Semliki Forest virus strongly reduces mosquito host defence signaling

The Alphavirus genus within the Togaviridae family contains several important mosquito-borne arboviruses. Other than the antiviral activity of RNAi, relatively little is known about alphavirus interactions with insect cell defences. Here we show that Semliki Forest virus (SFV) infection of Aedes albopictus-derived U4.4 mosquito cells reduces cellular gene expression. Activation prior to SFV infection of pathways involving STAT/IMD, but not Toll signaling reduced subsequent virus gene expression and RNA levels. These pathways are therefore not only able to mediate protective responses against bacteria but also arboviruses. However, SFV infection of mosquito cells did not result in activation of any of these pathways and suppressed their subsequent activation by other stimuli.


Introduction
The Alphavirus genus within the Togaviridae family contains several important mosquito-borne arboviruses including Chikungunya (Charrel et al ., 2007) and the equine encephalitis viruses (Weaver & Barrett, 2004). Replication of the prototype Sindbis and Semliki Forest viruses is well understood (Strauss & Strauss, 1994;Kaariainen & Ahola, 2002). The genome consists of a positive-stranded RNA with a 5 ′ cap and 3 ′ polyA tail. The 5 ′ two-thirds encodes the nonstructural polyprotein P1234, which is translated from the viral genome after virus entry and cleaved into four replicase proteins nsP1-4 (Merits et al ., 2001;Kim et al ., 2004;Lulla et al ., 2006). The structural polyprotein is encoded in the 3 ′ third of the genome, and is cleaved into capsid and glycoproteins after translation from a subgenomic mRNA. Cytoplasmic replication complexes are associated with cellular membranes (Salonen et al ., 2005).
Viruses and other intracellular pathogens have to overcome cellular host defences to replicate efficiently. In vertebrates, detection of viral signatures, for example dsRNA, is an important trigger of antiviral defences (Haller et al ., 2006;Randall & Goodbourn, 2008).
Little is known about how invertebrate cells react to arbovirus infections. Establishment of persistent infections in invertebrate cells, as opposed to cell death in vertebrate cells, suggests fundamental differences (Brown, 1984;Higgs, 2004). dsRNA is detectable in alphavirus-infected mosquito cells (Stollar et al ., 1972). Unlike in shrimp (Robalino et al ., 2007) and Lepidoptera (Hirai et al ., 2004), there is however no indication that control dsRNAs used in RNAi experiments stimulate antiviral responses in mosquitoes Franz et al ., 2006), or transfected mosquito cells (Caplen et al ., 2002). To date, RNAi is the best-studied cellular mechanism known to prevent or limit arbovirus replication in mosquitoes and mosquito cells Campbell et al ., 2008).
Recently, a heat shock protein was shown to downregulate O'nyong-nyong alphavirus replication in Anopheles gambiae cells (Sim et al ., 2007), while the Toll pathway controls Dengue virus infection (Xi et al ., 2008). Genomic studies have used gene arrays to analyse the transcriptome of mosquito and thrips larvae (Medeiros et al ., 2004;Sanders et al ., 2005;Sim et al ., 2005). In mosquito midguts infected with Sindbis virus (Sanders et al ., 2005), early Toll and (possibly) IMD pathway component upregulation, late JNK activation and RNAi ( dicer-2 ) downregulation were observed. Thrips larvae infected with the plant-infecting tomato spotted wilt bunyavirus show upregulation of Toll, IMD and JNK pathways (Medeiros et al ., 2004). These effects might not directly be related to virus infection, as complex secondary responses such as cytokines or other 'danger' signals may also be relevant in tissues.
As the interplay between mosquito cell signaling pathways and alphaviruses is not understood, we investigated the effect of SFV infection on mosquito host defence signaling in Aedes albopictus -derived U4.4 cells during the acute phase of infection. SFV did not activate but effectively inhibited STAT-, IMD-and Toll-mediated host defence signaling in mosquito cells and reduced host gene expression. This is similar to SFV infection of vertebrate cells, but the reduction in gene expression was less and the infection did not result in cell death. Stimulation of pathways involving STAT and IMD (but not Toll) responses before infection reduced viral gene expression and RNA synthesis, indicating that these pathways can activate antiviral activities.

SFV4 readily establishes infection of U4.4 mosquito cells
Early studies on SFV in mosquito cells showed that, like other arboviruses, SFV (strain unknown) infection of cultured Ae. albopictus cells begins with an acute phase of efficient virus production between 12 -24 h, and then enters a persistent phase during which only few cells (1-2%) produce virus (Davey & Dalgarno, 1974). Persistently infected mosquito cell cultures can be maintained for many passages. For Sindbis virus, the Ae. albopictus -derived U4.4 cell line closely resembles infectivity in the mosquito and has been used to study virus/host interactions (Riedel & Brown, 1979;Condreay & Brown, 1986;Condreay & Brown, 1988;Miller & Brown, 1992). To study innate defence mechanisms in SFV-infected U4.4 cells, we first determined whether SFV4 (derived from the SFV prototype cDNA clone) could establish infection of these cells. Indirect immunofluorescence was used to verify SFV4 infection. At a multiplicity of infection (m.o.i.) of 10, approx. 100% of cells were infected at 18 h post-infection (p.i.) (Fig. 1A). Infection was productive; high levels of virus production were observed between 0 and 12, and 12 and 24 h p.i. (Fig. 1B). Thereafter, virus production dropped to low levels as the culture entered the persistent phase of infection. This pattern of infection was similar to that described in the earlier experiments of Davey & Dalgarno (1974) and to that described for Sindbis virus-infected U4.4 cells (Mudiganti et al ., 2006). Cell numbers of infected and uninfected cultures increased at similar rate throughout the observation period of 96 h p.i. (Fig. 1C). Thus, SFV4 displays the expected characteristics of an arbovirus in these mosquito cells.

SFV4 infection down-regulates host gene expression in U4.4 cells
Interference with host gene expression is one mechanism by which alphaviruses interfere with host responses in vertebrate cells (Aguilar et al ., 2007;Breakwell et al ., 2007;Garmashova et al ., 2007). A previous report (measuring 3 H-uridine incorporation into cellular RNA) showed that host RNA synthesis is inhibited 1.5-to 1.7-fold in Sindbis virus-infected Ae. albopictus cells (Sarver & Stollar, 1977), and led us to investigate if there is also inhibition of host gene expression in SFV-infected mosquito cells.
Reporter genes expressed from constitutively active promoters have been shown to reflect host gene expression in arbovirus-infected vertebrate and mosquito cells (Kohl et al ., 2004;Leonard et al ., 2006). To analyse host gene expression, U4.4 cells were transfected with a RNA polymerase II-dependent reporter plasmid, pGL4.75 ( Renilla luciferase expression under the control of a CMV promoter), infected with SFV4 (m.o.i. 10) immediately post-transfection and Renilla expression measured during the acute phase of virus production. For comparison, Renilla expression was also measured in infected and uninfected mouse NIH 3T3 cells. As shown in Fig. 2A, SFV4 infection caused a moderate but consistent and significant reduction in Renilla activity in U4.4 cells (2-fold at 24 h p.i.), but a strong reduction in NIH 3T3 cells (7-fold at 12 h then 50-fold at 24 h p.i.) (Fig. 2B). As U4.4 and NIH 3T3 cells are infected at the same m.o.i. and all cells were infected, those differences are not due to cell numbers. Interestingly, reporter gene expression was reduced by the same magnitude at 48 h p.i., suggesting that biologically active virus proteins are still present as the culture becomes persistently infected.

SFV4 infection of U4.4 mosquito cells does not activate the STAT, IMD or Toll signaling pathways and activation of these pathways is strongly reduced in virus-infected cells
In arthropods, the STAT, IMD and Toll signaling pathways are known to be activated by bacteria and fungi. It is not known whether these pathways are activated by arboviruses. Dual luciferase reporter assays were used to determine the ability of SFV4 to activate these pathways. U4.4 cells were co-transfected with constitutively active internal reporter plasmid pAct-Renilla (expressing Renilla luciferase) and one of three plasmids encoding Firefly luciferase under the control of an inducible, pathway-responsive promoter. The Firefly luciferase-expression plasmids p6x2DRAF-Luc, pJL169 and pJM648 containing respectively promoters for STAT-, IMD-or Toll-inducible signaling pathways were used. Each of these dual luciferase reporter assays used a different Firefly plasmid generating different background levels of luciferase expression and different ratios of Firefly to Renilla luciferase expression (Fig. 3). For cells transfected with the STAT or IMD signaling pathway reporters, addition of heat-inactivated Escherichia coli activated both pathways (relative to PBS controls). In contrast, SFV4 infection did not activate either pathway and, consistent with the previously observed reduction in host gene expression (Fig. 2), background levels of both reporter genes were reduced. Addition of heat-inactivated bacteria to the virus-infected cells activated both the STAT and IMD pathways, demonstrating that these virus infected cells were still capable of responding to other pathogenic stimuli (Fig. 3); the magnitude of this response was however far less than that of uninfected cells. To activate the Toll pathway, which cannot be activated by E. coli or Staphylococcus aureus in U4.4 mosquito cell culture (our observations), an expression plasmid (pJL195) for a constitutively active Toll receptor (Toll Δ LRR) (Tauszig et al ., 2000) was cotransfected with the luciferase reporter plasmids. This strongly activated the Toll pathway, > 15-fold (Fig. 3), whereas empty insect cell expression vector (pIB-V5/His) did not. Again, SFV4 infection did not activate this pathway but reduced the background level of reporter genes. SFV4 infection of Toll Δ LRR-expressing cells dramatically reduced activation of the Toll signaling pathway. Internal Renilla controls were consistently expressed at slightly higher levels when the Toll pathway was activated; the reason for this is not known. Taken together, these results show that SFV4 infection of U4.4 cells does not activate the STAT, IMD or Toll pathways and that infection strongly reduces the level of signaling induced by these pathways; this is possibly due to down-regulation of host cell gene expression.

Activation of host defence signaling pathways prior to virus infection suppresses viral gene expression
That SFV4 infection of U4.4 cells does not activate the STAT, IMD or Toll signalling pathways could result from lack of triggering or effective suppression. Whether responses initiated by these signaling pathways are able to affect arbovirus infection is not known. To allow quantitative assessment of early and late virus gene expression, recombinant viruses expressing Renilla luciferase from either the replicase or the structural open-reading frame of SFV4 were constructed (Kiiver et al ., 2008). Renilla luciferase expressed from the genomic promoter was flanked by nsP2-protease cleavage sites at the C-terminus of nsP3 in SFV4(3H)-Rluc ; Renilla luciferase expressed from the subgenomic promoter was inserted between the capsid and the envelope glycoproteins and released from the latter by inclusion of a self-processing foot and mouth disease  To test the effects of immune signaling involving STAT and IMD on SFV gene expression, U4.4 cells were pre-incubated for 1 h with heat-inactivated E. coli or with PBS, then infected with SFV4(3H)-Rluc or SFV4-StRluc. As shown in Fig. 4B, bacterial activation of host defences prior to infection reduced virus gene expression both from the genomic and the sub-genomic promoters by around 50%. Immunostaining of parallel cultures demonstrated that the percentage of virus-infected cells in both the bacterial-treated and PBS control-treated cells were identical (data not shown), indicating that prior stimulation with bacteria did not affect virus entry. Expression of constitutively active Toll receptor (Toll ΔLRR) for 24 h, followed by infection with Renilla-expressing SFVs had no effect on virus gene expression at 12 h p.i. (Fig. 4C). Activation of Toll signaling before IMD/STAT signaling did not reverse the inhibitory effects of bacterial stimulation on viral gene expression (not shown).

Stimulation of host defences involving STAT/IMD reduces virus RNA replication rates
To determine whether viral RNA replication was affected by prior simulation with heat-inactivated bacteria (activating pathways involving STAT/IMD signaling), viral genome copies (which also serve as mRNA) were measured by real time quantitative PCR. As previously, U4.4 cells were stimulated with heat-inactivated E. coli for 1 h, followed by infection with SFV4 for 12 h, and total RNA was isolated. As shown in Fig. 5, bacterial stimulation reduced viral genome copy number by 40 -60% (compared to non-stimulated cells), similar to the reduction in viral gene expression, demonstrating that bacterial activation of cellular defences interferes with virus RNA replication. The magnitude of this effect is similar to the effect of type I interferon on SFV RNA synthesis in vertebrate cells (Mecs et al., 1967).

Discussion
Relatively little is known about arbovirus interactions with mosquito host defences. In vertebrate cells, alphavirus infections result in a reduction of host cell gene transcription which includes at least some host defence responses (Garmashova et al., 2006;Aguilar et al., 2007;Breakwell et al., 2007;Garmashova et al., 2007). Sindbis virus infection of an Aedes albopictus cell line different to that used in the present study reduced host RNA levels 1.5 to 1.7-fold (Sarver & Stollar, 1977). This suggests that transcriptional inhibition can also take place in alphavirusinfected mosquito cells. Our results show that alphavirus inhibition of mosquito cell gene expression occurs in SFV4-infected U4.4 cells, and is similar in magnitude to the previous observation.
Direct comparison of cellular gene expression in U4.4 and NIH 3T3 cells infected with SFV4 demonstrated that reduction of cellular gene expression was far greater in mouse (50-fold at 24 h p.i.) than mosquito cells (2-fold). Our studies extend previous work by demonstrating that SFV4-infected mosquito cells make only limited responses to stimulation of the STAT and IMD pathways with heat-inactivated bacteria, and Toll pathway-induced gene expression is also strongly reduced by SFV4 infection.
In U4.4 cells, there was no activation of the STAT, IMD or Toll pathways by SFV4 infection. Possibly this is because virus infection does not trigger these pathways. Alternatively, virus might trigger these pathways but no gene expression results due to down-regulation of host cell gene expression in virus-infected cells. Given that this down-regulation of gene expression was in the order of 2-fold, far less than that observed in vertebrate cells, and there was no indication of any upregulation of STAT, IMD and Toll pathways in virus-infected cells, it seems most likely that SFV infection does not trigger these pathways. However, we cannot rule out the possibility that SFV4 directly targets and inhibits these host defence pathways. Viruses can directly target cell defence pathways, many examples of this exist in virus/ vertebrate cell interactions (Randall & Goodbourn, 2008); in mosquito cells, STAT phosphorylation is inhibited in Japanese encephalitis arbovirus (Flaviviridae) infection (similar to mammalian cells) (Lin et al., 2004). Venezuelan encephalitis alphavirus capsid protein does not target nuclear import in mosquito cells, while it does in vertebrate cells to interfere with host gene expression; however the experiments described would not allow to detect subtle changes in gene expression, mode of action by capsid might differ, and/or different viral proteins might be involved (Atasheva et al., 2008). Interestingly though, if host defence pathways, including STAT and IMD but not Toll, were stimulated prior to SFV4 infection, reporter gene expression from both the virus replicase and structural open-reading frames was reduced (50%). Virus RNA levels were also reduced (40 -60%) in these cells providing an explanation, though the possibility that reduction in virus RNA levels was compounded by reductions in translation of these virus RNAs cannot be ruled out. In summary, although SFV4 infection probably does not trigger the STAT, IMD or Toll pathways, activation of host defence pathways which include STAT and IMD, but not the Toll pathway alone, does inhibit virus replication. It is possible that STAT and/or IMD pathways are the main mediators of antiviral effects against SFV4 in mosquito cells, though other pathways responsive to Gram-negative bacteria might also be involved. Other antiviral pathways might exist in mosquitoes, and this possibility needs to be further investigated. The pathogenic Microplitis demolitor bracovirus can not just inhibit Toll and IMD signaling, but also inhibits antimicrobial melanization reactions (Thoetkiattikul et al., 2005;Beck & Strand, 2007;Lu et al., 2008). Despite the evolutionary distance between Aedes and Drosophila, our results show that tools to study Drosophila microbe/cell interactions can also be successfully used in Aedes; this suggests some conservation of important physiological processes and might allow mosquito researchers to take further advantage of related invertebrate organisms. Figure 5. Activation of host cell defence signaling inhibits SFV4 RNA replication. U4.4 cells were treated with heat-inactivated Escherichia coli for 1 h (to stimulate signaling pathways involving STAT/IMD) or mock-treated, then infected with SFV4 (m.o.i. 1) for 12 h before RNA isolation. Viral genome copy numbers, which also act as viral polyprotein mRNA and are thus linked to viral gene expression, were measured by real time quantitative PCR targeting a region of nsP3. RNAs from three independent biological replicates (for both non-stimulated and stimulated cells) were reverse transcribed in triplicate; each reverse transcription reaction was then amplified in triplicate. Each bar represents the average of all measurements from a representative real time quantitative PCR experiment (see Experiemental procedures); error bars represent the standard deviation. Every experiment was repeated at least twice under the same conditions. It remains to be seen how this study compares to genomic studies suggesting (sometimes delayed) upregulation of host defence pathways in response to arbovirus infection in entire insects or specific target tissues (Medeiros et al., 2004;Sanders et al., 2005). IMD pathway activation in Sindbis virus-infected Aedes aegypti midguts is intriguing and might indeed have antiviral potential during natural infection (Sanders et al., 2005); however it remains to be seen what exactly activates this pathway. Direct analysis of infected cells, as in this study, suggests that host defence signaling is actively downregulated, but if activated prior to infection then virus gene expression and replication are reduced. Activation of these pathways in tissues might be a secondary reaction to dsRNA, cytokines, or other 'danger signals', as suggested for Drosophila X virus (Zambon et al., 2005). Tissue damage should not be excluded as alphaviruses can induce pathological changes in mosquitoes (Mims et al., 1966;Weaver et al., 1988;Scott & Lorenz, 1998;Bowers et al., 2003).
It is not yet clear how virus gene expression and replication can be affected by mosquito innate immunity, and how important these pathways are compared to the relatively slow RNAi responses. Mammalian defensins have antiviral activity at various levels (Klotman & Chang, 2006) and antiviral activities of mosquito antimicrobial peptides is possible.
In summary, host defence signaling, including the STAT, IMD and Toll pathways, is strongly reduced in SFV4-infected U4.4 cells, probably as a result of a general inhibition of host cell gene expression. These pathways are not activated in SFV4-infected cells, possibly because of this prior inhibition of host cell gene expression or alternatively because they are not triggered by this infection. Activation of pathways involving STAT/IMD but not Toll signaling prior to infection does nevertheless exert an anti-viral effect.
Virus-containing supernatants were clarified by centrifugation (3 ×, 30 min, 15 000 rpm) and viruses concentrated from supernatant on a 20% (w/v) sucrose/TNE buffer (pH 7.4) cushion by ultracentrifugation (25 000 rpm, 90 min, SW28 rotor). Pellets were resuspended in TNE buffer, and viruses titrated by plaque assay. Infection of mosquito and mammalian cells was performed at 28 °C or 37 °C for 1 h, respectively, at a m.o.i. of 10 plaque forming units (PFU) per cell (unless otherwise stated) in PBS/0.75% BSA. After infection complete medium was added to the cells. To establish growth curves, approx. 6.5 × 10 5 U4.4 cells/well (in 6-well plates) were infected at a m.o.i. of 10; medium to be titrated was taken off and replaced with fresh medium for intervals indicated. For growth comparisons of infected and noninfected cells, approx. 1.3 × 10 5 U4.4 cells/well (in 24-well plates) were infected at a m.o.i. of 10 and counted at times indicated.

Plasmids
Plasmid pGL4.75 (Promega, Southhampton, UK) contains the Renilla luciferase gene under control of the CMV IE promoter. Insect cell expression plasmid pIB-V5/His was obtained from Invitrogen. pAct-Renilla (Karsten et al., 2006) contains the Renilla luciferase gene under control of the Drosophila actin 5C promoter for constitutive expression. p6x2DRAF-Luc (Hombria et al., 2005) is a multimerised Drosophila STAT-responsive element with a Firefly luciferase reporter. IMD-pathway reporter pJL169 (Firefly luciferase under control of the Drosophila Attacin A promoter), Toll-pathway pathway reporter pJM648 (Firefly luciferase under control of the Drosophila Drosomycin promoter) and plasmid expressing constitutively active Toll receptor (pJL195; Toll ΔLRR) have been previously described (Tauszig et al., 2000).

Pathway stimulation with Escherichia coli
To prepare working stocks of E. coli strain JM109 (New England Biolabs, Hitchim, UK), 1 μl of bacterial stock solution was added to 5 ml LB medium (without antibiotics), incubated at 37 °C/18 h, and centrifuged (4 °C, 2500 rpm/10 min). Cells were resuspended in 0.5 ml PBS, and bacteria inactivated by heating the suspension at 80 °C for 10 min. Fresh stocks were prepared for each experiment (concentration approx. 9 × 10 5 bacteria/μl). For transfection/ stimulation experiments, 5 μl of the suspension (approx. 4.5 × 10 6 cells) were added into 2 ml complete medium/well, post transfection (and, where indicated, infection). Assuming that © 2008 The Authors Journal compilation © 2008 The Royal Entomological Society, 17, 647-656 volume of Aedes hemolymph is approx. 50 nl (Shapiro et al., 1986), the concentration of approx. 2250 inactivated bacteria per μl culture medium is below lethal concentrations in live Drosophila (Pham et al., 2007). To measure effects on viral gene expression, U4.4 cells (650 000 cells/well in 6-well plates) were stimulated (1 h) with heat-inactivated bacteria or mock stimulated as described above, followed by infection with recombinant SFV for 1 h. Reporter gene expression was analysed by luciferase assays at 12 h p.i.