key: cord-0723257-2mceonc1
authors: Python, Sylvie; Gerber, Markus; Suter, Rolf; Ruggli, Nicolas; Summerfield, Artur
title: Efficient Sensing of Infected Cells in Absence of Virus Particles by Blasmacytoid Dendritic Cells Is Blocked by the Viral Ribonuclease E(rns)
date: 2013-06-13
journal: PLoS Pathog
DOI: 10.1371/journal.ppat.1003412
sha: 8b11ed8edd1f59ac713ddcc3282ad5a98b8b68b7
doc_id: 723257
cord_uid: 2mceonc1

Plasmacytoid dendritic cells (pDC) have been shown to efficiently sense HCV- or HIV-infected cells, using a virion-free pathway. Here, we demonstrate for classical swine fever virus, a member of the Flaviviridae, that this process is much more efficient in terms of interferon-alpha induction when compared to direct stimulation by virus particles. By employment of virus replicon particles or infectious RNA which can replicate but not form de novo virions, we exclude a transfer of virus from the donor cell to the pDC. pDC activation by infected cells was mediated by a contact-dependent RNA transfer to pDC, which was sensitive to a TLR7 inhibitor. This was inhibited by drugs affecting the cytoskeleton and membrane cholesterol. We further demonstrate that a unique viral protein with ribonuclease activity, the viral E(rns) protein of pestiviruses, efficiently prevented this process. This required intact ribonuclease function in intracellular compartments. We propose that this pathway of activation could be of particular importance for viruses which tend to be mostly cell-associated, cause persistent infection, and are non-cytopathogenic.

Although representing a rare cell type of the immune system, plasmacytoid dendritic cells (pDC) are the most important source of systemic interferon (IFN) type I in the early phase of many virus infections, and as such a critical early alarm system against viruses [1, 2] . This is based on the ability to produce around 1000 times more IFN type I than any other cell type [1] . Accordingly, pDC possess the necessary cell biological features such as Toll-like receptor (TLR)7 and TLR9 and constitutive high levels of IFN regulator factor (IRF)-7 to sense viruses with high efficiency [2] . While it is evident from the literature that TLR7 is the most important sensor of RNA viruses and TLR9 for DNA viruses, it is less clear how the viral nucleic acids have access to these compartments and how the encapsulated viral nucleic acids get in contact with the TLR's. Recently, a novel process of pDC stimulation by infected cells independent of viral particles and their uptake by pDC has been described for hepatitis C virus (HCV) in which pDC sense infected cells in a cell contact-and TLR7-dependent manner [3, 4] . This process has been described to be more effective than stimulation of pDC with cell-free virions. Stimulation of pDC by infected cells has also been reported for HIV [5] . However, this process was blocked by neutralizing antienvelope antibodies [5] implying a different mechanism of RNA transfer to the TLR7 compartment as observed with HCV where stimulation is not blocked by virus neutralization [3] . For the latter virus, lipid rafts and tetraspanin-enriched membrane microdo-mains have been described to be involved in infected cell sensing by pDC [4] .

Considering the potential importance of this mechanism of pDC activation, we initiated this study with a focus on another member of the Flaviviridae, classical swine fever virus (CSFV) belonging to the pestivirus genus. CSFV is the causative agent of a viral hemorrhagic fever in pigs with disease characteristics resembling dengue hemorrhagic fever if pigs are infected with highly virulent CSFV strains. However, with low virulent strains chronic disease and persistent infections are observed [6] . CSFV, in contrast to HCV, has a particular tropism for cells of the macrophage and DC lineage and most efficiently infects pDC [7] [8] [9] .

Our results demonstrate that the basic characteristics of pDC stimulation by infected cells resemble those of HCV. In addition, using such cultures the present study identified a striking function of the enigmatic viral E rns , an essential structural protein with ribonuclease (RNase) activity [10] . The pestivirus genus is composed of major veterinary pathogens, the most important being CSFV and bovine viral diarrhea virus (BVDV). Interestingly, E rns is unique to pestiviruses. Despite its function as structural protein, E rns exists as soluble form secreted from infected cells and has been proposed to be involved in immune evasion of pestiviruses (for review, see [11] ). Removal of the RNase activity was demonstrated to result in virus attenuation [12, 13] and abrogation of the capacity of pestiviruses to establish immunetolerance and persistent infections after infection of fetuses [14, 15] .

However, it has been difficult to understand the mechanism of immune evasion using in vitro studies. Recombinant E rns degrades synthetic single-stranded and double-stranded RNA added to the cultures [16] [17] [18] but pestiviruses with or without RNase activity do not induce IFN type I in cell culture and replicate to the same titers as their wild type counterpart. In this study we have identified how E rns potently counteracts IFN-a induction in pDC. It represents the first example of a viral protein that prevents the stimulation of pDC by infected cells, and thus represents a novel pathway of viral evasion of the type I IFN system. Furthermore, it underlines the importance of stimulation of pDC by infected cells, rather than virions.

In accordance to previous studies [7, 8] , CSFV as well as virus replicon particles (VRP) lacking the E rns gene (VRPDE rns ) were poor stimulators of pDC , inducing between 0 and 550 IFN-a units per ml, dependent on the experiment. Interestingly, stimulation of pDC by co-culture with CSFV-infected or VRPDE rns -infected SK-6 cells induced up to 100-fold more IFN-a compared with direct infection of the pDC, with an optimum at 40'000 to 80'000 infected SK-6 cells per 2610 5 CD172a + enriched pDC ( Figure 1A and B). While no significant difference between direct CSFV and VRPDE rns stimulation was observed, CSFV-infected SK-6 cells stimulated an average of 5.1 more IFN-a when compared to direct stimulation by CSFV ( Figure 1C and D). This difference was even more evident when direct simulation with VRPDE rns was compared to stimulation by VRPDE rns -infected cells ( Figure 1E ). Interestingly, VRPDE rns -infected SK6 cells were in average around 8 times more stimulatory than CSFV-infected SK6 cells ( Figure 1F ).

In accordance to previous studies demonstrating that pDC were the only cell type able to respond to CSFV by IFN-a production [19] [20] [21] , pDC were the only source of IFN-a following stimulation with infected cells, as demonstrated by intracellular IFN-a staining which was only found in the CD4 high CD172a + pDC population. Furthermore, purified monocytes did not produce IFN-a in response to any of the stimuli tested (Supplementary Figure 1) . 

Plasmacytoid dendritic cells (pDC) represent the most potent producers of interferon type I and are therefore of major importance in antiviral defences. A TLR7-dependent induction of interferon-a in pDC by infected cells in the absence of virions has been demonstrated for hepatitis C virus. Here, we show that this pathway is also very efficient for classical swine fever virus, a pestivirus that is also a member of the Flaviviridae. Our data indicate a transfer of RNA from the donor cell to pDC in a cell-contactdependent manner requiring intact lipid rafts and cytoskeleton of the donor cell. Importantly, we demonstrate that the enigmatic viral E rns protein unique to pestiviruses efficiently prevents this pathway of pDC activation. This novel function of E rns is dependent on its RNase activity within intracellular compartments. The present study underlines the importance of pDC activation by infected cells and identifies a novel pathway of virus escaping the interferon system. Considering that E rns is required for pestiviruses to establish persistent infection of foetuses after transplacental virus transmission resulting in the development of immunotolerant animals, this report also points on a possible role of pDC in preventing immunotolerance after viral infection of foetuses.

The above results suggested that infected SK-6 cells would transfer viral RNA to pDC resulting in pDC activation. Considering the fact that VRP deliver self-replicating RNA which replicates for many days in SK-6 cells [22] , we tested if functional replicon RNA was transferred between SK-6 cells and pDC by determining the expression of the viral NS3 protein in pDC. NS3 is generated by post-translational processing of the CSFV precursor polyprotein. Detectable amounts of NS3 in cells can only be obtained with replication competent pestivirus genomes, i.e. full-length genomes and replicons. As shown in Figure 2 , after co-culture of pDC with VRPDE rns -infected SK-6 cells for 22 h, approximately 12-14% pDC expressed NS3, indicating either a transfer of intact full-length replicon RNA or viral NS3 protein between the cells. Interestingly, co-culture of CSFV-infected cells with pDC resulted in a higher degree of infection (94%) compared to direct infection by the virus (65%). Another observation was that, when infectious CSFV or VRP were present, the percentage of NS3-expressing pDC was higher when compared to the monocytes that were co-purified using the CD172a selection. It is also noteworthy that NS3 + monocytes were found after co-culture with VRP-infected SK-6 cells. The results presented in Figure 2 also highlight that there is no correlation between infectious titers, percentage of viral protein expressing pDC and IFN-a responses. An over 20 times higher IFN-a response was found when pDC were stimulated with VRP-infected SK-6 cells in which infectious virus was barely detectable. We confirmed these results by employing RT-PCR to quantify the number of viral genome copies in these cultures. pDC were stimulated either directly with VRPDE rns or with VRPDE rns -infected SK-6 cells, and then resorted from these cultures by CD172a MACS sorting. This demonstrated that around 0.1% of the total viral genome present in infected SK-6 cells was transferred to pDC/monocytes. Direct infection of enriched pDC was ,6 times more efficient in delivering RNA but induced much lower IFN-a levels ( Table 1 ).

In order to rule out a role for free virions in the induction of IFN-a, we measured the infectivity in VRPDE rns -infected SK-6 at the time of co-culture with pDC and found titers below 10 2 TCID 50 /ml. This corresponded to MOI of less than 10 24 TCID 50 /enriched pDC, suggesting that stimulation of pDC does not occur by infection of pDC with VRP. The lack of involvement of free virions in pDC stimulation by infected cells was further confirmed by addition of a neutralizing monoclonal antibody against the main glycoprotein E2 of CSFV or by addition of a porcine polyclonal hyper immune serum obtained from an infected pig which contained antibodies against both glycoproteins of the virus, E2 and E rns . The antibody preparations completely blocked direct activation of pDC by cell-free virions ( Figure 3A and B), but were unable to inhibit activation by infected cells, independent whether the stimulation used CSFV-or VRPDE rnsinfected SK-6 cells ( Figure 3C and D). With CSFV-infected SK-6 cells, the antibodies even enhanced pDC stimulation. Antibodies from naïve animals or an irrelevant monoclonal antibody had no effect (data not shown). Similarly, the neutralizing antibodies blocked infection of pDC after direct stimulation but were unable to inhibit the expression of NS3 in pDC after stimulation by infected cells (supplementary Figure S2) .

We further confirmed the absence of any virus particle in pDC stimulation by employing RNA-transfected SK-6 cells to stimulate pDC. To this end, SK-6 cells were transfected with in vitro transcribed RNA synthesized with plasmids encoding the E rnsdeleted genome of CSFV (pA187-DE rns , DE rns RNA) or a genome of CSFV devoid of all structural proteins (pA187D-Apa, DApa Figure 3E demonstrate that CSFV RNA-transfected SK-6 cells can activate pDC, although the levels of IFN-a were low, compared to VRP-infected SK-6 cells. This was explainable by the relatively low transfection efficiency of 15-20% in terms of NS3 + expression (data not shown). We also generated VRP with a deletion of E2 instead of E rns . SK-6 cells infected with such VRPDE2 were more potent in inducing IFN-a in pDC than VRPDE2 virions. Interestingly, SK-6 cells infected with VRPDE rns were more stimulatory than SK-6 cells infected with VRPDE2 ( Figure 3F ).

In order to determine the role played by TLR7 in sensing CSFV-infected cells, we used the immunoregulatory sequence 661 (IRS661) representing an oligodeoxynucleotide inhibitor of TLR7 which had been previously established for the human, murine and porcine immune systems [23, 24] . IRS661 at a concentration of 0.7 mM efficiently inhibited CSFV-and VRP-induced pDC activation ( Figure 4A ). With infected cells inducing much higher levels of IFN-a, the reduction caused by IRS661 was still over 80% with the highest inhibitor concentration ( Figure 4B ).

A striking observation was that in all experiments VRPDE rnsinfected SK-6 cells were clearly more efficient at inducing IFN-a compared to CSFV-infected SK-6 cells ( Figure 1F ). Considering that VRPDE rns lacks the E rns gene, we postulated a role for this viral protein in inhibition of this novel type of IFN-a induction and tested this by comparing SK-6 cells stably expressing E rns to the parent wild type SK-6 cells. Indeed, the SK-6(E rns ) cells infected with VRP or CSFV were very inefficient at inducing IFN-a in contact with pDC ( Figure 5A and B). This was not a result of different susceptibility to the viruses as the infection rate of the SK-6(E rns ) was 99%, similar to CSFV ( Figure 6F ). As expected, E rns expressed by the SK-6(E rns ) trans-complemented VRPDE rns to generate infectious particles resulting in a higher degree of NS3 + pDC (supplementary Figure S3 ). Considering that this must be associated with a higher viral RNA load in pDC, these results indicate a potent function of E rns in preventing the stimulation of pDC by infected cells. E rns has been shown to be mainly associated with intracellular membranes in particular of the ER, with almost no cell surface expression [25] . However, considering that approximately 16% is secreted to the extracellular space [26, 27] , we tested if E rns secreted by SK-6(E rns ) cells was responsible for the observed inhibition. To this end, supernatants of SK-6(E rns ) or SK-6 were added to co-cultures of VRPDE rns -infected SK-6 cells and pDC. As shown in Figure 5C , there was no evidence for any suppressive effect of soluble E rns in the supernatants.

We next determined the role of N pro , a well established type I IFN antagonist of CSFV in non-pDC and pDC targeting IRF3 and IRF7 transcription factors [8, 19, 28] . To this end, we compared the ability of SK-6 cells infected with VRPDE rns and VRPDE rns D 136 N expressing a non-functional mutant of N pro [22] , to activate pDC. Our results demonstrated that only E rns , but not N pro prevents the activation of pDC by infected cells ( Figure 5D ).

In order to confirm this function of E rns , we compared the effect of SK-6 and SK-6(E rns ) cells infected with two other RNA viruses, the picornavirus foot-and-mouth disease virus (FMDV) and the coronavirus transmissible gastroenteritis virus (TGEV), on pDC. When the stimulation used FMDV-or TGEV-infected SK-6 and SK-6(E rns ) cells for comparison, the responses with infected SK-6(E rns ) cells were 3 to 6-fold lower compared to SK-6 cells ( Figure 5 E and F). This was not caused by an inhibitory effect of E rns on virus replication (Supplementary Figure S4 ). In order to exclude a potential ''toxicity'' derived from E rns expressing cells we also tested the responses of pDC to CpG when stimulated in cocultures with SK-6 and SK-6(E rns ) cells, and found similar levels of IFN-a (Supplementary Figure S5 ).

In order to test the requirement of RNase activity for the above E rns function, we constructed an RNase-negative mutant of CSFV by deleting the histidin codon at position 346 [13] . The mutation abolished the RNase activity ( Figure 6A ), confirming previously published data [13] . Interestingly, RNase activity of E rns was detectable in cell extracts only ( Figure 6A ). While SK-6 cells infected with wild type CSFV induced under 500 U/ml of IFN-a, cells infected with the CSFV-E rns D346 mutant induced approximately 10-fold higher responses ( Figure 6B ). Furthermore, direct stimulation of pDC by virus also dramatically increased when the CSFV-E rns D346 mutant was employed. Nevertheless, the levels of IFN-a remained clearly under those induced by infected cells. For further confirmation that the inhibitory effect of E rns in pDC depends on the RNase-activity of E rns , we constructed lentivirustransduced SK-6 cell lines expressing the parent E rns [SK-6LV(E rns )] or an RNase-inactive mutant E rns [SK-6LV(E rns D346)]. As expected, the deletion of the histidine at position 346 completely abolished the RNase activity of E rns ( Figure 6C ). Again, RNase activity of the parent E rns was detectable only in the cell extracts. As expected, only SK-6 expressing RNase active E rns prevented IFN-a induction by VRP infection of the cell lines ( Figure 6D ). All SK-6 cell lines expressing both wild-type and mutant E rns had comparable levels of E rns , demonstrating that the striking differences observed in RNase and inhibitory activity for activation of pDC were not caused by lack of E rns expression. In addition, the levels of E rns in these cells were below those found after natural infection by wild-type CSFV ( Figure 6E ). Furthermore, all E rns -expressing cell lines were found to be fully susceptible to VRPDE rns infection as determined by NS3 expression ( Figure 6F ). The observed higher levels of NS3 expression in E rns expressing cells was a consequence of transcomplementation of VRPDE rns and generation of infectious virions in these cultures.

Considering that macrophages (MW) and endothelial cells (EDC) represent important target cells for CSFV, we also tested if infected MW and EDC, similar to SK-6 cells, were able to activate pDC. Indeed, VRPDE rns -infected and CSFV-infected MW were able to activate pDC while no responses were detectable with CSFV alone ( Figure 7A ). VRPDE rns were relatively inefficient at infecting MW (26% NS3 + versus 97% NS3 + with CSFV), which explains the lower IFN-a responses when compared to CSFV. The highest levels of IFN-a were induced by MW infected with the CSFV-E rns D346 mutant (83% NS3 + MW) confirming the functioning of E rns also in MW ( Figure 7A ). Neither the VRPDE rns , nor WT CSFV or the CSFV-E rns D346 mutant were able to induce IFN-a in MW (data not shown). We also further confirmed our findings using the immortalized porcine EDC line PEDSV.15. VRPDE rns -infected and CSFV-E rns D346-infected EDC were more potent at activating pDC when compared to CSFV-infected EDC ( Figure 7B ). In contrast to MW, the rate of endothelial cells infection was comparable with the viruses and VRPDE rns (VRPDE rns : 80% NS3 + , CSFV: 92% NS3 + and CSFV-E rns D346 mutant 92% NS3 + ).

Based on the above results we postulated that viral RNA is transported from the infected cells to the TLR7 compartment of pDC in a manner avoiding contact of the RNA with the extracellular space. Consequently, we were interested to investigate if membrane vesicles transporting viral RNA from the infected cell to the pDC could be involved in pDC activation by infected cells. To this end, we co-cultured pDC with VRP-infected SK-6 cells in transwell culture dishes using 0.4 or 1 mm pore sizes. Only pDC with direct contact to infected SK-6 cells produced large quantities of IFN-a ( Figure 8A and B) . Notably, pDC responses to the CpG control were in the same level of magnitude if the pDC were cultured in the insert or in the well of the plate (data not shown). When the same experiments were performed with CSFV-infected SK-6 cells, low IFN-a responses were also observed in all transwell conditions ( Figure 8B ). Considering that CSFV virions are only 60 nm, this response was probably mediated by direct stimulation of pDC with virions passing the membranes. Similar to VRP-infected or CSFV-infected SK-6 cells, infected MW and EDC were unable to induce IFN-a when separated from pDC using a transwell culture system (data not shown).

Drugs affecting cytoskeleton and membrane cholesterol prevent stimulation of pDC by infected cells. The above results indicate that infected SK-6 cells transfer RNA to pDC using a pathway which requires cell-to-cell contact dependent communication between infected cells and pDC. In order to further characterize this pathway we tested the impact of cytoskeleton inhibitors as well as the cholesterol-depleting chemical MbCD. The VPR-infected SK-6 cells were treated for the last 2 h of infection followed by removal of the inhibitor by a wash step before co-culture, to avoid direct effects of the drugs on pDC and to enable efficient virus replication before co-culture. As shown in Figure 9 , latranculin B, an inhibitor of microfilament formation [29] , as well as nocodazole, which depolymerizes microtubules [30] , almost completely prevented pDC activation. Furthermore, depletion of cholesterol from SK-6 cells by MbCD treatment prior to co-culture with pDC also blocked IFN-a responses of pDC, suggesting a potential role of lipid rafts in RNA transfer to pDC. To rule out a direct suppressive effect of the drugs on pDC functions we also stimulated pDC with CpG in presence of drugtreated uninfected SK-6 cells, and found no significant effects ( Figure 9B ). Furthermore, we demonstrated that the drug treatments did no negatively impact virus replication in SK-6 cells (Supplementary Figure S6 ).

During a virus infection, pDC will not only encounter virions but also virus-infected cells. The ability to sense the latter has the advantage of being able to sense infection before or without release of virions and also to better sense viruses which are particularly cell-associated and tend to cause persistent virus infections, such as HIV, HCV and pestiviruses. The present study underlines the importance of this by demonstrating that pDC stimulation by infected cells can be much more efficient than stimulation by virions. This is also emphasized by the identification of a viral protein that appears to have evolved to efficiently inhibit this pathway. After HIV [5] , HCV and Venezuelan equine encephalitis virus [3] , CSFV is now the fourth virus for which this mechanism of pDC stimulation has been found to be more potent than direct stimulation of pDC by virions. Although free CSFV was also able to stimulate pDC, the IFN-a responses were inferior to those induced by cell-free virions. Interestingly, also with FMDV, a virus which does not or very inefficiently stimulate pDC [31] , infected cells were able to stimulate pDC.

The basic characteristics of pDC stimulation by CSFV-infected cells are similar to those observed with HCV. It represents a TLR7-dependent process which cannot be blocked by neutralizing antibodies and does not require expression of viral glycoproteins [3] . This is in contrast to the situation for HIV [5] . Our data indicate that this pathway of pDC activation is dependent on cell contact, intact actin filaments, microtubules as well as cell membrane cholesterol indicating a role for lipid rafts, although future studies are required to directly demonstrate the functional interaction of these cellular components with viral proteins. We have also identified that the pestiviral E rns potently prevents pDC activation by infected cells. E rns possesses several remarkable features, of which the RNase activity is of particular interest, considering that it is expressed by an RNA virus. E rns has structural similarities with plant T2 RNases which have their optimal catalytic activity at an acidic pH [32] with a preference for cleaving single-stranded RNA [33, 34] . This would point on an activity within the endosomal compartment, which is also supported by our data indicating that genomic RNA does not appear to be degraded. The protein also has an unusual membrane anchor composed of an amphipathic helix without a typical membrane anchor [27, 35] , but with a retention signal ensuring its association with intracellular membrane compartments [25] . Whether this enables accumulation in the endosomal system with appropriate orientation needs to be investigated. We postulate that viral RNA may have access to the endosomal system via the autophagy process. These endosomes could be transferred to pDC by a pathway to be defined and then fuse with the TLR7 compartment. Alternatively, both cytoplasmic viral RNA and E rns could be autophagocytosed after transfer to pDC followed by fusion of autophagosomes with TLR7-containing endosomes [36] . Only at this acidic location the RNAse function would be highly active and rapidly degrade the viral RNA to reduce TLR7 triggering. Consequently, one of the first questions to be addressed in future studies is whether E rns is transported from SK-6 cells to pDC to degrade RNA in the TLR7 compartment of pDC. Certainly, E rns can act in pDC as our data is showing that CSFV with E rns lacking RNase activity induce much higher IFN-a responses compared to wild type CSFV when pDC are directly stimulated by virions. Strikingly, E rns can exert its inhibitory function when expressed independently of the viral context in the infected cells that stimulate the pDC. In addition, our results suggest that E rns functions by using intracellular rather than extracellular pathways, since SK-6(E rns ) supernatants did not suppress pDC activation by VRPDE rns -infected SK-6 cells. This is contrary to a role proposed for secreted E rns by several authors. Based on the observation that a minor part of the protein was found to be secreted from infected cells or cells expressing E rns [25] [26] [27] , recombinant E rns was tested and found to degrade synthetic single-stranded and double-stranded RNA added to the cultures [16] [17] [18] .

Considering that pDC are by far the most potent producers of IFN type I and of crucial importance in linking innate to adaptive immunity, our data shed light into one of the most fascinating aspects of pestivirus biology. These viruses cause persistent infections, in both cattle and pigs. When bovine fetuses are infected transplacentally by BVDV between the 2 nd and 4 th month of pregnancy, a time point at which the fetuses are not yet immunocompetent [37] , persistently infected calves are born CSFV with inactive RNase function of E rns (CSFV-E rns D346) is a better stimulator of pDC. Enriched pDC were activated by virus (hatched bars) or by coculture with infected SK-6 cells. The IFN-a responses to wild type CSFV and CSFV-E rns D346 are compared. D to F. SK-6 cells expressing RNase active E rns prevent IFN-a induction by VRPDE rns infection, while VRPDE rns -infected SK-6 cells expressing E rns D346 re-gain their ability to stimulate pDC. SK-6 cells were transduced by lentiviral gene delivery to express wild type E rns or E rns D346 with 75% transduction efficiency (E). The SK-6(E rns ), and the LVtransduced SK-6LV(E rns ) and SK-6LV(E rns D346) were infected with VRPDE rns and then tested for their ability to induced IFN-a in enriched pDC (D). The mean values of three replicates with standard deviations are shown calculated from data which is representative of three independent experiments. All E rns expressing cells lines were highly susceptible to infection by VRPDE rns as determined by viral NS3 expression (F). doi:10.1371/journal.ppat.1003412.g006 which are fully immunotolerant to the virus and cannot mount any adaptive immune responses against the virus. These persistently infected calves play a major role in the epidemiology of the disease by shedding virus. Similar observations have been reported for pigs after infection of pregnant sows with low virulent strains of CSFV before eradication of CSFV from most European countries and the U.S.A. [38] [39] [40] [41] [42] [43] [44] . This contrasts with the pathogenesis of highly virulent strains of CSFV which induce acute disease with high mortality which is associated with high systemic levels of IFN-a [45] . This appears contradictory to the present report but compared to other viruses the ability of CSFV to activate pDC is weak [9] . Even if pDC are stimulated by cells infected with wild type CSFV, IFN type I responses remain relatively weak compared to influenza virus. Our concept to explain this apparent contradiction to the in vivo situation with virulent strains of CSFV is based on the prominent tropism of CSFV for MW and DC, and its localization predominantly in lymphoid tissue. Highly virulent isolates of CSFV cause a very rapid and strong viremia, in which the virus reaches simultaneously all primary and secondary lymphoid tissues where relatively high numbers of pDC are localized. This situation could explain why such viruses are able to induce a potent IFN-a response despite the activities of N pro and E rns [9] .

Interestingly, it has been demonstrated that the virus not only needs a functional N pro , but also an E rns with active RNase to establish persistent infections in cattle [14] . N pro induces the degradation of IRF3 and thereby efficiently prevents IFN type I induction in all host cells including conventional DC, which have been induced to express IRF7 by IFN type I pre-treatment [19, 28] . However, N pro can only partially prevent IFN-a responses in pDC [8] and is unable to stop the much more potent activation of pDC by infected cells (this study). We thus propose that E rns has evolved to prevent this pathway of innate immune system activation, which is much more potent and therefore likely to be essential for the virus to be able to establish persistent infections, representing a main survival strategy of pestiviruses [46] .

Bleeding and care of donor pigs was carried out in accordance with EU standards and National laws (Tierschutzgesetz SR455). Specifically, approval of the protocol employed was obtained by the Animal Welfare Committee of the Canton of Bern, Switzerland (animal license BE26/11). Cell preparation and cell culture

The porcine kidney cell lines SK-6 [47] , PK-15 (LGC Standards-ATCC, Molsheim, France) and the porcine immortalized endothelial cells PEDSV.15 [48] (obtained from Dr. Jörg Seebach, University of Geneva, Switzerland) were propagated in Earle's minimal essential medium (MEM) substituted with 7% horse serum and in Dulbecco's minimal essential medium (DMEM) supplemented with 5% horse serum, nonessential amino acids and 1 mM Na-pyruvate, respectively. SK-6 cells stably expressing E rns of CSFV strains Alfort/187, termed SK-6(E rns ) were generated as described previously [49] . Baby Hamster Kidney (BHK) 21 cells were grown in Glasgow's minimum essential medium (Life Technologies) supplemented with 5% v/v fetal bovine serum (FBS, Biowest, Nuaillé, France). Peripheral blood mononuclear cells (PBMC) were obtained from blood of specific-pathogen-free pigs using ficoll-paque density centrifugation (1.077 g/L, Amersham Pharmacia Biotech). pDC were enriched as described previously [31] by cell sorting of CD172a + PBMCs using the magnetic cell sorting system (MACS) with LD columns (Miltenyi Biotec GmbH, Germany). This permits a 10fold enrichment of functional pDC to 2-5%. In some experiments, pDC were enriched by a first depletion of CD14 + monocytes followed by CD172a enrichment [24] , permitting a pDC enrichment of around 5-10%. Enriched pDC were cultured in DMEM with Glutamax, 20 mm b-mercaptoethanol (Life Technologies). 10% FBS was only added to when indicated. Porcine monocytes were isolated by CD14 + selection using MACS with LS columns. Porcine MW were generated from CD14 + PBMCs as previously described using a 3-day culture in DMEM supplemented with 10% autologous porcine serum [50, 51] .

CSFV strain vA187-1 was derived from the full-length cDNA clone pA187-1 [52] . Plasmid pA187-1 carries a full-length cDNA copy of the CSFV strain Alfort/187 and served as basis for all viruses and replicon cDNA constructs. The vA187-E rns D346 virus (referred to as CSFV-E rns D346 mutant), with a histidine deletion in E rns at position 346 of the viral polyprotein resulting in loss of RNase activity [13] was rescued by standard procedure [52, 53] from plasmid pA187-E rns D346. This latter construct was gener-ated from pA187-1 using PCR-based site-directed mutagenesis with oligonucleotide primers encompassing the deletion and PfuUltra DNA polymerase (Agilent), employing standard cloning techniques as previously described [54] . The plasmids pA187-DE rns carrying an in frame deletion of the E rns gene in the pA187-1 backbone, and pA187-D 136 N-DE rns carrying the same deletion and expressing a non-functional D 136 N mutant of N pro were described elsewhere [22, 49] . Plasmid pA187-E2del373 carrying an in frame deletion of the complete E2 gene in the pA187-1 backbone was also described previously [22, 49, 53] . Plasmid pA187-DApa encoding a CSFV replicon with a deletion of the structural protein genes C, E1, E rns and most of the E2 gene (i.e. the codons encoding amino acids 96 to 962 of the polyprotein) was described earlier [53] . In vitro transcribed replicon RNA was produced using SrfI-linearized pA187-DE rns , pA187-D 136 N-DE rns , pA187-E2del373 or pA187-DApa as described [22, 49, 53] . VRPDE rns carrying a genome with a complete deletion of the E rns gene were described previously, and produced by transfection of SK-6(E rns ) cells with A187-DE rns or A187-D 136 N-DE rns replicon RNA. The SK-6(E rns ) cells express the E rns protein required for the generation of VRPDE rns by trans-complementation [22, 49] . Similarly, VRPDE2 carrying the pA187-E2del373-derived replicon with a complete E2 deletion were rescued by transfection of SK-6(E2p7) cell line expressing the CSFV E2 and p7 proteins as described previously [22, 49, 53] . TGEV (TGEV; strain Perdue 115) was propagated in PK15 cells [55] . The FMDV type O UK/ 2001 isolate was grown in (BHK21) cells as described previously [56] . All virus titers were determined on SK-6 cells, PK-15 cells or BHK21 cells (for CSFV, TGEV and FMDV, respectively) by standard endpoint dilution and were expressed as 50% tissue culture infectious doses (TCID 50 ) per ml.

CSFV or VRP RNA was quantified using a published real-time RT-PCR [57] . Briefly, RNA was extracted using the Trizol method and RT-PCR performed with the SuperScript III Platinum One-Step qRT-PCR System (Life Technologies) using 7500 Real-time PCR System, Applied Biosystems. To determine the absolute number of RNA copies, vA187-1 RNA transcripts generated in vitro were employed. Generation of cell lines expressing E rns using lentiviruses A lentivirus (LV) expression system using plasmids obtained from the laboratory of Dr. Didier Trono (http://tronolab.epfl.ch/ Ecole Polytechnique Federale de Lausanne, Switzerland) or through Addgene (Cambridge MA, USA) [58, 59] was employed. For cloning of E rns and its RNase-knock-out mutant variant into the lentiviral transfer plasmid pWPT-GFP (Addgene) the pA187-1 and pA187-E rns D346 plasmids were used as template for PCR amplification using primers to insert the MIuI and Sall restriction sites excising the GFP in the pWPT-GFP vector. The PCR products were first inserted into the pJET vector (Fermentas). The cloned E rns genes of pJET-E rns and pJET-E rns D346 were verified by nucleotide sequencing and then excised with MIuI and Sall and ligated into the pWPT-GFP vector employing standard techniques and Stbl2 bacteria, resulting in pWPT-E rns and pWPT-E rns D346 respectively. All primer sequences and construction details can be obtained on request. In order to generate lentiviruses, HEK293T cells were transfected with the envelope plasmid (pMD2.G), the packaging plasmid (pCMV-R8.74) and the pWPT-E rns or pWPT-E rns D346 plasmid using standard calcium phosphate precipitation. Medium was changed after overnight incubation at 37uC and the supernatant harvested after 48 h, centrifuged (3506 g, 10 min) and filtered. The virus was purified and enriched by centrifugation on a 20% sucrose cushion at 100'0006 g for 90 min at 4uC. The cells were transduced twice with 1:100 dilutions of the purified lentiviruses in 1 ml serum free medium of a T25 cell culture flask followed by culture overnight at 37uC and medium change between the transductions. Transduction efficiency was found to be over 90% in terms of detectable anti-E rns expression by flow cytometry and found to remain stable over at least three passages.

For pDC phenotyping and isolation, monoclonal antibodies (mAb) against CD172a (mAb 74-22-15A), CD14 (CAM36A) and CD4 (mAb 74-12-4 and PT90A) were used. Hybridomas for mAb 74-22-15A and 74-12-4 were kindly donated by Armin Saalmüller (University of Veterinary Medicine, Vienna, Austria). The mAb PT90A and CAM36A were purchased from VMRD (Pullman, WA). The hybridomas for mAb HC/TC26 [60] and C16 [61] binding the CSFV glycoprotein E2 and nonstructural protein NS3 respectively were kindly provided by Irene Greiser-Wilke, Hannover Veterinary School, Hannover, Germany. E rns expression was demonstrated with mAb 140.1 (used at a 1:200 dilution, Prionics, Switzerland) as described previously [62] . For NS3 and E rns detection the cells were fixed and permeabilized with FIX&PERM solution (An der Grub Bio Research GmbH). As fluochromes, isotype-specific fluorescein isothiocyanate (FITC), Rphycoerythrin (RPE) (Southern Biotechnology Associates), RPE-Cy5 (Dako) and APC (Becton Dickinson, Basel , Switzerland) conjugates were used as described previously [63] .

For stimulation of enriched pDC by cells infected with CSFV and VRP, SK-6 cells were infected at an multiplicity of infection (MOI) of 5 TCID 50 /cell, cultured for 24 h and then washed four times to remove the inoculums. Infectivity was verified and found to be above 90% NS3 + cells following infection with CSFV or VRP at these conditions. The SK-6 cells were then added at 40'000 cells/well for 96 well plate (if not indicated otherwise) or at 200'000 cells/well for 24-well plates. Freshly isolated CD172a + enriched pDC were then added to the cultures at 200'000 cells/ well of 96-well plates or 1610 6 per well of 24-well plates. After another 22 h, supernatants were isolated for IFN-a ELISA and the cells for NS3 expression in some experiments. For stimulation of pDC by cells infected with FMDV or TGEV virus the cells were infected at the indicated MOIs, incubated for 90 min at 37uC and washed 4 times before addition of enriched pDC. For FMDV BHK21 cells were employed, for TGEV PK15 cells. In some experiments we employed 24-well plate transwell inserts with 0.4 mm or 1.0 mm pore sizes (Corning, Sigma Chemicals, Buchs Switzerland and Becton Dickinson, Basel Switzerland). All cultures with enriched pDC were done at 39uC, 6% CO 2 . As a positive control for pDC stimulation, direct CpG D32 at 10 mg/ml [63] was used.

A 50-mer RNA oligonucleotide probe complementary to nucleotides 12242-12193 of the vA187-1 genome sequence (GenBank accession number X87939.1) and carrying a Dyomics 781 modification at the 59 end (Dy-781-O1-RNA) was synthesized by Dr. Fabian Axthelm (Microsynth AG, Balgach, Switzerland). The Dy-781-O1-RNA probe was mixed at 40 nM final concentration with MEM containing 3610 23 U RNase A/ml as digestion control, with 50 mM TrisHCl pH 7.4 as negative control, and with the samples to be tested for RNase activity, and incubated for 1 h at 37uC. The treated probes were mixed with 2 volumes of 97% Formamide (Sigma) and separated on a 10% polyacrylamide and 35% urea gel in 133 mM TrisHCl, 45.5 mM boric acid and 3.2 mM EDTA. Image acquisition was performed with the Odyssey Infrared Imaging System (LI-COR).

As TLR7 inhibitor IRS661 (59-TGCTTGCAAGCTTG-CAAGCA-39) and a control oligonucleotide (Ctrl-ODN; 59-TCCTGCAGGTTAAGT-39) was used [23] . IRS661 and Ctrl-ODN were purchased from Eurofins MWG Operon (Ebersberg, Germany). The impact of various metabolic inhibitors on SK-6 cells was tested by addition of the inhibitor for the last 2 h of infection, in order to avoid interference with infection. Before addition of pDC, the inhibitors were removed and the cells washed three times. The following final concentrations were used: 1 mM latranculin B, 5 mM nocodazole or 20 mM methyl-b-cyclodextrin (MbCD). All chemicals were purchased from Sigma Chemicals.

IFN-a was quantified by enzyme-linked immunosorbent assay (ELISA) using the mAbs K9 and F17 (kindly provided by Dr. B. Charley, INRA, Jouy-en-Josas, France) as described previously [31] . For detection of IFN-a by intracellular staining mAb F17 with a previously published protocol was employed [63] .

Statistics P values were calculated by an unpaired t-test using the GraphPad Prism Software. Figure S1 pDC but not monocytes represent the source of IFN-a after stimulation by infected cells. pDC were purified by a combined depletion of CD14 + cells followed by CD172a enrichment. Monocytes were purified by CD14 cell sorting. Both pDC (10% purity) and monocytes (98% purity) were stimulated by MOCK or VRPDE rns -infected SK-6 cells for 20 h. Intracellular IFN-a staining was then performed by three-color flow cytometry and IFN-a in the supernatants was quantified by ELISA. The results are representative of two independent experiments. (TIF) Figure S2 Neutralizing antibodies prevent viral NS3 expression in pDC after stimulation by virions but not by infected cells. Left panels: Enriched pDC infected with VRPDE rns (MOI of 5 TCID50/cell) in absence (upper left panel) or presence (lower left panel) of neutralizing serum. Right panels: the pDC were co-cultured with MOCK-treated or VRPDE rns -infected SK-6 cells for 24 h, again in absence (upper panel) or presence neutralizing antibody. After 24 h, the cells were analyzed by threecolor FCM to determine the NS3 expression in pDC (defined as CD4 + CD172a low ). Blue histograms represent mock cultures, red histograms cultures with VRP. The percentage of NS3 + pDC is shown. (TIF) Figure S3 Viral protein expression in pDC after coculture with VRPDE rns -infected SK-6 or SK-6(E rns ) cells. Enriched pDC were co-cultured with MOCK-treated SK-6 cells, with CSFV-or VRPDE rns -infected SK-6 cells, or with CSFV-or VRPDE rns -infected SK-6(E rns ) cells for 20 h as indicated, and then analyzed by three-color FCM to determine the NS3 expression in pDC (defined as CD4 + CD172a low ). The percentage of NS3 + pDC is shown. Mean and standard deviation calculated from triplicate cultures are shown. The results are representative of three independent experiments. (TIF) 

Plasmacytoid dendritic cells: recent progress and open questions

IPC: professional type 1 interferon-producing cells and plasmacytoid dendritic cell precursors

Plasmacytoid dendritic cells sense hepatitis C virus-infected cells, produce interferon, and inhibit infection

CD81/CD9 tetraspanins aid plasmacytoid dendritic cells in recognition of HCV-infected cells and induction of IFNalpha

Innate sensing of HIV-infected cells

Pathogenesis and epidemiology of hog cholera

FcgammaRII-dependent sensitisation of natural interferon-producing cells for viral infection and interferon-alpha responses

Classical Swine Fever virus npro limits type I interferon induction in plasmacytoid dendritic cells by interacting with interferon regulatory factor 7

Viewpoint: Factors involved in type I interferon responses during porcine virus infections

Identification of a structural glycoprotein of an RNA virus as a ribonuclease

BVDV: A pestivirus inducing tolerance of the innate immune response

Recovery of virulent and RNase-negative attenuated type 2 bovine viral diarrhea viruses from infectious cDNA clones

Mutations abrogating the RNase activity in glycoprotein E(rns) of the pestivirus classical swine fever virus lead to virus attenuation

Bovine viral diarrhea virus: prevention of persistent fetal infection by a combination of two mutations affecting Erns RNase and Npro protease

Experimental transplacental transmission of hog cholera virus in pigs. IV. Virological and serological studies in newborn piglets

Role for bovine viral diarrhea virus Erns glycoprotein in the control of activation of beta interferon by double-stranded RNA

RNase-dependent inhibition of extracellular, but not intracellular, dsRNAinduced interferon synthesis by Erns of pestiviruses

The viral RNase E(rns) prevents IFN type-I triggering by pestiviral single-and double-stranded RNAs

N(pro) of Classical Swine Fever Virus Prevents Type I Interferon-Mediated Priming of Conventional Dendritic Cells for Enhanced Interferon-alpha Response

Classical swine fever virus interferes with cellular antiviral defense: evidence for a novel function of N(pro)

Interaction of classical swine fever virus with dendritic cells

Immunogenic and replicative properties of classical swine fever virus replicon particles modified to induce IFN-alpha/beta and carry foreign genes

Nucleic acids of mammalian origin can act as endogenous ligands for Toll-like receptors and may promote systemic lupus erythematosus

Interplay of foot-and-mouth disease virus, antibodies and plasmacytoid dendritic cells: virus opsonization under nonneutralizing conditions results in enhanced interferon-alpha responses

A new type of intracellular retention signal identified in a pestivirus structural glycoprotein

Processing of the envelope glycoproteins of pestiviruses

The carboxy-terminal sequence of the pestivirus glycoprotein E(rns) represents an unusual type of membrane anchor

Classical swine fever virus Npro interacts with interferon regulatory factor 3 and induces its proteasomal degradation

Cytochalasins block actin filament elongation by binding to high affinity sites associated with F-actin

Cytoplasmic dynein and dynactin are required for the transport of microtubules into the axon

Plasmacytoid dendritic cell activation by foot-and-mouth disease virus requires immune complexes

Crystal Structure of the Pestivirus Envelope Glycoprotein E(rns) and Mechanistic Analysis of Its Ribonuclease Activity

Classical swine fever virus glycoprotein E rns is an endoribonuclease with an unusual base specificity

RNase of classical swine fever virus: biochemical characterization and inhibition by virus-neutralizing monoclonal antibodies

Mutation of cysteine 171 of pestivirus E rns RNase prevents homodimer formation and leads to attenuation of classical swine fever virus

Autophagydependent viral recognition by plasmacytoid dendritic cells

Cytopathic bovine viral diarrhea viruses (BVDV): emerging pestiviruses doomed to extinction

A persistent hog cholera viremia in young pigs

THE ROLE OF IMMUNE TOLERANCE IN TRANS-MISSION OF HOG CHOLERA

Persistent hog cholera infection detected during virulence typing of 135 field isolates 580

A congenital persistent swine fever infection. II. Immune response to swine fever virus and unrelated antigens

Persistent infections of hog cholera: A review

Experimental transplacental transmission of hog cholera virus in pigs. I. Virological and serological studies

Atypical hog cholera infection: viral isolation and clinical study of in utero transmission

High IFN-alpha responses associated with depletion of lymphocytes and natural IFNproducing cells during classical swine fever

Pestiviruses: how to outmaneuver your hosts

Establishment, viral susceptibility and biological characteristics of a swine kidney cell line SK-6

Immortalized bone-marrow derived pig endothelial cells

Classical swine fever virus replicon particles lacking the E(rns) gene: a potential marker vaccine for intradermal application

Immunopathogenesis of classical swine fever: role of monocytic cells

Porcine dendritic cells generated in vitro: morphological, phenotypic and functional properties

Nucleotide sequence of classical swine fever virus strain Alfort/187 and transcription of infectious RNA from stably cloned full-length cDNA

Cytopathogenic and noncytopathogenic RNA replicons of classical swine fever virus

Oronasal vaccination with classical swine fever virus (CSFV) replicon particles with either partial or complete deletion of the E2 gene induces partial protection against lethal challenge with highly virulent CSFV

Induction of alpha interferon by transmissible gastroenteritis coronavirus: role of transmembrane glycoprotein E1 1234

Innate immune defenses induced by CpG do not promote vaccine-induced protection against foot-and-mouth disease virus in pigs

Validation of a real-time RT-PCR assay for sensitive and specific detection of classical swine fever

Multiply attenuated lentiviral vector achieves efficient gene delivery in vivo

Toll-like receptor 7 and MyD88 knockdown by lentivirus-mediated RNA interference to porcine dendritic cell subsets

Identification of conserved epitopes on a hog cholera virus protein

Heterogeneous expression of the non-structural protein p80/p125 in cells infected with different pestiviruses

Identification of classical swine fever virus protein E2 as a target for cytotoxic T cells by using mRNA-transfected antigen-presenting cells

Type-A CpG oligonucleotides activate exclusively porcine natural interferon-producing cells to secrete interferon-alpha, tumour necrosis factor-alpha and interleukin-12

We thank Dr. Fabian Axthelm (Microsynth AG, Balgach, Switzerland) for custom synthesis of the 59-Dyomics-781-modified RNA oligonucleotide, Nathalie Vielle for help with RNA extraction and Andreas Michel (IVI) for animal care and blood sampling.