Novel Staufen1 ribonucleoproteins prevent formation of stress granules but favour encapsidation of HIV-1 genomic RNA

The mechanisms governing the encapsidation of retroviral genomic RNAs have been the subject of intense research for many years. The principal determinants for genomic RNA (vRNA) encapsidation are an intact nucleocapsid (NC) domain in the precursor protein, pr55^Gag (termed Gag herein) and an intact packaging signal, psi, within the vRNA (De Guzman et al., 1998). Recent work demonstrates that long-range interactions, conformational changes and the dimeric state of vRNA dictate the readiness for encapsidation (D'Souza and Summers, 2005; Ooms et al., 2004; Paillart et al., 2004), and the recent structure determination might help shed light on this (Watts et al., 2009). For murine leukemia virus (MLV), a network of interactions promotes sequence- and structure-specific binding by NC to the psi RNA. The interaction between host proteins and vRNA might also provide a molecular switch by suppressing vRNA translation to favour encapsidation (Cimarelli and Luban, 1999). There is still some controversy about whether this type of mechanism exists, because HIV-1 vRNA that is translated might also be encapsidated (Butsch and Boris-Lawrie, 2000; Kaye and Lever, 1999; Poon et al., 2002). Many questions remain concerning how HIV-1 co-opts host protein function and machineries to traffic and encapsidate its RNA genome.

Staufen1 is a double-stranded (ds)RNA-binding protein with roles in RNA localization (St Johnston et al., 1991), translation (Dugre-Brisson et al., 2005) and mRNA decay (Kim et al., 2005). It is also a principal component of RNA transport ribonucleoproteins (RNPs) (Villace et al., 2004), stress granules (SG) (Thomas et al., 2005), APOBEC3G complexes (Chiu et al., 2006) and other types of RNPs (Jonson et al., 2007; Snee and Macdonald, 2009). Staufen1 is mobile in these contexts and its presence in several types of RNPs is considered an important characteristic for performing its multiple functions (Dahm et al., 2008). Recent in vivo work also supports a role for Staufen in synapse development (Vessey et al., 2008).

In humans, there are two genes that encode Staufen1 and Staufen2 (STAU1 and STAU2). Staufen2 is primarily found in neuronal cells and has roles in RNA biogenesis and trafficking (Duchaine et al., 2002; Monshausen et al., 2004). Staufen1 shares these functional characteristics but STAU1 is more ubiquitously expressed in human tissues. Moreover, each of these genes generates a primary RNA transcript that can be differentially spliced to produce identifiable isoforms. While some work has made in-roads into deciphering the functional differences of the Staufen2 isoforms, little information is known about the functional impact of the expression of Staufen1 isoforms. Nevertheless, their existence suggests multiple and possibly overlapping roles for these proteins in RNA biogenesis and utilization.

Evidence for a role for Staufen1 in HIV-1 replication derives from our observation that Staufen1 was incorporated in HIV-1 to levels that corresponded to the quantity of vRNA (Mouland et al., 2000). Staufen1 overexpression led to a several-fold increase in vRNA encapsidation and a corresponding decrease in HIV-1 infectivity of progeny virions (Mouland et al., 2000). In later work, we showed that Staufen1 was found in the HIV-1 RNP that contains vRNA and precursor Gag (Chatel-Chaix et al., 2004). The depletion of Staufen1 by siRNA led to defects in both virus production and infectivity (Chatel-Chaix et al., 2004). We also found that Staufen1 modulates Gag assembly (Chatel-Chaix et al., 2007), probably due to a direct interaction between the NC domain of Gag and a short region in the N-terminus of Staufen1 (Chatel-Chaix et al., 2008). The selectivity of Staufen1 to bind precursor Gag and vRNA suggested that Staufen1 performs its role in the context of a Staufen1-Gag-vRNA ternary complex and could be involved in the initial recognition and selection of vRNA during the assembly of HIV-1.

In this report, a novel Staufen1 RNP is described whose formation depends on HIV-1. Staufen1, vRNA and Gag colocalize in these Staufen1 HIV-1-dependent RNPs (SHRNPs), the size of which depends on existing Staufen1 levels in cells. A stable association between Staufen1 and vRNA was found in colocalization studies but Gag could nevertheless be displaced from SHRNPs. Staufen1 was found in various HIV-1- and stress-induced RNPs. A relationship between the abundance of virion-associated vRNA and Staufen1 and the apparent size of SHRNPs indicates that SHRNP assembly is an important event in the selection, trafficking and encapsidation of vRNA.

Our early work in which we performed co-immunoprecipitation analyses suggested that Staufen1 interacts with HIV-1 Gag in an RNA-independent manner and forms a ternary complex with the vRNA, excluding other HIV-1 mRNA species (Chatel-Chaix et al., 2004). We developed a sequential, two-step, affinity-based immunoprecipitation assay to provide further evidence that this complex exists. We mock-transfected cells or transfected cells with HIV-1 proviral DNA with or without a Staufen1-FLAG expression construct and harvested cells for western blotting and binding analyses. We readily detected Staufen1-FLAG and Gag by western blotting and both the vRNA and gapdh mRNA by RT-PCR in HIV-1-expressing cells (in INPUT). Cell extracts were first submitted to an immunoprecipitation directed to the FLAG epitope. Staufen1-FLAG was first eluted from FLAG-antibody coated beads in the presence of an excess of FLAG peptide. Staufen1-FLAG was found in the first eluate (Chatel-Chaix et al., 2004), and Gag co-purified only when Staufen1-FLAG was coexpressed with HIV-1 (Fig. 1A) attesting to the specificity of Gag immunoprecipitation with the anti-FLAG resin. To examine whether the vRNA was carried through and remained associated to Staufen1-Gag, an immunoprecipitation was then performed using an anti-Gag antibody. In the final eluate, a specific RT-PCR signal for vRNA was found only when both Staufen1-FLAG and HIV-1 were coexpressed in cells, but not when Staufen1-FLAG was excluded from the transfection. This demonstrated the specificity with which vRNA co-precipitates with the anti-FLAG and anti-Gag antibodies. The result provides corollary biochemical evidence that a Staufen1-Gag-vRNA ternary complex exists in HIV-1-producing cells. To confirm the existence of this ternary complex and to better characterize it, we examined the localization of Staufen1 in relationship to vRNA and Gag in situ by using combined immunofluorescence and fluorescence in situ hybridization analyses (IF/FISH) followed by laser scanning confocal microscopy (LSCM). In control (siNS-treated) HIV-1-expressing cells, Staufen1, Gag and vRNA exhibited colocalization in the cytoplasm and prominent colocalization at the plasma membrane. Imaging analyses demonstrated that these viral and host components are found as a ternary complex in HIV-1-producing cells (Fig. 1B). These observations extend our earlier work that showed that Staufen1 associates selectively with these viral components in cells (Chatel-Chaix et al., 2004; Mouland et al., 2000). Whereas the diameter of Staufen1-positive punctae in mock-treated (non-HIV-1-expressing) cells was determined to be less than 200 nm, in HIV-1-expressing cells Staufen1, Gag and vRNA were generally found in small punctae of approximately 0.6-0.7 μm (Fig. 1B,E).

We next examined the effects on HIV-1 vRNA and Gag localization of depleting cells of both Staufen1 isoforms by using si55/63, or of the high molecular weight Staufen1^63kDa alone by using si63. HeLa cells were mock-transfected or transfected with proviral DNA and non-silencing control siRNA (siNS), si55/63 or si63. Staufen1 depletions were always verified by RT-PCR and/or western blotting (Fig. 1C). RT-PCR analyses showed that the siRNAs specifically targeted Staufen1 mRNAs as expected, with si55/63 targeting the expression of Staufen1^55kDa and Staufen1^63kDa transcripts and si63 targeting uniquely the Staufen1^63kDa (T3) transcript. The depletions of Staufen1 by si55/63 or si63 were consistent, selective and highly efficient (depletions of 90-92%).

In our imaging analyses, we found an apparent doubling in the size of granules for the RNPs that were enriched in vRNA, Gag and the residual Staufen1 in si55/63-treated cells and these were principally found in juxtanuclear and cytoplasmic domains (Fig. 1D,E). Strikingly, in 90±5% (mean ± s.e.m., n>250 cells) of the cells, we only found Staufen1-containing RNPs in HIV-1-expressing cells (white arrowheads, Fig. 1D) and not in cells in which HIV-1 expression was not detectable (yellow arrowheads, Fig. 1D, si55/63 panels). Residual levels of Staufen1 and Gag appeared in large bright punctae due to accumulations of these proteins in these foci. We also examined SHRNPs in Jurkat T cells, and found that these were enlarged when Staufen1 was depleted by short hairpin (sh)RNA (shSTF; data not shown); however, the paucity of cytoplasmic space limited our conclusions from the results obtained using this cell type. The localization of Staufen1, vRNA and Gag in Staufen1^63kDa-depleted HeLa cells did not markedly change SHRNP appearance in greater than 90% of cells examined (>200 cells, five experiments; Fig. 1D, bottom panel). Identical findings for Staufen1 depletions and SHRNP formation were obtained at 24 and 48 hours post-transfection (Fig. 1E). These results demonstrate that larger intracellular RNPs form that contain vRNA and the residual Staufen1, primarily when both Staufen1 isoforms are silenced in cells. We cannot completely rule out the possibility that the strong signal intensities were due to coalescing RNPs or other organelles at these foci. Because Staufen1, vRNA and Gag colocalize in cytoplasmic punctae and at the periphery in HIV-1-producing cells, we therefore named these RNPs Staufen1 HIV-1-dependent RNPs (SHRNPs).

Results from our Manders' colocalization analyses were also consistent, such that while Staufen1 and vRNA colocalized appreciably to 20±9.1% (mean ± s.e.m.) in siNS-treated cells, there was a tripling to 58% of the proportion of Staufen1 colocalizing with vRNA in conditions of si55/63 depletion (Table 1). This increase also shows that there was little redistribution or loss of Staufen1 from the vRNA-Gag complex in depletion conditions.

Viral infection can generate a cellular stress response (Gale et al., 2000; McInerney et al., 2005) and Staufen1 is a component of SGs (Thomas et al., 2005). Recent data show that Staufen1 might temper stress responses in cells by controlling the equilibrium between the formation of polysomes and SGs (Thomas et al., 2009). To determine whether SHRNP assembly in Staufen1-depleted cells depends on translation activity, a characteristic of SG assembly, HeLa cells were transfected with HIV-1 proviral DNA and siRNAs as above and left untreated, or briefly treated with puromycin (Pur) or cycloheximide (CHX) to promote or inhibit SG assembly, respectively. IF/FISH co-analyses were performed to identify Staufen1, Gag or PABP1 (a marker for SG) and vRNA. On the one hand, Pur treatment induced PABP1- and Staufen1-containing SG and these were observed when HIV-1 was expressed and in Staufen1-depleted cells (Fig. 2, panels 2-4). On the other hand, a translation block at the level of elongation induced by CHX treatment caused a uniform distribution of PABP1 as expected. The distribution of Staufen1 was likewise dispersed in this condition when HIV-1 was expressed (Fig. 2, panels 5-7). These data indicate that SHRNP assembly is dependent on translation and appears to be in equilibrium, like SG, with cellular mRNA translational activity.

To determine whether vRNA was included in the Staufen1-PABP1-containing foci, we next examined its localization under these experimental conditions. Without drug treatment, HIV-1 expression did not elicit a stress response in cells because large PABP1+ or Staufen1+ SG foci were not detectable (data not shown and Fig. 3A, panel 2). In untreated cells, we observed identical distributions for Staufen1, Gag and vRNA as shown in Fig. 1D for siNS, si55/63 and si63 conditions, respectively (Fig. 3A, panels 2-4). Pur treatment resulted in the formation of 2- to 5-μm structures containing Staufen1 in mock-transfected cells (Fig. 3A, panel 5) and caused a near-complete recruitment of vRNA and Staufen1 to Pur-induced SGs (Fig. 3A, panels 6-8) containing PABP1 (because PABP1 is also colocalized with Staufen1; see Fig. 2, panels 2-4) in HIV-1-expressing cells. Strikingly, Gag was almost completely excluded from SGs in Pur-treated HIV-1-expressing cells, demonstrating that the Staufen1-vRNA association is maintained under Pur treatment. In Staufen1-depleted and Pur-treated cells, the residual Staufen1 was also recruited to PABP1-positive SG foci together with vRNA (Fig. 2, panel 4; Fig. 3A, panel 7). This suggested that the prominent SHRNPs in Staufen1-depleted cells (no drug added) can be remodelled and that Staufen1 and vRNA can be recruited from Gag-containing SHRNPs to constitute SGs. This is consistent with the ability of Staufen to shuttle from RNA trafficking granules to SGs in response to stress (Thomas et al., 2005). Thus, the Pur-induced SGs containing Staufen1, vRNA and PABP1 are probably bona fide SGs, and the association between Staufen1 and vRNA remains stable in the context of both SHRNPs and Pur-induced SGs. In the latter, however, Gag is not present, and this was reflected by our sequential affinity-immunoprecipitation assay (data not shown and Fig. 1A). Consistently, we did not find a significant change in the colocalization between Staufen1 and vRNA in untreated and Pur-treated cells (Table 1). These results indicate that the Staufen1-containing RNPs undergo remodelling and that the composition of a SHRNP is distinct from that of a Pur-induced SG.

When cells were treated with CHX to block translation elongation, highly dispersed distributions for PABP1 (Fig. 2), vRNA and Gag were found. Few SG formed in all conditions and the SHRNPs in Staufen1-depleted cells were less numerous (Fig. 3, panels 9-12). The brief treatments with Pur or CHX had little impact on steady-state levels of Gag protein (Fig. 3B). These results further demonstrate a dependence on translation for SHRNP formation.

The exposure of cells to arsenic (Ars) is known to induce the formation of SGs that contain Staufen1, PABP1 and other proteins such as HuR, G3BP and eIF3 (Anderson and Kedersha, 2008; Kedersha et al., 2002; Thomas et al., 2005). Because SHRNPs and SGs are related by virtue of the presence of Staufen1 in these cytoplasmic foci, we determined what the outcome was on Staufen1, Gag and vRNA localization by treating HIV-1-producing cells with Ars. In mock-transfected cells we first confirmed that Ars treatment induced visible SGs enriched in both Staufen1 and PABP1 (Fig. 4A). Strikingly, however, in cells expressing HIV-1 (i.e. staining for vRNA and Gag; Gag staining is not shown), we were not able to find typical Ars-induced SGs in either siNS or si55/63 conditions. Here, vRNA and Staufen1 were dispersed and accumulated at a juxtanuclear region in cells in about 95% of cells (Fig. 4B,C; white arrowheads; n=150-200). Thus, HIV-1 expression prevents the formation of SG foci that are induced by Ars. The extent to which Staufen1 and vRNA colocalized was not different to that found in untreated cells (Table 1), indicating that a redistribution of these SHRNPs had occurred. In 85% of transfected cells (n=200), large Staufen1 (and PABP1-positive, not shown) SGs were detectable in HIV-1-negative cells (e.g. yellow arrowheads in Fig. 4B-D), similar to those found in mock-treated cells (Fig. 4A, right panel). In addition, a deficit in Staufen1 did not impede SG formation per se, consistent with recent findings (Thomas et al., 2009), nor did it change the abundance of SGs. However, HIV-1 changed the characteristics and location of Staufen1-containing RNPs. These data also demonstrate that HIV-1-expressing cells are resistant to Ars-induced SG formation. Finally, PABP1 staining in these vRNA and Staufen1 RNPs was virtually absent compared with that in the surrounding cytoplasm, indicating that these RNPs did not bear this hallmark SG marker (Fig. 4E). This was in contrast to the inclusion of PABP1 (and the exclusion of Gag) in Pur-induced SGs (Fig. 3). The block to the formation of Ars-induced SGs by HIV-1 was confirmed using other SG proteins such as eIF3 and G3BP (Anderson and Kedersha, 2008) (supplementary material Fig. S1). Moreover, the HIV-1-induced block to SG formation supports the notion that SHRNPs are distinct in HIV-1 producing cells because they also exclude four SG marker proteins tested in this work (PABP1, G3BP, eIF3 and HuR) (Fig. 4 and supplementary material Fig. S2).

It is known that SG formation and translational inhibition after Ars treatment is dependent on eIF2α phosphorylation, the reduction of the translation ternary complex, and recruitment of several RNA-binding proteins (Kedersha et al., 1999). Some viruses have developed mechanisms to replicate during cellular stress even when cellular translation is shut off by eIF2α phosphorylation (McInerney et al., 2005). We analyzed the levels of phosphorylated eIF2α (eIF2α-P) in cells. This analysis revealed that Ars treatment induced eIF2α-P almost threefold (293%). Although HIV-1 alone did not induce eIF2α-P levels to any great extent (121%), HIV-1 did not markedly alter eIF2α-P levels in Ars-treated cells either (230% vs 293%), demonstrating that HIV-1 expression does not protect cells from SG formation by inhibiting eIF2α phosphorylation (Fig. 4F).

The processing body (PB) represents another type of RNP complex that plays a role in RNA storage and metabolism. PBs have been posited to be important for replication of Ty retrotransposons and for other RNA viruses (Beckham and Parker, 2008; Beliakova-Bethell et al., 2006). Staufen1 has also been found in Drosophila PBs but not in those in mammalian cells (Barbee et al., 2006; Thomas et al., 2009). We therefore attempted to clarify whether the Staufen1-containing SHRNPs were related to PBs by staining cells with a classical PB marker, DCP-1. HeLa cells were transfected with HIV-1 DNA and siRNAs as described above, and PBs were detected using an anti-DCP-1 antibody by indirect IF and LSCM. As shown in Fig. 5A, there was no significant colocalization between vRNA, Gag or Staufen1 and DCP-1 in either control or Staufen1-depleted cells (Fig. 5A). We then counted the number of PBs in cells in which we detected DCP-1 foci. In mock-transfected cells, the average number of PB foci in a single LSCM focal plane through the middle of the cell (as judged by confocal sectioning along the Z-axis) totalled 28±1 (mean ± s.e.m.) foci per cell. By contrast, the number of PBs dramatically reduced to 7±1 (mean ± s.e.m.), when we counted all cells expressing DCP-1 at low and high levels in HIV-1-expressing cells. In other imaging experiments, we observed few — if any — PBs in HIV-1-expressing cells labelled with vRNA (Fig. 5B). Likewise, the effects of HIV-1 on PB abundance are probably due to effects on PB assembly because the abundance of DCP-1 in cells was maintained (in both transfected HeLa and infected Jurkat T cells; Fig. 5B and data not shown).

Ars treatment of cells dramatically induces the number of visible PB foci in cells. In addition, it promotes the formation of Staufen1- and PABP1-positive SGs. The relationship between SHRNPs and PBs was therefore analyzed after stress induction by Ars. We could not find PBs at the juxtanuclear region in Ars-treated cells, which was where vRNA, Gag and Staufen1 were found (Fig. 5C, panels 1,2). In mock-transfected cells, PABP1 staining was dispersed and PBs were detectable (Fig. 5B, panels 3,4), whereas in mock-transfected cell treated with Ars, we observed PBs in close proximity to numerous PABP1-positive SGs (Fig. 5C, panels 5,6, indicated by yellow arrowheads), confirming that SGs and PBs are dynamically linked (Kedersha et al., 2005). Our results demonstrate that in HeLa cells, vRNA and Gag do not localize to PBs, and SHRNPs do not contain DCP-1 and therefore are probably not PBs. Cumulatively, our results suggest that HIV-1 impairs the formation of constitutive or stress-induced silencing foci. By regulating the equilibrium between SGs, PBs and other RNPs, HIV-1 could efficiently exploit the cellular machinery to ensure efficient virus assembly when cells are exposed to different types of stresses.

The depletion of both Staufen1 isoforms by si55/63 resulted in markedly diminished Gag expression levels (53±5%, mean ± s.e.m., five experiments; Fig. 6). Staufen1^63kDa depletion alone had little effect on Gag expression (84±2.5%, mean ± s.e.m., five experiments) and both treatments had little effect on gp120 (Env) expression levels in the context of pNL4-3 (Fig. 7 and data not shown). The effects on Gag expression cannot be explained by the activation of the Jak-Stat pathway and the expression of IFN-inducible genes like PKR, by changes in the cell cycle or by changes in Gag protein stability (data not shown). Because we have shown that Staufen1 can enhance translation of certain mRNAs (Dugre-Brisson et al., 2005) and can stabilize polysomes (Thomas et al., 2009), these changes are probably due to effects of Staufen1 depletion on HIV-1 vRNA translation.

Gag has a central role in assembly and virus structure and this is substantiated by its ability to bind and co-opt many cellular proteins, including Staufen1 (Chatel-Chaix et al., 2004). Our observation that Gag could be rapidly excluded from Staufen1-vRNA RNPs by a brief Pur treatment revealed that these RNPs are dynamic and plastic structures. We therefore determined whether Staufen1-vRNA RNPs could form in the absence of Gag expression and also questioned whether PABP1 could replace Gag when it is not expressed. In order to do this, we expressed a pNL4-3-based proviral DNA (pNL4-XX) in which Gag expression is prevented due to the introduction of two stop codons in the gag open reading frame. This vector expresses all HIV-1 RNAs, and aborted translation of the vRNA prevents the synthesis of Gag and Gag-Pol proteins but the translation of 4 and 2kb RNAs is not impaired (Ajamian et al., 2008). Following expression of pNL4-XX in siNS and si55/63 conditions, cells were harvested and processed for western blotting and IF/FISH co-analyses for Staufen1 and vRNA. gp120 (Env) was stably expressed but Gag and Gag-Pol were not, as expected (Fig. 7A). Some 26% of cellular Staufen1 colocalized with vRNA (Table 1). In cells in which Staufen1 was depleted, we noted again larger Staufen1 punctae in pNL4-XX-expressing cells and a doubling of the extent to which Staufen1 colocalized with vRNA (to 60±10%, mean ± s.d.), consistent with a lowered abundance of cellular Staufen1 and little increase in the total amount of Staufen1 colocalizing with vRNA (Table 1). This also indicated that the association of Staufen1 to vRNA was probably not significantly affected when it was in limiting supply. We did not detect any significant plasma membrane localization when Staufen1 was depleted in cells (Fig. 7B), indicating that Gag expression might be involved in the localization of SHRNPs. This would be consistent with our recent findings showing a dependence of Gag on HIV-1 RNP localization (Lehmann et al., 2009).

We then determined whether Gag could be included in the Staufen1-vRNA RNP if Gag expression was rescued in trans. In this case, Rev is synthesized from pNL4-XX and a gag RNA is derived from a Rev-dependent Gag expressor supplied in trans (Lingappa et al., 2006). Gag was only expressed when pNL4-XX was coexpressed and, in si55/63 conditions, the decrease in Gag synthesis was again observed (Fig. 7A). Confocal imaging analyses for Staufen1, vRNA and Gag revealed that whereas the colocalization between Staufen1 and vRNA was maintained, Gag was not recruited to Staufen1-vRNA RNPs when it was expressed in trans (Fig. 7C). Furthermore, the larger RNPs that were found to be devoid of Gag had low immune reactivity for PABP1 (Fig. 7C), and this was found in the majority of cells (70%) in three experiments. Therefore, in order for a SHRNP to form that contains Gag, there might be a requirement for cis translation of the gag mRNA (i.e. vRNA) and the co-translational recruitment of Gag to Staufen1-vRNA complexes. This could also prevent the formation of typical SG in HIV-1-expressing cells by excluding PABP1. This result also supports the notion that SHRNP formation requires ongoing translation.

Finally, we studied the consequences of Staufen1 depletion and the remodelling of SHRNPs on the HIV-1 replication cycle. Because Staufen1 is a virion-associated protein (Mouland et al., 2000; Kozak et al., 2006), we determined whether there were any compositional differences between purified HIV-1 derived from control (siNS) and Staufen1-depleted cells. Cells were harvested for western blot and RT-PCR analyses. Virus was purified from cell supernatants and quantified by a sensitive ELISA that detects the mature Gag product, p24; this result was routinely confirmed by western blotting analyses for virion-associated Gag (Fig. 9B and data not shown). vRNA was isolated from cells, and equal quantities of virus were quantified by semi-quantitative RT-PCR. Whereas cellular levels of vRNA remained relatively unchanged, si55/63 enhanced the abundance of vRNA in virus particles (Fig. 8A). This enrichment of virion-associated vRNA was reproducible in different settings, using different proviral strains and siRNA targets and being measured by different methods (data not shown). Indeed, using shRNA to deplete intracellular Staufen1 (Fig. 1C), we confirmed that vRNA encapsidation was enhanced over twofold during the expression of an isogenic protease-negative provirus (HxBRU Pr—; Fig. 8B; corresponding expression blots are shown in Fig. 9B). The increases in vRNA encapsidation in both si55/63 and shSTF conditions ranged from 1.7- to 4.5-fold (3.6±0.5-fold, mean ± s.e.m., eight experiments). Thus, the depletion of Staufen1 from HIV-1-expressing cells consistently results in enhanced vRNA encapsidation.

Because vRNA encapsidation might be coupled to dimerization (Ooms et al., 2004), we then tested whether Staufen1 influences vRNA dimerization. Despite an efficient knockdown of Staufen1 expression by 90% (the corresponding western blots for this experiment are shown in Fig. 6), the melting curves for vRNA dimers derived from siRNA-treated cells were found to be similar (Fig. 8C), ruling out an influence of Staufen1 on vRNA dimerization and stability.

To examine the protein content of virus in these conditions, we performed western blotting on extracts from cells and purified virus. Equal protein quantities (for cell extracts) and equal p24-equivalents of virus were loaded onto SDS-PAGE gels. Staufen1^55kDa was efficiently depleted by 90% in si55/63-treated cells when compared to mock- or siNS-treated cells. Cellular Gag levels were diminished to a similar extent in si55/63 conditions (Fig. 9A). si55/63 treatment led to a several-fold increase in the abundance of virus-associated Staufen1^63kDa as shown in western blots for Staufen1 (Fig. 9A). Actin levels in the viruses were constant regardless of treatment. The presence and absence of actin and heterogeneous nuclear (hn)RNP A2/B1 in purified virus served as positive and negative virion-associated protein controls, respectively (Liu et al., 1999; Mouland et al., 2001). In an independent experiment, the abundance of ribosomal L7 and the major nonsense-mediated decay protein, Upf1, two proteins that have been identified in Staufen1 complexes (Brendel et al., 2004; Kim et al., 2005) and the HIV-1 RNP (Ajamian et al., 2008), were also increased in highly purified virus from Staufen1-depleted cells (Fig. 9A, right panel). These results show that Staufen1 depletion induces a selective salvaging of the Staufen1^63kDa isoform in virus.

We also depleted Staufen1 by shSTF during the expression of a protease-negative provirus. In this experiment, Staufen1 expression was depleted by 85% (Fig. 9B). The depletion of both Staufen1 isoforms again promoted the preferential encapsidation of Staufen1^63kDa in virus (1.9-fold increase; Fig. 9B). Western blot analyses for virion-associated Gag (pr55^Gag) demonstrated equal loading of virus. Levels of Staufen1^55kDa detected in viral lysates did not markedly change and the relative abundance of the two Staufen1 isoforms in cells and virus was reversed. This result indicates that Staufen1^63kDa is available for encapsidation and that when Staufen1 expression is compromised, it is salvaged, like the vRNA, into de novo synthesized virus particles, probably via the formation of SHRNPs in these conditions.

In mammalian cells, Staufen proteins are found in RNPs of molecular weight 440-670 kDa that are thought to represent RNA trafficking granules and to be involved in the repositioning and translational activity of polysome-bound mRNAs (Brendel et al., 2004; Kohrmann et al., 1999; Mallardo et al., 2003). Our earlier observations that indicated a relationship between Staufen1 and the selection of vRNA by Gag are followed up in this study in which we now provide evidence for a role for Staufen1 in vRNA encapsidation. First, we demonstrate that HIV-1 recruits Staufen1 to form distinct, intracellular HIV-1-dependent SHRNPs and that these appear as larger foci in intracellular domains when Staufen1 is in limiting supply. Second, the striking correlation between the formation of larger, visible SHRNP foci and an increased abundance of vRNA in virions suggests that SHRNPs might represent vRNA-containing scaffolds through which vRNA is trafficked and encapsidated into assembling HIV-1. Staufen1-containing RNPs form with or without the presence of Gag and are bona fide RNPs rather than preformed viral particles.

The colocalization of Staufen1 with vRNA and Gag, and the increase in the apparent size of SHRNPs that correlated with vRNA encapsidation levels (Figs 1 and 8), both suggest that Staufen1 is a relevant host factor during the selection of HIV-1 vRNA during assembly. The observed changes in size and content of SHRNPs are probably due to the accumulation of vRNA, whose fate and trafficking depend on Staufen1. A notable analogy can be drawn: in cases in which decapping enzymes have been depleted, small and poorly visible PBs become supraphysiologic in size and accumulate RNA substrates (Franks and Lykke-Andersen, 2007). Whether the increased size of SHRNPs in Staufen1-depleted cells represents an increase in size of the Staufen1 RNP or represents the coalescing of multiple SHRNPs remains to be determined. However, when we observe SHRNPs, HIV-1 expression alone can induce compositional changes to Staufen1 RNPs in cells (Ajamian et al., 2008), and modulation of Staufen1 levels enhances insoluble Gag complexes (Chatel-Chaix et al., 2007). SHRNP localization was also found to be partly coincident with the perinuclear region where vRNA and Gag have been shown to physically interact (Poole et al., 2005) and to which targeting of vRNA has been shown to lead to enhanced vRNA encapsidation (Levesque et al., 2006). From a mechanistic standpoint, the formation of larger RNPs could explain the deregulation of assembly and enhanced encapsidation of vRNA, Staufen1^63kDa and other Staufen1-binding proteins found in the HIV-1 RNP such as Upf1 (Ajamian et al., 2008). In addition, it could explain the deleterious effects we have demonstrated on infectivity (Chatel-Chaix et al., 2004). This is not without precedence. For example, Resh and colleagues showed that Gag transits through intracellular bodies before being trafficked to the plasma membrane (Martinez et al., 2008), and the selective knockdown of cellular RNA-binding proteins blocks HIV-1 RNA transport at distinct intracellular loci (Levesque et al., 2006). This type of selectivity of a Staufen RNP was also found in Drosophila in the Sponge-Body (Snee and Macdonald, 2009). Moreover, the ‘T-body’ RNP harbours the Ty1 retrotransposon genomic RNA and a Gag-like protein and was suggested to serve as a scaffold for viral RNA trafficking and assembly (Malagon and Jensen, 2008). Furthermore, large intracellular RNPs form during flavivirus replication and these contribute to genomic RNA encapsidation (Emara and Brinton, 2007). Other RNPs that contain Staufen are relevant for RNA trafficking and for virus restriction (Chiu et al., 2006; Gallois-Montbrun et al., 2007; Kozak et al., 2006; Menager et al., 2009). The SHRNP probably represents an RNP that is specifically engineered by HIV-1 to constitute an intermediary compartment to which vRNA is trafficked for selection by Gag. Further study of the SHRNP will be important in order to understand in what context it is found in cells and to develop a more complete understanding of the fate of the genomic RNA during HIV-1 assembly.

Virus infection can induce a cellular stress response and the formation of SGs (Gale et al., 2000; McInerney et al., 2005) but, conversely, several viruses have developed mechanisms to counteract cellular stress responses that can impose severe limitations to viral replication. Although both SG and SHRNP assembly depend on active translation, the SHRNPs that we have identified after Staufen1 depletion as well as in cells with normal levels of Staufen1 are distinct from the known Staufen1-containing SGs identified in earlier work (Thomas et al., 2005) (supplementary material Fig. S2). Pur, however, leads to the remodelling of SHRNPs by displacing Gag from Staufen1-vRNA-containing RNPs, thus demonstrating that SHRNPs are plastic in nature as are other Staufen-containing RNPs (Snee and Macdonald, 2009). Knockdown of Staufen1 enhances SG formation in murine cells but SGs in this particular case are devoid of Staufen1, leading to the notion that this host protein monitors stress responses in cells without being crucial to SG assembly (Thomas et al., 2009). This result perhaps parallels what is actually occurring in HIV-1-expressing cells in which the assembly of SHRNPs is favoured over that of SG. The Staufen1-, vRNA- and Gag-containing SHRNPs did not contain PABP1, indicating the preference towards SHRNP formation even when oxidative stress was induced by Ars in HIV-1-producing cells (Figs 4, 5 and supplementary material Fig. S1). Consistently, HIV-1 protease cleaves PABP1 and this could provide further substantive evidence that HIV-1 needs to deactivate PABP1 function (Alvarez et al., 2006) to favour the assembly of SHRNPs during the late stages of replication.

How and why HIV-1 dramatically impairs the assembly of Ars-induced SGs (Fig. 4, 5 and supplementary material Fig. S1) but not that of Pur-induced SGs remain important questions. Ars and Pur elicit the formation of SG by distinct mechanisms. For example, Pur induces SG by disassembling polysomes whereas Ars induces SGs by eliciting eIF2α phosphorylation to result in a block to the recycling of the translation ternary complex (Farny et al., 2009; Kedersha et al., 2000). Because HIV-1 expression did not promote the phosphorylation of eIF2α in basal conditions, nor did it impair Ars-induced eIF2α phosphorylation (Fig. 4), the inhibition of SGs is probably not due to a block to eIF2α phosphorylation as reported for other viruses (Gale et al., 2000; Schneider and Mohr, 2003). Future work will help unravel how HIV-1 blocks SG assembly and whether other retroviruses have the same effect. The phenotype we observed in this study is similar to a stress-resistant phenotype characterized for Semliki Forest Virus (McInerney et al., 2005). Likewise, HIV-1 has evolved strategies to prevent a stress response that might lead to apoptosis and cell death (Dayton, 2008). We propose that SG formation is blocked by HIV-1 to prevent deleterious effects on viral replication and this might be achieved by preventing the recruitment of key components required for SG assembly.

Given that polysomes, SGs and PBs are believed to be in a dynamic equilibrium (Eulalio et al., 2007; Kedersha et al., 2000) and share several components [Staufen1 for instance (Kedersha et al., 2005)], we examined the relationship between SHRNPs and PBs, the latter representing cytoplasmic RNPs involved in mRNA metabolism and storage (Bruno and Wilkinson, 2006). Although it was only our intent to rule out that SHRNPs were not PBs, in this manuscript we are able to report two novel and salient findings. First, vRNA, Gag and even Staufen1 were not detectable in PB foci (Fig. 5; and data not shown) and, second, a dramatic decrease in the abundance of PBs was observed in HIV-1-expressing cells and often, PBs were not detectable (Fig. 5). The decrease in size and abundance of PBs has also been noted when expression levels of other cellular proteins are modified or when the translational apparatus is stabilized (Eulalio et al., 2007). Likewise, Staufen1-containing APOBEC3G complexes that could restrict HIV-1 do not cofractionate with PB components (Chiu et al., 2006), and intimations from recent work suggest that PBs might be deleterious to HIV-1 replication (Nathans et al., 2009). We speculate that HIV-1 expression suppresses PB formation or promotes their disassembly by sequestration and/or relocalization of cellular proteins involved in Ars-induced SG and PB assembly or integrity. Recent findings support this idea in that a correlation between the expression of heat-shock protein 72 and SG and PB assembly was drawn (Mazroui et al., 2007). Interesting parallels can also be drawn between our results and those of others in which SGs and PBs cannot form in flavivirus-infected cells due in part to interaction between viral and SG components (Emara and Brinton, 2007). HIV-1 has probably evolved strategies to protect the vRNA from translational silencing and/or degradation in PBs to allow for Gag expression and vRNA selection for encapsidation. HIV-1 inhibits siRNA and miRNA silencing pathways (Bennasser et al., 2005; Triboulet et al., 2007) that actually prevent PB formation (Eulalio et al., 2007), supporting our observation that HIV-1 co-opts RNA degradation machineries to favour viral expression (Ajamian et al., 2008).

The RNA trafficking function of Staufen1 relies on its association with a dimeric RNA in the context of an RNP in Drosophila (Ferrandon et al., 1997), although we could not find a link between vRNA dimerization and encapsidation levels (Fig. 8). Because dimerization and encapsidation are coupled processes in HIV-1 (Paillart et al., 2004), it is possible that Staufen1 influences encapsidation via its association to dimeric vRNA. Indeed, MLV genomic RNA selection is influenced by the capacity of Gag to bind vRNA (D'Souza and Summers, 2004). In line with this notion is a recent report that shows that the anti-viral effect of APOBEC3G is controlled by its capacity to bind viral RNA (Soros et al., 2007).

The depletion of Staufen1 led to a decrease in cellular Gag gene expression but this was not due to changes in steady-state vRNA levels (Fig. 8) or Gag protein stability (data not shown). Staufen1 might therefore act as a molecular switch to arrest translation and to favour vRNA encapsidation, a proposed role for EF1α described in an earlier study (Cimarelli and Luban, 1999). An effect on vRNA translation might rely on the ability of Staufen1 to associate to vRNA, because we show that in the presence and absence of Gag and during stress conditions, Staufen1 remains associated with vRNA (Figs 1, 2, 3, 4, 7; and data not shown). Alternatively, the close association of Staufen1 to Gag could play a role (Chatel-Chaix et al., 2004). Support for this idea is found in work that shows that Gag levels control the equilibrium between vRNA translation and encapsidation via translational inhibition of vRNA for HIV-1 (Anderson and Lever, 2006) and for the Idefx retrotransposon (Meignin et al., 2003). The dynamic link that has been established between translationally silent PBs and Staufen1 RNPs (like that between PBs and SG) (Zeitelhofer et al., 2008) supports the implication of Staufen1 RNPs in translational derepression. The formation of an HIV-1-specific RNP harbouring the components important for vRNA translation and encapsidation is an attractive idea. It will be interesting to determine the compositional differences between these RNPs and the efficiency with which vRNA is translated in SHRNPs.

To our knowledge, few studies have attempted to differentiate between the functions of the Staufen1 isoforms, although evidence indicates that they might be isoform-specific (Duchaine et al., 2000; Duchaine et al., 2002; Macchi et al., 2004). The selective depletion of the larger Staufen1^63kDa had little effect on Gag expression levels, on the formation of large SHRNPs or on vRNA encapsidation levels (Figs 1, 8; and data not shown). This isoform is generally less abundant than Staufen1^55kDa and a more-subtle phenotype might not be detected using our methods. Conversely, our siRNA and shRNA experiments also show that Staufen1^63kDa is preferentially recruited into HIV-1 particles when Staufen1 is depleted in cells. In light of the reproducibility of these results using several knockdown strategies, future work will be needed to understand how and why Staufen1^63kDa is preferentially salvaged into virions and the functional relationships that exist between the various protein isoforms.

This work leads to a model in which distinct RNPs form that contain vRNA, Gag, Staufen1 and, probably, other cellular proteins. However, very much like cellular silencing foci (Fig. 10A), in HIV-1 expressing cells these so-called SHRNPs are also in equilibrium with translating polysomes, because both CHX and Pur affect their assembly. As is the case for SGs, apparent SHRNP size is enhanced by limiting Staufen1 levels (Fig. 10B), probably by mechanisms that involve the stabilization of polysomes. Although SHRNPs could be considered as HIV-1-specific silencing foci, their formation would serve as a scaffold for efficient packaging of the two vRNA molecules selected per virus particle. The inhibition of cellular silencing foci including SGs and PBs is consistent with the ability of HIV-1 to titrate key cellular components in order to assemble functional, virus-specific SHRNPs.

Transfection of HEK293T and HeLa cells and details on the proviral constructs (HxBRU, pNL4-XX) and siNS and siRNA sequences were described previously (Ajamian et al., 2008; Chatel-Chaix et al., 2004; Mouland et al., 2000). siRNA duplexes and shRNAs were used to knockdown Staufen1 gene expression. The Staufen1 RNA is alternatively spliced to generate four transcripts. Transcript variant 3 (T3) is the longest and encodes the Staufen1^63kDa that contains an additional 81 amino acids in the N-terminus. The transcripts T1, T2 and T4 encode Staufen1^55kDa (Fig. 1C). si55/63 (Chatel-Chaix et al., 2004) and si63 (5′-TGAATGGGTCTACCTGCATTT-3′) were synthesized (Qiagen-Xeragon) and used at 10 nM. We also performed parallel experiments using additional siRNAs targeted against different regions of Staufen1 mRNA, and an alternative strategy to knockdown Staufen1 by shRNA. Silencing and non-silencing shRNA (shSTF and shNS) were expressed from an RNA polymerase-II-driven DNA expression vector, as described (Chatel-Chaix et al., 2007). The similarity of results with the multiple siRNAs utilized increases the likelihood that the observed phenotypes were the result of Staufen1 knockdown and not due to off-target effects (Scacheri et al., 2004). Gag expression was rescued in some experiments using a Rev-dependent Gag expressor (from Jaisri Lingappa, University of Washington, Seattle, WA) (Lingappa et al., 2006). Pur and CHX (Sigma-Aldrich) were used at 0.25 mg/ml for 4 hours before fixation. HeLa cells were incubated for 30 minutes with Ars (0.5 mM) at 37°C, washed and incubated for additional 30 minutes at 37°C in fresh medium and then fixed.

At 12-48 hours post-transfection, cells were lysed and equal amounts of protein from cleared post-nuclear supernatant (40 or 50 μg protein) were subjected to SDS-PAGE and western blotting as described (Chatel-Chaix et al., 2004).

IF and combined two-colour IF/FISH analyses were performed as described (Beriault et al., 2004; Levesque et al., 2006). LSCM and image analysis were performed using a Zeiss Pascal LSM5. LSM Image Browser software (Carl Zeiss, Germany) was used to estimate the size of SHRNPs. At least 50 SHRNPs were measured for each condition in at least three experiments. vRNA-containing RNPs that were bigger than 2 μm were excluded from the statistical analysis. PB numbers were calculated by counting DCP-1-positive foci in at least 50 cells from the 10-15 optical fields for each experimental condition. 6× magnifications (Fig. 7) were acquired by LSCM. For other images, higher pixel resolution images to reproduce more detail and subtle colour transitions were obtained by increasing the number of pixels for selected regions in Adobe Photoshop (Adobe). For combined IF/FISH analyses our initial protocol was modified as described (Lehmann et al., 2009): after FISH with a digoxigenin-labelled probe, vRNA was visualized by staining with biotinylated anti-digoxigenin (Sigma-Aldrich), following by anti-biotin monoclonal antibody conjugated with Alexa Fluor 488 (Invitrogen). The vRNA was visualized in a colour of choice (green or red) due to the Alexa Fluor moiety on this secondary antibody. LSCM was performed at 1024×1024 pixel resolution using the same equipment and settings. Colocalization analyses were performed using Manders' coefficient as described (Ajamian et al., 2008) from at least 15 cells per experimental condition in three experiments. Intensity plots were generated by Imaris software (Bitplane) in the measurement pane through the centre focal plane, as described elsewhere (Lehmann et al., 2009).

A rabbit anti-Staufen1 RLS1 antibody, that recognizes both Staufen1 isoforms was used in IF and western analyses (Thomas et al., 2005). This does not react against Staufen2 in immunofluorescence or western blot analyses. Antibodies to Gag (CA/p24, pr55^Gag), MA/p17 polyclonal sheep serum (NIH AIDS Reference and Reagent Program from Michael Phelan), GAPDH, hnRNP A2/B1, anti-tubulin and calnexin are described elsewhere (Beriault et al., 2004; Chatel-Chaix et al., 2004; Levesque et al., 2006; Mouland et al., 2000). Anti-Upf1 and anti-DCP1 were provided by Jens Lykke-Anderson (University of Colorado, Boulder, CO). Anti-PABP1 was provided by Nahum Sonenberg (McGill University, Montreal, Canada), anti-HuR and anti-Ras-GAP SH3 domain binding protein (G3BP) were provided by Imed Gallouzi (McGill University, Montreal, Canada) and anti eIF3η was purchased from Santa Cruz Biotechnology. Anti-actin, anti-eIF2α-P, anti-pan-eIF2α were purchased from Abcam; anti-L7, anti-EF1α, anti-PABP1 and anti-GAPDH were purchased from Novus Biologicals, Upstate, Sigma-Aldrich and TechniScience, respectively. A goat anti-gp120 (Env) was obtained from the NIH (#385, from Kathelyn Steimer). The secondary antibodies were anti-rabbit-Alexa-Fluor-594, anti-mouse-Alexa-Fluor-488, -594 or -647 and anti-sheep-Alexa-Fluor-647 (Invitrogen).

Dimerization analysis was performed using RNA extracted from 5 μg p24-equivalents virus as described (Laughrea and Jette, 1997).

Total RNA or viral RNA was extracted by Trizol reagent and subjected to SuperScript One-Step RT-PCR (Beriault et al., 2004; Chatel-Chaix et al., 2004). Virus was quantified by p24-ELISA and equal amounts of virus were used in subsequent RNA extraction. RNA was treated with DNAse I, repurified and used in RT-PCR. Staufen1 primers were designed to amplify portions of the T3 and T2 transcripts to generate 540 bp and 270 bp PCR products, respectively: sense primer (nt 140-169) and antisense primer (nt 670-699) of the Staufen1 cDNA (NM017453). gapdh RNA was quantified by RT-PCR as described (Levesque et al., 2006).

HeLa cells were mock-transfected with FLAG epitope-tagged Staufen1 expressor (Staufen1-FLAG; (Chatel-Chaix et al., 2008) or transfected with proviral DNA (pNL4-3) and Staufen1-FLAG. At 30-36 hours post-transfection, total cell lysates were collected in a Nonidet P40 lysis buffer (Chatel-Chaix et al., 2007). An aliquot representing 3% of the cell lysate (input) was used for western blotting for Staufen1-FLAG, Gag and GAPDH. 1 mg protein was immunoprecipitated with FLAG-agarose beads (Sigma-Aldrich) and then eluted with excess FLAG peptide (Sigma-Aldrich) according to the manufacturer's instructions. 10% of this eluate was assessed by western blotting to measure the recovery of Staufen1-FLAG and then Gag was immunoprecipitated from the first eluate with affinity-purified mouse anti-p24 monoclonal antibody (Ajamian et al., 2008). The immune complex was then eluted by 0.1 M glycine-HCl (pH 2.5). The RNA was purified using Trizol LS (Invitrogen). 10% of each immunoprecipitate was collected for western blotting analysis. 1 μg of input total cellular RNA and 50% of that isolated from the immunoprecipitate were used for RT-PCR. RT-PCR was performed for vRNA and gapdh mRNA as described above.

We thank Imed Gallouzi for critical reading of the manuscript and helpful discussions and Jens Lykke-Andersen, Nahum Sonenberg, Imed Gallouzi, William Rigby, Koren Mann, David Ott, Jaisri Lingappa, Kathelyn Steimer, Michael Phelan and the NIH AIDS Reference and Reagent Program for reagents and antibodies. Anne Monette and Lara Ajamian are recipients of Canadian Institutes of Health Research (CIHR) Doctoral fellowships and A.J.M. was supported by a CIHR New Investigator and FRSQ Chercheur-boursier Senior career awards. This work was supported by grants from the Canadian Foundation for Innovation and the CIHR (MOP-38111, MOP-56794, OPC-83178) to A.J.M.