FLAVIVIRUS NS4A-INDUCED AUTOPHAGY PROTECTS CELLS AGAINST DEATH AND ENHANCES VIRUS REPLICATION

Flaviviruses include the most prevalent and medically challenging viruses. Persistent infection with flaviviruses of epithelial cells and hepatocytes that do not undergo cell death is common. Here, we report that, in epithelial cells, up-regulation of autophagy following flavivirus infection markedly enhances virus replication and that one flavivirus gene, NS4A, uniquely determines the up-regulation of autophagy. Dengue-2 and Modoc (a murine flavivirus) kill primary murine macrophages but protect epithelial cells and fibroblasts against death provoked by several insults. The flavivirus-induced protection derives from the up-regulation of autophagy, as up-regulation of autophagy by starvation or inactivation of mammalian target of rapamycin also protects the cells against insult, whereas inhibition of autophagy via inactivation of PI3K nullifies the protection conferred by flavivirus. Inhibition of autophagy also limits replication of both Dengue-2 and Modoc virus in epithelial cells. Expression of flavivirus NS4A is sufficient to induce PI3K-dependent autophagy and to protect cells against death; expression of other viral genes, including NS2A and NS4B, fails to protect cells against several stressors. Flavivirus NS4A protein induces autophagy in epithelial cells and thus protects them from death during infection. As autophagy is vital to flavivirus replication in these cells, NS4A is therefore also identified as a critical determinant of flavivirus replication.

Over 2.5 billion people live in areas at high risk for Dengue and related flavivirus infections [1]. Dengue hemorrhagic fever (DHF), a severe complication present in ~5% of cases, claims more lives annually than all other hemorrhagic fevers combined [2]; much of the damage is caused by death of infected cells. The fate of infected cells depends on cell type. Although flavivirus induces apoptosis of neurons and macrophages, infected hepatocytes and epithelial cells do not die. We find that flavivirus upregulates autophagy in MDCK renal epithelial cells and other cell types and subsequently protects them. We further identify non-structural protein NS4A of both Dengue-2 and Modoc viruses as the sole viral mediator of autophagy upregulation and protection against death. Upregulation of autophagosomes by either live virus infection or NS4A expression depends on PI3K and is important for replication of flavivirus in renal epithelial cells.
Flaviviruses often persist in liver and kidney following the acute phase of infection, suggesting that infected cells evade destruction by the host immune response, likely by flavivirusinduced upregulation of autophagy in these cells. In vitro infection of hepatocytes and fibroblasts leads to upregulation in autophagy and protection against death [3][4]. More relevant, Dengue-2 viruses replicate within the hepatocyte autophagosomes [5], and inhibition of autophagy attenuates virus replication [3]. We extend these findings by identifying the nonstructural viral protein NS4A as the virus-encoded protein that upregulates autophagy and thus protects the host cell against death, providing a well protected host cell for long-term replication of virus.
The flavivirus genome encodes ten genes as a ~11kDa polyprotein precursor that binds to ER membrane after translation at the rough ER: three structural proteins (capsid [core] protein, pre-membrane protein and envelope glycoprotein), and seven nonstructural genes (NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5) that mediate viral replication, assembly and evasion of the host immune system. The active proteins are cleaved from the ER-bound viral polyprotein by cellular proteases signalase and furin and viral protease NS3/2B. NS3 is both viral protease (with required cofactor NS2B [6]) and viral ATP-dependent helicase [7][8]; NS5 is an RNA-dependent RNA polymerase and methyl transferase [9] responsible for virus genome replication, while NS1 is possibly a part of the viral replication complex [10]. The small hydrophobic flavivirus proteins (NS2A, NS4A, NS4B) remain the most poorly characterized. NS2A is required for the assembly of new flavivirus virions, NS4A associates with the virus replication complex and induces ER membrane rearrangements (discussed below), and NS4B is an antagonist of interferon. Flavivirus NS4A is a ~16 kDa membrane-associated protein consisting of four transmembrane helices and an N-terminal cytosolic region. Once the viral genome is translated in the ER, NS4A is cleaved in the cytoplasm by viral NS2B-3 protease complex. The transmembrane domain of NS4A is cleaved by host signalase in the ER lumen [11]. Mature NS4A then promotes membrane rearrangements by inducing curvature of the ER membrane [12][13][14]. NS4A also co-localizes with dsRNA. NS4A and NS1 must interact for viral RNA synthesis to be efficient, suggesting that NS4A is the docking site for the replication complex [12,15]. NS4A also associates with host cell polypyrimidine tract-binding protein (PTBP) to indirectly bind dsRNA [16], stabilizing dsRNA viral genomic intermediates.
While it is clear that NS4A is involved in virus replication, the mechanisms are uncertain. In this report we identify flavivirus NS4A as the sole mediator of flavivirus-induced protection against cell death, induced by NS4A-mediated upregulation of autophagy signaling. This upregulation of autophagosomes can be induced by either whole virus or NS4A, depends on PI3K, and is important for flavivirus replication in renal epithelial cells.

Experimental Procedures
Cell and Virus culture and treatment MDCK, 293T, Vero, HeLa and Swiss Webster primary macrophage cells were maintained in Dulbecco's Minimum Essential Media (DMEM) with 10% FBS, 1% PenStrep at 37°C under a 5% CO 2 atmosphere.
Cells were plated to approximately 85% confluency and allowed to attach overnight in PenStrep-free cell-type appropriate culture media supplemented with 10% FBS.
Posttransfection, cells were incubated at 37°C for at least 16 hours before collection or treatment to allow plasmid expression. Dengue-2 virus was generously provided by Dr. Garcia-Sastre and Modoc virus (strain M544) was obtained from the American Type Tissue Collection (ATTC). For expansion of Dengue-2 stocks, sub-confluent C636 mosquito cells were infected with virus and incubated at 28°C for 6 days. For expansion of Modoc virus stocks, the virus was applied to sub-confluent Vero cells and incubated at 37°C for 7 days. In both cases, the culture media of the infected plates was agitated and collected, then mixed 2:1 with flavivirus freeze media [0.75% Bovine serum albumin fraction V in 0.12 M NaCl, 0.05 M H 3 BO 3 , at pH 9.0], and stored at -80°C. Viral load of Dengue and Modoc stock solutions was then determined by plaque assay as described by Davis [17]. Assessment of cell viability Cell were infected and exposed to toxins at 24hpi, incubated for an additional 24 h at 37°C, 5%CO 2 , collected by trypsin digestion, and stained with 0.4% trypan blue in 1x PBS. We have previously shown a direct correlation between trypan blue exclusion and other viability assays [18]. Live (white) and dead (blue) cells, were counted on a hemocytometer, with cell viability expressed as percent dead cells above mock. Camptothecin (CPT) (# C9911, Sigma-Aldrich) was applied at 50-100 M final concentration; Staurosporine (STS) (# S5921, Sigma-Aldrich) at 20 M and Cycloheximide (CHX) (# 01811, Sigma-Aldrich) at 150 M. Influenza A was applied to cells at MOI=5, and the cells were incubated for 1 hour at room temperature, washed once with 1x PBS and covered with maintenance media. When appropriate, autophagy was inhibited using Wortmannin, which inhibits both class I and class III PI3K's, and 3-methyl-adenine (3MA), which specifically inhibits class III PI3K signaling. Autophagy was upregulated by starvation in subculture media lacking FBS for 24 hours prior to and following infection or rapamycin treatment for 1 hour prior to and following infection. In all cases, cells were incubated with toxin for 24 hours prior to collection.

Transmission Electron Microscopy (TEM).
Reagents were purchased from Electron Microscopy Sciences (Fort Washington PA). MDCK cells were plated at a density of 1.5 x 10 6 , infected with influenza virus at a MOI of 5 with or without treatment. After 24 hpi, cells were fixed in 2.5% glutaraldehyde in 0.2 M cacodylate buffer, pH 7.4 for 20 min at room temperature (RT) on the plate. Cells were then scraped off and centrifuged in conical tubes at 1,200 x g for 5 min. After washing in cacodylate buffer cells were post-fixed with 1% osmium tetroxide in the same buffer for 1 h at RT and washed again. Cells were then dehydrated through ascending ethanol concentrations (50% -100%) for 10 min each and embedded in Agar 100 epoxy resin. Sections were counterstained with lead citrate and uranyl acetate. High-resolution images were taken with a Philips 208 electron microscope. Western blot, immunocytochemistry and cytochemistry Cells were infected and treated as described above. At 48hpi, cells were scraped and washed with ice-cold 1xPBS before whole lysate proteins were collected in RIPA buffer and quantified using the BioRad protein assay and an Ultrospec III spectrophotometer (Pharmacia LKB, Sweden). Western blot analysis was performed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) as described by Lin [18], using primary antibodies: 1:500 anti-LC3II (# L7543; Sigma-Aldrich) and 1:1000 anti-tubulin (# sc9104; Santa Cruz Biotech) as a loading control. Positive signals were detected using ECL (# RPN2132; GE Healthcare) and visualized using Amersham Hyperfilm ECL photoradiographic film (# 28906835; GE Healthcare). For immunocytochemical and cytochemical analysis, cells were seeded onto flame-sterilized glass cover slips, allowed to attach overnight, and infected and treated as above. At 48 hpi, cells were washed with 1xPBS and fixed with fresh, ice-cold 3% paraformaldehyde for 10 min, washed once and stored in 1xPBS in the dark at 4°C. Cells were then stained with 1:50 mouse anti-Flavi E antibody (Di-4G2-15) and 1:2000 Phalloidin-TRITC (# P1951; Sigma-Aldrich). 1:500 antimouse IgG-AlexaFluor 488 (# A11008; Invitrogen) was used as the secondary antibody. expressing cells were embedded immediately following fixation. Cells were observed by confocal microscopy (Leica, Germany). Assessment of viral replication Cell culture media samples of Modoc and Dengue-2 infected cells were collected at various times and total viral RNA was extracted using a viral RNA extraction kit (# 52904; Qiagen), following manufacturer's protocol.
cDNA was then generated from extracted viral RNA using a Superscript III first strand synthesis kit (# 18080-400; Invitrogen), following manufacturer's protocol. 1 g of cDNA was amplified by qRT-PCR in 20 μl reactions using a LightCycler FastStart DNA Master SyberGreen 1 Kit (# 03515869001; Roche), with primers for Modoc (fwd: GATTCAGGATGGCCCAAGAATC, rvs: AGTAGGAAGGTGGACAGAATGA), and Dengue-2 (fwd: TTAGAGGAGACCCCTCCC, rvs: TCTCCTCTAACCTCTAGTCC), using a Roche LightCycler 2.0 real-time PCR machine. Approximate viral titer (pfu/mL) was determined by quantifying viral RNA in serial dilutions of stock virus at a known concentration and generating a standard curve for comparison analysis. Relative RNA release in each cell type was compared to mock infected cells and presented as (infectedmock). For confirmation of qRT-PCR data, cell culture supernatant was also taken for replication analysis by plaque assay. Confluent BHK cells in 12-well plates were infected with serial dilutions of supernatant for 2 h at 37°C prior to overlay with plaquing media (45%: EMEM, 5%: FBS, 50%: 2% Low Melting Point Agarose) and incubation at 37°C for 4 days. Agar was then removed and cells stained with crystal violet solution. Plaques were counted and virus titer was. Each sample was run in triplicate, and error bars indicate 1 S.D.

Assessment of Flavivirus-Induced Autophagy
To examine flavivirus-induced autophagy, LC3 localization during infection was determined as described by Kabeya [19]. Briefly, cells were plated onto heat-sterilized glass cover slips in 35 mm plates and LC3-GFP was transiently expressed by transfection with a C2-LC3-GFP construct (provided by Guido Kroemer, Institut Gustave-Roussy, Villejuif, France), using Lipofectamine 2000, and incubated for 16 h to allow LC3-GFP expression. Cells were then infected as described above. At 24 hpi, cells were fixed with ice-cold 3.5% paraformaldehyde (Fisher, Cat# 04042) for 10 min, and rinsed once with 1x PBS. Cells were then embedded by Gel Mount (Biomeda, #M-01), and observed by confocal microscopy. Generation of a punctate GFP expression pattern is indicative of LC3 translocation and autophagosome formation. Transfection efficiency of at least 50% was confirmed using fluorescence microscopy (data not shown). Mock-infected cells were also analyzed to ensure that LC3-GFP expression alone did not cause autophagy. Autophagy upregulation was also examined as described above in HeLa cells stably expressing LC3-GFP.

Modoc virus does not kill infected epithelial cells or fibroblasts
The impact of flavivirus on cell fate is specific to cell type; macrophages and neurons die following infection, but hepatocytes and fibroblasts survive. Several authors have noted persistent flavivirus infection of the renal epithelium [20][21][22][23][24][25][26], indicating that renal epithelial cells also survive infection. To understand the particular virus-host interactions involved in renal flavivirus infection, we chose MDCK renal epithelial cells as our main model for in vitro study. As Modoc virus has been proposed as a research model for flavivirus infection in vivo, we first characterized Modoc-induced cell death of MDCK renal epithelial cells. As shown in Figure  1A, Modoc virus does not kill MDCK cells (<10% death, even after 288 hpi using undiluted stock virus (MOI=1190).
We next characterized the ability of Modoc virus to kill several other types of cell (Swiss Webster primary mouse embryonic fibroblasts (MEF's), C57Black primary MEF's, Vero and MDCK cells) and assayed for death at 24, 48, 96 and 144 hpi to establish that this effect was not specific to MDCK cells. As shown in Figure 1B, no significant levels of death (<10%) are observed in any of the cell types tested as late as 144 hpi at 10 MOI, demonstrating that Modoc virus does not kill epithelial cells or fibroblasts in vitro.

Modoc virus and Dengue virus infect and replicate in MDCK renal epithelial cells To ensure that MDCK cells support Dengue-2 and
Modoc virus infection and replication, we evaluated by immunocytochemistry and qRT-PCR the capability of each virus to infect and replicate in MDCK cells. For immunocytochemistry, cells were fixed at 48 hpi and stained for flavivirus E protein with antibody Di-4G2-4-15 (with AlexaFluor 488), and with phalloidin-TRITC to visualize cell boundaries and observed by confocal microscope. At 48 hpi, mock-infected cells show no positive staining for flavivirus protein ( Fig.  2A). Both Dengue-2 ( Fig. 2B) and Modoc ( Fig.  2C) virus are abundant in vesicles primarily within the perinuclear region of infected cells, a staining pattern consistent with previous findings of localization in the endoplasmic reticulum and Golgi bodies of infected cells [27]. Following infection with Dengue-2 or Modoc virus, over half the cells within infected populations are positive for flavivirus E protein (60% and 54%, respectively -data not shown), confirming that these viruses are capable of infecting and replicating in MDCK cells.
To quantify flavivirus replication in MDCK cells, supernatant samples from Modoc and Dengue-2 infected cells were collected at 48 hour intervals until 144hpi. Viral RNA was extracted from each sample and subjected to qRT-PCR. For both Dengue-2 and Modoc viruses, primers for the NS4A gene were used and qRT-PCR was performed as described in Materials and Methods.
As shown in Figure 2D, robust replication of both Dengue-2 and Modoc virus is evident by 48 hpi. A maximum virus titer of 10 7 pfu/mL is reached by Dengue-2 virus at 144 hpi, while the maximum titer generated by Modoc at 144hpi is 10 8 pfu/mL. These data confirm that MDCK cells support in vitro infection and replication of both Dengue-2 and Modoc virus, and provide a basis for the use of MDCK.

Dengue-2 and Modoc virus kill macrophages while protecting epithelial cells and fibroblasts
As killing of macrophages by flavivirus has previously been reported [23], we verified in our laboratory the ability of Modoc or Dengue-2 virus to kill or protect macrophages. We obtained primary macrophages by peritoneal perfusion of adult Swiss Webster mice. As shown in Figure   3A, infection of macrophages with either Dengue-2 or Modoc virus results in ~22% cell killing (p<0.001 for both), roughly equal to that caused by exposure to 75M CPT. Conversely Dengue-2 or Modoc virus do not kill primary Swiss Webster mouse embryonic fibroblasts (obtained from embryonic day 10.5 embryos). In fact, Dengue-2 infection caused a 59% decline, and Modoc a 70% decline in cell death caused by CPT (Fig. 3B) (p<0.0001, p=0.0005 respectively), demonstrating that infection with either virus protects these cells. We also tested several other cells for killing or protection after flavivirus infection: two human cell lines [293T (human kidney fibroblasts) and HeLa (human carcinoma)] and a primate cell line [Vero (green monkey kidney fibroblast)]. As shown in Figure 3C, infection with either Dengue-2 or Modoc virus does not kill 293T cells. Instead, Dengue-2 infection in 293T cells results in a 46% decline in death induced by CPT (p<0.001), while Modoc infection leads to a 42% decline (p<0.001). Likewise, neither virus kills HeLa cells. In HeLa cells (Fig. 3D), Dengue-2 infection leads to a 32% decline in CPT-induced cell death (p=0.001), and Modoc infection causes a 45% decline (p=0.001). Infection in Vero cells (Fig. 3E) by Dengue-2 leads to a 92% decline in cell death caused by CPT (p=0.04), while Modoc infection in these cells causes an 85% decline in death by CPT (p=0.03). These results demonstrate that cell fate following flavivirus infection depends on cell type, and indicate that pro-survival signaling is upregulated following infection in certain cell types.

Dengue-2 and Modoc infection in MDCK renal epithelial cells protects against death induced by several stimuli
The fact that Modoc virus does not kill epithelial cells or fibroblasts from several mammalian species, even at high MOI and late hpi, suggests that the virus can activate prosurvival signaling. We therefore tested whether Dengue-2 or Modoc virus could protect cells against agents that would normally activate apoptosis.
Cells were infected with either Dengue-2 or Modoc virus and incubated at 37° C until 24 hpi, at which time cells were exposed to toxin and held a further 24 h prior to collection at 48hpi and trypan blue exclusion assay.
We analyzed whether Dengue-2 or Modoc could protect MDCK cells against CPT (70 M), STS (20 M), CHX (150M) and Influenza A WSN/33 virus infection (MOI=5). As shown in Figure 3F, both Dengue-2 and Modoc virus protect MDCK cells against death by all insults examined. Pre-infection with Dengue-2 virus leads to a 39% decline in death induced by CPT (p<0.001), a 40% decline in that induced by STS (p<0.002), a 46% decline in death caused by CHX (p<0.0001), and a 31% decline in that caused by Influenza A (p=0.002) at 48 hpi when deaths are compared to mock-infected cells. Pre-infection with Modoc virus results in an 83% decline in CPT induced cell death (p<0.002), a 72% decline in STS induced death (p<0.001), a 51% decline in CHX induced cell death (p=0.004), and a 36% decline in Influenza A induced cell death (p=0.002) at 48 hpi. Thus, infection with either Dengue-2 or Modoc virus is sufficient to protect MDCK cells against death induced by a variety of insults. The generalized ability of infection to protect against such a wide range of insults suggests activation of a powerful and broad cell process.

Flavivirus infection induces PI3K-dependent autophagy
in MDCK epithelial cells Upregulation of autophagy often leads to PI3Kdependent pro-survival signaling [28][29][30][31]. We confirmed that both Modoc and Dengue-2 viruses induce autophagy during infection. To this end, we examined LC3-GFP translocation following infection with Dengue-2 and Modoc virus in both HeLa and MDCK cells. LC3 remains in a cytosolic, inactive distribution until activated during autophagy upregulation. This cytosolic distribution results in a diffuse LC3-GFP expression pattern. Upon activation, LC3 is cleaved into LC3-I and LC3-II. LC3-II then translocates to and accumulates at autophagosomal membranes, resulting in a punctate GFP distribution.
We first infected HeLa cells stably expressing LC3-GFP with either virus and fixed them at 24 hpi for observation by confocal microscopy. Mock-infected HeLa cells show little LC3-GFP punctation, indicating low levels of autophagy typical of healthy cells (Fig. 4A). Infection with either Dengue-2 ( Fig. 4B) or Modoc (Fig. 4C) virus induces a substantial increase in the number of cells containing punctate LC3-GFP, indicating an upregulation of autophagy following infection with either virus.
To confirm these results in our MDCK system, we next transfected MDCK cells with a C2-LC3-GFP plasmid and infected with either virus. At 48hpi, cells were fixed and mounted for observation by confocal microscopy. At 48hpi mock-infected cells display only limited LC3-GFP punctation, indicating little activation of autophagy (Fig. 4D), while both Modoc (Fig. 4E) and Dengue-2 ( Fig. 4F) infection produces pronounced LC3-GFP punctation, indicating upregulation of autophagosome formation. While only 17% of mock-infected cells display punctate LC3-GFP (indicating normal housekeeping activity), nearly 70% of both Modoc and Dengue-2 virus-infected cells display LC3-GFP punctation (Fig. 4J), demonstrating that both viruses induce autophagy.
Autophagosome formation is completely abolished following treatment with the PI3K inhibitor wortmannin in both Modoc (Fig.  4H) and Dengue-2 ( Fig. 4I) infected cells, but is not altered in mock-infected cells (Fig. 4G). PI3K inhibition by this method reduces the number of infected cells that display punctate LC3-GFP to 17%, similar to mock levels of GFP punctation (Fig. 4J). These results support a model by which PI3K-dependent autophagy results from infection in both Modoc and Dengue-2 viruses, and are consistent with previous findings using Dengue-2 virus [29].
Absolute experimental confirmation of autophagy upregulation has historically relied on the observation of increased numbers of doublemembrane autophagosomes by electron microscopy. We therefore fixed and collected both Dengue-2 and Modoc virus infected cells at 48hpi for electron microscopy. Analysis of subcellular morphology of mock-infected cells reveals double-membrane vacuoles reminiscent of classic autophagosomes (Fig4K, boxes), representing normal housekeeping functions typical of healthy cells. Both Dengue-2 (Fig. 4L, boxes) and Modoc virus (Fig. 4M, boxes) infected cells display an increase in the number of double-membrane autophagosomes, demonstrating an upregulation of autophagy following flavivirus infection.
To complete our analysis we collected whole protein lysates from Dengue-2 infected cells at 0, 3, 6, 12, 24, 36 and 48hpi, and probed these lysates for cleaved LC3 by western blot. The appearance of LC3-II is often used as an indicator of active autophagy although the amount seen in cells may also increase if clearance is slowed [28]. Minimal LC3-ii is seen in mock-infected cells at 0 and 48hpi, reflecting normal housekeeping autophagy (Fig. 5A) However, in Dengue-2 infected cells, observed LC3-II increases steadily until 36hpi. The decline in cleaved LC3 by 48hpi potentially derives from completion of the autophagic cycle but, since these cells were in late-stage infection, the issue is more complex and was not pursued further here. -tubulin was used as a loading control (Fig. 5B). The total LC3 protein, as judged by western blot, at least doubles in response to Dengue-2 or NS4A (Fig. S1). Our results from microscopy and western blot suggest that autophagy is upregulated in MDCK cells following both Dengue-2 and Modoc virus infection.

PI3K-dependent autophagy is required for protection induced by flavivirus infection
Dengue-2 and Modoc virus both affect survival of MDCK and several other cells in a similar manner, suggesting similarities in the biochemistry of protection.
Flavivirus infection triggers upregulation of autophagy [3,5] and autophagy is known to protect against insults that trigger the activation of apoptosis [30][31][32]. We therefore hypothesized that the protection observed following infection derives from activation of autophagy. We sought to identify the effect of key autophagy-related proteins on flavivirus-induced protection in epithelial cells. Wortmannin inhibits Class I and III PI3K activity, subsequently inhibiting autophagy. Wortmannin treatment does little to cell survival in infected or un-infected cells, whereas CPT treatment kills 49% of mockinfected cells above control within 24 hours of exposure (Fig 6A). However, infection with either Dengue-2 or Modoc virus provides protection against death, with Dengue-2 infection causing a 76% decline (p=0.0004) and Modoc infection causing a 98% decline (p=0.0016) in CPT-induced death. Inhibition of PI3K by wortmannin treatment eliminates the protective response invoked by Modoc and Dengue-2 virus infection, allowing CPT-induced death to reach 50% above control in infected populations, killing these cells as effectively as uninfected cells.
As wortmannin inhibits both Class I and Class III PI3K proteins, we next inhibited specifically Class III PI3K proteins during infection using the Class III PI3K inhibitor 3-methyl adenine (3MA), and repeated the assay. As shown in Figure 6B, Modoc and Dengue-2 infection and 3MA treatment alone do not kill cells, and levels of death for each are comparable to mock-infected cells at ~10%. As in previous experiments, CPT treatment induces ~50% death in mock-infected cells, while pre-infection with either Dengue-2 or Modoc virus leads to a 70% reduction in death induced by CPT when compared to mock-infected, CPT-treated cells (p=0.0004, p=0.0007, respectively). However, treatment with 3MA leads to the total abolition of both Dengue-2 and Modoc virus-induced protection against CPTinduced death. The data indicate that Class III PI3K activity is a regulatory factor for protection against cell death and suggest more specifically that autophagy signaling may be responsible for the ability of the virus to trigger a response leading to protection.

Autophagy-related proteins Beclin-1 and ATG5 are required for flavivirus-induced protection
These results demonstrate that Class III PI3K signaling is required for Modoc and Dengue-2 induced protection. Thus PI3K protects against cell death, suggesting that the ability of the virus to trigger autophagy leads to protection. To confirm this, we characterized Modoc and Dengue-2 induced protection in two different autophagy deficient cell lines: Beclin +/and ATG5 -/-MEFs.
Beclin (also called ATG6) interacts with PI3Ks to initiate autophagosome formation. We examined CPT-induced death after flavivirus infection in Beclin knockdown cells, which are heterozygous for the Beclin gene and have a reduced capability to form autophagosomes. WT 4-1 and Beclin +/cells were infected with Modoc or Dengue-2 virus and subsequently treated at 24 hpi with CPT. At 48hpi, cells were collected and percent cell death was analyzed by Trypan blue exclusion. As shown in Figure 6C, CPT kills 34% above control of treated wild-type cells, while both viruses were able to protect against cell death with Dengue-2 and Modoc viruses generating a 50% (p<0.0001) and 24% (p=0.0002) decline in cell death when compared respectively to mock. In mock-infected Beclin +/cells, CPT kills more effectively, resulting in 52% cell death above control. Moreover, in these cells, virus-induced protection was eliminated, with Dengue-2-and Modoc-infected cells actually exhibiting more cell death after CPT treatment (p<0.0001, p=0.0001, respectively) when compared to infected CPT-treated wild-type cells. These figures were essentially equivalent to the amount of cell death provoked by CPT in the absence of virus. Thus Beclin activation, a key event in the autophagy response, is essential to the protection afforded by Modoc and Dengue-2 virus.
We also analyzed the effect of ATG5 knockout in mouse embryonic fibroblasts. ATG5 is an autophagy related protein that acts downstream of Beclin-1 and is essential for formation of the isolation membrane. Wild-type and ATG5 KO MEFs were infected with either Dengue-2 or Modoc virus, incubated for 24 h, treated with CPT for an additional 24 h and collected for analysis of trypan blue exclusion. As shown in Figure 6D, CPT kills 31% above control of wild-type cells, while infection with either virus protects against death. Dengue-2 and Modoc infection both totally block CPT-induced cell death in wild-type cells (p<0.001, p=0.001, respectively). In mock-infected ATG5 KO cells, CPT induces 37% cell death above control. Both Dengue-2 and Modoc infection fail completely to protect ATG5 KO cells against CPT, with CPTinduced death reaching levels similar to those seen in mock infections.
Thus flavivirus-induced protection depends on ATG5. These results are consistent with those for Beclin +/-, suggesting a role for autophagy in flavivirus-induced protection against death.

Upregulation of autophagy prior to flavivirus infection does not enhance the protective effect
To determine if flavivirus-induced protection results from the infection or derives from infection-induced autophagy, we examined flavirus-induced protection while upregulating autophagy. Prior to infection, autophagy was upregulated by starvation or by addition of rapamycin (which inhibits mTOR therefore stimulating autophagy). As shown in Figure 6E we found that starvation alone has no effect on cell survival and can protect cells from death induced by CPT. This protection remains the same in starved cells infected with Dengue-2 or Modoc virus. Thus autophagy, not infection itself, is responsible for Modoc and Dengue-2 induced protection against death.
Upregulation of autophagy using rapamycin treatment yielded similar results. For this experiment we increased the concentration of CPT to 100 M, to ensure that the protection we see following infection with either Dengue-2 or Modoc virus is genuine. As shown in Figure 6F, mock infected cells without rapamycin exhibited 74% cell death after 24 hours of CPT. Dengue-2 and Modoc pre-infected cells each showed a ~85% decline in death (p<0.0001 for both) when compared to mock-infected CPT-exposed cells.
(Slightly different conditions produced a modestly higher kill rate, but the pattern was consistent in all experiments.) Rapamycin protected all cells, mock-infected or infected with Modoc or Dengue-2, against CPT-induced death, each yielding only ~8% cell death. Since Modoc and Dengue-2 infection, rapamycin and starvation can each elicit autophagy and protection against CPT-induced death, the mechanism by which autophagy is upregulated does not determine the level of protection. The data suggest that the protection elicited by these two viruses depends on the ability of the host cell to activate autophagy.

Autophagy enhances Dengue-2 and Modoc virus replication
Others have demonstrated that autophagy upregulation is important to Dengue-2 replication in hepatocytes [3], and that components of the virus replication complex colocalize with autophagosomal membranes [5]. As reported above a similar autophagy response is activated in epithelial cells during flavivirus infection.
We therefore evaluated flavivirus replication by plaque assay in MDCK cells while inhibiting autophagy with wortmannin and 3MA. Each sample was run in triplicate, and error bars indicate ±1 SD. For plaque assay results, virus titer is expressed as plaque forming units (pfu) per mL.
Plaque assay of Modoc virus replication during inhibition of autophagy by 3MA revealed that Modoc virus is also dependent upon PI3Kmediated autophagy for optimum replication. Modoc virus replication is evident in control cells (Fig. 7B), generating 2.3 x 10 6 pfu/mL at 96hpi. Inhibition of autophagy by 3MA treatment reduces Modoc viral titer to 6.7 x 10 5 pfu/mL by 96hpi, a 72% reduction in viral replication (p=0.02).
We also analyzed Modoc virus replication by RT-PCR. Autophagy was inhibited by 3MA and virus RNA was extracted from each supernatant sample at 96 hpi and subjected to qRT-PCR analysis. As shown in Figure 7C, total Modoc virus RNA in the supernatant of autophagy-inhibited cells is significantly reduced compared to control at 96 hpi (p=0.02), indicating that replication of Modoc virus is enhanced by autophagy, a result similar to those using Dengue-2 virus. (CP value represents the number of cycles before a transcript is recognized. Thus higher numbers reflect lower initial value of RNA. Here the difference is approximately 2 -5 , or a reduction to approximately 3% of control.) These results provide substantial evidence that autophagy enhances both Dengue-2 and Modoc virus replication in renal epithelial cells. The distant relationship of these two viruses also suggests that upregulation of autophagy is a general characteristic of flavivirus infection in epithelial cells.

Protection against cell death is mediated by flavivirus NS4A Having established links among flavivirus-induced protection, autophagy and flavivirus replication during infection with both
Dengue-2 and Modoc viruses, we sought to identify key viral proteins triggering these effects. We transfected MDCK cells with one of each of the ten viral genes of Dengue-2 virus, and the NS2A, NS4A, and NS4B genes (the small hydrophobic proteins) of Modoc virus, and allowed 20 hours for protein expression. We confirmed expression of transfected plasmids by western blot (Figure S2). We then assayed for protection against cell death induced by CPT as above. Live virus infection was done in tandem as a control for comparison. As in the previous experiments, CPT kills 45% of control cells while Dengue-2 and Modoc viruses both protect against CPT-induced death (Figs. 8A, 8B,  These results demonstrate that NS4A expression from effectively mimics the protection afforded by either Dengue-2 or Modoc live virus infection. NS4A-induced protection is not specific to CPT-induced death and protects the cells from cell death induced by STS as well. As shown in Figure 8C, CPT kills 57% of MDCK cells and STS kills 52% of mock-infected cells. Expression of empty plasmid control had no effect on either CPT or STS induced death, allowing 55% and 53% cell death, respectively. Expression of NS4A from either virus reduces CPT-caused death by 70% (Dengue-2 p=0.0006; Modoc p=0.0004) and STS-caused death by 30% (Dengue-2 p=0.0012; Modoc p=0.0012). Inducing autophagy with rapamycin protected cells regardless of the presence or absence of NS4A from either virus, allowing almost no CPT-induced death. As with live Modoc and Dengue-2 virus infections, inhibiting autophagy using either Wortmannin or 3MA eliminated NS4A-expression-induced protection against death by CPT and STS, allowing amounts of death similar to that of mock infection, at ~50% in each case. These results demonstrate that expression of either Modoc or Dengue-2 NS4A is sufficient to protect the cells and indicate that NS4A-induced autophagy is responsible for NS4A-induced protection against death.
NS4A expression induces LC3 cleavage and translocation As reported above, infection with either Modoc or Dengue-2 live virus stimulates autophagy, which protects infected cells against many toxins. We therefore determined whether NS4A expression also induces autophagy. We examined LC3 cleavage by western blot in NS4A expressing cells. Whole-cell lysates of both live virus-infected and NS4A-transfected cells were obtained and western blot analysis was performed using a LC3-specific primary antibody. As shown in Figure 9A, mock-infected cells exhibit low levels of LC3 cleavage at 48hpi. In contrast, both live Modoc and Dengue-2 infected cells, as well as Dengue-2 and Modoc NS4A-expressing cells show upregulation of LC3 cleavage by western blot. -tubulin was used as a loading control (Fig.  9B). These results indicate that NS4A expression from either virus is sufficient to induce autophagy, and suggest that NS4A-induced autophagy is responsible for conferring protection against death.
To confirm that NS4A alone is responsible for LC3 cleavage and autophagy in these cells, we collected whole-protein lysates from cells transiently individually expressing NS4A or control proteins NS1, NS2A, NS4B or empty plasmid backbone. As shown in Figure 9C, each condition caused a small amount of LC3 cleavage, likely due to cellular stress induced by our transfection protocol as indicated by the LC3 cleavage seen in cells treated with the transfection agent (Lipofectamine 2000) but lacking plasmid (+L2000). However, LC3 cleavage is markedly increased in NS4A expressing cells, indicating that among the proteins examined NS4A activity alone holds the ability to induce LC3 cleavage in MDCK cells. -tubulin was used as a loading control (Fig.9D). Together with the results above, our data indicate that NS4A activity is unique among the flavivirus proteins in its ability to induce autophagy and cell protection during flavivirus infection.
We also examined LC3 translocation in HeLa cells stably expressing LC3-GFP following both live virus infection and NS4A expression. As shown in Figure 9E, mock-infected cells display limited (housekeeping) LC3-GFP punctation. Consistent with our previous results, both Dengue-2 (Fig. 9F) and Modoc (Fig. 9G) virus induce LC3-GFP punctation in a large subset of infected cells, indicating an upregulation in autophagosome formation. Expression of empty plasmid vector induces little LC3-GFP punctation (Fig. 9H). Consistent with our results above, expression of both Dengue-2 NS4A (Fig 9I) and Modoc NS4A (Fig. 9J) induces LC3-GFP punctation similar to live virus infection. Our results demonstrate that NS4A from both Modoc and Dengue-2 viruses is sufficient to induce autophagy, and indicate that NS4A-induced autophagy is responsible for flavivirus-induced protection against death.
Flaviviruses cause apoptosis in neurons and macrophages [40][41][42][43][44][45], resulting in much of the pathology associated with in vivo infection. Dengue-2 replicates to very high titer in these cells (which are the primary sites of virus replication during acute infection) before the cells die. As infected macrophages cross epithelia within the body (including the hepatic and renal tubular epithelium), they transmit the virus to surrounding epithelial cells and fibroblasts. In contrast to neurons and macrophages, these latter cells do not die following infection. Instead they survive and often become persistently infected [21][22]46]. Consistent with these results, we show that in vitro infection of macrophages with either Dengue-2 or Modoc virus results in cell killing similar to that of CPT at 48 hpi, while infection with either virus fails to kill renal epithelial cells and fibroblasts (MDCK, 293T, Vero, SW primary MEF, C57/Bl primary MEF) even when the cells are infected at extremely high titer (MOI=~1190) and the cells are evaluated after 12 days (288 hpi, the duration of the experiment).
One explanation for the lack of death in these cells is that our experimental in vitro protocol did not allow sufficient virus uptake and/or replication. We therefore assessed the infectivity and replication of Dengue-2 and Modoc virus in MDCK cells using immunocytochemistry and RT-PCR, and found that both viruses infect and replicate to high titer in MDCK cells. Therefore, while renal epithelial cells and fibroblasts are susceptible to flavivirus infection and allow efficient virus replication, infection does not kill the cells. This report therefore provides further evidence that cell fate following flavivirus infection is specific to cell type, as both Dengue-2 and Modoc viruses kill macrophages while sparing renal epithelial cells and fibroblasts.
Dengue-2 infection induces autophagy [3], and the autophagy is important for flavivirus replication in infected hepatocytes [5]. Consistent with these results, we find that infection of both MDCK renal epithelial cells and HeLa cells with either Dengue-2 or Modoc virus induces PI3Kdependent autophagy, and chemical inhibition of autophagy by either Wortmannin or 3MA in MDCK cells leads to a 27% reduction of extracellular virus titer. These results demonstrate that autophagosome formation enhances flavivirus replication in epithelial cells, similar to previously published results using hepatocytes [3,5]. As the liver and kidneys are both prominent sites of virus replication during persistent flavivirus infections, it is tempting to speculate that the cell-type specific activation of autophagy in these cells plays a major role in the initial establishment of persistent infection in vivo.
A common consequence of autophagy signaling is the upregulation of pro-survival signaling and protection from death [30][31][32]. To evaluate the link between flavivirus-induced autophagy and protection against death following infection, we assessed the protection offered by flavivirus infection following inhibition of autophagy by both chemical and genetic means. Inhibition of autophagy by Wortmannin or 3MA treatment eliminates any protective effect that Dengue-2 or Modoc virus confers on infected cells. Inhibition of autophagosome formation by Beclin-1 knockdown or ATG5 knockout also removes the protection against death conferred by infection with either virus.
These results demonstrate that flavivirus-induced protection is dependent upon upregulation of autophagy following infection.
When autophagy is upregulated prior to infection by either starvation or rapamycin treatment, cells are protected irrespective of flavivirus infection, providing evidence that it is specifically autophagy, and not other virus-mediated events, that provides protection against death following infection. By expressing each virus gene individually, we determined that NS4A from either Dengue-2 or Modoc virus is the only flavivirus protein sufficient to protect against death by CPT. As with live virus infection, NS4A expression upregulates autophagy and induces broad-spectrum protection, shielding cells against both CPT-and STS-induced death. Expressing NS4A protects cells against CPT nearly as well as infecting the cells with live virus. Inhibition of PI3K signaling by Wortmannin and 3MA eliminates autophagy induced protection, while inducing autophagy by rapamycin confers protection regardless of NS4A expression. As with whole virus infection, we also find that of several Dengue-2 non-structural genes tested, NS4A expression alone induces PI3K-dependent autophagy and subsequent protection against death.
We conclude that flavivirus NS4A expression during whole virus infection leads to upregulation of autophagy.
NS4A-induced autophagy subsequently protects against death during infection. We do not yet know how the NS4A protein regulates autophagy.
As autophagosomes figure prominently in flavivirus replication within these cells, flavivirus NS4A is thus identified as a potential target for development of anti-viral drugs. We are currently exploring the specific biochemical virus-host interactions that govern NS4A-induced upregulation of autophagy following flavivirus infection, and searching for clues as to the mechanism(s) behind its cell-type specificity.
Dengue-2 gene constructs. We thank Dr Stephane Boissinot of CUNY Queens College for the Modoc gene constructs, and Dr Guido Kroemer of Institut Gustave-Roussy, Villejuif, France, for the LC3-GFP construct. We also thank Ms Narges Zali, and Mr Pierre Co, Mr Alain Goldman and Ms Michelle Sahli for technical support. Dr Areti Tsiola of the Queens College Core Facility for Imaging, Molecular and Cellular Biology provided technical expertise. We especially thank Dr Richard Lockshin for his critical review of this manuscript prior to submission. This work was supported in part by funding from the NIH (MARCUSTAR), grant 2T34GM070387-03, to Zahra. Zakeri and by a PSC-CUNY award to Zahra Zakeri.         Individual expression of any of the ten Dengue-2 structural and nonstructural genes does not lead to significant death in MDCK renal epithelial cells. Dengue-2 NS4A is the only gene that protects against cell killing by CPT, similar to live virus infection. NS2A and NS5 each allow slightly more death by CPT than mock-infected cells. Individual expression of empty plasmid vector, C, prM, E, NS1, NS2B, NS3 or NS4B has no effect on CPT-induced death. (B) Infection with live Modoc virus infection or individual expression of Modoc NS4A also protects cells against CPT-induced death. Expression of NS2A, NS4A or NS4B does not lead to significant cell killing in MDCK cells (<10%). Modoc NS4A expression protects in a manner similar to live virus infection. Modoc NS2A expression allowed slightly more killing than in mock-infected cells. Expression of the empty plasmid vector or Modoc NS4B had no effect on CPT-induced killing. (C) NS4A expression from either Dengue-2 or Modoc virus protects against a DNA damaging agent (CPT) and a kinase inhibitor (STS). Upregulation of autophagy by rapamycin treatment protected against death by CPT regardless of expression of NS4A from either virus. Inhibition of autophagy by 3MA treatment abolishes Dengue-.2 and Modoc NS4A-induced protection. Inhibition of autophagy by wortmannin also eliminates protection against STS-induced cell death induced by NS4A from either virus.

Figure 9. Flavivirus NS4A-induced protection in epithelial cells is dependent upon autophagy. (A)
Western blotting using a specific antibody for cleaved LC3II reveals an increase in LC3II cleavage following both live Dengue-2 and Modoc infections, as well as after individual expression of NS4A from either virus. (B) -tubulin was used as a loading control. (C) Whole-protein lysates were collected from MDCK cells transiently expressing Dengue-2 non-structural proteins NS1, NS2A, NS4A and NS4B at 24 hours post-transfection, as well as cells transfected with empty plasmid and cells exposed to the transfection agent as controls. All conditions stimulate a moderate amount of LC3 cleavage, likely owing to stress occurring during our transfection protocol as indicated by LC3 cleavage after exposure to the transfection agent alone. NS4A expression leads to a substantial increase in LC3 cleavage compared to controls and expression of other Dengue-2 non-structural genes, indicating NS4A expression alone is uniquely sufficient to upregulate autophagy in MDCK cells. (D) -tubulin was used as a loading control.