A Rab11- and Microtubule-Dependent Mechanism for Cytoplasmic Transport of Influenza A Virus Viral RNA

Influenza A virus is a ubiquitous pathogen of homeothermic vertebrates with the capacity to cause seasonal epidemics and occasional, potentially devastating outbreaks in humans (55). The virus has a single-stranded RNA genome split into eight segments that encode between 10 and 12 polypeptides depending on the virus strain (61). The viral RNA (vRNA) segments are encapsidated into ribonucleoprotein (RNP) particles by the viral RNA-dependent RNA polymerase and a single-strand RNA-binding protein, NP (40). Once delivered to the cell nucleus at the start of infection, they act as independent units for the purposes of transcription and replication of the virus genome. Virus assembly subsequently takes place at the apical plasma membrane of infected cells; here, the three viral membrane proteins (HA, NA, and M2) join with the M1 matrix protein, small amounts of the NEP/NS2 protein, and the RNPs to form new enveloped virus particles through a process of budding. This assembly phase of the virus life cycle is not especially well understood but is thought to result from a cascade of protein-protein interactions between the viral components, as well as probable RNA-RNA interactions between specific vRNAs that ensure a complete genome is incorporated (27, 36). Unlike many other enveloped viruses, the formation of a mature virus particle does not depend on the cellular ESCRT pathway (11, 15, 58) but requires the function of the small GTPase Rab11 (10) and the viral M2 protein (48).

A further aspect of the viral life cycle that is imperfectly understood concerns the intracellular trafficking of the viral genome. The nuclear export of newly synthesized RNPs involves sequential interactions between the RNPs, M1, and NEP and the cellular nuclear export factor Crm1 (8). However, the mechanisms that then transport RNPs across the cytoplasm to the sites of budding at the apical plasma membrane are poorly defined. Individual RNPs range in size from ∼2.5 to 6 MDa (depending on the length of the vRNA segment), and thus simple diffusion is unlikely to operate over long distances. On general principles, as well as precedents from other viruses (41), the involvement of one or more arms of the cytoskeleton seems likely. Although NP has been defined as an actin-binding protein (19) and shown to partially colocalize with actin at the plasma membrane (5, 53), direct evidence for a role in RNP transport is lacking. Instead, an intact cortical actin web seems to be required to support filamentous viral budding (47, 53). Cytoplasmic RNPs have been suggested to colocalize with microtubules (34), but although (like actin) tubulin has been found to copurify with RNPs and be present in virions (31, 52), direct evidence for a role in trafficking is sparse. In general, pharmacologic interference with the cytoskeleton (after virus entry) has relatively slight effects on influenza A virus replication outside certain drugs that also interfere with viral protein synthesis (4, 22, 34, 45, 47, 50, 53). Thus, this aspect of the virus life cycle has many unresolved questions.

Here, we provide positive evidence that the transport of the viral genome from the nucleus to the cell periphery does indeed involve the microtubule network. Furthermore, we provide a new model for efficient RNP transport that involves the piggybacking of RNPs onto the vesicular transport network at the pericentriolar RE via an interaction between PB2 and Rab11.

The influenza virus genome is not thought to exist normally as naked RNA but rather as encapsidated RNP particles containing one copy of a trimeric RNA polymerase and multiple copies of the nucleoprotein (40). Accordingly, the majority of studies that have investigated the localization of the viral genome in cells by microscopy have utilized indirect immunofluorescence against NP as a proxy marker for vRNA. However, our recent development of FISH protocols for the direct detection of influenza virus RNAs (2, 26, 42) allow more direct visualization of genome trafficking.

We carried out a time course experiment in which MDCK cells were fixed and stained at various times postinfection (p.i.) for vRNA from segments 1 and 4. Early in infection, obvious vRNA staining of the nucleoplasm but not the nucleoli was apparent (Fig. 1A, 5 h). At later times, staining became increasingly biased toward the cytoplasm, such that by 12 h p.i. many nuclei appeared empty of vRNA (Fig. 1A). This time-dependent shift from predominantly nuclear to predominantly cytoplasmic localization is broadly consistent with that expected from the immunofluorescent staining of NP (9, 21, 30). However, at intermediate times postinfection, many cells also showed bright perinuclear foci of vRNA (Fig. 1A, arrows). Similar perinuclear accumulations of vRNA were seen in other cell types, including A549 cells (see later) and 293T cells (Fig. 1B). The perinuclear accumulation of NP has been noted previously (10, 34), and consistently with this information, the structures seen here contained both vRNA and NP, although the FISH staining of the viral genome generally was proportionally more intense (Fig. 1B). The examination of Z-stack animations of serial planes of focus taken throughout the body of the cells showed that these perinuclear foci contained more than one segment and tended to be around the midline of the cell or slightly higher, but they were below the level of the plasma membrane (see Movie S1 in the supplemental material). When multiple cells were examined from replicate experiments and scored according to whether perinuclear vRNA staining was present, it was apparent that these structures appeared soon after the onset of vRNA nuclear export and reached a maximum around 7 h p.i., and although they sometimes persisted throughout infection, they generally became less prominent later in infection (Fig. 1C).

To investigate this perinuclear accumulation further, we first asked whether it was a feature of influenza viruses in general or specific to the PR8 strain. We also tested whether the perinuclear body represented a specific cellular location by staining the microtubule organizing center (MTOC) with anti-γ-tubulin. As before, 293T cells infected with PR8 virus contained a prominent perinuclear focus of vRNA staining at 6 h p.i. that, under high magnification, was seen to consist of punctate dots surrounding the MTOC (Fig. 2). This was a common occurrence, as 45 of 51 infected cells examined from two independent experiments displayed this pattern of localization. Similar structures that in all cases surrounded the γ-tubulin staining also were seen in cells infected with the human H3N2 Udorn strain as well as with a low-pathogenicity Duck/Eng H4N6 avian virus and the H3N8 equine influenza virus strain Eq/Nkt (Fig. 2). The same outcome was obtained when the experiment was repeated with MDCK cells (data not shown).

The consistent localization of vRNA around the MTOC suggested its interaction with one or more of the similarly localized cellular organelles; this was investigated by double staining for components of cellular vesicular transport pathways. Little colocalization (Pearson correlation coefficient of 0.068 ± 0.008, from four fields of view) was seen between perinuclear vRNA and GM130, a marker for the cis-Golgi network (Fig. 3A). A similar result (Pearson coefficient of 0.179 ± 0.06, n = 3) was obtained when TGN46 was used as a marker for the trans-Golgi compartment, and although both vRNA and the cellular protein distributed as a pattern of perinuclear spots, the examination of high-magnification images showed that these patterns were distinct and largely separate (Fig. 3A, inset). A similar lack of colocalization was obtained when EEA1, a marker for early endosomes, or calnexin, a marker for the endoplasmic reticulum, were compared to vRNA (Pearson coefficients of 0.037 ± 0.03, n = 3, and 0.087 ± 0.05, n = 3, respectively; image data not shown). We have shown previously that cytoplasmic NP colocalizes with Rab11, including at a perinuclear structure (10). Unfortunately, the epitopes seen by our Rab11 antisera did not survive the FISH process (data not shown), making it impossible to test whether vRNA also colocalized with the endogenous protein. However, when a transfected GFP-tagged constitutively active Rab11 (GFP-Rab11 CA) (16) was used as a marker for the RE, a high degree of colocalization was seen with the perinuclear foci of vRNA (Fig. 3B). The Pearson coefficient for vRNA (CA Rab11; 0.541 ± 0.05, n = 4) was significantly higher (P < 0.0001) than the overlap seen for TGN46 or the other cellular markers. This degree of colocalization also was maintained when more dispersed cytoplasmic foci of vRNA were examined at a later time point (Fig. 3B, 12 h).

The colocalization of vRNA with Rab11-positive structures near the MTOC soon after the onset of genome replication and nuclear export suggested that cytoplasmic vRNA trafficking involves the cellular vesicular transport pathway and therefore the microtubule network. Accordingly, we tested the effects of the pharmacological disruption of these processes on vRNA localization. In untreated MDCK cells, the viral genome was predominantly cytoplasmic late in infection, concentrating at the apical plasma membrane, while NP showed a similar distribution but with additional bright foci in the nuclei (Fig. 4A). Treatment with the microtubule-depolymerizing agent nocodazole led to the loss of the apical polarization of vRNA and NP, and instead optical sections through the midline of the cells showed their pronounced accumulation at lateral membranes. A similar outcome of aberrant vRNA localization to lateral membranes was seen with demecolcine, another drug that destablizes microtubules, as well as when cells were treated with paclitaxel, a microtubule-stabilizing agent, but instead of localizing to lateral membranes it formed prominent aggregates toward the cell periphery. The disruption of vesicular trafficking with the ionophore monensin also altered vRNA trafficking, leading to a persistence of the prominent perinuclear foci of vRNA staining normally seen earlier in infection, as well as a degree of NP nuclear retention. We also examined the effect of the drugs on virus replication by titrating released virus from the cell supernatants. This showed modest but consistent decreases of around 4-fold after drug treatment (Fig. 4B). The Western blot analysis of selected virus polypeptides showed no major effect on viral protein accumulation (Fig. 4C). Overall, these data are consistent with cytoplasmic vRNA trafficking involving the microtubules and the cellular vesicular transport pathway.

To further investigate the hypothesis that RNP cytoplasmic trafficking involves the microtubule network, we established a system in which the movement of RNPs could be investigated in living cells using GFP-tagged NP. Because of the difficulty of introducing a large protein tag such as GFP into the viral genome, we turned to a plasmid-based minireplicon system in which the three polymerase proteins (3P), NP, and synthetic vRNAs are transcribed from cDNA (35, 38). Although our GFP-NP construct can functionally replace WT NP in the minireplicon setting (29), RNPs were reconstituted by transfecting a mixture of tagged and untagged NP to reduce the density of the GFP tag. Also, since the nuclear export of the viral genome requires both M1 and NEP proteins (8), we supplied segments 7 and 8 as vRNA templates, thus ensuring the expression of these proteins. To validate this system, we first tested for the successful reconstitution of RNPs by biochemical means. When lysates from transfected cells were incubated with GFP-affinity beads, PB2 coprecipitated with GFP-NP but not GFP (Fig. 5A, compare lanes 4 and 6), suggesting RNP formation. Confirming this, vRNA also was detectable in the bound fraction of lysates from cells containing 3P and GFP-NP, but not 3P and GFP or 2P (lacking PB2) and GFP-NP (Fig. 5B). We next examined the distribution of GFP-NP and vRNA in fixed cells. In control transfections lacking the PB2 subunit and thus a functional polymerase complex, the majority of GFP-NP was nuclear and relatively little signal was seen for segment 7 vRNA, with only a few nuclear foci of staining evident that potentially resulted from plasmid rather than viral transcription (Fig. 5C, image i). In contrast, when the full 3P-NP complex was present, bright nuclear and cytoplasmic vRNA staining was evident, with the latter often concentrating in a perinuclear structure as well as smaller, more dispersed foci throughout the cytoplasm (Fig. 5C, images ii to vi). When GFP-NP was included with the 3P-NP complex, the strong colocalization of the vRNA and GFP signal was seen both at the perinuclear region (image iii) and for the dispersed cytoplasmic puncta (image iv, inset). Similarly, when GFP-NP was replaced with a GFP-tagged constitutively active (CA) Rab11 mutant, good colocalization was seen between vRNA and the fluorescent Rab11 protein both at the perinuclear region (image v) and for the smaller cytoplasmic granules (image vi). Thus, the RNP reconstitution assay successfully recapitulated the perinuclear accumulation of cytoplasmic vRNA seen in authentic viral infection, and furthermore it confirmed good cytoplasmic colocalization between GFP-NP and vRNA as well as between GFP-Rab11 CA and vRNA.

We next used this system to investigate the trafficking of viral RNPs in living cells by collecting time-lapse movies. Cells displaying the prominent perinuclear accumulation of GFP-NP (which thus likely contained successfully reconstituted RNPs) also showed the presence of many additional, smaller foci of NP in the cytoplasm (Fig. 6A). Many of the small GFP-NP foci remained approximately static (except for Brownian-type movement, which is defined here as random) during the observation period, but some exhibited the classic saltatory pattern of intermittent but rapid movement expected for microtubule-based transport. (see Movies S2 and S3 in the supplemental material; example tracks and selected frames are shown in Fig. 6A). Other particles showed more processive (defined as visible for at least five consecutive frames) but generally slower movement. When the speeds of multiple particles exhibiting the various classes of movement were measured, those that displayed random Brownian motion moved at an average of 0.12 ± 0.07 μm/s, while those exhibiting processive movement traveled at 0.25 ± 0.11 μm/s. In contrast, particles showing saltatory and/or bidirectional movement traveled at much faster speeds of up to nearly 3 μm/s with an average of 0.81 ± 0.40 μm/s (Fig. 6B and C). The large perinuclear accumulation remained mostly stationary, although its shape continually changed throughout observation and smaller particles could be seen to apparently both merge with and depart from it (see Movie S2 in the supplemental material).

To test the involvement of microtubules in RNP movement, transfected cells were treated with a panel of drugs that interfere with microtubule and/or vesicular transport and imaged up to 3 h later. From this, it was apparent that monensin treatment had a relatively slight (although statistically significant) effect on RNP movement (Fig. 6D), in contrast to its dramatic effect on vRNA trafficking in infected cells (Fig. 4). Under the shorter timescale of the live-cell imaging experiments, it is possible that monensin blocks the formation of new RE vesicles without disturbing the trafficking of preexisting ones (33). However, all drugs targeting microtubules caused a statistically significant, nearly complete loss of the fast, saltatory class of movement (Fig. 6D; an example of a nocodazole-treated cell is shown in Movie S4 in the supplemental material). These data further support the hypothesis that RNP cytoplasmic trafficking is mediated in part by the microtubule network.

Based on the observation that the number of cells displaying an obvious perinuclear accumulation of vRNA increased and then decreased during the infection cycle (Fig. 1C), we hypothesized that RNPs localized there first before distributing to the cell periphery to be assembled into budding virions. Therefore, we used fluorescence recovery after photobleaching (FRAP) to examine whether GFP-tagged NP in the perinuclear compartment dynamically exchanges. A bleach area large enough to cover the entire perinuclear compartment was defined (Fig. 7A) to ensure that FRAP recovery represented the exchange of GFP-NP with the cytoplasm rather than an internal rearrangement of protein in the compartment or movement in the z axis. Relatively slow but consistent recovery of fluorescence within the perinuclear body was observed, on average to 10.8% ± 1.3% of the initial value during a period of 85 s (Fig. 7A to C; also see Movie S5 in the supplemental material), indicating measurable but limited exchange with the perinuclear compartment during the period of a minute or so. Thus, the perinuclear compartment is not static, but neither is there rapid exchange with GFP-NP in the cytoplasm; this finding is compatible with the directed transport of RNPs rather than free diffusion.

Since cytoplasmic NP and vRNA colocalized well with endogenous and CA forms of Rab11 in fixed cells (10) (Fig. 3 and 5), we tested whether this extended to living cells by transfecting cells with plasmids to reconstitute cytoplasmic GFP-tagged RNPs as before and additionally with an otherwise WT RFP-tagged Rab11. An excellent degree of colocalization between cytoplasmic puncta of GFP-NP and RFP-Rab11 was seen (Fig. 6E). Furthermore, time-lapse imaging showed that many of the structures double labeled with the two fluorophores displayed the characteristic rapid, saltatory movement in which particles sometimes reversed direction (see Movie S6 in the supplemental material).

We next tested whether the colocalization and cotrafficking of NP, vRNA, and Rab11 in cells reflected an interaction that could be detected biochemically. Cells were transfected with GFP-Rab11 CA or, as negative controls, with GFP-Rab11 DN, GFP-Rab8 CA, or GFP alone. GFP-Rab11 DN does not colocalize with NP in infected cells, while Rab8 is not known to play a role in the influenza virus life cycle (10). Following virus superinfection, cell lysates were prepared and analyzed by Western blotting for GFP and selected viral proteins before or after fractionation over GFP-affinity beads. The P proteins, NP, M1, M2, and various GFP polypeptides were detected as expected in the cell lysates (Fig. 8A, lanes 1 to 5). GFP-containing polypeptides also were retained as expected on the anti-GFP beads (lanes 6 to 10). Only trace quantities of M1 or M2 were detected in any of the bound fractions. However, NP and the P proteins were readily detectable in the sample containing GFP-Rab11 CA, while only background amounts were seen in samples containing GFP-Rab11 DN or GFP-Rab8 CA proteins (Fig. 8A, compare lane 7 to lanes 6, 9, and 10). The quantification of the amounts of the viral proteins bound under various conditions from replicate experiments showed that the protein components of the viral RNP were consistently precipitated by GFP-Rab11 CA at around 10 times the level obtained with GFP alone or with GFP Rab11 DN, Rab8 CA, or Rab4 CA (Fig. 8B), strongly suggesting that RNPs interact with the GTP-bound form of Rab11. In confirmation of this, vRNA also was directly detected in samples affinity purified with GFP-Rab11 CA but not the other GFP polypeptides (Fig. 8C). To determine which component of the influenza virus RNP interacts with GTP-bound Rab11, we cotransfected the individual RNP subunits with GFP-Rab11 CA or GFP and performed GFP-affinity pulldowns. No detectable amounts of NP, PB1, or PA were precipitated under these conditions despite their presence in the starting cell lysates, whereas PB2 was specifically bound by GFP-Rab11 CA but not by GFP alone (Fig. 8D). When replicate experiments were quantified, levels of PB2 bound by GFP-Rab11 were around 6-fold higher than those precipitated by GFP alone, while the amounts of PB1, PA, and NP were similar for both ligands (Fig. 8E). These data are consistent with the microscopy data showing the colocalization of viral RNPs with WT or CA forms of Rab11 (Fig. 3, 5, and 6) but not the DN form (10) and support the hypothesis that RNPs traffic via an Rab11-dependent pathway potentially mediated by an interaction with the PB2 subunit of the polymerase.

We recently showed that the siRNA-mediated depletion of Rab11 in 293T cells caused defects in influenza A virus budding without obvious alterations to NP localization (10). However, we did not examine vRNA localization directly, and 293T cells have a small cytoplasmic volume, which may have masked more subtle effects on NP trafficking. We therefore examined the effect the loss of Rab11 had on vRNA distribution in A549 cells, which have a much larger cytoplasm/nuclear ratio. Cells were depleted of Rab11a and Rab11b isoforms by siRNA treatment as before (10); Western blot analysis of Rab11a and selected viral proteins confirmed the effective knockdown of Rab11 without major effect on viral gene expression (Fig. 9A). When replicate experiments were quantified, an average depletion of around 90% was obtained. We then examined the effect this had on RNP localization across a time course of virus infection. In cells treated with control siRNA oligonucleotides, vRNA and NP displayed time-dependent changes in localization similar to those seen in other cell types: early nuclear accumulation (with NP but not vRNA, tending to concentrate toward the nuclear periphery), followed by increasing cytoplasmic accumulation at later time points (Fig. 9B). The cytoplasmic phase of RNP localization included the initial accumulation of vRNA at a perinuclear site (Fig. 9B, 7 h, arrowheads), which in some cells apparently was associated with filamentous or tubular structures (Fig. 9B, 7 h, inset). At later time points, the perinuclear vRNA staining was lost, being replaced by many large (≥1-μm) aggregates dispersed throughout the cytoplasm, which mostly double stained for both vRNA and NP (Fig. 9B, 10- and 16-h time points, and C). In cells depleted of Rab11, vRNA still accumulated in the nucleus at an early time point and was successfully exported to the cytoplasm thereafter (Fig. 9D). However, the pattern of RNP localization in the cytoplasm was considerably altered, as the initial perinuclear accumulation and the subsequent large cytoplasmic structures were absent, being replaced instead with a much more dispersed, finer-grained pattern of staining (Fig. 9C and D). Thus, the depletion of Rab11 leads to altered vRNA trafficking in A549 cells.

Finally, we asked whether NP and Rab11 interacted in infected A549 cells, since the majority of our previous data came from 293T cells and overexpressed Rab11. Attempts to detect an interaction between endogenous Rab11 and RNPs by pulldown assay were unsuccessful (from A549 or 293T cells; data not shown). We therefore turned to the Duolink PLA in situ cell-based fluorescence proximity assay, in which closely apposed (∼40 nm or less) primary antibodies are detected by the amplification of intermolecular ligation events of oligonucleotides covalently attached to secondary antibodies (54). When cells were double stained by this procedure for NP and either Rab11 or Rab4 (as a negative control), the numbers of detected events were around 70-fold higher in infected cells than in mock-infected cells for Rab11 but were only around 15-fold higher for Rab4 (Fig. 9E). In cells first depleted of Rab11, however, levels of signal from Rab11 and NP but not Rab4 and NP were drastically reduced, indicating the specificity of the reaction. Thus, while not proving a physical interaction, NP and Rab11 normally are in close proximity during infection.

To provide information on where the NP-Rab11 interaction occurred in A549 cells, we performed conventional indirect immunofluorescence. In uninfected A549 cells, endogenous Rab11 was difficult to detect by this method, producing faint punctate staining throughout the cytoplasm as well as the more distinct labeling of a perinuclear structure likely to be the RE (Fig. 10). Unexpectedly, Rab11 staining was markedly more intense in infected cells. At 7 h p.i., brighter punctate staining throughout the cytoplasm was apparent, and in many cells, there was prominent perinuclear staining with obvious colocalization with NP (Fig. 10, arrows). Late in infection, large cytoplasmic aggregates of Rab11 appeared that again colocalized strongly with NP (Fig. 10, 16 h p.i.). The virus-induced increase in the intensity of Rab11 staining did not result from increased amounts of the protein, as judged by Western blot analysis (Fig. 9A), but it was not artifactual cross-reactivity with another antigen, since infected cells depleted of Rab11 by siRNA treatment displayed very low levels of staining (Fig. 10). One possibility is that the more concentrated protein facilitates detection, although an infection-dependent unmasking of the Rab11 C-terminal sequence recognized by the antibody also is possible. The depletion of Rab11 did, however, again lead to the altered cytoplasmic distribution of NP as previously observed for vRNA (Fig. 9D and 10). Overall, therefore, RNPs colocalized with Rab11 in A549 cells, and the loss of Rab11 altered their pattern of cytoplasmic localization. This supports the hypothesis that the cytoplasmic trafficking of the influenza virus genome involves Rab11.

Our aim was to better define the trafficking mechanism(s) responsible for the transport of the influenza A virus genome from the nuclear to plasma membranes. Using a combination of static and live-cell imaging techniques, we provide evidence for two distinct patterns of vRNA localization in the cytoplasm (perinuclear and dispersed cytoplasmic granules), as well as evidence that RNPs traffic in part using the microtubule-based vesicular transport system, and that the adaptor linking viral and cellular components for this process is Rab11. We interpret our data as supporting a two-step model for the cytoplasmic trafficking of the viral genome, in which after leaving the nucleus, RNPs first accumulate at the pericentriolar recycling endosomal compartment through an interaction with Rab11 and then piggyback onto the microtubule network via vesicles destined for the cell periphery (Fig. 11).

The perinuclear accumulation of NP and vRNA has been noted before (10, 28, 34). Here, we extend this observation by showing that it is a time-dependent phenomenon occurring midway through infection, that it can be seen in living cells, and that it occurs with several strains of virus in different cell types, making it likely that it reflects an important feature of the viral life cycle. We also identify the location as the pericentriolar recycling endosomal compartment. We base this on the failure of vRNA to meaningfully colocalize with other cellular vesicular markers found in this region of the cell, including cis and trans compartments of the Golgi network, as well as markers for early endosomes or aggresomes (data not shown). Evidence in favor of this is that vRNA and Rab11a overlap was significant, while GFP-NP-tagged RNPs also colocalized near the MTOC with transferrin (data not shown). Both Rab11a and transferrin serve as markers for the pericentriolar RE compartment (20, 43, 56). The apparently tubular localization pattern seen for vRNA in a subset of A549 cells also is consistent with localization at the RE (25). However, we disagree with previous conclusions that perinuclear RNPs colocalized with markers for the TGN (28, 34).

We propose that vRNAs accumulate at the pericentriolar RE first before moving to the periphery of the cell. This aspect of the model is based on the observations that vRNA localization here is seen soon after the onset of genome replication and export and that it tends to fade later in infection, suggesting a dynamic rather than static association. Further supporting this, the observation of GFP-tagged NP particle movement in living cells was consistent with the movement of RNPs both into and out of the perinuclear body during relatively short time scales. It also is supported by the persistence of vRNA staining in the perinuclear region after vesicular transport in infected cells was blocked by treatment with monensin. The model also is consistent with the known directionality of Rab11 in directing vesicular trafficking from the RE to the plasma membrane (43, 56, 59).

We hypothesize that an interaction between PB2 and Rab11 recruits the viral genome to the pericentriolar RE based on the finding that all components of an RNP can be specifically coprecipitated by the GTP-bound form of Rab11 from infected cells but only PB2 as an individual protein, and that in the absence of normal levels of Rab11 RNPs do not concentrate at the perinuclear site. Also, their cytoplasmic localization in general is altered. The simplest hypothesis is a direct protein-protein interaction between PB2 and Rab11, but at present we cannot rule out bridging interactions with other cellular proteins, such as a Rab11-family interacting protein.

In a refinement of a recent proposal that RNP trafficking involves the vesicular transport system (28), we also postulate that RNP transport to the cell periphery is at least partly based on an attachment to Rab11-positive vesicles that move via microtubules, which is consistent with the known transport of such cellular cargo (3, 7, 13). In evidence of microtubule-based movement of RNPs, the speeds and saltatory character exhibited by many GFP-NP particles in time-lapse movies of living cells are diagnostic as well as consistent with the values measured for the movement of intracellular enveloped vaccinia virus on microtubules (44, 57). We also point to the disruption of normal vRNA localization seen after treatment with specific inhibitors of the microtubule network. Similar perturbations of cytoplasmic NP localization after treatment with inhibitors of microtubule function have been reported previously (34). The apparent misdirection of vRNA to the lateral membranes of MDCK cells after nocodazole treatment also is consistent with a study that found a loss of polarity in virus budding under similar circumstances (45). It is worth noting that nocodazole treatment has been shown previously to disperse Rab11-positive vesicles, resulting in an accumulation at the lateral membranes (3, 13). Evidence for RNP trafficking piggybacking on vesicles is less direct, as at present we cannot formally exclude a mechanism in which RNPs are targeted to microtubules without an intervening lipid structure. However, we think this hypothesis is reasonable given the extensive colocalization between NP, vRNA, and both WT and CA forms of Rab11 and the fact that the GTP-bound form of Rab11 would be expected to be membrane bound. It also is consistent with the observation that monensin treatment throughout infection inhibits vRNA movement away from the RE.

We would not argue that all RNP transport occurs via the model just described. While we measured saltatory movement consistent with microtubule-based trafficking for some RNP particles, others exhibited diffusional-type movement, similarly to that previously observed for microinjected RNPs (6). Other GFP-NP particles exhibited short-range processive movement with an average speed of around 0.2 μm/s, which is consistent with actin-based motility (44). These observations, together with the relatively minor effects of microtubule-disrupting drugs on virus replication and the diffuse localization of vRNA in Rab11-depleted cells, make it plausible that redundant mechanisms apply in transporting vRNPs to the plasma membrane. We hypothesize that microtubules are required for the efficient long-range delivery of RNPs, while other modes of transport (i.e., diffusion or along microfilaments) also operate, potentially over shorter ranges and/or in different areas of the cell. A degree of functional redundancy may explain the relatively minor effects cytoskeletal poisons have on the overall titer of virus replication in cell culture.

Our model for vRNA trafficking may have parallels with certain other viruses. Notably, the capsid assembly of certain retroviruses has been proposed to take place at the pericentriolar RE (1), and this region of the cell also may be used as a way point by several large DNA viruses (60). Most notably, however, our findings are strikingly similar to those recently reported for the nonsegmented negative-sense Sendai virus, in which the colocalization of viral RNPs and Rab11a as well as the microtubule-based movement of the RNPs was shown (14).

Previous work in our laboratory (10) showed that Rab11 was required for virus budding. However, the question of whether Rab11 was directly involved in mediating membrane scission or indirectly linked (by transporting a viral or cellular factor essential for pinching off) was not resolved in our earlier work. We did not observe a notable NP-trafficking defect in Rab11-depleted 293T cells (10), in contrast to the greatly altered distribution of NP and vRNA seen in depleted A549 cells here. A549 cells, as human lung cells with considerable cytoplasmic volumes, represent a more authentic model to study viral trafficking. The discrepancies seen in some circumstances between the indirect immunofluorescent staining of NP and FISH staining of vRNA (e.g., Fig. 1B, 4-h time points, and Fig. 9, 4- and 16-h time points) also raise the possibility of separate pools of free NP in the cytoplasm and nucleus. Nevertheless, the role of Rab11 in influenza RNP trafficking presented here is consistent with the theory that Rab11 is indirectly involved in viral budding. One possibility is that Rab11 also is required for the efficient delivery of M2 to the apical plasma membrane (48). However, the possibility that the trafficking of the viral genome is specifically required to allow efficient virus assembly and/or budding is an attractive hypothesis and is one that represents an area of future research.

We thank Colin Crump, Folma Buss, Stephen Ferguson, Janet Daly, and Miguel Seabra for the gift of reagents and Fumitaka Momose for helpful discussion and sharing unpublished data.

This work was supported by a grant from the MRC (no. G0700815) to P.D. E.A.B. is supported by a studentship from the Gates Cambridge Trust. E.K.C.R., A.F., and R.M. are supported by studentships from the Wellcome Trust.