Inhibition of host translation by virus infection in vivo

Abstract

Infection of cultured cells with lytic animal viruses often results in the selective inhibition of host protein synthesis, whereas viral mRNA is efficiently translated under these circumstances. This phenomenon, known as “shut off,” has been well described at the molecular level for some viruses, but there is not yet any direct or indirect evidence supporting the idea that it also should operate in animals infected with viruses. To address this issue, we constructed recombinant Sindbis virus (SV)-expressing reporter mRNA, the translation of which is sensitive or resistant to virus-induced shut off. As found in cultured cells, replication of SV in mouse brain was associated with a strong phosphorylation of eukaryotic initiation factor (eIF2) that prevented translation of reporter mRNA (luciferase and EGFP). Translation of these reporters was restored in vitro, in vivo, and ex vivo when a viral RNA structure, termed downstream hairpin loop, present in viral 26S mRNA, was placed at the 5′ end of reporter mRNAs. By comparing the expression of shut off-sensitive and -resistant reporters, we unequivocally concluded that replication of SV in animal tissues is associated with a profound inhibition of nonviral mRNA translation. A strategy as simple as that followed here might be applicable to other viruses to evaluate their interference on host translation in infected animals.

The interference of animal viruses with host translation was first documented in the 1960s in human fibroblasts infected with poliovirus (1) Further studies revealed that the halt of host translation (or “shut off”) was a general phenomenon observed in cells infected with lytic RNA and DNA viruses (2–6). Of the three steps in protein synthesis, viruses mainly affect the initiation step by hijacking or modifying the activity of key eukaryotic initiation factors (eIFs) to ensure an efficient translation of viral mRNAs and the simultaneous decline of host translation (5–8). The main targets of viruses are components of the cap-binding complex (eIF4F) that are required for the recruitment of ribosomes to mRNAs. Thus, some picornaviruses (e.g., poliovirus and rhinovirus) and lentiviruses (e.g., HIV-1) express proteases that proteolyze the eIF4G (the largest component of eIF4F) and polyA binding protein to dismantle cellular cap-dependent translation, whereas viral translation continues by the presence of internal ribosome entry site elements in viral mRNA that allow the recruitment of ribosomes in a cap-independent manner (2, 9–14). Other picornaviruses (e.g., encyphalomyocarditis virus), rhabdoviruses (vesicular stomatitis virus or VSV), and adenovirus decrease the activity of eIF4E (the cap-binding protein of eIF4F complex) by promoting its dephosphorylation and activation of its inhibitors 4E-BP1 and -2 (15–17). In other cases, such as rotavirus and influenza, viral proteins hijack eIF4G to redirect it toward viral factories where viral mRNA is being translated (18, 19).

Another important point of translation control in infected cells relies on the activity of eIF2 that brings the initiator Met-tRNA to 40S ribosomes (20–23). In response to viral infection, host dsRNA-activated kinase (PKR) phosphorylates eIF2 in an attempt to block general translation (both cellular and viral) (24–27). However, most viruses avoid this by expressing products that prevent the activation of PKR in infected cells (reviewed in ref. 28). A remarkable exception to this are the members of alphavirus group (Sindbis and Semliki Forest virus). Thus, complete phosphorylation of eIF2 was found in cultured mouse fibroblasts infected with these viruses, so that only viral mRNA is translated under these circumstances (22, 29). The presence of a secondary structure, termed downstream hairpin loop (DLP) and located 27 nts downstream from the initiator AUG in viral 26S transcripts, allows this mRNA to be translated by an eIF2-independent mechanism in infected cells (22, 30).

The shut-off phenomenon has been extensively studied in cultured cells, but not in infected animals, so that evidences for virus-induced shut off in vivo are still lacking. In vitro experiments often require somewhat artifactual conditions, such as the use of highly susceptible cell lines to virus and high multiplicity of infection, two conditions that are difficult to find during virus replication in animals. Moreover, the analysis of virus-induced shut off in vivo raises technical difficulties to measure de novo protein synthesis in a single animal's cell infected with the virus. Many viruses also interfere with other steps of host gene expression, such as transcription, mRNA transport, or stability (3, 29, 31–35), making it very difficult to attribute a reduction in the synthesis of host proteins (or of a given host protein used as reporter) to the sole effect on translation.

We show here that, as occurs in vitro, replication of Sindbis virus (SV) in mouse brains is linked to a phosphorylation of eIF2 that was detected in infected neurons. We show indirect, but strong evidence that shut off of host translation also occurs in animals infected with viruses.

Results

Engineered Reporter mRNAs That Mimic Translation of Cellular and Viral mRNA in SV-Infected Cells.

We previously reported that as consequence of PKR-induced eIF2 phosphorylation, only viral 26S mRNA that bears DLP structure is translated in alphavirus-infected cells (Fig.1A and see ref. 22). Like the rest of cellular mRNAs, heterologous mRNA expressed from a recombinant SV was no longer translated in infected mouse fibroblasts because of eIF2 phosphorylation (Fig. 1 C and F and see ref. 22). Translation of reporter mRNA was easily restored when viral DLP (90 nts in length) structure was placed at the 5′ end of a coding sequence of these mRNAs, allowing a translation as efficient as that of viral 26S mRNA (Fig. 1E). We reasoned that, in terms of translation in SV-infected cells, these reporter mRNAs lacking or containing DLP structures might behave as bona fide cellular and viral mRNAs, respectively. Moreover, because these reporter mRNAs are transcribed from a viral promotor, differential expression should reflect exclusively differences in the rate of translation. Thus, virus-induced shut off in vivo could be easily inferred by comparing the expression of reporter genes in animals infected with these two types of viruses (SV-reporter vs. SV-DLP reporter). To validate our experimental approach, we first carried out a detailed characterization of recombinant viruses in cultured cells (Fig.1B). The synthesis of luciferase and EFGP was strongly inhibited in 3T3 cells infected with recombinant SV-Luc and SV-EGFP, respectively, as compared with their counterparts that bear the viral DLP structure (SV-DLP Luc and SV-DLP EGFP). No differences in expression were detected in PKR^o/o cells, showing that phosphorylation of eIF2 hampered translation of nonviral mRNAs as described before (22). Next, we quantified the extent of translational exclusion of EGFP mRNA in SV-infected cells compared with cellular mRNAs (e.g., β-actin). Thus, we metabolicaly labeled infected cells with [³⁵S]-Met/Cys to analyze de novo protein synthesis of EGFP, virus capsid protein (SV C) and cellular β-actin. In parallel, we also analyzed the steady-state levels of their corresponding mRNAs by northern-blot (Fig. 1E). Infection with all recombinant SV viruses induced a strong inhibition of β-actin synthesis (>90%) without affecting the levels of its corresponding mRNA in a significant way (Fig. 1E). Similar amounts of EGFP and DLP-EGFP mRNAs accumulated in infected cells, but only DLP-EGFP mRNA was translated. Thus, translation of both EGFP and β-actin mRNAs was inhibited to a similar extent (>90%) in infected cells. In contrast, we estimated that translation of DLP-EGFP mRNA was comparable to 26S mRNA that encodes the structural proteins of virus (including C protein). Translation of EGFP, however, was almost completely abrogated when the DLP structure was disrupted by point mutations that destroyed its secondary structure in SV-ΔDLP EGFP virus, showing that DLP was essential for translational resistance to eIF2 phosphorylation.

Fig. 1.

In vitro characterization of recombinant SV expressing shut off-sensitive and -resistant reporter mRNAs. (A) Flowchart showing the main translational alteration in SV-infected cells (see text for details). (B) Schematic diagram of recombinant SV expressing reporter mRNAs. The genomic organization of SV RNA is shown, including the natural and duplicate subgenomic promotors (blue) that drive the synthesis of 26S mRNA encoding the viral structural proteins and the reporter mRNAs (luciferase or EGFP), respectively. Arrows show the transcription start site from each promotor. A DLP structure included in the first 90 nts of the 26S mRNA coding sequence was also placed in the indicated reporter mRNAs. In SV-ΔDLP EGFP the secondary structure of DLP was disrupted by point mutations as described previously (22). (C) Western-blot analysis of recombinant SV in wild-type (PKR^+/+) and PKR^o/o 3T3 cells. Cells were infected with the indicated virus at a multiplicity of infection of 25 pfu per cell and analyzed at 6 h postinfection by Western blot with the indicated antibodies. Note that the placement of 90 nts of the coding sequence of the C protein that includes the DLP increased the size of EGFP and delayed its electrophoretic mobility. (D) Immunofluoresence (IF) of SV expressing the indicated versions of EGFP in wild-type 3T3 cells. Micrographs were taken at 6 h postinfection. (E) De novo translation of cellular and viral-expressing mRNA in infected cells. Cells were infected with the indicated virus and labeled with [³⁵S]Met/Cys at 5.5 h postinfection for 30 min. Proteins were analyzed by SDS/PAGE and autorradiography. Bands corresponding to β-actin, SV C, and EGFP were quantified by densitometry, corrected for the number of methionines and cysteines and expressed as percentage of control (mock for β-actin and DLP-EGFP for EGFP). Parallel infections were used for extracting total RNA and Northern blot analysis was performed against the indicated mRNAs. (F) Luciferase activiy of PKR^+/+ and PKR^o/o cells infected with the indicated viruses. Samples were analyzed at 6 h postinfection and luciferase activity was measured as described in Materials and Methods. The SD from three independent experiments is shown.

Shut Off Induced by SV Infection of Mouse Brain.

We next infected mice with recombinant viruses to compare the reporter activity in whole organs (luciferase) or in single-infected cells (EGFP). SV shows a marked neurotropism in mice, infecting neurons of the neocortex and hippocampal regions of the brain (36–38). Inoculation of mice with recombinant viruses by the intranasal route resulted in a rapid replication in brains over a period of 4 d postinfection, yielding 10⁶ to 10⁷ pfu per brain. First, we cryosectioned brains of infected animals and the resulting slices were subjected to immunofluorescence (IF) with anti-SV C and antiphospho eIF2α. At 3 d postinfection, viral antigens were detected in groups of neurons in anterior ventral regions of brain, as well in basolateral areas corresponding to the pyriform cortex (Fig. 2A). At 4 to 6 d postinfection, viral antigens were detected in areas of the somatosensorial and motor cortex, as well in the hippocampus, suggesting that virus entered via the olfatory bulb to further spread out to upper regions of the cortex. Interestingly, a prominent label of phospho-eIF2α was detected only in regions of virus replication in wild-type animals. We found that up to 90% of areas expressing viral antigens immunoreacted to phospho-eIF2α antibodies, showing that replication of SV in animals is also associated with inactivation of eIF2. No phospho-eIF2α staining was detected in PKR^o/o-infected animals, showing that PKR quinase is also responsable for eIF2 phosphorylation during SV replication in vivo.

Fig. 2.

Phosphorylation of eIF2 in mice infected with SV and inhibition of nonviral translation. (A) Representative IF micrographs of coronal brain sections from wild-type and PKR^o/o mice infected with SV at 3 d postinfection. Adjacent sections were incubated with anti-SV C or anti-phosphoeIF2α antibodies. Of the 238 replication foci scored from three wild-type-infected animals, 214 showed strong staining of phosphoeIF2α (89%) (Right), whereas no eIF2phosphorylation associated to SV replication was detected in PKR^o/o mice. No immunoreaction of antiphosphoeIF2α antibodies was detected away from replication foci in any wild-type mouse analyzed. (B) Expression of Luc, but not of DLP-Luc, was inhibited in brains of wild-type animals infected with recombinant viruses. Mice were infected with the indicated viruses and brain homogenates were prepared at the indicated times to quantify viral yields (Left) and luciferase activity (Right). (C) Translation of Luc mRNA was restored in PKR^o/o animals infected with SV-Luc.

Replication of SV-Luc and SV-DLP Luc in mice, measured by viral yields, was identical (Fig. 2B). However, only luciferase activity was detected in wild-type mice infected with SV-DLPLuc. These differences in luciferase activity between SV-luc and SV-DLP Luc were even more marked than in cultured cells (Fig. 1F), showing that translation of nonviral mRNA was severely impaired in mouse brain neurons of wild-type animals. As expected, no differences in luciferase activity were detected among SV-Luc and SV-DLP Luc viruses in PKR^o/o mice (Fig. 2C). We next analyzed the expression of EGFP in brain neurons of mice infected with SV-EGFP and SV-DLP EGFP at the peak of virus replication (3 d). Although replication of SV-EGFP and SV-DLP EGFP was indistinguishable as judged by IF staining of viral antigens, we found strong differences in EGFP expression among brains of animals infected with these two viruses (Fig. 3A). For SV-DLP EGFP, about 50% of cells that immunoreacted with anti-SV C antibodies expressed EGFP, whereas only 5 to 10% of cells infected with SV-EGFP showed detectable EGFP expression (Fig. 3). Moreover, the few cells expressing EGFP from SV-EGFP showed a fluorescence intensity lower than their counterparts infected with SV-DLP EGFP.

Fig. 3.

Inhibition of EGFP expression, but not of DLP-EGFP, in single neurons infected with recombinant virus in vivo and ex vivo. (A) Brains of infected animals were analyzed at 3 d postinfection for simultaneous EGFP fluorescence and anti-SV C reactivity. Representative micrographs with scale bars are shown. Eighty neurons expressing viral antigens from each virus were scored, and 32 of them showed EGFP fluorescence for SV-DLP EGFP virus (40%), whereas only 4 neurons infected with SV-EGFP showed green fluorescence (5%) (Lower). (B) SV replication and EGFP expression in rat hippocampal slices infected with the indicated virus and analyzed at 1 d postinfection. Samples were processed as described above and 372 neurons expressing viral antigens from SV- DLP EGFP and 1,098 from SV EGFP were scored for statistical analysis (Lower).

Shut Off also Operates ex Vivo.

Organotypical explants can be easily derived from rat hippocampus and maintained in culture for a variety of purposes, including electrophysiological studies (39, 40). Moreover, hippocampal slices can be transduced with nonreplicative derivates of Sindbis and Semliki Forest viruses for the expression of foreigner genes (41, 42). It was interesting, therefore, to test whether shut off also happened in explanted brain slices after in vitro infection with SV. Thus, slices were incubated with preparations of SV-EGFP or SV-DLP EGFP viruses (10⁴ pfu each) and analyzed by IF one day later. Both viruses spread rapidly throughout the explant, infecting an elevated number of neurons, most of them with a pyramidal shape. Notably, a dramatic difference in number and fluorescent intensity of neurons expressing EGFP was found among slices infected with SV-EGFP and SV-DLP EGFP (Fig. 3B). Virtually all cells infected with SV-DLP EGFP simultaneously expressed EGFP (94%), whereas only a very small proportion of neurons infected with SV-EGFP virus showed EGFP fluorescence (2%). Slices were also incubated with antiphosphoeIF2α antibodies, showing that ex vivo infection with SV also triggered eIF2α phosphorylation, as occurred in vitro and in vivo (Figs. S1 and S2).

Discussion

We are unique in showing that replication of a virus in an animal tissue resulted in the inactivation of a translation initiation factor (eIF2). Although we ourselves, as well as other investigators, had already reported a strong activation of PKR that led to a complete phosphorylation of eIF2 in cultured cells infected with SV and Semliki Forest virus, there was no experimental evidence supporting the idea that such an event happened in infected animals. A detailed examination of brains from infected animals revealed that virtually all groups of neurons expressing viral antigens also immunoreacted to antiphosphoeIF2α antibodies. This finding was notable and showed that replication of SV in animals is intimately linked to PKR-mediated eIF2 phosphorylation. Accordingly, PKR expression in mouse brain has been found particularly high in the neocortex and hippocampus (http://www.brain-map.org/), where SV replication was easily detected. Interestingly, phosphorylation of eIF2 in cortical neurons has been reported to occur during ischemic stress and other pathological situations, such as Alzheimer's and Huntington diseases (43–47).

By means of recombinant viruses expressing engineered reporter mRNAs, we present indirect but solid evidence that translation of nonviral mRNA is strongly inhibited in infected mouse-brain neurons. Virus-expressing reporter genes were used in earlier studies to track replication and spreading of the virus to different organs of infected animals (37, 41, 48). However, to date, the shut off phenomenon has not been addressed in vivo, probably because of the technical difficulties that such a study raises. Our approach is based on the assumption that translation of reporter mRNAs used here faithfully reflected translation of host and viral mRNAs in infected cells. All results obtained supported this. First, translation of EGFP and luciferase mRNA was inhibited to a similar extent as β-actin and the majority of cellular mRNAs in infected cells. Second, the placement of a DLP structure at the 5′ end of the EGFP coding sequence restored translation to a level comparable to that of translation in viral 26S mRNA. Third, the use of a viral promotor that drives the synthesis of reporter mRNA allowed direct measurement of the effect of virus replication on translation, obviating the perturbations that Sindbis and other viruses exert on cellular transcription (3, 31–33, 49). In fact, infection with alphavirus also results in a halt of host transcription that can be separated from the translation shut-off phenomenon (49). Despite this, we found that the steady-state levels of abundant host RNA, such as ribosomal or β-actin mRNA, did not significantly decrease at 6 h postinfection, when translation of host mRNA was completely inhibited (Fig.1).

A similar experimental strategy to that described here might be applicable to the study of shut off in other viruses, where the interference with host translational machinery has been well clarified. This approach requires, however, a previous knowledge of molecular tricks that allow the mRNA of a given virus to be translated in an environment of general translational inhibition. This process has been well described for picornavirus and roughly clarified for VSV, influenza, adenovirus, and rotavirus but not for others, such as poxvirus (6, 16, 17, 19, 50, 51). Thus, the low dependence of VSV, adenovirus, and influenza for the cap binding-protein eIF4E might be used to create reporter mRNA with different translational capabilities in cells infected with these viruses (52–55). A limitation of the strategy described here is that reporter genes should be placed under subgenomic promotors in RNA or DNA viruses to create a transcriptional independence, which excludes picornavirus and other viruses that initiates transcription exclusively from the end of the genomic strand.

The demonstration that shut off also takes place in vivo, at least for alphavirus, could have profound implications for a better understanding of virus-host interactions in infected animals. Moreover, the ability of viruses to block host translation might be critically regulating their pathogenic potential by preventing the synthesis of proteins with antiviral function, such as interferons and other inflammatory cytokines.

Finally, the influence of viral DLP on translation of mRNA in SV-infected cells could improve the expression of foreigner genes from SV-derived vectors, which are widely used to transduce primary neurons and organotypical explants of brain animals (41, 42).

Materials and Methods

Animals and Cell Lines.

Wild-type (Charles River) and PKR knock-out animals [kindly provided by J. C. Bell, University of Ottawa, Canada, (56)] from 129sv strain were used. Four-week-old females were infected by the intranasal route with 5 × 10⁶-10⁷ pfu of SV. 3T3 cells derived from wild-type and PKR^o/o animals (27) and BHK21 were grown in DMEM supplemented with bovine serum (3T3) and fetal serum (PKR^o/o and BHK21), as described previously (22). Mouse embryonic fibroblasts derived from 129sv mice were prepared following standard protocols (23).

All animal experiments were performed in compliance with the standards of our institution's bioethics and biosecurity committees and adhering to the “Protection of animals used for scientific purposes” (Spanish National Royal Decree RD 223/88).

Construction of Recombinant Viruses.

Recombinant viruses expressing luciferase or EGFP mRNAs were constructed in the pT7SV-2p plasmid, an infectious cDNA clone of the Sindbis virus that carries a second subgenomic promotor at the 3′ of genomic mRNA, and which has been designed to express foreigner genes (57). The construction of SV-EGFP has been described (22). For SV-Luc, the luciferase coding sequence was amplified by PCR with the following primers: 5′ Luc GGGCGCTAGCGGATCCA ATGGAAGACGCCAAAAAC and 3′ Luc: CGCCGCTAGCTTACAATTTGGACT TTCCGCC. PCR products were digested with NheI enzyme and cloned into the XbaI site of pT7SV-2p. For SV-DLP EGFP, we amplified by PCR a DNA fragment containing the DLP of SV fused in frame to the EGFP coding sequence from plasmid p5′CEGFP-N1 (22). The primers used were: 5′C SV GCGCGCTAGCATGAA TAGAGGATTC and 3′ EGFP CGCGCTCTAGATTACTTGTACAGCTCGTC. The resulting PCR product was cloned into the XbaI site of pT7SV-2p as described above. For SV-ΔDLP EGFP, the template for PCR amplification was p5′CΔDLP EGFP-N1 plasmid carrying point mutations in DLP region that disrupted the secondary structure of RNA as described before (22). For SV-DLP Luc, the PCR fragment of luciferase coding sequence described above was cloned into p5′CEGFP-N1 plasmid using BamHI and XbaI enzymes. Then, a PCR amplification was done using 5′C SV and 3′ Luc primers and the resulting fragment was cloned into pT7SV-2p plasmid as described before. All constructions were verified by sequencing. Infectious RNAs were generated in vitro by transcription with RNApol T7 and electropored in BHK21 cells, as described previously (22). Viruses were collected 2 to 3 d later when the cytophatic effect was massive and purified by ultracentrifugation [29 K for 4 h in an SW-28 Ti rotor (Beckman Coulter)] through a sucrose cushion at 4 °C. The resulting viral preparation showed titres of 5 × 10⁸-10⁹ pfu/mL and a high degree of genetic homogeneity (Fig. S1).

Immunofluorescence.

For IF of tissues, brains of infected mice were extracted 3 d postinfection and fixed overnight with 4% PFA at 4 °C and then hydrated with 30% sucrose for 48 h. Brains were cryosectioned at 15 μm, postfixed with PFA at room temperature for 15 min, and permeabelized with 0.2% Triton X-100 in PBS for 30 min. Sections were treated with ammonium chloride, blocked in 5% BSA-PBS and incubated overnight at 4 °C with primary antibodies: anti-C SV (1:300), antiphosphoeIF2α (Cell Signaling, 1:200). Sections were washed three times for 15 min with PBS-0, 1 Triton X-100, and incubated with secondary antibodies coupled to Alexa 555 or Alexa 594 for 1 h. Finally, sections were washed as described above, mounted, and photographed in a Leika confocal microscope. For IF of culture cells growing on coverslips, the protocol was similar, except that primary antibodies were incubated 2 h at room temperature. Western blot was carried out essentially as described previously (22).

Metabolic Labeling of Proteins.

Cells growing in 24-well plates were infected with a multiplicity of infection of 25 pfu per cell and 5.5 h later labeled with 25 μCi/mL of [³⁵S]-Met/Cys (20) for 30 min in medium lacking methionine. After washing with cold medium, monolayers were lysed in a sample buffer, boiled, and analyzed in a 12% SDS/PAGE followed by fluorography with 1 M salicylate solution and exposure to x-ray film.

Luciferase Assays.

Brains of mice infected with luciferase-expressing viruses were homogenated in PBS and extracted with 1 volume of 2× luciferase lysis buffer (KH₂PO₄ 15 mM, MgSO₄ 15 mM, EGTA 4 mM, DTT 4 mM and T-X100 1%). After centrifugation at 10,000 × g for 5 min; 20 μL of lysates were used to measure luciferase activity.

Infection of Organotypic Slices from Rat Hippocampus.

Hippocampal slices from 6-day-old rats were prepared as described (40) and maintained in culture for 1 week before infection with 10⁴ to 10⁵ pfu of the indicated virus. A 2-μL drop of virus preparation was applied on slices twice, and the drops were allowed to drain away between applications. Virus replication and spreading were checked every 24 h by living examination of EGFP fluorescence. IF analysis was identical to described above, except for incubations were done on floating sections.