IRF-3 Activation by Sendai Virus Infection Is Required for Cellular Apoptosis and Avoidance of Persistence

Abstract

Here, we report that specific manipulations of the cellular response to virus infection can cause prevention of apoptosis and consequent establishment of persistent infection. Infection of several human cell lines with Sendai virus (SeV) or human parainfluenza virus 3, two prototypic paramyxoviruses, caused slow apoptosis, which was markedly accelerated upon blocking the action of phosphatidylinositol 3-kinases (PI3 kinases) in the infected cells. The observed apoptosis required viral gene expression and the action of the caspase 8 pathway. Although virus infection activated PI3 kinase, as indicated by AKT activation, its blockage did not inhibit JNK activation or IRF-3 activation. The action of neither the Jak-STAT pathway nor the NF-κB pathway was required for apoptosis. In contrast, IRF-3 activation was essential, although induction of the proapototic protein TRAIL by IRF-3 was not required. When IRF-3 was absent or its activation by the RIG-I pathway was blocked, SeV established persistent infection, as documented by viral protein production and infectious virus production. Introduction of IRF-3 in the persistently infected cells restored the cells' ability to undergo apoptosis. These results demonstrated that in our model system, IRF-3 controlled the fate of the SeV-infected cells by promoting apoptosis and preventing persistence.

The host response to virus infection is a complex process. In vivo, the immune system, both innate and adaptive, plays major roles in determining the outcome of an infection. But these outcomes as well as survival of the infected cells and the efficacy of virus replication are initially determined by host-virus interactions at the cellular level, a topic that can be experimentally addressed using cells in culture. It is clear that the fate of the infected cell, such as death or survival, acute or persistent infection, transformation or normal growth control, is determined not only by the viral gene products but also by the products of hundreds of cellular genes whose expression and functions are modulated by the infection process. Thus, both the virus and the cell are equally important partners in determining the fate of infection at the cellular level. An important choice made at this level is whether the infected cell produces progeny viruses, dying in the process, or lives. If the cell can survive infection, it has the opportunity to become persistently infected. In this paper, we demonstrate that the choice between lytic and persistent infection by paramyxoviruses is made by the action of one cellular protein, IRF-3.

The Paramyxoviridae includes major human and animal pathogens, such as measles virus, mumps virus, respiratory syncytial virus, parainfluenza viruses (PIVs), and Newcastle disease virus (18). Sendai virus (SeV) is an extensively studied member of the Respirovirus genus, which also includes human PIV3 (hPIV3). The V, C, and W proteins of paramyxoviruses play major roles in combating the innate immune system of the host and, consequently, replication and virulence. Mutant viruses that do not encode these proteins are highly attenuated in vivo. Paramyxoviruses can establish persistent infections in vitro and in vivo. In some cases, infectious virions are produced, and in others nucleocapsids are passed on to daughter cells during cell division. For infection with SeV, the general outcome is apoptotic death of infected cells. However, viral mutants have been generated that can establish persistent infection. One such mutant contains the L1618V mutation in the L protein (13). Other temperature-sensitive mutants, which can establish persistent infection, have mutations in the M and HN proteins (13) or the P protein (13). The reciprocal situation, in which wild-type SeV causes either apoptosis or persistent infection in the same cell line, has not been reported in the literature. Thus, little is known about cellular factors that determine the survivability of an infected cell.

The innate immune responses to virus infection are often initiated by Toll-like receptors; alternatively, cytoplasmic double-stranded RNA (dsRNA)-recognizing RNA helicases RIG-I and Mda5 can initiate signaling (17). For SeV, RIG-I is the primary initiator of signaling (17). The transcription factors IRF-3, NF-κB, and AP-1 are activated and, consequently, transcription of hundreds of cellular genes, including the interferon (IFN) genes, is induced. Many of the same genes that are induced by IRF-3 are also induced by interferon (9). Expression profiling of genes induced by SeV infection of various cell lines has generated important mechanistic insights (5). As expected, some genes, such as A20, are not induced in the absence of functional NF-κB, whereas others, such as ISG56, require IRF-3 for their activation. Induction of a third class of gene, such as Noxa, is independent of either NF-κB or IRF-3. Interestingly, we observed that IRF-3 could also inhibit the induction of some genes by SeV infection (5). These genes constitute a subset of NF-κB-driven genes, including the one that encodes the interesting protein A20, which negatively regulates both NF-κB signaling and IRF-3 signaling (4, 21). IRF-3 is expressed in all cell types, although at vastly different levels. It is activated by phosphorylation at multiple sites, followed by dimerization and translocation to the nucleus, where it binds to IFN-stimulated response elements in the promoters of target genes. The protein kinases TBK1/IKKɛ are responsible for mediating IRF-3 phosphorylation (7, 30). Our studies have revealed that additional phosphorylation is mediated by phosphatidylinositol 3-kinase (PI3K) for Toll-like receptor 3 (TLRtin3)-dependent signaling, but not for RIG-I-dependent signaling (29). Importantly, IRF-3 activated by RIG-I signaling, but not by TLR signaling, is polyubiquiated and degraded by the proteasome. Thus, the complexity of phosphorylation and activation of IRF-3 are inducer dependent. A large number of cellular genes, the viral stress-inducible genes, are induced by IRF-3.

Programmed cell death, or apoptosis, is caused by sequential activation of death pathways which are initiated by widely different signals but converge on the same end points, such as DNA fragmentation, membrane property changes, cleavage of specific cellular proteins, and ultimately cellular disintegration (2, 3, 28). The two distinct pathways mediating this process are initiated by two distinct caspases: the intrinsic pathway initiated by caspase 9 and the extrinsic pathway initiated by caspase 8. Both lead to the activation of the effector caspases, caspase 3 and caspase 7. Because there is cross talk between the intrinsic and the extrinsic pathways, it is often difficult to identify the origin of the apoptotic process activated in a virus-infected cell. For example, in SeV-infected cells, both caspase 8 and caspase 9 are activated (1). It has been suggested that caspase 9 activation in these cells is mediated by a novel pathway independent of Apaf-1 and that this activation is needed for apoptosis of SeV-infected mouse embryo fibroblasts. It is not clear whether the same conclusions apply to more physiologic target cells of this virus, namely, airway epithelial cells. There are strong suggestions in the literature that apoptosis in SeV-infected cells may be initiated by the activation of the transcription factor IRF-3. Weaver et al. (34) showed that viral or dsRNA-induced apoptosis is independent of p53 and IFN, but it depends on the activation of IRF-3. In this context, they also showed that a dominant-negative mutant of IRF-3 blocked this process. Reciprocally, a constitutively active mutant of IRF-3 causes apoptosis without virus infection. Under these conditions, both caspase 8 and 9 inhibitors had protective effects on the cells, implicating both intrinsic and extrinsic pathways in this system (12). PI3Ks participate in many aspects of cellular physiology, including growth regulation, signal transduction, differentiation, and oncogenic transformation (33). Many of the cellular effects of PI3K activation are mediated by downstream protein kinases that are activated in cascades. The AKT/PKB pathway is the best-known branch of PI3K signaling that regulates cell survival and apoptosis (26). The involvement of the NF-κB pathway in this action remains controversial (26). Recently, it has been reported that Akt activity is also required for optimal replication of paramyxoviruses and other nonsegmented negative-strand viruses (32).

MATERIALS AND METHODS

Cell lines, inhibitor pretreatment, and transfections.

HT1080, 2f-SR, U4C, P2.1, P2.1.17, and A549 cells have been previously described (24, 27). All cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS) and antibiotics. The media for 2f-SR cells and P2.1.17 cells were further supplemented with G418 (400 μg/ml) and puromycin (1 μg/ml), respectively. To generate the two new cell lines, HT1080 cells were transfected with plasmids by using FuGENE6 (Roche) following the manufacturer's protocol. Cells were selected for G418 resistance and screened for protein expression. HT1080/RIG-IC cells express the dominant-negative C-terminal helicase domain of RIG-I and lack the CARD domain (8). HT1080/siIRF-3 cells express shRNAs that target IRF-3 for silencing by the RNA interference (RNAi) pathway. Generation of the targeting vector is described below. BEAS-2B cells were cultured in F-12K medium supplemented with 10% FBS, 1% l-glutamine, 10 mM HEPES, and antibiotics. Where indicated, cells were pretreated with the PI3 kinase inhibitor LY294002 (20 to 50 μM; Alexis Biochemicals) or vehicle control (methanol) for 30 min. Caspase inhibitors (100 μM; Enzyme Systems Products, Livermore, CA) or a dimethyl sulfoxide (DMSO) control were added 60 min before viral infection. All caspase inhibitors were FMK conjugated.

Generation of an IRF-3 RNAi targeting vector.

Using the PCR short hairpin activated gene silencing method, we generated two shRNA expression cassettes that targeted nucleotides 3 to 32 and 1333 to 1363 in the 5′ and 3′ untranslated regions of the IRF-3 mRNA. Each cassette was driven by the U6 snRNA promoter, which is transcribed by RNA polymerase III and is constitutively expressed in vivo. The cassettes were cloned in tandem into the pcDNA3 backbone to allow for selection.

Virus infections.

Sendai virus (Cantell strain) was obtained from Charles Rivers SPAFAS (Preston, CT). For infections, cells were washed two times with virus infection medium, Dulbecco's modified Eagle's medium supplemented with 2% FBS, and then placed in a minimal amount of virus infection medium. SeV was added at a concentration of 80 hemagglutinating units/ml. Cells were incubated with virus for 1 h with gentle agitation every 10 minutes. The virus was removed, and cells were washed twice with complete medium. The cells were placed in complete medium until they were harvested. In experiments where PI3 kinase and caspase inhibitors were used, cells were washed and placed in the virus infection medium prior to addition of specific inhibitors. After virus attachment, inhibitors were added back to the complete medium for the remainder of the incubation period. Where indicated, anti-TRAIL blocking antibody (RIK-2; eBioscience) was also added at this step. Sendai virus was inactivated using a UV Stratalinker 2400 (Stratagene) for 30 or 60 s. The hPIV3 virus was propagated in CV1 cells (22). Virus infections were carried out as described for Sendai virus with a multiplicity of infection of 2.

Persistently infected lines were generated by infecting the parental cell lines with Sendai virus as described previously. Cells were passed every 3 to 4 days, and the remaining cells were saved for protein analysis. To determine viral titers, equivalent numbers of cells were plated and allowed to attach overnight. Cells were washed, and complete medium was added for 24 h. The medium containing virus was removed and stored at −80°C. On the day of virus titer determinations, the conditioned medium was thawed and incubated in a sonic water bath for 30 s and treated with trypsin for 30 min at 37°C. Serial dilutions of the virus were made and placed on confluent monolayers of LLCMK2 cells for 90 min with gentle agitation every 30 min. The virus inoculum was removed, and the cells were overlaid with medium containing 0.5% agar. After 3 days the agar was removed, and the cells were washed with phosphate-buffered saline. Virus colonies were visualized by incubating the monolayer with a 0.1% suspension of chicken red blood cells (Colorado Serum Company) for 20 to 30 min. The monolayer was then washed with phosphate-buffered saline, and the hemabsorbed plaques were scored.

Immunofluorescence.

Cells were plated on coverslips in six-well dishes at least 16 h before treatment. For apoptosis analysis, cells were stained using the DeadEnd fluorometric terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) system (Promega). IRF-3 staining has been previously described (27). The Sendai virus antibody was raised against the entire Sendai virus virion and was obtained from Atsushi Kato (16). Coverslips were mounted with antifade agent containing 4′,6′-diamidino-2-phenylindole (DAPI; Vector Labs).

Western analysis.

Whole-cell extracts were prepared as previously described (10). Proteins were separated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene difluoride (PVDF). Antibodies for C-PARP, caspase 8, caspase 9, JNK, and phosphorylated JNK (Thr 183/Tyr 185) were obtained from Cell Signaling (Beverly, MA), and Western analysis was performed following the manufacturer's protocol. The antibody against actin was obtained from Sigma. The IRF-3 and Sendai virus C protein antibodies were gifts from Michael David (14) and Atsushi Kato (16), respectively.

RNA analysis.

mRNA was isolated with RNA-Bee following the manufacturer's instructions (Tel-Test Inc., Friendswood, TX). RNase protection assays were performed with the RPA III kit (Ambion). The 561 probe protected nucleotides 1342 to 1511 of the 561 message, while the actin probe has been previously described (6).

For quantitative real-time PCR, RNA samples were treated with DNA-free (Ambion), and their purity was confirmed by PCR. First-strand cDNA synthesis was performed with random hexamer primers using the SuperScript III kit (Invitrogen). Real-time PCR was performed with Sybr green PCR core reagents (Applied Biosystems) with an annealing temperature of 60°C. TRAIL-specific primers corresponded to nucleotides 330 to 351 and 508 to 534 of the TRAIL mRNA and spanned exons 2 and 3. Samples were normalized to message for ribosomal protein L32, which targets nucleotides 246 to 265 and 499 to 518 of the mRNA (35).

RESULTS

Characteristics of cellular apoptosis by Sendai virus infection.

We have been using the human rhabdomyosarcoma cell line HT1080 for analyzing the signaling pathways that lead to the induction of virus stress-inducible genes. In this context, we observed that LY294002 (LY), an inhibitor of PI3 kinase, inhibited gene induction by the dsRNA/TLR3 pathway. Because the TLR3 pathway and the SeV-activated RIG-I pathway are partially overlapping, we wondered whether LY would also inhibit gene induction by SeV infection. In the course of these experiments, we made the unexpected observation that SeV-infected LY-treated cells underwent very rapid apoptosis. Usually, productive infection of cells in culture with SeV causes slow apoptosis which, depending on the cell line, takes 2 to 4 days for completion, but almost all LY-treated cells were lysed within 6 h after infection, whereas no lysis was observed in uninfected cells that were treated with LY or in those that were infected but not treated with LY (Fig. 1A, left panel). In another assay for apoptosis that measured cleavage of PARP, giving rise to the C-PARP fragment, the same conclusion was confirmed (Fig. 1B). The apoptosis was very rapid; as early as 4 h after infection, C-PARP was detected (Fig. 1C). These results clearly demonstrated that if PI3 kinase was inhibited, SeV-infected cells were subject to rapid apoptosis. To examine the physiological significance of our observation, we tested for apoptosis in two airway epithelial cell lines, A549 and BEAS-2B, which are natural targets of SeV infection. Rapid apoptosis was observed in both cell lines only in response to both infection and LY treatment (Fig. 1D and E). Similar observations were made with the clinically important paramyxoviruses, human parainfluenza virus 3 (Fig. 1F), and respiratory syncytial virus (data not shown). Thus, we established that the observed phenomenon was true for infection of physiologically relevant cells with several paramyxoviruses.

FIG. 1.

PI3K inhibition enhances apoptosis of cells infected with Sendai virus or hPIV3. (A) HT1080 cells were pretreated with the PI3 kinase inhibitor LY294002 (LY; 20 μM) or vehicle control for 30 min. Cells were then mock infected or infected with Sendai virus (80 hemagglutinating units/ml) as indicated. Six hours after virus addition, cells were fixed, permeabilized, and stained for DNA fragmentation using a TUNEL protocol. The same field of cells was imaged for phase contrast and TUNEL staining. (B) HT1080 cells were treated as described for panel A. After 6 hours, whole-cell extracts were prepared, and proteins (30 μg) were separated by SDS-PAGE and transferred to PVDF. A Western blot assay was performed with an antibody specific for the cleaved PARP protein (C-PARP). A Western blot assay against actin controlled for protein loading. (C) Cells were pretreated with the LY inhibitor and infected with the Sendai virus as described for panel A. Extracts were prepared at the indicated times, and Western blot assays for cleaved PARP and actin were performed as described for panel B. (D and E) A549 (D) and BEAS-2B (E) cells were pretreated with the PI3K inhibitor LY, infected with SeV, and stained as described for panel A. (F) A549 cells were pretreated with LY and stained as described for panel A. Cells were mock infected or infected with hPIV3 (multiplicity of infection, 2) as indicated.

Next, we investigated several characteristics of the above phenomenon. By subjecting the virions to UV inactivation, we demonstrated that in our system, viral gene expression was necessary for causing both apoptosis (Fig. 2A) and the induction of viral stress-inducible genes, such as ISG56 (Fig. 2B). Two major death pathways, initiated by activated caspase 8 and caspase 9, are used for inducing apoptosis in many virus-infected cells. We observed that in SeV-infected cells both caspases were activated, as indicated by their cleavage (Fig. 3A). Caspase 8 activation was noticeable at 4 h after infection (Fig. 3B), similar to PARP cleavage (Fig. 1C). As expected, the pan-caspase inhibitor z-VAD inhibited apoptosis of infected cells (Fig. 3C, middle right panel). An inhibitor of caspase 8, z-IETD, also blocked apoptosis (Fig. 3C, left lower panel), but an inhibitor of caspase 9, z-LEHD, failed to do so (Fig. 3C, right lower panel). These results indicated that although both caspase 8 and caspase 9 were activated in SeV-infected LY-treated cells, only the caspase 8 pathway was responsible for inducing apoptosis.

FIG. 2.

Viral gene expression is needed for apoptosis. A. HT1080 cells were pretreated with LY294002 for 30 min. Cells were then mock infected (Con) or infected with Sendai virus that had been inactivated by UV treatment for 0, 30, or 60 seconds. Cells were lysed 6 h p.i., and Western blot assays to determine PARP cleavage were performed as described in the legend for Fig. 1B. B. HT1080 cells were treated as described for panel A. RNA was harvested 4 h p.i., and 561 and actin mRNA levels were determined by RNase protection assays.

FIG. 3.

Apoptosis is induced by the caspase 8 pathway. A. HT1080 cells were treated with LY294002 and Sendai virus as previously described. Whole-cell lysates were prepared, and Western blot assays were performed with antibodies against caspase 8 and caspase 9. The positions of the full-length (57- and 47-kDa) and cleaved (43/41- and 37/35-kDa) caspase 8 and caspase 9 are indicated. B. LY294002 pretreatment and Sendai virus infection of HT1080 cells were performed as described in the legend for Fig. 1. Cells were lysed at the indicated time points, and a caspase 8 Western blot assay was performed as described for panel A. C. HT1080 cells were pretreated with 100 μM of the general caspase inhibitor z-VAD, the caspase 8 inhibitor z-IETD, the caspase 9 inhibitor z-LEHD, or the DMSO solvent control 1 hour before virus addition. The PI3 kinase inhibitor LY294002 (50 μM) was added to all cells 30 min before virus addition. Cells were fixed and stained 6 h after infection.

Because of the observed effects of LY on virus-infected cells, we anticipated that SeV infection caused activation of the PI3 kinase pathway, which would lead to AKT phosphorylation. Indeed, phosphorylated AKT was noticeable as early 2 h after infection (Fig. 4A). Similarly, JNK1 and JNK2 were phosphorylated and activated in virus-infected cells; LY treatment did not affect JNK activation (Fig. 4B). LY treatment also did not block IRF-3 activation by SeV infection as measured by the nuclear translocation of IRF-3 (Fig. 4C); the nuclear IRF-3 was transcriptionally active, inducing ISG56 in both LY-treated and untreated cells (Fig. 4D). In another experiment, a dominant-negative mutant of the p110 catalytic subunit of PI3 kinase was expressed to block its activity. As expected, in both SeV-infected and mock-infected cells, AKT phosphorylation by PI3 kinase was inhibited by the expression of the mutant protein (Fig. 4E, middle panel). Consequently, the SeV-infected cells, but not the mock-infected cells, showed early apoptosis (Fig. 4E, upper panel). The above results demonstrated that blocking PI3 kinase activity did not globally affect the cellular response to SeV infection, but the effect was specific for activating the caspase 8-mediated apoptotic pathway.

FIG. 4.

Activation of PI3K is not required for activation of the JNK and TBK1 pathways. A. HT1080 cells were infected with Sendai virus, and whole-cell extracts were prepared at the indicated times. Proteins were separated by SDS-PAGE and transferred to PVDF. Western blot assays were performed with antibodies against AKT and activated, phosphorylated AKT (P-AKT). B. HT1080 cells were pretreated with the PI3 kinase inhibitor LY or vehicle control and subsequently mock infected or infected with Sendai virus as indicated. Whole-cell extracts were prepared 6 h p.i., and Western blot assays were performed with antibodies against JNK and activated, phosphorylated JNK (P-JNK). C. HT1080 cells were pretreated with LY294002 or vehicle control for 30 min and subsequently mock infected or infected with Sendai virus as indicated. Two hours postinfection, cells were fixed, permeabilized, and stained for IRF-3 using a specific antibody. D. HT1080 cells were pretreated with LY294002 and infected with Sendai virus. RNA was harvested 4 h after virus addition. 561 and actin mRNA levels were determined by RNase protection assays. E. HT1080 cells were transfected with the expression vector of an inactive catalytic subunit of PI3K (p110KD), these cells were mock infected or infected with SeV as indicated, and PARP cleavage and P-AKT were analyzed at 8 h p.i.

Requirement of IRF-3, but not NF-κB or IFN signaling, in mediating viral apoptosis.

SeV infection induces the synthesis of type I interferons by activating the transcription factors NF-κB and IRF-3. By using the HT1080 mutant cell line U4C, which lacks functional Jak1 and hence cannot respond to either type I or type II IFNs, we investigated whether IFNs have any role in mediating viral apoptosis. LY-treated U4C cells were very rapidly killed by SeV infection, demonstrating that IFN signaling was irrelevant for the effect (Fig. 5A). At this time, only 6 h after infection, there was no noticeable death of cells that had not been treated with LY, although they all died much later. The possible involvement of NF-κB-mediated gene induction was similarly explored in a genetically modified cell line, 2f-SR, which expresses a high level of a mutant IκB that cannot be phosphorylated and dissociated from NF-κB; consequently, NF-κB-driven genes are not induced in 2f-SR in response to virus infection or other stimuli. SeV infection killed LY-treated 2f-SR cells very efficiently, demonstrating that NF-κB action was not needed for the proapoptotic effects (Fig. 5B). Interestingly, rapid apoptosis of 2f-SR cells still required both virus infection and LY treatment, indicating that in wild-type (WT) cells LY treatment does not induce apoptosis by blocking activation of NF-κB by AKT.

FIG. 5.

Functional Jak/STAT and NF-κB pathways are not needed for apoptosis. A. U4C cells, which lack JAK1 and the interferon response, were pretreated with the PI3 kinase inhibitor LY294002 for 30 min. Cells were then mock infected (control) or infected with Sendai virus in the absence or presence of LY as indicated. Six hours after virus addition, cells were fixed, permeabilized, and stained. B. 2f-SR cells, which constitutively express the IκB superrepressor and subsequently fail to activate NF-κB, were treated as for panel A.

In contrast to the above results, we observed that viral apoptosis absolutely required the presence of IRF-3. P2.1 cells were derived by us from U4C cells, by mutagenesis; these cells express a very low level of IRF-3, and SeV cannot induce ISG56, an IRF-3 target gene, in these cells. P2.1 cells were also refractory to apoptosis by virus infection (Fig. 6A), as were HT1080 cells depleted of IRF-3 by small interfering RNA (siRNA) treatment (data not shown). Expression of exogenous IRF-3 in P2.1 cells (P2.1.17 cell line) restored cell killing by SeV (Fig. 6B), demonstrating that a low level of IRF-3, and not any other defect, made P2.1 cells resistant to apoptosis.

FIG. 6.

IRF-3 is required for the induction of cell death during Sendai virus infection. A. P2.1 cells were pretreated with the PI3 kinase inhibitor LY294002 and subsequently mock infected or infected with Sendai virus as indicated. B. P2.1.17 cells were treated with LY294004 and infected with Sendai virus as for panel A.

Among many cellular genes induced by virus-activated IRF-3 is TRAIL, a known proapoptotic gene. Hence, it was worth investigating whether IRF-3 was exerting its proapoptotic effect through TRAIL induction. The results shown in Fig. 7 indicate that this was not the case. Induction of TRAIL mRNA was relatively slow (Fig. 7A). As expected, TRAIL was induced in WT cells, but not P2.1 cells, in response to SeV, but surprisingly, LY treatment strongly inhibited TRAIL induction by virus infection (Fig. 7B). Finally, TRAIL antibody, added to the culture medium, failed to block viral apoptosis (Fig. 7C). This antibody has been shown to block IFN-β-induced apoptosis of ovarian cancer cells, which is mediated by TRAIL (25).

FIG. 7.

TRAIL is not required for apoptosis by Sendai virus. A. HT1080 cells were infected with Sendai virus as described in Materials and Methods. Cells were collected at the indicated times, and RNA was isolated. The amount of TRAIL message was determined by quantitative reverse transcription-PCR. B. HT1080 and P2.1 cells were pretreated with LY or vehicle control and infected with Sendai virus. Cells were harvested after 6 h, and RNA was isolated. TRAIL message was determined as for panel A. C. HT1080 cells were preincubated with LY294002 (50 μM) and subsequently infected or mock infected with Sendai virus. After infection, cells were washed and placed in complete medium containing LY294002. Anti-TRAIL blocking antibodies or DMSO vehicle control was added to the medium. Cells were fixed after 24 h.

Establishment of viral persistence upon abrogation of IRF-3 activation.

In the next series of experiments, we inquired into the fate of cells that did not undergo apoptosis upon SeV infection. We used three cell lines for this purpose: P2.1 cells, which express little IRF-3, HT1080/siIRF3 cells, in which IRF-3 expression had been ablated by siRNA treatment, and HT1080/RIG-IC cells, in which IRF-3 activation by the RIG-I pathway has been blocked by expression of the dominant-negative inhibitor of RIG-I, RIG-IC. The cells were infected with SeV in the absence of LY treatment. The HT1080/RIG-IC cells were not killed by SeV infection and could be continually passaged. The cells remained infected with the virus, as indicated by the presence of SeV proteins in passage 15 cells (P15), as observed with immunofluorescence (Fig. 8A, right lower panel). From HT1080/siIRF-3 infected cells, two clonal lines were established. Both clones expressed high levels of SeV C protein at P9, as revealed by Western analysis (Fig. 8C, upper panel). The same was true for HT1080/RIG-IC cells at P4 and P5 (Fig. 8C, middle panel) and P2.1 cells at P12 through P16 (Fig. 8C, lower panel). In contrast, within 3 days U4C cells or HT1080 cells, both of which express normal levels of IRF-3, were completely killed by virus infection, even in the absence of LY treatment. In U4C cells, the viral C protein level increased up to 16 h and then started decreasing (Fig. 8B). These results demonstrated that avoidance of apoptosis allowed establishment of persistent infection. Reintroduction of IRF-3 to persistently infected HT1080/siIRF3 restored cell killing as indicated by PARP cleavage (Fig. 8D). It was possible to carry out the experiment shown in Fig. 8D because the siRNAs to IRF-3 were directed to the untranslated regions of IRF-3 mRNA. Thus, an mRNA without the untranslated regions expressed from transfected IRF-3 vectors was not targeted by the siRNA, and IRF-3 was efficiently expressed. The above results showed that IRF-3 served as the regulatory switch between apoptosis and persistence. Finally, after long passages, infectious virus production by the three persistently infected cell lines was measured based on viral titers in the culture medium. All three lines produced significant levels of infectious virions in the culture medium, demonstrating the establishment of true persistence (Table 1).

FIG. 8.

Abrogation of the RIG-I/IRF-3 pathway leads to persistent infection. A. HT1080/RIG-IC/SeV cells (P15) or their matching mock-infected control cells were plated on coverslips and stained with an antibody against Sendai virus. Nuclei were visualized by DAPI staining. B. U4C cells, which lack JAK1 and the interferon response but express normal levels of IRF-3, were infected with SeV, total proteins extracts were separated by SDS-PAGE, and Western blot assays were performed with an antibody against the Sendai virus C protein at the indicated times after infection. C. Cell lysates were prepared from the three persistently infected cell lines or their matched mock-infected controls. Proteins (40 μg) were separated by SDS-PAGE, and Western blot assays were performed with an antibody against the Sendai virus C protein. D. HT1080/siIRF-3 clone 1 cells were transfected with an IRF-3 expression plasmid. As the RNAi targeting vector recognizes the 5′ and 3′ untranslated regions, IRF-3 cDNA expressed from a plasmid is not repressed. After 3 hours, the transfection reagents were removed, and the cells were washed and placed in complete medium. Cell lysates were prepared after 2 days, and a C-PARP Western blotting was performed as described for Fig. 1B.

TABLE 1.

Absence of IRF-3 activation establishes persistence

Persistently infected cell line	Characterization	Passage no.	Viral titer (PFU/ml)
HT1080/siIRF-3/SeV	IRF-3 ablated	21	2.4 × 103
HT1080/RIG-Ic/SeV	Dominant-negative RIG-I	20	2.1 × 105
P2.1/SeV	Very low IRF-3	22	5.0 × 104

DISCUSSION

Synthesis of our new observations with those in the literature led to the formulation of the working model presented in Fig. 9. Infection with SeV, and presumably other paramyxoviruses, causes the activation of a proapoptotic arm and an antiapoptotic arm of cellular responses. The proapoptotic arm requires viral gene expression and the activation of IRF-3 by the cytoplasmic RIG-I pathway; activated IRF-3 induces the expression of many cellular genes, one or more of which causes apoptosis. In the antiapoptotic arm, infection causes PI3K activation and resultant Akt activation, which is required for robust viral gene expression and replication. We hypothesize that a viral or a cellular antiapoptotic factor, whose synthesis or activation needs Akt, provides the antiapoptotic signal in the infected cells. In the normal situation, the dynamic equilibrium of cell survival shifts from anti- to proapoptotic as the infection progresses. We have tested many of the salient features of this model in this study. Without the presence of IRF-3 or its activation, the proapoptotic arm is eliminated; infected cells did not undergo apoptosis at all, even when PI3K was inhibited. For triggering the antiapoptotic arm, activation of PI3K and Akt was essential. When the brake provided by this arm was taken away by inhibiting PI3K, the cells were killed very rapidly, but only if IRF-3 was activated. Thus, the absence of PI3K activity only accelerates the apoptotic process without changing its requirements. Although in many experiments we have used the LY inhibitor to enhance apoptosis to observe it at 6 h postinfection (p.i.), essentially similar results would have been obtained in untreated infected cells, but at a much later time of 48 h p.i. This point is convincingly made in experiments shown in Fig. 8 and Table 1, in which no LY treatment was used but all of the infected WT cells died nonetheless. In contrast, IRF-3-deficient cells became persistently infected and produced infectious virus continuously.

FIG. 9.

SeV-induced apoptotic pathway. Infection of cells with SeV activates IRF-3, which in turn induces expression of virus stress-inducible genes and causes host cell apoptosis. In addition, virus infection also activates the PI3 kinase pathway, which is needed for replication of virus and generation of antiapoptotic signal in the host cells. Blocking the PI3 kinase pathway by chemical inhibitors induces early apoptosis of the host cells. A suitable balance between the apoptotic and antiapoptotic signals dictates the fate of the infected cell; however, the apoptotic signal is determined by the activation of IRF-3.

Cellular apoptosis caused by viruses is thought to contribute to the spread of the infection by facilitating the release and dissemination of progeny virions. For the benefit of the virus, the apoptosis must await viral replication and assembly of new virions. However, as a defense mechanism of multicellular hosts against widespread viral infection, premature apoptosis of initially infected cells is an effective antiviral strategy. Thus, the timing of apoptosis of infected cells can be a crucial determinant of the fate and the severity of the overall infection process in an organism. Often this process is regulated by antiapoptotic viral gene products. Many viruses produce both pro- and antiapoptotic proteins in a temporally regulated fashion so that cells are killed at the right time, after viral replication and before the host immune response. Our study indicates that PI3 kinase activity is an important sensor for mediating this level of regulation: its functional inactivation caused premature apoptosis of the infected cells. The involvement of the PI3 kinase pathway in cellular apoptosis by other viruses has been reported. Cytomegalovirus major immediate-early proteins and simian virus 40 large T antigen block apoptosis by activating the PI3 kinase pathway and the downstream target, Akt, and treating cells with LY eliminated the antiapoptotic effects of the DNA viral proteins. It is clear that viral gene expression is required for triggering the rapid apoptosis, because UV inactivation of the infectious virions blocked this effect. We did not investigate the nature of the viral gene product responsible for the rapid apoptosis, but there is information in the literature regarding how the viral gene product, trailer RNA (trRNA), negatively regulates apoptosis (13). The SeV trRNAs are short transcripts generated during abortive gene replication (13). Viral apoptosis is blocked by trRNA, which sequesters a cellular proapoptotic RNA-binding protein, TIAR. Conversely, overexpression of TIAR enhances SeV-mediated apoptosis. Sun et al. (32) reported that PI3K activity is needed for many steps of virus replication; it will be interesting to examine whether it is needed for trRNA synthesis as well.

We have used genetically modified cell lines to analyze the signaling pathways required for the rapid apoptosis. One of the cell lines was the U4C line, which is derived from HT1080 cells but lacks functional JAK 1, a tyrosine kinase essential for signaling by type I and type II IFNs (23). Even though these cells cannot respond to IFN-α, IFN-β, or IFN-γ, they were killed very efficiently by SeV infection and LY treatment, demonstrating that the virus-induced IFNs are not required for triggering the rapid apoptosis and JAK 1 itself is not involved in the apoptotic signaling pathway. Because viral infection activates the NF-κB pathway, which is known to protect cells from apoptosis (20), it remained a possibility that interference with this pathway was causing rapid apoptosis. This possibility was attractive, because PI3 kinase has been shown to be involved in mediating complete activation of NF-κB in response to the cytokine interleukin-1 (31). In the presence of LY, interleukin-1 can activate the release of NF-κB from its inhibitor, IκB, and the released protein can bind its cognate DNA in a gene, but it cannot activate transcription of the gene. For acquiring its full activating potential, the released NF-κB needs to be phosphorylated by a kinase that, in turn, requires the PI3 kinase pathway for its activation. Thus, it was possible that in the presence of LY, SeV infection could not activate the antiapoptotic NF-κB pathway and hence promote rapid apoptosis. To examine this possibility, we resorted to another genetically manipulated cell line, 2f-SR. This cell line overexpresses a nonphosphorylatable mutant of IκB, the superrepressor of NF-κB (27). Consequently, in these cells, activation of IκB kinases by any stimulus does not release NF-κB from IκB, making them functionally NF-κB null cells. If the NF-κB pathway were involved in protecting cells from rapid apoptosis by SeV infection, these cells would be hypersensitive to apoptosis. This was not the case; neither SeV nor LY alone killed these cells, although together they were effective. These results demonstrated that the observed effects of PI3 kinase were not mediated by its ability to block NF-κB activation.

In contrast to the absence of any roles of the IFN or the NF-κB signaling pathway in the process of viral apoptosis, the IRF-3 pathway had an essential role. To establish this role, we took advantage of two other cell lines established in our laboratory. P2.1 cells, which were derived from U4C cells, are not only deficient in IFN signaling but are also devoid of responses to dsRNA (19). The defect in one branch of the dsRNA signaling pathway in P2.1 cells was traced to a very low level of IRF-3 because of its rapid degradation (27). In P2.1.17 cells, restoration of IRF-3 by its ectopic expression restored the IRF-3 signaling pathway without alleviating the defects in other dsRNA signaling pathways, including NF-κB, JNK, and p38 (27). Consequently, IRF-3-dependent genes were efficiently induced by dsRNA or SeV in P2.1.17 cells but not in P2.1 cells. As shown in Fig. 6, apoptosis followed the same pattern; the P2.1 cells were resistant to the process, whereas the P2.1.17 cells were efficiently killed. This experiment provided conclusive genetic evidence that IRF-3, which is known to be activated by SeV infection, is absolutely required for the observed apoptosis. The protein kinases JNK1 and JNK2 were equally activated by SeV infection in the presence or absence of LY, and IRF-3 itself was equally activated, as judged by its nuclear translocation, under the two conditions. Transcriptional induction of the 561 mRNA, which encodes the viral stress-inducible protein p56 (11), requires the activation of IRF-3 (27). We have previously reported (29) that IRF-3 activation by the TLR3 pathway requires PI3K activity; without it, IRF-3 is only partially phosphorylated and cannot induce 561 mRNA. Obviously, this was not the case for the SeV-activated RIG-I pathway, because LY did not inhibit 561 mRNA induction. Thus, the observed effect of LY on apoptosis is most probably not mediated by affecting IRF-3 phosphorylation.

The biochemical pathways leading to cellular apoptosis are complex. The cellular proteases, caspases, are often involved in this process (13). This was true for the observed apoptosis in response to SeV infection; a universal inhibitor of all caspases, z-VAD, blocked the process. The caspases function in cascades, two branches of which are initiated independently by caspase 8 and caspase 9 (13). Both branches converge at caspase 3, the executioner caspase, by cleaving and activating it. The same pathways appear to operate in the SeV-induced apoptosis process. The MCF-7 cell line, which lacks functional caspase 3 because of disruption of its gene (15), was resistant to Sendai virus plus LY-mediated apoptosis (data not shown), demonstrating the need for caspase 3. The process was also blocked by the inhibition of caspase 8 but not caspase 9. Thus, it appears that the caspase8/caspase 3 pathway is used for mediating the observed rapid apoptosis. In contrast, although caspase 9 is also activated, as indicated by its cleavage, its action is not needed for this process. Further investigation will be required to fully delineate the mechanism of activation of the caspase 8 pathway by SeV infection.

There are several possible mechanisms by which IRF-3 might be inducing apoptosis of SeV-infected cells. It was apparent that “activation” of IRF-3 by the RIG-I pathway was required, because in cells expressing IRF-3 and RIG-IC there was no apoptosis upon virus infection. Activated IRF-3 is known to induce expression of many cellular genes. Products of one or more of these genes could be proapoptotic, but the possible involvement of one of the most attractive candidates, TRAIL, was ruled out by our experiments. If this mechanism is correct, more investigation will be needed to identify the responsible IRF-3-induced proapoptotic protein. An alternative mechanism is that IRF-3 blocks the induction of specific antiapoptotic genes, a model supported by our observation that IRF-3 can block induction of selected NF-κB-driven genes, including many antiapoptotic genes (5). Finally, it remains possible that the proapoptotic function of IRF-3 is mediated by its role in a process other than regulation of gene induction. A relevant interesting fact is that IRF-3 activation by the TLR3 pathway, in contrast to the RIG-I pathway, does not lead to apoptosis (unpublished observation). Hence, it remains possible that additional cellular or viral factors are required for manifesting the apoptotic effect of IRF-3.

As experimentally demonstrated here, the absence of IRF-3 led to viral persistence. This observation has important physiological significance, because IRF-3 is rapidly degraded upon SeV infection. Hence, it is likely that a few infected cells, in which the IRF-3 level is sufficiently low because of its degradation, can escape apoptosis and be persistently infected. The cell population will be self-perpetuated, because progeny expressing high IRF-3 will be eliminated from the population, as mimicked in our experiment with exogenously expressed IRF-3 in persistently infected cells. Another relevant factor is that although IRF-3 is constitutively expressed in all cells and tissues, the levels of expression are highly different. For example, in T cells, IRF-3 expression is about 30 times higher than that in kidney cells (unpublished observation). Thus, it is possible that in a SeV-infected organism, different cell types will be differentially susceptible to IRF-3-mediated apoptosis.

It was somewhat unexpected that in all cell lines tested here, inhibition of apoptosis was sufficient for establishing persistent infection. The cell lines lacking IRF-3 or its activation were not only infected with SeV but also produced infectious virus continuously. What connects blockage of apoptosis and establishment of persistent infection remains to be explored. Another interesting issue to address is the status of the IRF-3-independent innate response of persistently infected cells. Are they in a perpetually activated state with, for example, high NF-κB activity, or is their state at the basal level of activity manifested in uninfected cells? Now that we have the means of switching between persistent and lytic infections in the same cell-virus combination, many of these relevant questions will be experimentally approachable.