Expression of the Vaccinia Virus Antiapoptotic F1 Protein Is Blocked by Protein Kinase R in the Absence of the Viral E3 Protein

Previous studies have shown that E3-deficient vaccinia virus triggers apoptosis of infected cells. Our study demonstrates that this proapoptotic phenotype stems, at least in part, from the failure of the mutant virus to produce adequate quantities of the viral F1 protein, which acts at the mitochondria to directly block apoptosis. Our data establish a regulatory link between the vaccinia virus proteins that suppress the innate response to double-stranded RNA and those that block the intrinsic apoptotic pathway.

ABSTRACT

Poxviruses encode many proteins with the ability to regulate cellular signaling pathways. One such protein is the vaccinia virus innate immunity modulator E3. Multiple functions have been ascribed to E3, including modulating the cellular response to double-stranded RNA, inhibiting the NF-κB and IRF3 pathways, and dampening apoptosis. Apoptosis serves as a powerful defense against damaged and unwanted cells and is an effective defense against viral infection; many viruses therefore encode proteins that prevent or delay apoptosis. Here, we present data indicating that E3 does not directly inhibit the intrinsic apoptotic pathway; instead, it suppresses apoptosis indirectly by stimulating expression of the viral F1 apoptotic inhibitor. Our data demonstrate that E3 promotes F1 expression by blocking activation of the double-stranded RNA-activated protein kinase R (PKR). F1 mRNA is present in cells infected with E3-null virus, but the protein product does not detectably accumulate, suggesting a block at the translational level. We also show that two 3′ coterminal transcripts span the F1 open reading frame (ORF), a situation previously described for the vaccinia virus mRNAs encoding the J3 and J4 proteins. One of these is a conventional monocistronic transcript of the F1L gene, while the other arises by read-through transcription from the upstream F2L gene and does not give rise to appreciable levels of F1 protein.

IMPORTANCE Previous studies have shown that E3-deficient vaccinia virus triggers apoptosis of infected cells. Our study demonstrates that this proapoptotic phenotype stems, at least in part, from the failure of the mutant virus to produce adequate quantities of the viral F1 protein, which acts at the mitochondria to directly block apoptosis. Our data establish a regulatory link between the vaccinia virus proteins that suppress the innate response to double-stranded RNA and those that block the intrinsic apoptotic pathway.

INTRODUCTION

Innate immunity in vertebrates serves to rapidly detect and eliminate potential sources of infection. The innate immune system responds to a broad array of features common to pathogens and allows rapid clearance of these pathogens without necessarily recruiting the slow and energetically demanding adaptive immune system (1). Double-stranded RNA (dsRNA) is an inevitable by-product of most viral replication strategies, and its recognition by the cell results in the activation of numerous cellular signaling pathways, including those that lead to apoptosis (2–4). Viruses have evolved diverse mechanisms to antagonize the innate immune response, especially the dsRNA response (3, 4). Poxviruses are a family of large dsDNA viruses. One of the best-characterized members of the poxvirus family is vaccinia virus (VACV). VACV has long been known to actively inhibit the interferon (IFN) response in infected cells, and it is now clear that VACV encodes a wide variety of proteins that interfere with host responses, including IFN signaling, the dsRNA response, and apoptosis (5–8).

The vaccinia virus E3 protein is essential for virus replication in a wide range of cells (9). E3 suppresses the interferon response by preventing activation of the dsRNA-activated protein kinase R (PKR). PKR is one of four host protein kinases capable of phosphorylating and inactivating translation initiation factor eIF2α, leading to global inhibition of translation (2). E3-mediated inhibition of PKR is required to maintain efficient protein synthesis over the course of the virus infection cycle. E3 also suppresses interferon induction in some cells by inhibiting phosphorylation of IRF3 and IRF7 (10, 11). E3 is a 190-amino-acid protein with well-conserved N-terminal DNA-binding and C-terminal RNA-binding domains. Many early in vitro studies documented that the C-terminal RNA-binding domain is necessary and sufficient for PKR inhibition and virus replication (12–14). These studies showed that vaccinia virus lacking E3 (ΔE3L) or the RNA-binding domain of E3 (E3LΔ26C) are interferon sensitive and replication defective in many cell lines; in contrast, a mutant virus retaining the RNA-binding domain but lacking the amino-terminal DNA-binding domain (E3LΔ83N) is interferon resistant and replicates normally in cultured cells. Nevertheless, the E3LΔ83N virus is significantly attenuated compared to wild-type virus in an animal model of infection, documenting that the N-terminal domain plays a yet-to-be defined role in viral pathogenesis (15, 16). In HeLa cells, E3 masks virtually all dsRNA synthesized by the virus (17). In the absence of E3, early and intermediate viral transcripts are detected while late mRNAs and proteins are absent, likely because PKR blocks translation of intermediate mRNAs; in addition, extensive degradation of rRNA is detected, due to activation of the 2′-to-5′ oligoadenylate synthetase/RNase L system (7, 17). In addition to suppressing the host response to dsRNA, it has been reported that E3 can act as an inhibitor of apoptosis (14, 18, 19).

Apoptosis is a critical component of the host's antiviral response (20). By committing suicide, an infected cell effectively halts viral replication and limits the spread of infection. It is therefore not surprising that many viruses encode proteins that inhibit or delay apoptosis (21, 22). Mitochondria are a major junction in the cellular apoptotic response, and many viruses, including poxviruses, have targeted the Bcl-2 family of proteins to inhibit cell death at the level of mitochondria. Not only do viruses inhibit Bak and Bax activity in a fashion similar to cellular antiapoptotic Bcl-2 family members, but many also do so using proteins that are structurally related to Bcl-2 proteins (23–26). One such viral Bcl-2-like antiapoptotic protein is vaccinia virus F1, which localizes to the outer mitochondrial membrane and protects the infected cell against a variety of apoptotic stimuli (27–29). F1 blocks the release of cytochrome c and prevents the loss of mitochondrial membrane potential (28) by inhibiting the activation of the proapoptotic activity of Bax and Bak. F1 interacts with Bak, and this interaction prevents Bak oligomerization (28, 30, 31). F1 also indirectly inhibits the activation and oligomerization of Bax, likely through its interactions with the Bcl-2 homology domain 3 (BH3)-only protein Bim (19, 30) and/or Noxa (32). Thus, in virus-infected cells F1 functionally replaces cellular Mcl-1 to act as an antiapoptotic modulator. In addition to inhibiting apoptosis, F1 inhibits the activity of inflammasomes (33). Cells infected with vaccinia virus mutants lacking F1 undergo apoptosis at intermediate and late times postinfection (29, 30) through activation of Noxa (19), and ectopically expressed F1 is sufficient to protect cells from apoptosis in the absence of other viral proteins (19, 27).

Two previous reports have documented that vaccinia virus mutants lacking the E3 protein (VACVΔE3) trigger apoptosis during infection (18, 19). Further analysis revealed that apoptosis stems from activation of Noxa (32), similar to the phenotype of F1-deficient virus. Moreover, the proapoptotic phenotype of VACVΔE3 is eliminated by knocking down PKR (35), indicating that PKR plays an essential role. These observations raise the possibility that VACVΔE3 triggers apoptosis at least in part because activated PKR limits the expression of viral apoptotic inhibitors such as F1. Here, we provide evidence supporting this hypothesis by showing that the F1 protein does not accumulate during infection in the absence of E3 and that the proapoptotic phenotype of VACVΔE3 is suppressed by providing F1 in trans. We also show that two transcripts of the F1L gene are produced during infection; one of these arises by read-through transcription from the upstream F2L gene and does not give rise to appreciable amounts of F1 protein.

RESULTS

F1 protein does not accumulate during VACVΔE3L infection.

As reviewed in the introduction, VACV mutants devoid of E3 trigger apoptosis through the intrinsic pathway (17, 19), a phenotype strikingly similar to that displayed by mutants lacking the F1 protein, a direct inhibitor of apoptosis (29–31, 36). As previously suggested, it is possible that E3 and F1 act together in the same pathway, and thus both are required to inhibit apoptosis (19). However, this suggestion is difficult to reconcile with the ability of F1 to prevent apoptosis in the absence of any other VACV proteins (27, 29). An alternative possibility is that E3 is required for expression of the F1 protein during infection. Although F1 is an early protein (37) and although many early proteins are expressed in the absence of E3 (38), a recent report demonstrated that expression of the early D9 mRNA decapping enzyme requires E3 (39), providing a precedent for E3-dependent early gene expression. We therefore examined the effect of deleting E3 on expression of the F1 protein.

HeLa, RAW, Jurkat, and 293T cells were mock infected or infected with VACV or VACVΔE3L for 6 h. Cell lysates were then probed for F1 protein by Western blot analysis (Fig. 1A). The viral I3 protein was used as an infection control as previous work has demonstrated that it is expressed during VACVΔE3L infections (35), and β-tubulin was used as the loading control. As expected, F1 protein was readily detected in all cell types infected with wild-type VACV; in contrast, F1 could not be detected in any of the cells infected with VACVΔE3L. These data suggest that E3 is required for efficient F1 expression in a wide variety of cell lines. In contrast, accumulation of the dUTPase encoded by the neighboring F2L gene was not greatly affected by deleting E3 (Fig. 1B). Note that the F2 antibody used gives a weak signal in uninfected cells, likely due to cross-reaction with a cellular protein.

FIG 1.

(A) F1 is not detected during VACVΔE3L infection. The indicated cells were mock infected, infected with VACV, or infected with VACVΔE3L at an MOI of 5. At 6 h postinfection the cells were lysed in RIPA buffer, and cell lysates were immunoblotted with anti-F1 antibody to analyze the expression of F1 and with anti-I3 to determine the levels of infection. β-Tubulin was used as a loading control. (B) F2 is expressed in VACVΔE3L-infected cells. HeLa cells were mock infected, infected with VACV, or infected with VACVΔE3L at an MOI of 5. At 6 h postinfection the cells were lysed and immunoblotted with anti-F2 antibody to analyze the expression of the F2 protein and with anti-I3 to determine the levels of infection. β-Tubulin was used as a loading control. WB, Western blotting.

As a further test of the hypothesis that accumulation of F1 requires E3, we asked whether ectopic expression of E3 restores F1 expression during VACVΔE3L infection. To this end, HeLa cells were transfected with plasmids encoding enhanced green fluorescent protein (EGFP) or an EGFP-E3L fusion protein under the cytomegalovirus (CMV) promoter. After 18 h, the transfected cells were infected with either VACV or VACVΔE3L for 6 h. The infected cells were then subjected to Western blot analysis (Fig. 2A). In HeLa cells expressing EGFP and infected with VACVΔE3L, only low levels of F1 were detected by Western blotting. In contrast, expression of F1 was partially restored in HeLa cells expressing EGFP-E3L. This result argues that the failure of VACVΔE3L to efficiently express F1 stems from the absence of E3 rather than from an unanticipated secondary mutation located elsewhere in the viral genome.

FIG 2.

(A) Ectopic expression of E3 rescues the expression of F1. HeLa cells were transfected with either pEGFP (vector) or EGFP-E3L. At 18 h posttransfection, the cells were infected with VACV or VACVΔE3L at an MOI of 5. At 6 h postinfection the cells were lysed in RIPA buffer and immunoblotted with antibodies against F1, I3, GFP, and β-tubulin. (B) Activated PKR inhibits the expression of F1 in the VACVΔE3L mutant. HeLa and HeLa PKR KO cells were mock infected or infected with VACV or VACVΔE3L virus at an MOI of 5. At 6 h postinfection the infected cells were lysed in RIPA buffer. The cell lysates were subjected to Western blot analysis using anti-F1, anti-I3, and anti-β-actin antibodies.

Taken in combination, the data presented in this section demonstrate that accumulation of the F1 protein during infection requires the viral E3 protein.

PKR inhibits expression of F1 by the VACVΔE3L mutant.

Zhang et al. have shown that knocking down PKR eliminates the proapoptotic phenotype of VACVΔE3 and restores late gene expression and virus replication (35). These observations led us to query the role of PKR in the suppression of F1 expression during VACVΔE3L infection. HeLa and HeLa PKR knockout (PKR KO) cells generated by CRISPR/Cas9-mediated gene editing were mock infected or infected with VACV or VACVΔE3L and scored for F1 expression (Fig. 2B). We found that F1 expression was restored to wild-type levels in the HeLa PKR knockout cells infected with VACVΔE3. These results demonstrate that PKR restricts F1 expression during VACVΔE3L infection.

Ectopic expression of F1 prevents apoptosis triggered by VACVΔE3L.

The results described above are consistent with the hypothesis that the proapoptotic phenotype of VACVΔE3L stems from the failure of this mutant virus to express the F1 protein. If this is indeed the case, then providing F1 in trans should prevent the apoptosis that occurs during VACVΔE3L infection. To test this prediction, we transfected HeLa cells with plasmids expressing EGFP or an EGFP-F1 fusion protein under the CMV promoter. At 16 h posttransfection, the cells were mock infected or infected with VACV or VACVΔE3L. Apoptosis was then measured by terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assay 24 h later (Fig. 3). Low numbers of TUNEL-positive cells were observed in HeLa cells that had been transfected with pEGFP or pEGFP-F1L and subsequently mock infected. Similar results were obtained for HeLa cells transfected with pEGFP or pEGFP-F1L and infected with VACV. In contrast, 55% of HeLa cells that had been transfected with pEGFP and infected with VACVΔE3L stained TUNEL positive. However, in HeLa cells transfected with pEGFP-F1L and infected with VACVΔE3L, the number of TUNEL-positive cells fell to 25% of the total of screened cells (Fig. 3). This result demonstrates that F1 protein provided in trans suppresses apoptosis triggered by VACVΔE3L infection. The transfection efficiency in these experiments was approximately 80%; untransfected cells therefore likely contribute to the residual levels of apoptosis in cell populations transfected with pEGFP-F1L. In addition, it is possible that the EGFP tag used partially interfered with F1 function. Overall, these data are consistent with the hypothesis that VACVΔE3L induces apoptosis at least in part because it does not express adequate quantities of F1.

FIG 3.

Ectopic expression of F1L is sufficient to prevent apoptosis triggered by VACVΔE3L. HeLa cells were transfected with pEGFP or pEGFP-F1L. At 16 h posttransfection, the transfected cells were mock infected or infected with VACV or VACVΔE3L at an MOI of 5. The level of apoptosis was assessed by TUNEL assay. The percentage of apoptotic cells (TUNEL positive) was determined by flow cytometry (n = 3, P ≤ 0.05).

Two distinct transcripts of the F1L gene are detected in both the presence and absence of E3.

As reviewed in the introduction, activated PKR blocks mRNA translation by phosphorylating eIF2α. Thus, PKR can inhibit protein accumulation without reducing the levels of the corresponding mRNA. To determine if F1 mRNA is a direct target of PKR-mediated translational repression, we asked if F1 mRNA accumulates normally in the absence of E3. HeLa cells were mock infected or infected with VACV or VACVΔE3L at a multiplicity of infection (MOI) of 5 for 3 h, and total RNA was harvested and analyzed by Northern blot hybridization using a double-stranded ³²P-labeled probe corresponding to the entire F1L open reading frame (ORF). Somewhat surprisingly, the probe hybridized to two distinct bands of approximately 1.4 and 0.9 kb (Fig. 4A). In repeated trials, the longer transcript was ca. 7-fold more abundant than the shorter. Both RNA species were detected in VACVΔE3L-infected cells, where they accumulated with roughly the same kinetics as during wild-type infection (Fig. 4B). The F1L ORF is 681 bp in length, and hence the shorter transcript is approximately the size expected for F1 mRNA, taking into account the length of the 3′ poly(A) tail (∼200 nucleotides [nt]). Although the double-stranded probe used could potentially detect transcripts arising from either DNA strand, we provide direct evidence below that the shorter transcript indeed encodes the F1 protein (see Fig. 9).

FIG 4.

(A) Two transcripts hybridize to the f1l probe. HeLa cells were mock infected or infected with VACV or VACVΔE3L at an MOI of 5. As a loading control, RNAs were stained with SYBR gold. Subsequently, RNA was transferred onto a positively charged nylon membrane via vacuum blotting and hybridized to ³²P-labeled probes specific for f1l that were generated using the F1 gene PCR amplicon. (B) F1 transcript time course during VACV and VACVΔE3L infection. HeLa cells were mock infected or infected with VACV or VACVΔE3L at an MOI of 5. Total RNA was isolated at 1, 2, 3, 4, and 5 h postinfection. RNA was transferred onto a positively charged nylon membrane via vacuum blotting and hybridized to ³²P-labeled probes specific for f1l.

FIG 9.

(A) Schematic diagram of the F2L truncation mutant VACVF2L_trunc. A YFP/GPT expression cassette containing a transcription termination signal was inserted after the third codon in the F2L gene. (B) VACVF2L_trunc expresses F1 protein. HeLa cells were either mock infected or infected with VACV or VACVF2L_trunc virus at an MOI of 5. At 6 h postinfection, the infected cells were lysed in RIPA buffer. The cell lysates were subjected to Western blot analysis using anti-F1, anti-F2, anti-I3, and anti-β-actin antibody. (C) VACVF2L_trunc expresses F1 mRNA. HeLa cells were mock infected or infected with VACV, VACVΔF1L, VACVF2L_trunc or VACVΔE3L at an MOI of 5. Total RNA was isolated at 3 h postinfection and electrophoretically separated in 1% agarose formaldehyde gels, applying 10 μg of total RNA per lane. Subsequently, RNA was transferred onto a positively charged nylon membrane via vacuum blotting and hybridized to ³²P-labeled probes specific for F1L and F2L.

When considering possible explanations for the origin of the longer transcript, we speculated that it might arise by read-through transcription from the upstream F2L (dUTPase) gene which is transcribed in the same direction (Fig. 5A). Indeed, although a canonical early transcription termination signal (T₅NT) (40, 41) is located immediately downstream of the F1 ORF, no such signal is present downstream of F2 or within the F1 ORF. Previous work has documented an analogous situation at the J3/J4 locus, where no T₅NT signal is present between the J3 and J4 ORFs, leading to production of a J3-J4 read-through transcript (42, 43). Consistent with the read-through hypothesis, a double-stranded probe corresponding to the F2 ORF hybridized to a single band which comigrated with the larger transcript detected by the f1l probe in RNA samples isolated from cells infected with wild-type and E3-null VACV Copenhagen (VACV and VACVΔE3L, respectively) and with wild-type VACV Western Reserve (VACVwr) (Fig. 5B). Moreover, this f2l-reactive band was replaced by a ca. 600-nt species in cells infected with VACVΔF1L, which bears a deletion/substitution mutation that replaces a portion of the F1L ORF with an EGFP expression cassette driven from a synthetic poxvirus early/late promoter (28). The truncated F2 transcript produced by VACVΔF1L likely arises via transcriptional termination at the run of 12 successive T residues within the synthetic promoter sequence (44). Overall, these data suggest that the long transcript encompasses both the F2L and F1L ORFs, while the short RNA bears only F1L coding sequences. We also detected a faint ca. 1.5-kb band with the f1l probe in cells infected with VACVΔF1L (Fig. 5B); this signal likely stems from the ca. 350 nt of homology between the probe and the residual F1 locus in the mutant virus. However, the precise nature of this transcript remains unclear.

FIG 5.

The longer transcript contains F2L sequences. (A) Diagram of the F1L and F2L regions of the VACV genome. The F1 and F2 open reading frames are depicted in yellow, and the extent of the deletion in VACVΔF1L is shown. The f1 and f2 probes used in the Northern blot experiments depicted in panel B are shown. CDS, coding sequence. (B) HeLa cells were infected with either VACV (VACVCop), VACVwr, VACVΔE3L, or VACVΔF1L at an MOI of 5. Total RNA was isolated at 3 h postinfection and electrophoretically separated in 1% agarose formaldehyde gels by the application of 10 μg of total RNA per lane. For a loading control, RNAs were stained with SYBR gold. Subsequently, RNA was transferred onto a positively charged nylon membrane via vacuum blotting and hybridized to ³²P-labeled probes specific for f1l and f2l (depicted in panel A).

Further evidence that the long transcript spans the F2L and F1L genes emerged from transcriptome sequencing (RNA-seq) analysis. While analyzing RNA-seq data obtained from cells infected with either wild-type or E3-null VACV mapped onto the viral genome, we noted that most viral gene boundaries are clearly demarcated by abrupt changes in RNA-seq read coverage. For example, the boundaries between the K7R and F1L, F2L and F3L, F3L and F4L, and F4L and F5L genes are clearly evident, presumably marking the transcription start and termination sites of the transcripts of these genes (Fig. 6B, black arrows). In contrast, there were no obvious changes in the coverage depth of the RNA-seq reads that mapped to the F1L and F2L regions of the viral genome, and consequently no clear F2L-F1L gene boundary was evident (Fig. 6A). These observations argue that the long F1L transcript detected by Northern blot analysis initiates at the F2L promoter and ends at the F1L transcription termination site. The (much less abundant) short transcript detected by Northern blot analysis is not resolved in our RNA-seq data, presumably because it represents only a small fraction of the transcripts of this region. We consider it likely that the short transcript initiates at, or shortly upstream of, the F2L-F1L gene boundary and ends at the F1L transcription termination site.

FIG 6.

RNA-seq coverage over the region of F1L to F4L of the VACV genome. RNA extracted from HeLa cells infected with VACV virus at an MOI of 5 for 1.5 h was subjected to RNA-seq analysis. The figure shows the RNA-seq read coverage over the region extending from the 3′ end of the K7R gene to the 3′ end of the F5L gene. Abrupt changes in read density corresponding to gene boundaries are marked with arrows. (A) Read coverage in the region of K7R to F2L. (B) Read coverage in the region of F2L to F5L.

Overall, these data argue that the F2L-F1L region of VACV gives rise to two 3′ coterminal mRNAs, driven from separate promoters (Fig. 7). Such 3′ coterminal mRNA families are common in RNA and DNA viruses, and, as described above, an example has been previously described in VACV.

FIG 7.

Schematic diagram showing siRNA targets. The figure depicts the proposed structure of the shorter and longer F1 transcripts and diagrams the sequences targeted by the siRNA used in the experiment shown in Fig. 8.

F1 protein is translated from the shorter mRNA.

In almost all cases where viruses produce overlapping 3′ coterminal mRNAs, each of the mRNAs in a given 3′ coterminal family is translated to yield only one protein, that encoded by the 5′ open reading frame. If this is the case for the VACV F1L 3′ coterminal transcripts, then one predicts that the larger transcript is translated to yield F2 protein while the smaller is used to produce F1. We tested these predictions using an RNA interference (RNAi) approach. We designed two sets of small interfering RNAs (siRNAs), each targeting distinct regions of the long mRNA (Fig. 7). One set, comprised of F2L-siRNA-1 and F2L-siRNA-2, targets sequences near the 5′ end of the long mRNA and should not affect the short transcript. The second set (F1L-siRNA-1 and F1L-siRNA-2) targets the shared 3′ region of the long and short mRNAs. HeLa cells were transfected with the targeting siRNAs or scrambled controls and then infected with VACV. Cell lysates were scored for F2 and F1 protein levels by Western blot analysis (Fig. 8). As previously noted, the F2 antibody used gives a weak signal in uninfected cells. Both of the F2L siRNAs strongly reduced F2 protein levels relative to levels with the scrambled controls but had little or no effect on F1 expression (Fig. 8A). In contrast, the F1L-siRNAs inhibited expression of both proteins (Fig. 8B). These results indicate that the long RNA gives rise to F2 protein, while the short mRNA is the major source of F1.

FIG 8.

(A) f2l-siRNA inhibits the expression of F2 but not F1. HeLa cells were transfected with two distinct f2l-siRNAs or the scrambled siRNA for 12 h. Posttransfection, the HeLa cells were either mock infected or infected with VACV at an MOI of 5. At 6 h posttransfection the cells were harvested and lysed in RIPA buffer. The cell lysates were subjected to immunoblotting using the anti-F1L antibody, anti-F2L antibody, and anti-I3L antibody. (B) f1l-siRNA inhibits the expression of F1 and F2. HeLa cells were transfected with two distinct f1l-siRNAs or the scrambled siRNA control for 12 h. Posttransfection, the HeLa cells were either mock infected or infected with VACV at an MOI of 5. At 6 h posttransfection the cells were harvested and lysed in RIPA buffer. The cell lysates were subjected to immunoblotting using the anti-F1L antibody, anti-F2L antibody, and anti-I3L antibody.

Disruption of the long mRNA has no effect on F1 expression.

As a further test of the hypothesis that the short mRNA is the major source of F1 protein, we examined the effect of truncating the long mRNA within the F2L ORF. To this end, we inserted a yellow fluorescent protein-guanosine phosphoribosyltransferase (YFP/GPT) expression cassette bearing a 3′ VACV transcription terminator (45) after the third codon of the F2L ORF, yielding VACVF2L_trunc (Fig. 9A). This manipulation should terminate transcripts initiated at the F2 promoter within the inserted cassette. As expected, Western blot analysis revealed that the mutation reduced the F2 signal to background levels in infected HeLa cells, without greatly altering F1 protein levels (Fig. 9B). Moreover, the mutation eliminated the long transcript without altering the short mRNA (Fig. 9C). Taken together, these data document that the short transcript is the major source of F1 protein and that it does not arise through processing of the long transcript.

DISCUSSION

We have drawn three major conclusions based on the data presented in this paper. First, expression of the VACV apoptosis inhibitor F1 requires the viral E3 protein, a dsRNA-binding protein and PKR antagonist. In the absence of E3, F1 mRNA accumulates normally, but little if any F1 protein is detected. However, efficient F1 protein expression can be rescued by inactivating PKR. Taken together, these data document that translation of F1 mRNA is efficiently restricted by PKR. This finding was somewhat surprising as F1 has been viewed as an early protein, and many VACV early proteins are expressed in the absence of E3 (17, 35). However, a recent report documented that expression of the early D9 mRNA decapping enzyme also requires E3 (39). Currently, it remains unclear why F1 and D9 mRNAs display enhanced sensitivity to translational inhibition by PKR relative to that of other early mRNAs. In the case of F1, the timing of RNA accumulation does not appear to be a key factor as F1 and F2 mRNAs accumulate with comparable kinetics (Fig. 4B), yet F2 protein accumulates efficiently in the absence of E3 while F1 does not (Fig. 1). This observation suggests that one or more structural features of F1 mRNA render its translation unusually sensitive to PKR. Further studies are required to uncover the molecular basis for this phenotype.

Our second conclusion is that the proapoptotic phenotype of VACVΔE3 likely stems, at least in part, from the failure of this virus to produce adequate levels of the F1 protein. Evidence for this conclusion includes our findings that F1 expression requires E3 (Fig. 1) and that ectopic expression of F1 suppresses apoptosis induced by VACVΔE3 (Fig. 3), along with the previous observation that E3-null virus triggers apoptosis through the same Noxa-dependent pathway that is activated in the absence of F1 (32). The conclusion is fully in accord with the findings that (i) mutations that inactivate the dsRNA binding domain of E3 provoke apoptosis during infection while lesions in the N-terminal DNA binding domain do not (35) and (ii) the phenotype of proapoptotic E3 mutants is suppressed by inactivating PKR (35).

Our results document an intriguing regulatory link between the major VACV suppressors of dsRNA responses and apoptosis. It is interesting to speculate that this link might play an important role in viral immune evasion during in vivo VACV infections by ensuring that any cells that express suboptimal levels of E3 succumb to apoptosis, reducing the levels of antiviral cytokines produced in response to the unmasked dsRNA. This, in turn, could enhance virus replication and spread in the infected tissue. It will therefore be important to determine if the expression of F1 orthologues in other orthopoxviruses is subject to similar regulation.

Finally, we have obtained evidence that two 3′ coterminal transcripts span the F1 ORF: the longer appears to initiate at the F2 promoter while the shorter is likely driven from a promoter upstream of the F1 ORF (Fig. 7). A similar situation has been previously described for VACV J3 and J3 mRNAs, which share a 3′ end (42, 43). The biological significance of this arrangement remains uncertain. Our data indicate that the shorter transcript is the major source of F1 protein under the conditions of our experiments (Fig. 8); however, it remains possible that the longer transcript is also able to give rise to significant levels of F1 protein under other circumstances, using a noncanonical mechanism of translational initiation. In this context, it is possibly relevant that many other orthopoxviruses, such as ectromelia, monkey pox, and horse pox, also lack a clear predicted early transcriptional terminator between their F2 and F1 orthologous ORFs (data not shown), raising the possibility that these viruses produce analogous 3′ coterminal transcripts from this region. Further studies are required to test this possibility.

MATERIALS AND METHODS

Cells and viruses.

HeLa, human embryonic kidney 293T (HEK293T), BGMK, RAW, and BHK-21 cells were grown in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies) supplemented with 10% fetal bovine serum (FBS) (Life Technologies), 2 mM l-glutamine (Life Technologies), 50 U/ml penicillin (Life Technologies), and 50 μg/ml streptomycin (Life Technologies). A HeLa PKR knockout clone (HeLa PKR KO1) has been described previously (46) and was grown in the same medium. Jurkat cells (ATCC) were maintained in RPMI 1640 medium (Life Technologies) supplemented with 10% FBS, 50 U/ml penicillin, 50 μg/ml streptomycin, and 100 μM β-mercaptoethanol (BioShop Canada, Inc.). Vaccinia virus (VACV) strain Copenhagen (VACVCop) was provided by G. McFadden (University of Florida, Gainesville, FL). VACVCop lacking F1L (VACVΔF1L) has been previously described (28), and VACVCop devoid of E3L (VACVΔE3L) was provided by B. Jacobs (Arizona State University, Tempe, AZ) (47). Vaccinia virus strain Western Reserve (VACVwr) was provided by D. Evans (University of Alberta, Edmonton, Alberta, Canada). All vaccinia viruses except VACVΔE3L were propagated in BGMK cells as described previously (48). VACVΔE3L was propagated in BHK-21 cells as described previously (47).

VACVF2L_trunc was generated using a previously published method (45). A DNA fragment containing the F3L ORF and the first 20 bp of the 5′ end of the F2L ORF was amplified using primers 5′-GTATATCTCATCGGTGGATGGATGAACAATGAAAT-3′ and 5′-AGCTTCACCACCATGTTGAACATAAAACTAATATTTTATT-3′, which bear terminal XhoI and HindII sites, respectively. The fragment (fragment 1) was then digested with XhoI and HindIII and gel purified. Next, a DNA fragment containing the remainder of the F2L ORF was amplified using 5′-AATATTAACTCACCAGTTAGATTTGTTAAGGAAACTAAC-3′ and 5′-CTATTGTTTATTATCTAAGTCCTGTTGATCCAAAC-3′, bearing terminal BamHI and NotI sites, respectively. The fragment (fragment 2) was digested with the respective enzymes and gel purified. Fragments 1 and 2 were then sequentially cloned into the pGDLoxP vector to generate pGDLoxP-f3-f2. The pGDLoxP-f3-f2 was linearized with BglII, and 10 μl of the linearized plasmid was transfected into HeLa cells using Lipofectamine 2000 (Invitrogen). The transfected HeLa cells were then infected with VACV at an MOI of 0.01. Forty-eight hours postinfection, the cells were harvested and sorted using YFP expression using a BD FACSAria III instrument. VACVF2L_trunc was then isolated via multiple rounds of fluorescent-focus purification.

Detection of DNA fragmentation by TUNEL.

DNA fragmentation was assessed by using the terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) method (Roche Diagnostics). HeLa cells (2 × 10⁶) were infected at an MOI of 5 for 6 h. Apoptosis was stimulated using 2 μM staurosporine (STS) for 6 h. Cells were fixed with 2% (wt/vol) paraformaldehyde in PBS for 10 min at room temperature and TUNEL stained according to the manufacturer's instructions. The cells were analyzed on a BD LSRFortessa flow cytometer. In each experiment 30,000 cells were analyzed. TUNEL-positive cells were counted through the FL-2 channel equipped with a 561-nm filter (586/15-nm band pass), and data were analyzed with FlowJo (Tree Star, Inc.).

Cell lysis and sample preparation.

Transfected or infected cells were washed in phosphate-buffered saline (PBS) and lysed in radioimmunoprecipitation assay (RIPA) buffer for 15 to 20 min at 4°C or on ice. DNA in the lysates was sheared using a 28-gauge hypodermic needle. The lysates were cleared by centrifugation at 20,000 × g for 5 min at 4°C. All lysates were mixed with 5× protein sample buffer prior to storage at −20°C.

SDS-PAGE.

Prior to SDS-polyacrylamide gel electrophoresis (SDS-PAGE), the samples were boiled at 98°C for 20 min. The protein samples were separated in 7.5%, 10%, or 12.5% acrylamide gels by SDS-PAGE using a Bio-Rad Mini-Protean Tera cell system (Bio-Rad). A PageRuler prestained protein ladder (Fermentas) was used to approximate the sizes of the reduced protein bands. Protein samples were separated at 80 to 150 V in SDS-PAGE running buffer until adequate resolution was achieved.

Western blotting.

Proteins that were resolved by SDS-PAGE were transferred to Immobilon-FL polyvinylidene difluoride (PVDF) membranes (Millipore), and the membranes for infrared imaging were blocked in Odyssey blocking buffer (Li-Cor) and Tris-buffered saline plus Tween (TBST) (1:1) for 1 h at room temperature. Primary antibodies were incubated overnight at 4°C, and secondary antibodies (Alexa Fluor 680 goat anti-rabbit [Invitrogen] and Alexa Fluor 790 goat anti-mouse [Invitrogen]) were incubated for 1 h at room temperature. All antibodies were diluted in TBST. Proteins were detected using an Odyssey Infrared Imaging System (Li-Cor). The images were analyzed using Li-Cor ImageStudio Lite software. Antibodies used for immunoblotting were as follows: mouse anti-β-tubulin (ECM Bioscience), mouse anti-α-actin (Sigma-Aldrich), rabbit anti-β-actin (Sigma-Aldrich), mouse anti-I3L (provided by D. Evans, University of Alberta, Edmonton, Alberta, Canada), and rabbit anti-F1L serum (29). A rabbit anti-vaccinia virus F2L antipeptide antiserum was raised against CRFVKETNRAKSPTR peptide in rabbits by ProSci, Poway, CA.

RNA extraction.

HeLa cells were infected with vaccinia virus (VACV), VACVΔE3L, VACVΔF1L, or VACVΔF2L for the times indicated in the figure legends. Postinfection, the total RNA was extracted using RNAzol RT (Sigma) reagent according to the manufacturer's instructions. The mock-infected or infected HeLa cells (2 × 10⁶) were lysed in 1 ml of RNAzol RT solution and frozen at −80°C overnight or for up to 1 week. The samples were thawed on ice, and RNA was harvested according to the manufacturer's instructions. The RNA pellet was resuspended in 50 μl of water (DNase/RNase free) (Gibco).

RNA-seq.

For the RNA-seq experiment, HeLa cells were infected with VACV. The cells were lysed in RNAzol RT solution at 0.5, 1, and 2 h postinfection. The RNA was extracted as mentioned earlier. RNA samples were qualified using an Agilent Technologies 2100 Bioanalyzer (Total RNA Nano kit); all samples displayed RNA integrity numbers (RIN values) of >9. RNA concentration was obtained from duplicate measurements using the Qubit platform. Approximately 920 ng of RNA was used for library generation according to the manufacturer's protocol. After an initial poly(A) selection step, the RNA was fragmented for 8 min and converted to cDNA (first- and second-strand synthesis). Stocks (10 nM) of each library were prepared using the average fragment length and concentration for each library. Pooled libraries (25 microliters at ∼2.6 nanograms/microliter) were submitted for sequencing (2 by 125 bp) on an Illumina HiSeq 2500 instrument at The Centre for Applied Genomics (TCAG), Toronto, Canada. The RNA-seq data were analyzed using the Geneious for RNA-seq assembler in Geneious, version R9.

Northern blotting.

We generated random-primed probes of the VACV F1L and F2L genes for Northern blot analysis. The template for F1L was obtained by cleaving the F1L gene from the pEGFP-F1L plasmid using BamHI and XhoI. Similarly, the F2L template was obtained by cleaving the F2L gene from the pcDNA3.1-F2L plasmid using BamHI and XhoI. The template was combined with 2 μg/μl random hexamers, heated at 95°C for 5 min, and then cooled on ice. The template/hexamers were then incubated with oligonucleotide labeling buffer (44 mM Tris-Cl [pH 8.0], 4.4 mM MgCl₂, 9 mM 2-mercaptoethanol, 17.6 μM each dATP, dGTP, and dTTP, 181 mM HEPES [pH 6.6], 20 μCi of [α-³²P]dCTP [PerkinElmer]), bovine serum albumin (New England BioLabs), and Klenow for 30 min at 37°C. The reaction mixture was diluted 1:1 with 1× Tris-acetate-EDTA (TAE), phenol-chloroform extracted once, and passed through an illustra Nick column (17-0855-01; GE). All radiolabeled products were quantitated using an LS6500 scintillation counter (Beckman).

Ten micrograms of RNA samples was combined with 3 μl of 10× morpholinepropanesulfonic acid (MOPS) buffer, 5 μl of 37% formaldehyde, and 15 μl of deionized formamide. The samples were heated at 55°C for 15 min and then immediately moved to ice. The samples were then subjected to electrophoresis through a 1% agarose gel containing 6% formaldehyde. Electrophoresis was carried out at approximately 8 V/cm for 2 h in 1× MOPS buffer containing 6% formaldehyde. The gel was stained with SYBR gold (Invitrogen), diluted in 1× MOPS buffer, for 30 min. The stained gel was visualized using an FLA-5100 imager (Fujifilm). RNA was then transferred to a Zeta-Probe membrane (Bio-Rad) in 10× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and then UV cross-linked using the auto-cross-link option (Stratalinker 2400; Stratagene).

Church buffer (49) was used to prehybridize the Zeta-Probe membrane at 65°C in a hybridization oven for 4 h. The prehybridization buffer was removed and replaced with 7 ml of Church buffer. To this 1,000,000 cpm/ml of ³²P-labeled probe was added. The probe was allowed to hybridize at 65°C for 13 to 17 h. The membranes were then washed two times with Church wash buffer 1 (49) for 20 min each. This was followed by two washes with Church wash buffer 2 (49) for 20 min each. The Zeta-Probe membrane was subjected to autoradiography using Kodak X-OMT AR film.

Accession number(s).

The FASTQ files from the RNA-seq experiments have been uploaded to the NCBI database under BioProject accession number PRJNA468252.