RIG-I Enhanced Interferon Independent Apoptosis upon Junin Virus Infection

Junin virus (JUNV) is the etiological agent of Argentine hemorrhagic fever (AHF), a human disease with a high case-fatality rate. It is widely accepted that arenaviral infections, including JUNV infections, are generally non-cytopathic. In contrast, here we demonstrated apoptosis induction in human lung epithelial carcinoma (A549), human hepatocarcinoma and Vero cells upon infection with the attenuated Candid#1 strain of, JUNV as determined by phosphatidylserine (PS) translocation, Caspase 3 (CASP3) activation, Poly (ADP-ribose) polymerase (PARP) cleavage and/or chromosomal DNA fragmentation. Moreover, as determined by DNA fragmentation, we found that the pathogenic Romero strain of JUNV was less cytopathic than Candid#1 in human hepatocarcinoma and Vero, but more apoptotic in A549 and Vero E6 cells. Additionally, we found that JUNV-induced apoptosis was enhanced by RIG-I signaling. Consistent with the previously reported role of RIG-I like helicase (RLH) signaling in initiating programmed cell death, we showed that cell death or DNA fragmentation of Candid#1-infected A549 cells was decreased upon siRNA or shRNA silencing of components of RIG-I pathway in spite of increased virus production. Similarly, we observed decreased DNA fragmentation in JUNV-infected human hepatocarcinoma cells deficient for RIG-I when compared with that of RIG-I-competent cells. In addition, DNA fragmentation detected upon Candid#1 infection of type I interferon (IFN)-deficient Vero cells suggested a type I IFN-independent mechanism of apoptosis induction in response to JUNV. Our work demonstrated for the first time apoptosis induction in various cells of mammalian origin in response to JUNV infection and partial mechanism of this cell death.


Introduction
Arenaviruses are bisegmented, negative sense RNA viruses with enveloped virions that use an ambisense coding strategy [1]. The large segment of the arenavirus genome encodes for a RNAdependent RNA polymerase (L) with endonuclease cap snatching activity [2] and a small RING finger protein (Z) with matrix-like functions [3]. The small segment encodes for the viral nucleoprotein (NP), endowed with a 39 to 59 exoribonuclease activity [4], and the glycoprotein precursor (GPC). Co-translational cleavage of GPC by the cellular signal peptidase produces a 58 amino acid stable signal peptide (SSP), and subsequent posttranslational cleavage by the cellular site 1 protease produces peripheral virion attachment (GP1) and fusion-active transmembrane (GP2) proteins [5]. The members of the Arenaviridae family are divided into two serologically and geographically distinct groups: New and Old World arenaviruses [6]. Five members of the New World clade B arenaviruses (Junin, Guanarito, Sabia, Machupo and Chapare), Cytopathic effect (CPE) in vitro has been reported only in response to infections by non-pathogenic arenaviruses. On the other hand, arenaviruses associated with hemorrhagic diseases in humans, including JUNV, are generally considered to be noncytopathic viruses [21][22][23]. The non-pathogenic clade B arenavirus Tacaribe, but not the pathogenic Romero strain of JUNV, was documented to induce pronounced CPE in Vero cells and conglomeration of human blood purified monocytes [22]. Exposure of PS, a phospholipid component kept on the innerleaflet of cell membranes in normal cells, on the surface of transformed mouse monocytes/macrophages infected with Pichinde virus (a New World arenavirus that is non-pathogenic for humans) has been described [24]. A recent report [23] documented an absence of apoptosis induction in Vero E6 cells infected with the Romero strain of JUNV. The lack of apoptosis in these cells was proposed to be mediated by the caspase decoy function of Romero NP [23].
A safe and effective live-attenuated JUNV vaccine (Candid#1) is licensed in Argentina and has been used with success within the JUNV endemic area to prevent AHF [25]. However, the documented genetic and virulence heterogeneity of Candid#1 [26], and the lack of understanding of the mechanisms underlying Candid#1 attenuation pose great barriers to its acceptance in the United States. Compared with its parental, as well as with other virulent JUNV strains, Candid#1 contains multiple amino acid changes in GP, NP and L that hinder the identification of the genetic markers of attenuation [27][28][29][30][31][32].
We have recently documented an induction of type I IFN in response to pathogenic, Romero, and attenuated, Candid#1, strains of JUNV infection in human lung epithelium carcinoma cells (A549) [33]. We also showed that siRNA-mediated downregulation of RIG-I or IRF3 production in A549 cells resulted in drastic reduction of STAT1 phosphorylation and IFN-stimulated gene (ISG) induction in response to Candid#1 infection. These data revealed RIG-I as a primary trigger of type I IFN signaling in response to JUNV infection in A549 cells [33]. We also observed that the ISG response was substantially reduced at the mRNA level in A549 cells infected with the virulent Romero strain compared with that of Cadid#1-infected cells, indicating that Candid#1 might be a more potent stimulator of the RIG-I/IRF3 signaling pathway than Romero. However, both viruses were resistant to the antiviral effect of type I or II IFN pre-treatment in Vero cells indicating that the attenuation of Candid#1 is not related to a higher susceptibility to the antiviral status induced by these cytokines [33].
Initiation of apoptosis in response to viral infection [34] has been linked to RLH signaling independent of type I IFN pathway activation [35][36][37][38][39]. Specifically, IFN-I-independent, caspase-9, Apaf-1-dependent signaling trough RIG-I/MDA5-MAVS-mediated induction of Puma and Noxa transcription [36,40]; or MAVS mediated IFN-I, IRF3, NF-KB-independent, caspase-3, -9-dependent [35]; or IFN-I, NF-KB-independent, RIG-I-TRAF3/ TRAF2/TRAF6-mediated IRF3 interaction with Bax protein [38,39] has been shown to activate mitochondrial apoptotic pathway in response to dsRNA or viral infection. We observed that infection with Candid#1 JUNV induces CPE in primate cell lines. These findings led us to hypothesize that Candid#1 infection induces a RLH-mediated type I IFN-independent apoptosis. In this study we have used different experimental approaches to test our hypothesis. Here we demonstrated apoptosis induction in two human cell lines and Vero cells upon infection with the attenuated Candid#1 vaccine strain of JUNV as determined by PS translocation, CASP3 activation, PARP cleavage and/or chromosome DNA fragmentation. Furthermore, to investigate the potential role of apoptosis induction in Candid#1 attenuation or JUNV pathogenesis we assayed cell death induction in response to the pathogenic Romero as well. We found that the pathogenic Romero strain of JUNV was less cytopathic than Candid#1 in human hepatocarcinoma and Vero cells as determined by DNA fragmentation. However, stronger apoptosis induction was observed in A549 and Vero E6 cells in response to Romero infection. Additionally, we found that JUNVinduced apoptosis was enhanced by RIG-I signaling but independent of production of type I IFN.

Viruses
Romero (GenBank accession nos. AY619640 and AY619641) [44] and Candid#1 (GenBank accession no. U70801) strains of JUNV were obtained from Dr. Thomas G. Ksiazek (Centers for Disease Control and Prevention, Atlanta, GA) and Dr. Robert Tesh (The World Reference Center for Emerging Viruses and Arboviruses (WRCEVA), University of Texas Medical Branch, Galveston, TX), respectively. Virus stocks were prepared using Vero cells. Cell debris in supernatants were filtered out through 0.45 mm HV Durapore Membrane Stericup sterile vacuum filtration system (Millipore Corporation, Billerica, MA). Cleared supernatants were concentrated through 30 min centrifugation at 32206g using Amicon Ultra-15 Centrifugal Filter Unit PLHK Ultracel-PL Membrane, 100 kDa (Millipore Corporation, Billerica, MA). All work with JUNV Romero was performed at the University of Texas Medical Branch BSL-4 facilities (Robert E. Shope Laboratory or the Galveston National Laboratory) in accordance with institutional health and safety guidelines and federal regulations [45].

Plaque Assay on Vero Cells
Cells were plated in 12-well multiwell plates. Ten-fold supernatant dilutions were added to cell monolayers for 1 h at the 37uC, 5% CO 2 . Following incubation, cells were overlaid with MEM containing 0.6% gum tragacanth, 1% fetal bovine serum, and 1% Penicillin-Streptomycin (10,000 U/mL). Following 8 days incubation, cells were fixed with 10% formaldehyde and stained with crystal violet.

Cell Viability Determination
Cells (1610 4 ) were mock-infected or infected with JUNV in quadruplicates or triplicates. Cell viability was assayed using Cell Growth/Viability Determination Kit MTT Based (Sigma-Aldrich, St. Louis, MO) according to the manufacturer's instruction.

Statistical Analysis
Data were analyzed by two-way or three-way ANOVA using SigmaPlot 12.0 (Systat Software, Inc., San Jose, CA).

A549 Cells Undergo Apoptosis during Candid#1 Infection
A549 cells were mock-infected or infected with Candid#1 and cell viability was determined by an MTT-based viability assay. Viability of Candid#1-infected cells was decreased compared with that of mock-infected controls starting from 2 days p.i. and reached over 42% reduction by day 4 p.i. (Fig. 1A). The highest level of Candid#1-induced reduction in cell viability correlated with a peak production of infectious virus progeny as determined by plaque assay (Fig. 1A).
Next we examined whether the reduced survival observed in Candid#1-infected cells was caused by apoptotic cell death. For this we assessed PS flipping from the inner to the outer layers of cell membrane, which was detected via Annexin V binding. In addition, we examined cell membrane integrity by staining with an amine reactive L10120 viability dye. Stained cells were analyzed by FACS. Transition from early apoptotic (Annexin V+ and L101202) to late apoptotic state (Annexin V+ and L10120+) is characteristic of apoptotic cell death [46]. In mock-infected samples, the percentages of early (Fig. 1Bi, ii) and late (Fig. 1Bi, iii) apoptotic cells remained largely unchanged from day 1 to day 4 p.i. and were 6.7% and 0.3% on average, respectively. Induction of early apoptosis in Candid#1 infected cells was detected 2 days p.i. (Fig. 1Bi, ii). Likewise, the percentage of late apoptotic cells in Candid#1-infected samples increased from 5.2% at day 3 p.i. to 18.3% at 4 days p.i. (Fig. 1Bi, iii).
Fragmentation of chromosomal DNA (mono-and oligonucleosome formation) occurs at the late stage of apoptosis as a result of DNA cleavage by activated endonucleases [47]. We quantified cytoplasmic histone-associated-DNA-fragments through ELISA. Induction of DNA fragmentation in Candid#1infected cells was 1.2-, 2.2-and 1.9-fold of that in mock-infected cells at days 2, 3 and 4 p.i., respectively (Fig. 1C), a finding consistent with the reduced cell viability observed in virus infected samples (Fig. 1A).
Consistent with the increased level of Annexin V staining and DNA fragmentation, we also detected in Candid#1-infected cells the large fragment of cleaved executioner CASP3 and cleaved PARP, a CASP3 downstream substrate (Fig. 1D) by western blotting using cleaved CASP3 and PARP antibody as described [33]. Treatment with camptothecin was included as a positive control of apoptosis induction.
These data indicate that infection of human lung epithelium carcinoma cells with the attenuated strain of JUNV induces apoptosis.

RIG-I-enhanced Apoptosis in Response to JUNV Infection
Recently we reported that infection with JUNV, both the pathogenic Romero and live-attenuated Candid#1 vaccine strains, activated the RIG-I/IRF3 signaling pathway as well as IFN-type I signaling in A549 cells [33]. Previously, RLH signaling has been linked to apoptosis induction [35][36][37][38][39]. We, therefore, were interested in examining a possible role of RIG-I signaling in apoptosis induction during Candid#1 and Romero infection. For this we first used a siRNA-based approach to down-regulate expression levels of RIG-I and IRF3 in A549 cells as described [33]. Initially we explored the feasibility of this approach using only Candid#1. At 1.5 days post siRNA transfection, cells were mock-infected or infected with Candid#1. Consistent with previous findings [33] both mock-and Candid#1-infected cells transfected with IRF3-specific siRNA exhibited efficient silencing of IRF3 expression (70-80% reduction) at 1 and 2.5 days p.i. as determined by western blotting (Fig. 2A). Similarly to our published data [33], RIG-I expression was induced by Candid#1 1 at 1 and 2.5 days after infection, since RIG-I is an IFNstimulated gene. The induction was observed even in the cells transfected with RIG-I-specific siRNA (Figure 2A). Induction of RIG-I expression in response to Candid#1 infection occurred at the time as apoptotic changes became detectable; that made it experimentally difficult to examine the extent of RIG-I contribu- tion to the apoptosis induction. Therefore, we decided not to perform the same experiment with Romero virus in the BSL-4 laboratory. Nevertheless, even the transient knockdown of RIG-I, and to the lesser extent of IRF3, resulted in increased cell viability of Candid#1-infected cells. At 2.5 days p.i. cell viability was 55.461.8 and 67.162.4% in IRF3 and RIG-I knockdown cells, respectively, versus 48.661.4% in Candid#1-infected cells transfected with control siRNA (Fig. 2B). Increased cell survival in RIG-I and IRF3 knockdown Candid#1-infected cells was observed despite 9.3-and 3.7-fold higher virus production, respectively, as compared with that of cells transfected with the control siRNA (Fig. 2C). This observation suggests that the increased cell viability was not due to a reduced viral replication.
Next, we examined whether Romero induced levels of cell apoptosis similar to those observed in Candid#1-infected cells, and whether RIG-I signaling influenced also apoptosis in Romeroinfected cells. To achieve a long-term down-regulation of RIG-I expression we generated A549 cells stably transduced with a lentivirus expressing either a RIG-I-targeting shRNA (RIG-I KD) or a control non-targeting shRNA (Control KD). To confirm target knockdown, RIG-I KD and Control KD cells were transfected with Poly(I:C) and cell lysates were examined by western blotting. Induction of RIG-I expression upon poly(I:C) treatment was detected in Control KD but not in RIG-I KD cell lysates (Fig. 3A).
RIG-I KD and Control KD lines were infected with Candid#1 or Romero JUNV or mock-infected, and assessed for cell apoptosis by determining levels of DNA fragmentation. At 4 days p.i. we observed increased levels of DNA fragmentation in Control KD cells infected with either Candid#1 (1.8-fold) or Romero (1.5-fold) compared with RIG-I KD cells infected with the corresponding viruses (Fig. 3B). Reduced DNA fragmentation in the infected RIG-I KD cells was observed despite a 5.8-fold increased production of infectious progeny over that observed in infected Control KD cells at 4 days p.i. (data not shown).
In order to exclude the possibility that the apoptotic response to JUNV infection was cell type-specific, we also mock-infected or infected with Candid#1 or Romero JUNV human hepatocarcinoma cells with a functional (Huh7) or non-functional (Huh7.5) form of RIG-I [43,48]. DNA fragmentation level was significantly higher in JUNV-infected Huh7 cells than in Huh7.5 cells (4.3-and 3.7-fold for Candid#1 and 3.3-and 2.6-fold for Romero virus) at 3 and 4 days p.i., respectively (Fig. 3C). Production of infectious virus was similar in both cell lines (data not shown).
Together our findings indicate that an active RIG-I signaling pathway enhances apoptosis during JUNV infection of human cells.

Type I IFN-independent Apoptosis Induction in Response to JUNV Infection
We have shown an induction of type I IFN signaling upon JUNV infection of human cells [33]. Type I IFN signaling has also been linked to induction of apoptosis in response to viral infection [49]. To determine the role of type I IFN in JUNV-induced apoptosis, we analyzed DNA fragmentation and virus production in two type I IFN-deficient cells of non-human primate origin: Vero [41] and its clone Vero E6 [42]. Cells were mock-infected or infected with Candid#1 or Romero. Production of infectious progeny of both Candid#1 and Romero viruses was similar in both cell lines (data not shown). Infection with Candid#1 virus induced cytoplasmic nucleosomes in Vero cells starting at day 2 p.i. Induction of DNA fragmentation in Candid#1-infected Vero cells was 5.8-, 15.2-and 9.2-fold higher than that in mock-infected cells at days 2, 3 and 4 p.i., respectively. Romero infection of Vero E6 cells led to detectable cytoplasmic nucleosome formation (1.9and 5.3-fold of that in mock-infected Vero E6 cells at days 3 and 4 p.i., respectively). Though, DNA fragmentation in these cells was delayed (3-4 days p.i.) and reduced, relative to that of Candid#1 infection of Vero cells (Fig. 3D). Detection of increased levels of cytoplasmic nucleosomes in type I IFN-deficient Vero cells upon Candid#1 and VeroE6 cells upon Romero infection suggests that type I IFN production is not required for JUNV-induced apoptosis. In contrast, no DNA fragmentation was detected in Romero-infected Vero cells or Candid#1-infected Vero E6 cells (Fig. 3D).

Discussion
CPE in vitro has been documented previously in mammalian cells infected with non-pathogenic arenaviruses Tacaribe [22] and Pichinde [24]. In agreement with that in the current study we have demonstrated an induction of cell death in human cells upon infection with vaccine strain of JUNV, Candid#1. Moreover, for the first time, we confirmed the apoptotic nature of this cell death using four different experimental approaches: 1) PS flipping (detected via Annexin V binding) from the inner to the outer membrane layer of cell that at the same time exclude viability dye is indicative of early apoptosis; 2) Transition over time from early (Annexin V positive, viability dye negative) to late apoptotic state (Annexin V and viability dye positive) provides confirmation of cell death via the apoptotic pathway [46]; 3) Fragmentation of cell DNA as a result of nuclease activation, which is a hallmark of late apoptosis [47], results in mono-and oligo-nucleosome formation that can be quantified using ELISA and 4) Detection of the cleaved CASP3 and PARP is a well-accepted surrogate of activation of the apoptotic cell death pathway. Together these multiple measurements strongly suggest that JUNV Candid#1 infection induces cellular apoptosis. In addition, we detected DNA fragmentation in human hepatocarcinoma and non-human primate Vero cells infected with Candid#1 JUNV.
Moreover, we observed DNA fragmentation in three types of mammalian cells (human lung epithelial carcinoma, human hepatocarcinoma and non-human primate Vero E6 cells) in response to infection with pathogenic Romero strain and no detectable mono-and oligo-nucleosome formation in Romeroinfected Vero cells. The magnitude and kinetics of apoptosis induction in Huh7 and Vero cells were stronger upon infection with attenuated strain of JUNV. It seems conceivable that an induction of apoptosis upon Candid#1 infection in cells of mononuclear lineage, JUNV primary target [50][51][52][53][54] or parenchymal cells, may contribute to the host antiviral response by limiting virus replication and spread, as well as increasing clearance and immunogenicity of infected cells [46,55,56]. For instance, immunogenicity of apoptotic cancer cells has been attributed to the exposure of calreticulin, an endoplasmic reticulum chaperon, on the cell surface during early apoptosis. TLR4 on immature DCs recognizes calreticulin, stimulating antigen processing and presentation. The release of high-mobility group box 1 (HMGB1) chromatin-binding protein to the extracellular space during late apoptosis has the same effect [55,56]. Moreover, mouse macrophages have been shown to specifically phagocytose apoptotic mouse thymocytes with PS on the outer leaflet of the plasma membrane [46]. At the same time, a pathogenic role of apoptosis induction in response to viral infections has been documented [57,58]. In macaque, guinea pig and type I and II IFN receptor deficient mouse models of Argentine hemorrhagic fever multiple tangible body macrophages have been detected in spleen of infected animals [59][60][61]. Germinal center tingible body macrophages contain stainable condensed chromatin fragments of phagocytized, apoptotic cells [62]. Moreover, chromatolysis and pyknosis in neurons suggestive of neuronal apoptosis and/or necrosis was detected in a study of 10 autopsy cases of AHF [63]. These observations do not indicate that infected cells undergo apoptosis, however, they suggest a possible pathogenic role of apoptosis in JUNV infection.
IFN-I independent RLH-mediated induction of apoptosis in response to dsRNA, RNA and DNA viruses has been documented [35][36][37][38][39]. Likewise, deficiencies in RLH or apoptotic pathways often result in enhanced viral replication or pathogenicity in cultured cells and animal models [34,39]. Accordingly, siRNAmediated down-regulation of RIG-I and IRF3 expression increased viability of Candid#1-infected A549 cells despite enhanced viral production. Transient effect of siRNA knockdown and the ISG nature of RIG-I could have contributed to the moderate increase we observed in cell viability and virus production in infected cells. We also found drastically reduced DNA fragmentation in RIG-I deficient A549 RIG-I KD and Huh7.5 cells infected with JUNV relative to that of the corresponding infected RIG-I competent controls. Our data indicate that RIG-I contributes to induction of the programmed cell death in response to JUNV infection. Supporting type I IFN independent mechanism of apoptosis induction in response to JUNV infection, we detected DNA fragmentation in Candid#1or Romero-infected type I IFN-deficient Vero or VeroE6 cells, respectively.
Our observation of detectable levels of DNA fragmentation in Romero-infected Vero E6 cells appears to contradict the recent report [23], which shows the lack of apoptosis in Romero virus infected Vero E6 cells. These seemingly conflicting findings may be related to the sensitivity of the assays used to evaluate apoptosis. In the aforementioned study the absence of apoptosis in Romeroinfected Vero E6 cells was based on the lack of visible chromatin condensation by microscopic examination after DAPI staining and undetectable PARP/CASP3 cleavage as determined by western blotting [23]. In contrast, we have used an ELISA-based assay that allowed us to detect and quantify formation of mono-and oligonucleosome in low number of cells (assay detection limit is 1000 cells). These data illustrate the need of using several alternative and carefully characterized assays to study the interplay between JUNV infection and host apoptosis response.
In spite of conservation of proposed CASP cleavage motifs [23] in NP of Candid#1, infection with this virus induced cytoplasmic nucleosomes in Vero and human hepatocarcinoma cells. Additionally, we documented apoptosis induction in human lung epithelial carcinoma cells in response to Candid#1 infection using four different experimental approaches. Similarly, we detected pronounced DNA fragmentation in two human cell lines upon Romero infection. Only in Vero E6 cells we observed a lack of considerable apoptosis upon Candid#1-infection. Likewise, DNA fragmentation was significantly reduced in Romero-infected Vero E6 cells relative to that of Candid#1-infected Vero cells. These results suggest that absence/reduction of apoptosis induction in JUNV-infected Vero E6 cells could be a clone-specific phenomenon. It is possible that a host factor required for Candid#1 induced apoptosis is present in the majority of Vero cells, but absent in the sub-population of cells that were selected as Vero E6 clone. Similarly to our data, CPE in response to Middle East Respiratory Syndrome coronavirus infection was significantly delayed and less profound in Vero E6 cells compare with that of Vero cell [64]. Future studies of molecular defect(s) in Vero E6 should help to address the inability of these cells to undergo apoptosis upon infection with JUNV.
In summary, we have presented evidence of apoptosis induction in response to JUNV infection. Vaccine strain of JUNV Candid#1 induces more potent apoptosis than virulent Romero strain in human hepatocarcinoma and Vero cells, but not in A549 or Vero E6 cells. We have also shown that RIG-I, independent of type I IFN, enhances apoptosis in response to JUNV infection. Future studies should address the detailed molecular mechanism underlying JUNV induced apoptosis and its contribution to virus attenuation or pathogenesis.