Processing of genome 5' termini as a strategy of negative-strand RNA viruses to avoid RIG-I-dependent interferon induction.

Innate immunity is critically dependent on the rapid production of interferon in response to intruding viruses. The intracellular pathogen recognition receptors RIG-I and MDA5 are essential for interferon induction by viral RNAs containing 5′ triphosphates or double-stranded structures, respectively. Viruses with a negative-stranded RNA genome are an important group of pathogens causing emerging and re-emerging diseases. We investigated the ability of genomic RNAs from substantial representatives of this virus group to induce interferon via RIG-I or MDA5. RNAs isolated from particles of Ebola virus, Nipah virus, Lassa virus, and Rift Valley fever virus strongly activated the interferon-beta promoter. Knockdown experiments demonstrated that interferon induction depended on RIG-I, but not MDA5, and phosphatase treatment revealed a requirement for the RNA 5′ triphosphate group. In contrast, genomic RNAs of Hantaan virus, Crimean-Congo hemorrhagic fever virus and Borna disease virus did not trigger interferon induction. Sensitivity of these RNAs to a 5′ monophosphate-specific exonuclease indicates that the RIG-I-activating 5′ triphosphate group was removed post-transcriptionally by a viral function. Consequently, RIG-I is unable to bind the RNAs of Hantaan virus, Crimean-Congo hemorrhagic fever virus and Borna disease virus. These results establish RIG-I as a major intracellular recognition receptor for the genome of most negative-strand RNA viruses and define the cleavage of triphosphates at the RNA 5′ end as a strategy of viruses to evade the innate immune response.


Introduction
The efficacy of the innate immune response against virus infections is highly dependent on a rapid production of interferons and other cytokines. To achieve this, all nucleated cells are endowed with a cytoplasmic signaling cascade leading from the detection of intruded virus particles to the expression of cytokines, including type-I interferons (IFN-a/b) [1,2]. IFN-a/b directly induce an antiviral state that alleviates the virus burden but also initiate adaptive immune responses [3]. Viruses that block expression of IFN-a/b are therefore often highly virulent [4,5].
An important group of pathogens causing emerging and reemerging diseases are viruses with a negative-stranded RNA genome (NSVs). Prominent examples are influenza viruses (family Orthomyxoviridae), Ebola virus (Filoviridae), rabies virus (Rhabdoviridae), Nipah virus (Paramyxoviridae), Lassa virus (Arenaviridae), and several members of the Bunyaviridae family e.g. Rift Valley fever virus, Hantaan virus, or Crimean-Congo hemorrhagic fever virus (Table 1). These pathogens can cause rapid, systemic and often fatal illnesses which are characterized either by a fulminant pneumonia and multi-organ failure or by a severe hemorrhagic fever [6][7][8][9].
NSV particles consist of a lipid envelope and a genome which is encapsidated by the viral nucleocapsid protein and polymerase. The genome is present either as one continuous strand of RNA (nonsegmented NSVs), or is divided into two, three, or up to eight segments. Upon infection of a host cell, the NSV genome first needs to be transcribed into mRNAs in order to furnish the proteins necessary for virus replication. In the course of these events, the incoming nucleocapsids must be unwrapped and the genomic RNA exposed. Cells, in turn, are able to sense the infection and activate an innate immune response. They immediately synthesize cytokines, most prominently the antivirally active type I interferons, but also other pro-inflammatory cytokines and chemokines [10].
The main intracellular sensors to recognize viral RNA structures and trigger cytokine synthesis are the RNA helicases RIG-I and MDA5 [11][12][13]. Until very recently, it was assumed that the only viral agonist of these pathogen recognition receptors (PRRs) is double-stranded RNA (dsRNA) which is generated as a by-product of genome replication. However, we have recently found that NSVs are exceptional in not producing substantial amounts of dsRNA [14]. Instead, the genomic single-stranded RNA (ssRNA) bearing a 59 triphosphate group was shown to be sufficient to activate RIG-I dependent IFN-a/b induction [15,16]. The importance of RIG-I for recognizing genomic RNA has been demonstrated for influenza virus, rabies virus and vesicular stomatitis virus [15,16], but for none of the other NSVs.
Here, we tested genomic RNA isolated from highly pathogenic NSVs to investigate whether activation of RIG-I is a common denominator. Our results not only establish RIG-I as the main intracellular sensor for the genome of both nonsegmented and segmented NSVs, but also show that some viruses are able to escape RIG-I-dependent cytokine induction by a hitherto unrecognized stealth strategy.

Results
Genomic RNAs from highly virulent nonsegmented NSVs activate the IFN system We tested the genomic RNAs of Zaire Ebola virus (ZEBOV, family Filoviridae) and Nipah virus (NiV, family Paramyxoviridae) for the capability to activate an innate immune response. RNA isolated from influenza A virus (FLUAV) particles was used as positive control. As assessed in a reporter assay ( Fig. 1A and B), transfection of human 293T cells with RNAs isolated from ZEBOV and NiV particles (vRNA) resulted in strong activation of the IFN-b promotor, in a manner similar to FLUAV vRNA (Fig. 1C). The constitutively active SV40 promoter, by contrast, was not activated by vRNAs, indicating specificity for the IFN system. Furthermore, RT-PCR analysis showed that treatment of cells with vRNAs resulted in a transcriptional upregulation of the IFN-b gene as well as the IFN-stimulated genes IP-10, ISG56 and OAS1 (Fig. 1D), thus confirming our results with the reporter system. We concluded from these data that genomic RNAs of ZEBOV and NiV are potent elicitors of an IFN response.  IFN induction depends on RIG-I and the 59 triphosphate group of genomic RNAs MDA5 and RIG-I are the main intracellular PRRs driving an antiviral innate immune response against RNA viruses [13]. To investigate which of these factors is required for the recognition of ZEBOV and NiV RNA, we established RNA interference (RNAi) in 293T cells, using short hairpin RNAs (shRNAs) expressed by retroviruses. Initially, for each target two different retroviral shRNA constructs were tested for their efficiency to knock down gene expression. In Fig. 2A it is demonstrated that the shRNA constructs MDA5 #1 and RIG-I #2 significantly and specifically reduced target expression levels, whereas MDA5 #2 and RIG-I #1 had more modest effects. Therefore, in all subsequent experiments, we used MDA5 #1 and RIG-I #2 for gene-specific knockdowns. As a next step, we pretreated cells with these constructs, incubated them for 5 days to establish RNAi, seeded them again and transfected them with viral RNAs. The IFN-b promoter assay was used as a representative measure of cytokine induction. Fig. 2B shows that IFN induction by ZEBOV and NiV vRNAs was affected when RIG-I was downregulated. By contrast, MDA5 knockdown had no negative effect but sometimes increased IFN induction to some -statistically not significant -extent.
RIG-I-dependent responses to FLUAV vRNA were previously shown to critically depend on the presence of a 59 triphosphate group [16]. In agreement with this, genomic RNAs from ZEBOV and NiV particles lost their IFN-inducing activities when treated with shrimp alkaline phosphatase (SAP). This effect could be inhibited by EDTA, excluding an unspecific effect of SAP on IFN induction (Fig. 2C).
These data suggest that the 59triphosphate group is the decisive molecular pattern triggering a RIG-I-dependent IFN response to genomic RNAs of ZEBOV and NiV.

Genomic RNAs of segmented NSVs differ in their ability to trigger IFN induction
We extended our studies to NSV groups containing a segmented genome. As highly relevant representatives we investigated Lassa virus (LASV), a member of the Arenaviridae family, and several members of the Bunyaviridae family, namely Rift Valley fever virus (RVFV, genus phlebovirus), Hantaan virus (HTNV, genus hantavirus), and Crimean-Congo hemorrhagic fever virus (CCHFV, genus nairovirus). The RNAs of LASV and RVFV activated the IFN-b promoter in a RIG-I-and 59 triphosphate-dependent manner as observed for the nonsegmented NSVs in the previous experiments ( Fig. 3A and B). Surprisingly, however, RNAs isolated from HTNV and CCHFV particles did not trigger IFN induction (Fig. 3C). To rule out trivial explanations for the lack of stimulatory activity, we confirmed the identity and integrity of the bunyavirus genomic RNAs by RT-PCR analysis (see below) and denaturing formaldehyde agarose gel electrophoresis (Fig. 3D). Thus, in stark contrast to our findings with the other NSVs, the genomic RNAs of bunyaviruses belonging to the genera hantavirus and nairovirus are devoid of an IFN-inducing activity.

Non-inducing viral RNAs contain a 59 monophosphate
The absence of IFN induction by hantaviral and nairoviral RNA suggests a particular strategy to avoid RIG-I activation. As all initial products of RNA polymerases contain a triphosphorylated 59 end [17], we hypothesised that those viruses which possess noninducing genomic RNAs have removed the 59 triphosphate group. In fact, for HTNV it was previously shown that the first nucleotide of the genome 59 end is cleaved off by a viral endonuclease activity (''prime and realign''), resulting in a monophosphorylated 59 end [18]. To investigate whether CCHFV may follow a similar strategy, we employed a 59-39 exonuclease which specifically digests ssRNA having a 59 monophosphate [19]. As control, we incubated RVFV RNA which should be protected from enzymatic degradation due to its 59 triphosphate group. Fig. 4A shows that RVFV RNA can still be detected after treatment with the 59 monophosphate-specific RNase (panel 1), whereas HTNV RNA is degraded (panel 2). A similar RNase sensitivity was observed with RNA from CCHFV particles (panel 3). Thus, the inability of viral RNAs to activate RIG-I-dependent IFN induction correlates with the presence of a 59 monophosphate.
An unusual mechanism of RNA replication was also proposed for one other member of the NSVs, Borna disease virus (BDV) [20,21]. Interestingly, RNA isolated from BDV particles was unable to activate the IFN promoter, whereas RNA of measles virus (MV) was a good inducer, as expected (Fig. 4B). Moreover, BDV RNA, but not MV RNA, was undetectable after treatment 293T cells were treated with retroviral shRNA constructs directed against either RIG-I or MDA5, and cotransfected with expression constructs for HA-tagged MDA5 (left panels) or GFP-fused RIG-I (right panels). Western blot analysis using antibodies against the respective fusion tags is shown. Detection of cellular b-tubulin was used as an internal control. (B) Effect of shRNA knockdowns on IFN induction by viral RNAs, using the reporter constructs and RNA transfection protocols as described for Fig. 1A. The negative control shRNA construct (CTRL) targets the heat shock 70 interacting protein and was tested to have no effect on IFN induction (data not shown). (C) Genomic RNAs from ZEBOV and NiV were either mock treated, treated with SAP, or treated with SAP in the presence of the phosphatase inhibitor EDTA. IFN-b reporter assays and RNA transfections were performed as described for with the 59 monophosphate-specific RNase (Fig. 4C). Thus, BDV, the sole representative of the Bornaviridae family, avoids RIG-I activation in a manner similar to HTNV and CCHFV, namely by generating a 59-terminal monophosphate during the course of genome replication.

Non-inducing viral RNAs are not bound by RIG-I
We wondered whether non-inducing vRNAs would differ from inducing vRNAs in the ability to bind RIG-I. To investigate this, we performed an RNA pulldown assay, using GFP-RIG-I coupled to Sepharose beads (see Materials and Methods). These GFP-RIG-I beads were incubated with vRNAs of either RVFV, HTNV, CCHFV, or BDV. After extensive washing, RNAs were isolated from the GFP-RIG-I beads and subjected to RT-PCR analysis.    1) shows that the 59 triphosphate-containing vRNA of RVFV was precipitated by the GFP-RIG-I beads, as expected. SAP treatment abrogated RIG-I binding (data not shown), thus confirming the dependency of the interaction on the 59-terminal triphosphate group. In contrast to RVFV, particle RNAs of HTNV, CCHFV and BDV were not bound by RIG-I (Fig. 5, panels 2 to 4). These data support the notion that processing of RNA 59 ends enables viruses to escape RIG-I signaling and, hence, innate immune activation.
Taken together, our results demonstrate that the genomic RNAs from several highly pathogenic NSVs trigger RIG-I-dependent cytokine induction via their 59-terminal triphosphate group, whereas MDA5 plays no apparent role. In addition, we have uncovered that members of the Bunyaviridae and Bornaviridae have evolved a unique strategy to escape this potent pathogen sensing mechanism by removing the 59-terminal triphosphate group from their genome.

Discussion
Viruses with a negative-stranded RNA genome are responsible for a wide range of diseases with significant medical and economical impact. They are divided into groups having a nonsegmented genome (Paramyxoviridae, Filoviridae, Rhabdoviridae, Bornaviridae) and those having a segmented genome (Arenaviridae, Bunyaviridae, Orthomyxoviridae). The genome replication strategy of NSVs involves primer-independent initiation of RNA synthesis by a single nucleotide. Due to this mechanism, all NSVs were believed to contain a triphosphate group at the 59 ends of their genomic RNAs, with the notable exception of hantaviruses for which a ''prime and realign'' mechanism was described [18]. However, to our knowledge, IFN-inducing 59 triphosphates have only been proven for the RNAs of rabies virus and vesicular stomatitis virus (family Rhabdoviridae) and FLUAV (family Orthomyxoviridae) [15,16]. Here, we undertook a systematic study of all remaining NSV families and found that HTNV, CCHFV and BDV contain 59 monophosphates whereas all other NSVs tested indeed had triphosphates. These analyses suggest that the 59 triphosphate group may be a major determinant of the massive cytokine responses which are observed for infections with ZEBOV, NiV, LASV, and RVFV [9,22]. By contrast, a recent study involving hantavirus-infected patients showed that type I IFN levels did not increase during the course of disease [23]. Moreover, IFN upregulation by CCHFV or BDV infection were not reported so far.
In the recent years, an increasing number of viral proteins have been described which are effectively downregulating the induction of IFN and other cytokines, the so-called IFN antagonists [4,5]. Our results using genomic RNA from HTNV, CCHFV and BDV establish the generation of 59-terminal monophosphates as a novel strategy to avoid the activation of the IFN system. Interestingly, the mechanisms of 59 end processing differ between the viruses. HTNV employs prime-and realign to generate perfect complementary 59 and 39 genome ends able to form a panhandle structure, whereas BDV uses genome trimming resulting in a 39end overhang [18,20]. The strategy by which CCHFV processes its genome is currently unknown. However, viral RNA polymerases are only able to initiate RNA synthesis with a purine, but not with a pyrimidine [24], and CCHFV has 59-terminal pyrimidine residues [25]. Moreover, for nairoviruses a previous study had suggested that polymerase slippage occurs during RNA synthesis [26]. These observations further support our claim that, for genome replication, nairoviruses utilize a similar prime and realign mechanism involving removal of the initiating 59 nucleotide as hantaviruses do. The fact that the largely unrelated bunyaviruses and bornaviruses apply apparently different means to reach the same goal strongly suggests an independent evolution towards genomic 59 monophosphates, driven by the selection pressure of RIG-I and the antiviral IFN system. Strikingly, both HTNV and BDV are known to establish persistent infection in their natural hosts [21,27], and for CCHFV there is no evidence that it causes overt disease in carrier animals [7]. Possibly, the absence of a RIG-I ligand on the NSVs genome may facilitate the establishment of persistent or inapparent infection.
All NSV groups which were found to possess 59 triphosphates, namely paramyxoviruses, rhabdoviruses, filoviruses, arenaviruses, orthobunya-and phleboviruses and orthomyxoviruses are known to encode IFN antagonists acting on the level of IFN induction [4,[28][29][30][31][32][33][34]. Moreover, the fact that ssRNA with a triphosphorylated 59 end is an excellent trigger of RIG-I implies that an increase in the number of genome segments contained within infecting particles requires an increasingly strong and early anti-IFN function as a counterbalance. Indeed, orthobunyaviruses and phleboviruses, which have a tripartite genome, as well as the influenza A virus which has eight segments express potent IFN antagonists [4,29,35]. Interestingly, these segmented NSVs dedicate particular non-structural proteins (NSs for Bunyaviruses, NS1 for influenza virus) to silence the IFN system, whereas twosegmented and nonsegmented NSVs mostly employ structural proteins for this purpose [30][31][32][33]. This observation may implicate that for NSVs with few genome segments and hence few RIG-I ligands it is sufficient to endow structural proteins with an additional IFN-antagonistic function, whereas a higher number of genome segments requires either strong and specialized factors like NS1 and NSs, or the removal of the RIG-I ligands. Moreover, the fact that NSVs containing 59 triphosphates elicit strong cytokine responses despite the presence of IFN-antagonistic proteins indicates that the balance between induction and viral inhibition is disturbed at some point during infection, possibly by a surplus of non-functional genomes such as produced by defective interfering particles [36] or by infection of cells which are resistant to the action of viral IFN antagonists.
In addition to the removal of the RIG-I ligand, hantaviruses modulate IFN induction by a weakly active NSs protein [37] and by an additional function of their glycoproteins [38,39]. Likewise, the BDV P protein interferes with IFN induction [40], whereas for CCHFV no IFN antagonistic protein has been identified so far. The necessity to actively abrogate IFN induction despite the absence of the 59triphosphate group may be due to the presence of other viral structures which trigger an innate immune response, e.g. viral nucleocapsids or traces of dsRNA which were shown to play a role in NSV cytokine activation [41][42][43][44][45]. Moreover, enveloped virus particles by themselves can activate IRF-3 in a TLR-and RIG-I-independent manner [46,47], implicating that mammalian hosts can respond to several independent hallmarks of virus infection.
In summary, our study of the negative-stranded group of RNA viruses revealed that genome recognition by RIG-I is indeed a widespread phenomenon, but that hantaviruses, nairoviruses and bornaviruses avoid induction of antiviral responses by processing their RNA termini. These findings may help to better understand viral pathogenesis and particular cytokine profiles observed during infections.

Plasmid constructs
The firefly luciferase reporter construct for monitoring IFN-b promoter activation (p125Luc) was kindly provided by Takashi Fujita, Kyoto University, Japan [48]. The control plasmid pRL-SV40 (Promega) contains the Renilla luciferase gene under control of the constitutively active SV40 promoter. Expression plasmids pHA-MDA5 encoding human MDA5 fused to an N-terminal HA tag and pGFP-RIG-I expressing human RIG-I fused to the C terminus of EGFP were described previously [16].

Preparation of virus particles
Supernatants from infected cells were harvested at day 2 (CCHFV), day 3 (FLUAV, RVFV, NiV), day 4 (ZEBOV, MV), day 7 (LASV), day 10 (HTNV), or day 60 (BDV) post-infection. For preparation of virus particles, cell culture supernatants were first pre-cleared from cell debris by centrifuging at 1500 rpm for 10 min and 4uC. Then, virus particles were concentrated either by ultracentrifugation at 24,000 rpm for 2 h and 4uC (RVFV, MV, BDV), ultracentrifugation at 45,000 rpm for 3.5 h at 4uC (CCHFV, HTNV), or by PEG precipitation (FLUAV, NiV, ZEBOV, LASV). IFN-inducing properties are not affected by the method of virus concentration (data not shown). For PEGmediated particle precipitation, 175 ml of pre-cleared cell culture supernatant were supplied with 12.25 g PEG 8000 and 4.02 g NaCl and stirred overnight at 4uC. Then, samples were transferred into 50 ml Falcons and centrifuged at 6,000 rpm for 45-60 min at 4uC. The pellet was dissolved in 1.5 ml of TriFast reagent (Peqlab). RNA isolation was performed as described by the manufacturer.

Transient transfections and reporter gene assays
Subconfluent 293T cell monolayers grown in 12-well dishes were transfected with 0.25 mg p125-Luc and 0.05 mg pRL-SV40 in 100 ml OptiMEM (Gibco-BRL) containing 0.9 ml Fugene HD (Roche). After 6 h at 37uC, the liposome-DNA mixture was removed and cells were transfected either with 1 mg of viral RNA or with 5 mg of poly(IC) using 3 ml Metafectene (Biontex) per mg RNA prepared in 100 ml OptiMEM. At 18 h post transfection, cells were harvested and lysed in 100 ml of Passive Lysis Buffer (Promega). An aliquot of 20 ml lysate was used to measure luciferase activities using the dual luciferase assay (Promega).

shRNA knockdowns
Gene silencing by RNAi was achieved using shRNAs expressed by retroviruses (shRNAmir constructs, Open Biosystems). The construct which efficiently targets MDA5 had the catalog number RHS1764-9494563 (MDA5 #1) whereas RIG-I was targeted with RHS1764-9499511 (RIG-I #2). The control shRNA construct targets human Hsp 70 interacting protein and contained the sequence RHS1764-9500871. For establishing RNAi, 293T cells grown in 6-well dishes to 80% confluency were transfected with 1 mg each of pVpack-VSV G, pVpack-GP (Stratagene) and the shRNA construct, subcultured, incubated with Puromycin at day 3 post-transfection, and then incubated for another 2 to 3 days before usage.

Enzymatic treatments of RNA
To remove 59-terminal phosphates, purified RNAs were treated with 1 U of shrimp alkaline phosphatase (SAP) according to the protocol of the manufacturer (Roche). After incubation at 37uC for 10 min in the presence or absence of the inhibitor 1 mM EDTA, the enzyme was heat-inactivated at 65uC for 15 min.
Terminator TM (Epicentre Biotechnologies) is a 59-39 exonuclease able to digest ssRNA with a 59 monophosphate. To determine the presence of a 59 monophosphate on genomic RNAs isolated from virus particles, 500 ng of RNA were either mock treated or incubated with 1 U of Terminator exonuclease for 4 h at 30uC. The RNA was then purified using RNAeasy Mini Kit (QIAGEN), eluated in nuclease-free H 2 O and treated with 1 U DNase (Fermentas). To check for the presence of RNA, RT-PCR analysis was performed as indicated above.

RNA pulldown analysis
Human 293T cells grown in 6-well dishes to 60% confluency were transfected with 1 mg pGFP-RIG-I per well. After 40 h of incubation, cells were washed with phosphate-buffered saline and lysed for 10 min at room temperature in 200 ml of RIPA buffer (50 mM Tris/HCl pH 7.5, 150 mM NaCl, 1% NP-40, complete protease inhibitor mix (Roche)). In parallel, 20 ml of a 50% slurry of protein G Sepharose in RIPA buffer were preadsorbed with 5 ml polyclonal rabbit anti-GFP antiserum (Invitrogen) and incubated for 2 h at 4uC. Then, the Sepharose beads were washed three times with RIPA buffer and incubated for 2 h at 4uC with the lysate equivalent of half a 6-well plate of transfected 293T cells. The immunoprecipitates were washed three times with RIPA buffer and then incubated with 1 mg of viral RNA for a further 2 h at 4uC. After three washing steps using RIPA buffer, the volume of the Sepharose beads was adjusted with H 2 O to 100 ml and RNA was extracted using the RNeasy kit (Qiagen). RNA pulldown was analysed by RT-PCR using random hexamers for RT and virusspecific primers for PCR.