|Home | About | Journals | Submit | Contact Us | Français|
The downregulation of translation through eIF2α phosphorylation is a cellular response to diverse stresses, including viral infection, and is mediated by the GCN2 kinase, protein kinase R (PKR), protein kinase-like endoplasmic reticulum kinase (PERK), and heme-regulated inhibitor kinase (HRI). Although PKR plays a major role in defense against viruses, other eIF2α kinases also may respond to viral infection and contribute to the shutdown of protein synthesis. Here we describe the recessive, loss-of-function mutation atchoum (atc) in Eif2ak4, encoding GCN2, which increased susceptibility to infection by the double-stranded DNA viruses mouse cytomegalovirus (MCMV) and human adenovirus. This mutation was identified by screening macrophages isolated from mice carrying N-ethyl-N-nitrosourea (ENU)-induced mutations. Cells from Eif2ak4atc/atc mice failed to phosphorylate eIF2α in response to MCMV. Importantly, homozygous Eif2ak4atc mice showed a modest increase in susceptibility to MCMV infection, demonstrating that translational arrest dependent on GCN2 contributes to the antiviral response in vivo.
The activation of innate immune sensors of viral infection, such as the nucleic acid-sensing Toll-like receptors and RIG-I-like receptors, induces the production of inflammatory cytokines and type I interferons (IFN). IFNs induce an antiviral state that depends on the expression of genes driving, in addition to innate and adaptive immune responses per se, cellular activities that promote an environment detrimental to virus survival and proliferation. For example, viruses depend on the host translational machinery to synthesize their proteins, and mammalian cells downregulate translation as one means of countering viral infection.
GCN2 kinase, protein kinase R (PKR), protein kinase-like endoplasmic reticulum kinase (PERK), and heme-regulated inhibitor kinase (HRI) phosphorylate the eukaryotic translation initiation factor eIF2α at conserved serine residue 51 (40). When phosphorylated, eIF2α binds and functionally sequesters the guanine nucleotide exchange factor eIF2B, thereby decreasing levels of GTP-bound eIF2 required for translation initiation (21, 32). Three of the four eIF2α kinases respond to distinct environmental stresses: PERK to misfolded proteins in the endoplasmic reticulum (ER stress) (14, 39), HRI to heme deprivation and oxidative and heat stresses in erythroid tissues (13, 23), and GCN2 to amino acid deprivation, UV irradiation, and proteasome inhibition (9, 19, 41, 44). The fourth eIF2α kinase, PKR, is induced by type I IFN and activated by double-stranded RNA (dsRNA) (10), which is derived from dsRNA viruses and synthesized as an intermediate during the replication of single-stranded RNA viruses and dsDNA viruses. Thus, eIF2α phosphorylation by PKR leads to a global block of protein synthesis that in turn hinders viral protein production. The existence of an eIF2α kinase specialized to sense and respond to viral infection underscores the effectiveness of this antiviral strategy. Indeed, many viruses have evolved countermeasures to PKR signaling (6, 15, 24, 29, 33, 37).
A few reports also have implicated eIF2α phosphorylation by PERK and GCN2 in antiviral responses. PERK is activated during herpes simplex virus 1 infection, likely as a result of accumulated viral proteins in the ER (7) and during vesicular stomatitis virus infections through an unknown mechanism (3). GCN2 is directly activated in vitro by the binding of the Sindbis virus genomic RNA to the histidyl-tRNA synthetase-like domain of GCN2, and GCN2-deficient mice infected with Sindbis virus display elevated viral titers in the brain relative to those of wild-type mice (4). Whether GCN2 or PERK is involved in defense against other viruses is not known.
We carried out a genetic screen to identify genes important for the early control of DNA viruses, represented by mouse cytomegalovirus (MCMV) and human adenovirus, or RNA viruses, represented by influenza. Using thioglycolate-elicited peritoneal macrophages from N-ethyl-N-nitrosourea (ENU)-mutagenized mice, we identified a mutant strain, atchoum (atc), with macrophages that became infected by MCMV and human adenovirus with increased frequency relative to that of wild-type macrophages. The phenotype was ascribed to a missense mutation of the gene encoding GCN2. The atc mutation abrogated the phosphorylation of eIF2α in response to MCMV infection. Moreover, the mutation increased susceptibility to a sublethal inoculum of MCMV in homozygous mice.
The Eif2ak4atchoum, Ifnar1m1Btlr (macro-1; 3822164), Ifnar2m1Btlr (macro-2; 3841008), Stat1m1Btlr (domino; 3619019), Tlr9m1Btlr (CpG1; 3038816), Unc93b13d (3619211), and Myd88poc (pococurante; 3641255) alleles were generated on a C57BL/6J background by N-ethyl-N-nitrosourea (11) and are described at http://mutagenetix.utsouthwestern.edu/ (Mouse Genome Informatics database accessions numbers are indicated in parentheses). Tlr3−/− and Tlr7−/− mice were from Richard Flavell (Yale University, New Haven, CT) and Shizuo Akira (Osaka University, Osaka, Japan), respectively. Ifnar1−/− and Ifngr1−/− mice were from Jonathan Sprent (Garvan Institute of Medical Research, Darlinghurst, New South Wales, Australia). C57BL/6J mice used for mutagenesis and C57BL/10J mice were from The Jackson Laboratory; all other C57BL/6J mice were from the breeding colony of The Scripps Research Institute (La Jolla, CA). All animals were housed in The Scripps Research Institute Animal Facility. All procedures using animals were in accordance with guidelines of the Institutional Animal Care and Use Committee.
Macrophages were induced in mice by the intraperitoneal (i.p.) injection of 1.5 to 2 ml 4% (wt/vol) Brewer's thioglycolate medium powder (BBL Microbiology Systems) in distilled water 4 days prior to isolation. Macrophages were collected by lavage, concentrated by centrifugation, and resuspended in PEC medium (5% [vol/vol] heat-inactivated fetal bovine serum [Atlanta Biologicals], 200 IU/ml penicillin, 200 mg/ml streptomycin in HEPES-buffered saline). Mouse embryonic fibroblasts were prepared from day 13.5 embryos using a standard protocol.
MCMV-green fluorescent protein (GFP) (27) was a gift from Chris Benedict (La Jolla Institute of Allergy and Immunology, La Jolla, CA). It was propagated on mouse embryonic fibroblasts and purified as described previously (22). MCMV used for in vivo infections was prepared and maintained as described elsewhere (http://mutagenetix.utsouthwestern.edu/protocol/protocol_rec.cfm?pid=5). The nonreplicating hAd5-F16-GFP is the human adenovirus 5 serotype expressing the fiber protein from serotype 16 and tagged with GFP (16). hAd5-F16-GFP was a gift from Glen Nemerow (The Scripps Research Institute) and was grown in 293 cells and purified by CsCl gradient. The titer was determined as described previously (16, 38). PR8 influenza A virus was prepared and maintained as described previously (http://mutagenetix.utsouthwestern.edu/protocol/protocol_rec.cfm?pid=18).
The infection of macrophages and the measurement of infected cells was performed as described previously (http://mutagenetix.utsouthwestern.edu/protocol/protocol_rec.cfm?pid=10). Briefly, viruses were added at the indicated dosages to 105 macrophages per well in a 96-well plate and incubated at 37°C with 5% CO2 in a humidified incubator for 24 h (MCMV-GFP or PR8) or 72 h (hAd5-F16-GFP). Mice were infected with MCMV by the intraperitoneal injection of the indicated dose. The number of MCMV PFU in culture supernatant of wild-type or Eif2ak4atc/atc fibroblasts was determined by plaque assay in NIH 3T3 cells as described previously (31).
The stimulation of macrophages with dsDNA and the measurement of type I IFN production was performed as described elsewhere (http://mutagenetix.utsouthwestern.edu/protocol/protocol_rec.cfm?pid=6). Briefly, 0.1 μg of dsDNA and 0.25 μl Lipofectamine 2000 (Invitrogen) in a total of 50 μl of Opti-MEM medium (Gibco) was added to 5 × 104 macrophages per well in a 96-well plate and incubated at 37°C with 5% CO2 in a humidified incubator for 16 h. Following incubation, the concentration of type I IFN in the supernatant was assayed using an interferon-stimulated response element (ISRE)-driven luciferase reporter bioassay. Type I IFN in the supernatant of hAd5-F16-GFP-infected macrophages was measured 72 h postinfection using the same protocol.
Bulk segregation analysis (BSA) was performed as described previously (42) using F2 mice grouped into mutant and wild-type groups based on the percentage of macrophages infected by hAd5-F16-GFP. DNA for whole-genome sequencing was prepared from two different atc homozygous mice as described previously (2), and SOLiD sequencing was performed using three slides according to SOLiD 3 (2 slides) or SOLiD 4 (1 slide) instructions provided by the manufacturer (Applied Biosystems). SOLiD data were analyzed as described previously (2).
Mice were genotyped by sequencing PCR products, amplified from genomic DNA, across the particular mutation site using a 3730xl sequencer (Applied Biosystems). PCR primers were the following: Eif2ak4atc (forward, 5′-AATTGGCTGGGACGGTGTCAAG-3′; reverse, 5′-GGAAGCACTTTAAATGCTCGCCAC-3′); Ifnar1macro-1 (forward, 5′-AGAACAGCTTGCCACTTCACTGG-3′; reverse, 5′-GCAGAGAAGCCTTAGCCTTAGAAGAAC-3′); Ifnar2macro-2 (forward, 5′-TTGATACCACAGCGGAAGGTGAGC-3′; reverse, 5′-AACCATAGGCGGGACACATTAACTG-3′). Sequencing primers were the following: Eif2ak4atc (forward, 5′-GCTGTAACTTAGTGTACAGGGAC-3′; reverse, 5′-CATCGGTAGTGAGCACTCTACAG-3′); Ifnar1macro-1 (forward, 5′-CCCAGGGTAGCTTCAAACTTATG-3′; reverse, 5′-TCACAAAGTTCCTGGGTAGC-3′); Ifnar2macro-2 (forward, 5′-CCTCTCTGCTTAGAGGACAGATG-3′; reverse, 5′-CAAGGCCGTTTCCTGAATATG-3′).
Total RNA was prepared from thioglycolate-elicited peritoneal macrophages using TRIzol (Invitrogen). The RNA was reverse transcribed using a RETROscript first-strand synthesis kit (Ambion). Five products were amplified using the following PCR primers: forward, 5′-GAGAGCTATTCGCAGCGACAGG-3′; reverse, 5′-TGTCAAGCTAGTGATTTCCAAACGTTCC-3′. The PCR products were purified using a QIAquick PCR purification kit (Qiagen) and sequenced using BigDye terminator v3.1 on a 3730xl sequencer (Applied Biosystems) using the following primers: forward, 5′-ATTTACGGCTCGGACTTCCAG-3′; reverse, 5′-CTCTGAATCTCGTGGAGGATTTC-3′.
Macrophages or fibroblasts (105) were lysed in Laemmli sample buffer. Proteins were separated by SDS-PAGE and immunoblotted using the antibodies to GCN2 (3302; Cell Signaling Technology), eIF2α (9722; Cell Signaling Technology), eIF2α (phospho-Ser51) (9721; Cell Signaling Technology), and β-actin (sc-47778; Santa Cruz Biotechnology).
Age- and sex-matched mice were injected i.p. with 2 × 106 IU of a recombinant Semliki Forest virus (SFV) vector encoding the model antigen, β-galactosidase (rSFV-βGal), in a total volume of 200 μl of sterile 0.9% saline. Ten days later, the mice were injected i.p. with 50 mg of NP28-AECM-Ficoll (Biosearch Technologies) in a total volume of 200 μl of sterile 0.9% saline. On day 14 after immunization with rSFV-βGal (and day 5 after immunization with NP-Ficoll), mice were anesthetized with isoflurane and blood from the retro-orbital plexus of each animal was collected in serum separator tubes (Becton Dickinson). For the detection of specific antibodies, polyvinyl chloride microtiter 96-well round-bottom plates (Costar) were coated overnight at 4°C with 2 μg/ml β-galactosidase (Roche) or 5 μg/ml NP23-bovine serum albumin (Biosearch Technologies). Plates were washed in Dulbecco's phosphate-buffered saline (dPBS) without calcium and magnesium and blocked with 5% milk. Serum samples were serially diluted in 1% milk. Plates were incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgM or IgG (Southern Biotechnology) diluted in 1% milk. Plates were developed with SureBlue TMB microwell peroxidase substrate and TMB stop solution (KPL) and read at 450 nm on a MAXline Emax microplate reader (Molecular Devices). Background levels for the assay were determined by incubating sera pooled from immunized wild-type mice on uncoated wells.
We screened peritoneal macrophages isolated from third generation (G3) C57BL/6J mice carrying ENU-induced mutations for susceptibility to ex vivo infection by viruses (MCMV-GFP  or the mouse-adapted influenza virus strain PR8) or a nonreplicating viral vector (hAd5-F16-GFP ). The number of macrophages infected by MCMV or hAd5-F16 vector was monitored by flow cytometry to detect GFP-labeled cells; infection by PR8 was determined by the reactivity of macrophages to hemagglutinin (HA) antibody, which also was monitored by flow cytometry. After incubation with MCMV-GFP or PR8 for 24 h or with hAd5-F16-GFP for 72 h, the number of infected macrophages from C57BL/6J and G3 mice bearing ENU-induced mutations in homozygous and/or heterozygous form was analyzed.
To validate the effectiveness of the screen, we tested macrophages from mice with known mutations (Fig. 1A). Macrophages from homozygous Stat1domino (dom) (8) mice were highly permissive to infection by all three viruses, confirming the known requirement for type I IFN signaling in viral control (17, 28). In contrast, normal percentages of macrophages from Tlr3−/−, Tlr7−/−, Tlr9CpG1/CpG1 (34), Unc93b13d/3d (35), or Myd88poc/poc (20) mice were infected by MCMV-GFP, hAd5-F16-GFP, or PR8, indicating that redundant mechanisms for sensing these viruses exist in macrophages, or that requirements for sensing these viruses differ between macrophages and other cell types. IFN-γ signaling also was dispensable for the control of MCMV-GFP, hAd5-F16-GFP, or PR8 by macrophages, likely because macrophages are not a significant source of IFN-γ.
Macrophages from a total of 4,500 G3 mice were infected with MCMV-GFP, hAd5-F16-GFP, and PR8 (Fig. 1B). Two strains, designated macro-1 and macro-2, produced macrophages that were infected by all three viruses with increased frequency relative to that of C57BL/6J macrophages (Fig. 1B). Both the macro-1 and macro-2 phenotypes were transmitted recessively. Macrophages from a third strain, atc, were infected with increased frequency by hAd5-F16-GFP and MCMV-GFP (Fig. 1B). The permissiveness of atc macrophages to hAd5-F16-GFP was similar to that of macrophages from Stat1dom/dom mice, whereas permissiveness to MCMV-GFP was intermediate between wild-type and Stat1dom/dom levels. Infection by PR8 was equally well controlled by atc and C57BL/6J macrophages. The atc phenotype also was recessive. As another measure of viral control, the amount of type I IFN produced by the infected macrophages was measured in the culture supernatant. In response to dsDNA, type I IFN production by macro-2 macrophages was reduced compared to that produced by wild-type cells (Fig. 1C). In contrast, hAd5-F16-GFP induced normal type I IFN production by atc macrophages (Fig. 1C).
Because of the importance of type I IFN signaling to the innate antiviral response, the genes encoding STAT1 (Stat1), Isgf3g (Irf9), JAK1 (Jak1), Tyk2 (Tyk2), and the two chains of the type I IFN receptor (Ifnar1 and Ifnar2) were sequenced in macro-1 and macro-2 mice. An adenine-to-cytosine transversion in exon 8 of Ifnar1, causing the missense mutation T341P in IFNAR1, was identified in macro-1 mice (Fig. 2A). An adenine-to-guanine transition in exon 2 of Ifnar2, causing the replacement of the start methionine with valine, was found in macro-2 mice (Fig. 2B). The susceptibility of macro-1 and macro-2 macrophages to infection by MCMV, human adenovirus, and influenza is consistent with a deficiency of IFNAR signaling, as is the reduced dsDNA- and adenovirus-dependent type I IFN production, which is known to be regulated by a positive feedback loop in which IFN-β and IFN-α4 induce the expression of the transcription factors IRF7 and IRF8 (25, 36). The detection of IFNAR1 and IFNAR2 deficiencies validated the effectiveness of the screen.
The atc phenotype was distinct from those of mice with Ifnar1 or Ifnar2 deficiency, and therefore the mutation was mapped by bulk segregation analysis (BSA). Animals with the atc phenotype (C57BL/6J-atc) were outcrossed to C57BL/10J mice, and affected F1 mice were backcrossed to atc homozygotes (42). Using 40 F2 mice with mutant phenotypes and 9 with normal phenotypes, BSA indicated the strongest linkage of the atc mutation with a marker at position 103735349 on chromosome 2 (synthetic logarithm of odds [LOD], 6.0) (Fig. 3A). Among nucleotides covered at least once within the 38 Mb surrounding the marker with peak linkage (19 Mb upstream and downstream), 327 mutations were identified by SOLiD whole-genome sequencing, of which 281 were successfully reexamined by conventional capillary sequencing. A single mutation was validated at position 118226389, 14.5 Mb from the marker with peak linkage (Fig. 3B). The mutation corresponds to a thymine-to-cytosine transition of the sixth nucleotide of intron 2 of Eif2ak4, encoding GCN2 (Fig. 3C). The cDNA sequencing of the four largest Eif2ak4 transcripts demonstrated that the mutation invariably results in the skipping of exon 2, in some cases along with the skipping of other exons (Fig. 3D). The most abundant transcript lacked exons 2, 3, and 4. No GCN2 expression was detected in Eif2ak4atc/atc macrophages by immunoblotting (Fig. 3E).
We evaluated the function of the GCN2atc protein upon the stimulation of Eif2ak4atc/atc mouse embryonic fibroblasts with UV radiation, a known activator of GCN2 (9, 18). As expected, UVB resulted in the increased phosphorylation of eIF2α in C57BL/6J fibroblasts but not in Eif2ak4atc/atc fibroblasts (Fig. 3F), indicating that the atc mutation causes the complete loss of GCN2 function, at least where eIF2α phosphorylation is concerned.
To determine whether eIF2α phosphorylation by GCN2 is part of the antiviral response to DNA viruses, we compared eIF2α phosphorylation (eIF2α-P) levels by immunoblotting in C57BL/6J and Eif2ak4atc/atc fibroblasts 0, 8, 16, 24, 32, and 40 h after infection with MCMV. C57BL/6J fibroblasts displayed minimal eIF2α-P at each time point up to 32 h but increased eIF2α-P 40 h after MCMV infection. In contrast, no eIF2α-P was detectable in Eif2ak4atc/atc fibroblasts at any time point (Fig. 4A). In a separate experiment, we measured viral titers at the same time intervals in the culture supernatant of wild-type and Eif2ak4atc/atc fibroblasts. Supernatant from Eif2ak4atc/atc fibroblasts displayed titers similar to those of wild-type cell supernatant at all time points examined (Fig. 4B). These findings suggest that eIF2α phosphorylation by GCN2 is not required to prevent virus replication and shedding for at least 32 h after infection; an effect on viral titer may lag behind the observation of defective eIF2α phosphorylation in Eif2ak4atc/atc cells, which first occurs 40 h postinfection.
We then tested whether the deficiency in eIF2α-P affects susceptibility to MCMV in vivo. C57BL/6J and Eif2ak4atc/atc mice were injected intraperitoneally with 2 × 105 PFU/mouse of MCMV and observed for sickness. During the 2-week period following infection, 11 of 68 Eif2ak4atc/atc mice died, whereas all 39 of the C57BL/6J controls remained healthy (P = 0.0041) (Fig. 4C). Thus, GCN2 deficiency increases susceptibility to MCMV infection in mice as it does in cultured macrophages.
The deaths of Eif2ak4atc/atc mice in response to MCMV infection all occurred on or before 9 days postinfection, which is consistent with the hypothesis that susceptibility stems from a defect in the innate immune response. However, the expression of Eif2ak4 has been identified as part of a transcriptional signature highly correlated with the strength of the adaptive immune response, in particular the CD8+ T-cell response, to the yellow fever vaccine YF-17D (30). We therefore tested the antibody responses of Eif2ak4atc/atc mice but found that they mounted normal IgG responses to the T-dependent antigen β-galactosidase and normal IgM responses to the T-independent antigen NP-Ficoll (Fig. 4D). This finding, together with the fact that Rag1-deficient mice survive MCMV infection without incident for several weeks after inoculation, supports the interpretation that MCMV susceptibility in Eif2ak4atc/atc mice is caused by an innate immune defect.
Although the downregulation of protein synthesis as an antiviral response generally is attributed to PKR activation, eIF2α phosphorylation occurs in response to certain viruses even in the absence of PKR function (1, 12, 43). The responses of GCN2- or PERK-deficient mice to viral infections have been reported in only a few publications (3, 4, 7). By screening macrophages ex vivo, we identified an Eif2ak4 mutation that caused increased susceptibility to MCMV and human adenovirus infections. The Eif2ak4atc mutation is a single-nucleotide substitution at the sixth position of Eif2ak4 intron 2, which abrogated the splicing of exon 2 to exon 3 and resulted in the production of transcripts with aberrant splicing from exon 1 to exon 3, exon 4, exon 5, or exon 6. Splicing from exon 1 to exon 4, exon 5, or exon 6 maintains the correct translational reading frame. We detected no GCN2 protein in cells from Eif2ak4atc/atc mice, suggesting that the translated products are unstable and degraded.
Experiments focused on MCMV showed that eIF2α was far less efficiently phosphorylated in Eif2ak4atc/atc fibroblasts than wild-type fibroblasts in response to infection. A previous study (5) reported no difference in eIF2α phosphorylation between wild-type and Eif2ak3−/− Eif2ak4−/− (PERK−/− GCN2−/−) fibroblasts 16 h after infection with MCMV. At this time, eIF2α-P levels were equivalent to baseline levels of phosphorylation observed in uninfected cells. These findings are consistent with our data showing the absence to minimal presence of eIF2α-P in wild-type and Eif2ak4atc/atc fibroblasts for at least 32 h after infection, which is similar to levels in uninfected cells. By 40 h after MCMV infection, however, eIF2α-P was detected in wild-type cells but not in Eif2ak4atc/atc cells. Thus, GCN2-dependent eIF2α phosphorylation during the antiviral response to MCMV occurs at least 32 h postinfection. The relatively late action of GCN2 during the innate antiviral response may explain the modest susceptibility of atchoum homozygous mice to MCMV in vivo.
The genomes of both MCMV and human adenovirus encode factors that function to counter the blockade of translation imposed by PKR-dependent eIF2α phosphorylation. MCMV proteins m142 and m143 bind to PKR, possibly in conjunction with dsRNA, to prevent PKR activation (5). The noncoding virus-associated RNA molecule I (VAI RNA) of adenovirus binds to PKR and blocks its activation (26), while E1B-55K and E4orf6 proteins prevent PKR activation through a ubiquitin ligase-dependent mechanism (33). Our findings clearly demonstrate that GCN2 deficiency increases susceptibility to MCMV infection in vitro and in vivo, albeit only slightly, and raise the possibility that viral mechanisms also have evolved to oppose eIF2α phosphorylation specifically by GCN2.
The mechanism by which GCN2 is activated upon MCMV infection remains unknown. The restricted availability of amino acids, UV irradiation, and proteasome inhibition are known activators of GCN2, and of these, we hypothesize that viral infection is most likely to restrict amino acid availability, in that the pool of free amino acids within the cell may be depleted as a result of rapid viral protein synthesis during MCMV and adenovirus infection. This may stimulate GCN2 to downregulate further translation, affecting both cellular and viral proteins. The size of the viral genome does not appear to influence GCN2 activation, as the genomes of MCMV and human adenovirus differ in size by nearly 200 kb. The kinetics and the specific mechanisms by which a virus takes control of cellular machinery may be determining factors for GCN2 activation. Understanding the mechanism of GCN2 activation by MCMV infection may also provide insight into the DNA virus-specific host defense requirement for GCN2. Defense against Sindbis virus, an RNA virus, has been shown to depend on GCN2, which is activated by two noncontiguous regions of the viral genome resembling uncharged tRNA, the natural ligand for GCN2 (4). Such structures apparently are absent from the influenza genome, which also fails to drive levels of protein synthesis sufficient to activate GCN2 by amino acid depletion.
We are grateful to Shilpi Verma and Chris Benedict for conducting the plaque assays.
This work was supported by National Institutes of Health grant 2P01AI070167-06.
Published ahead of print 23 November 2011