alix et al., 2006; Randall and Goodbourn, 2008) and expressed in several forms (see Fig. 1) (Chenik et al., 1995; Pasdeloup et al., 2005). Significantly, owing to the presence or absence of NESs and NLSs, the different gene products are able to accumulate differentially in the nucleus and the cytoplasm (Moseley et al., 2007a; Pasdeloup et al., 2005) and can thereby affect IFN signalling at several stages by affecting the functions of signalling and effector proteins of the IFN system located in different compartments. Specifically, RPP has been reported to inhibit phosphorylation of cytoplasmic IRF3, to interact with nuclear and cytoplasmic STAT1, which affects subcellular localisation and DNA binding, and to affect nuclear PML-NBs (Blondel et al., 2002; Brzozka et al., 2005; Brzozka et al., 2006; Chelbi-Alix et al., 2006; Vidy et al., 2005; Vidy et al., 2007). Thus, RPP counters the IFN system by the combined action of several distinct mechanisms. Consistent with this, our data (Fig. 7E) indicate that IFN antagonism by P3 involves a significant MT-dependent component, as well as contribution by other mechanism(s) that are not disabled by MT disruption, to produce full inhibition of IFN signalling. Interestingly, our data also indicate that MT-dependent IFN antagonism by RPP is dependent on its oligomeric state, which correlates with a role for oligomerisation in regulating RPP-MT association (see below), and indicates that the mechanisms of IFN antagonism mediated by RPP may be dynamically regulated.

It is intriguing that although RNA viruses such as rabies virus ostensibly have no need to interact with the nucleus, nuclear trafficking has been reported as a feature of numerous RNA-virus-encoded proteins, including known IFN antagonists (e.g. Billecocq et al., 2004; Breakwell et al., 2007; Chelbi-Alix et al., 2006; Ghildyal et al., 2009; Montgomery and Johnston, 2007; Nishie et al., 2007; Rawlinson et al., 2009; Shaw et al., 2005; Shaw et al., 2004) and the capacity of viruses to regulate nuclear import and/or export functions of cellular proteins is a key mechanism in IFN antagonism (Randall and Goodbourn, 2008). The latter includes various strategies to prevent STAT nuclear accumulation and activation of transcription, including STAT degradation, inhibition of dimerisation, alteration or inhibition of phosphorylation, and sequestration into high molecular weight complexes (Randall and Goodbourn, 2008). The demonstration in the present report that IFN antagonists can exploit cellular MT cytoskeleton-dependent mechanisms to inhibit STAT1 nuclear accumulation highlights the diversity and intricacy of viral strategies to surmount the innate immune response. Given the potency of this mechanism to prevent IFN-dependent signalling (Fig. 7E), it seems likely that similar strategies will be used by other viruses, and might be implicated in previously reported cases of IFN-antagonist-mediated inhibition of STAT1 signalling.

In this report, we have also demonstrated the importance of dimerisation in MT association of RPP. It has been reported that certain cellular proteins associate with MTs as dimers (Short et al., 2002), indicating that oligomerisation might be involved in MTAS function. That stable MT association of RPP is dependent on oligomerisation is indicated by the fact that it is prevented by deletion of the endogenous self-association region of RPP (producing the truncated form, RPP_139-297), but can be reconstituted by the replacement of the self-association region with a heterologous dimerisation domain. Intriguingly, we observed previously that RPP_139-297 undergoes MT-facilitated nuclear import, which is dependent on the DLC-AS (Moseley et al., 2007b). This indicates that the oligomeric state of the protein might regulate the nature of the interaction of RPP with MTs, providing a dynamic mechanism to modulate RPP nucleocytoplasmic trafficking. This is the first demonstration that a protein can harbour the capacity to undergo both MT-inhibited- and MT-facilitated nuclear import, indicating that the MT network has a more dynamic and important role in nucleocytoplasmic trafficking than previously thought. RPP also contains at least two NESs, an NLS in the globular CTD structure (Mavrakis et al., 2004; Moseley et al., 2007a; Pasdeloup et al., 2005), and, as shown here, a second sequence within the region 54-172 might effect nuclear accumulation of RPP (Fig. 3A). There are multiple phosphorylation sites in the RPP sequence (Fig. 1), and protein kinase C has been shown to regulate NES/NLS-mediated transport of the CTD (Moseley et al., 2007a), and might affect dimerisation or MT association, because several phosphorylation sites are found within RPP_54-172 and in the CTD (Fig. 1). The high conservation of these trafficking and regulatory sequences among the P-proteins of rabies-virus-related lyssaviruses (Mavrakis et al., 2004) suggests that subcellular localisation of RPP is an essential and highly regulated aspect of viral pathogenicity.

In summary, we have identified a novel mechanism of IFN antagonism by human pathogenic virus, whereby a viral protein can alter the relationship of a transcription factor with the cytoskeleton and thereby affect its nuclear import and activation of transcription. Future research in our laboratory will focus on characterising the mechanism(s) regulating switching between the dual modes of RPP association with MTs and its potential as a target for antiviral drugs.

RPP used in this study is derived from CVS-strain rabies virus. Constructs for the expression in mammalian cells of GFP-P1 (RPP_1-297), GFP-P3 (RPP_54-297), GFP-RPP_139-172, GFP-RPP_139-297, GFP-RPP_174-297, GFP-RPP_1-172 and GFP-RPP_54-172 were generated using PCR from the P1 DNA template and cloned in frame C-terminal to the GFP sequence of pEGFP-C1 or the mCherry sequence of pmCherry-C1 (Shaner et al., 2004) using a BamHI-BglII strategy (Moseley et al., 2007a; Moseley et al., 2007b). P3 (D₁₄₃/Q₁₄₇-A) was produced by overlap PCR mutagenesis (Moseley et al., 2007a; Moseley et al., 2007b).

To produce GFP-Fv1-fused proteins, the coding sequence for Fv1 was PCR amplified from the pC4Fv1E vector (Ariad Pharmaceuticals, Cambridge, MA) and cloned into pEGFP-C1 in-frame C-terminal to the GFP sequence via the BamHI-BglII cloning strategy to produce the pEGFP-Fv1 vector. Sequential BamHI-BglII cloning (Moseley et al., 2007a; Moseley et al., 2007b) was performed to insert the coding sequences for RPP_139-172, RPP_139-297, RPP_139-297(D₁₄₃/Q₁₄₇-A) and RPP_174-297 into the pEGFP-Fv1 vector in-frame C-terminal to the GFP-Fv1 sequence.

The plasmid encoding STAT1-dsRed2 was a gift from Yi-Ling Lin, Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan, Republic of China (Lin et al., 2006). The plasmid for expression of dynamitin-mCherry has been described previously (Roth et al., 2007). The plasmid for expression of α-tubulin mCherry was from the Tsien laboratory, Howard Hughes Medical Institute and Department of Pharmacology, University of California at San Diego, La Jolla, CA (Shaner et al., 2004).

Cos7, HeLa, Vero, BSR and U-373-MG cells were cultured at 37°C, 5% CO₂ in DMEM medium with 10% FCS. For CLSM, cells were grown on coverslips to 80% confluency and transfected using Lipofectamine 2000 by the manufacturer's protocol, before live-cell CLSM as previously described (Alvisi et al., 2005; Moseley et al., 2007a; Moseley et al., 2007b; Roth et al., 2007). To disrupt or stabilise the MT network, cells were treated with NCZ (Sigma) for 4 hours at 5 μg/ml or TAX (Sigma) for 4 hours at 1 μg/ml (Moseley et al., 2007b; Roth et al., 2007). To activate IFN signalling for CLSM analysis, cells were treated with 2000 U/ml recombinant human IFNα (PML InterferonSource) or 3 ng/ml recombinant human IFNγ (Invitrogen) (added 2 hours before analysis). For inhibition of nuclear export mediated by CRM1, 2.8 ng/ml LMB (a gift from Minoru Yoshida, University of Tokyo, Tokyo, Japan) was added to cells for 4 hours before analysis.

For indirect immunofluorescence of tubulin, cells were fixed with 4% paraformaldehyde, before permeabilisation and labelling with anti-β-tubulin antibody (Cytoskeleton) followed by Alexa-Fluor-568-coupled secondary antibody (Invitrogen) as previously described (Moseley et al., 2007b; Roth et al., 2007). Alternatively, cells were fixed with methanol (10 minutes at –20°C) before labelling (Zhai et al., 1996). For indirect immunofluorescence of STAT1, cells were washed with PBS before fixation with 3.7% formaldehyde in PBS (10 minutes at room temperature) followed by 90% methanol (5 minutes at room temperature) before blocking (1% BSA, 2 hours, room temperature) and staining with anti-STAT1 (BD Biosciences, 610185) or anti-STAT1-P (Santa Cruz Biotechnology, sc-7988) followed by Alexa-Fluor-568-coupled secondary antibody (Invitrogen).

Routine CLSM for single-colour fluorophore live-cell analysis was performed using a Bio-Rad MRC-600 CLSM with a 40× water-immersion objective (NA 0.8) and heated stage (Alvisi et al., 2005; Lam et al., 2002; Moseley et al., 2007a; Moseley et al., 2007b; Poon et al., 2005; Roth et al., 2007); high-resolution CLSM was performed using an Olympus Fluoview 1000 with 60× oil-immersion lens (NA 1.4), and multicolour fluorophore imaging used the Olympus Fluoview 1000 (60× oil-immersion lens) or Leica SP5 (63× glycerol-immersion lens), with heated stage for live-cell imaging (Moseley et al., 2007b; Roth et al., 2007). In multicolour imaging, sequential scanning was performed to prevent cross-talk.

Analysis of digitised confocal files (single sections) was performed using ImageJ 1.62 public domain software (NIH) to calculate the ratio of nuclear to cytoplasmic fluorescence corrected for background (Fn/c) for individual cells; mean Fn/c values ± s.e.m. were calculated for samples of cells (n>30), and Graph Pad Instat software was used for statistical analysis, as previously described (Moseley et al., 2007a; Moseley et al., 2007b; Roth et al., 2007).

To induce dimerisation of protein in live cells, we used the Ariad inducible homodimerisation kit (Ariad Pharmaceuticals, Cambridge, MA). Proteins containing the Fv1 inducible homodimerisation domain were expressed in Vero cells for 18-24 hours before addition of the cell-permeable dimerisation reagent (DR) AP20187 (Ariad) at a concentration of 100 nM for 1 hour before CLSM analysis as above.

MT and associated proteins were extracted from Cos-7 cells using the MT/Tubulin In Vivo Assay Kit (BK038, Cytoskeleton), as described (Davis et al., 2005). Cos7 cells transfected to express GFP-P3 and GFP-T-ag-NLS (residues 110-135 of the SV40 large-T antigen corresponding to the NLS) or GFP as negative controls, were lysed using MT stabilising buffer containing 100 μM GTP, 1 mM ATP, and 10 μl per ml of buffer of protease inhibitor cocktail for 10 minutes at 37°C. The lysates were subjected to ultracentrifugation (55,000 g for 30 minutes at 37°C) to separate the supernatant (containing soluble proteins and non-polymerised tubulin) from the MT fraction containing MT-associated proteins. Before resuspension, the pellet was washed three times. The fractions were analysed by western blotting using anti-GFP (1:1000, Roche) or anti-β-tubulin (1:500, Cytoskeleton) followed by HRP-linked secondary antibodies, before analysis using enhanced chemiluminescence (Perkin Elmer).

For luciferase assays of IFN signalling, U-373-MG cells in 12-well plates were transfected with 0.75 μg pRL-TK and 2.5 μg pISREluc, with or without 2.5 μg plasmid encoding GFP, GFP-P3, GFP-Fv1 or GFP-Fv1-RPP_139-297. At 48 hours after transfection, cells were treated with or without 2000 U/ml IFN-α for 6 hours before harvesting and measurement of Firefly and Renilla luciferase activity according to the manufacturer's protocol (Dual-Luciferase Reporter Assay System; Promega). In some assays, cells were pretreated with NCZ (5 μg/ml added 1 hour before addition of IFN) and/or DR (100 nM for 1 hour before addition of IFN). Relative expression levels were calculated by dividing the Firefly luciferase values by those of the Renilla luciferase. In some cases, cells transfected with pRL-TK and pISREluc were infected 24 hours after transfection with rabies virus (CVS strain) at a multiplicity of infection (MOI) of 5 PFU/cell for 24 hours before treatment with IFN or NCZ as above.