Features of RNA from virus-infected cells which are important for induction of

Contribution of the terminal 5`-triphosphate of RNA for RIG-I signaling

Previous studies have indicated that the 5`-triphosphate is a structural feature responsible for IFN-α-inducing activity of in vitro transcribed RNA and flu vRNA in monocytes, bone marrow-derived DCs and HEK293 cells transfected with IFN-ß reporter.

Dephosphorylation of the 5`-triphosphate end completely abrogated the IFN-α response in the aforementioned cells. However, pDCs showed no decrease in IFN-α secretion upon dephosphorylation of oligonucleotides. Triphosphate RNA-mediated IFN-α induction neither required endosomal maturation nor TLR7 (Hornung et al. 2006; Pichlmair et al.

2006).

Our findings showed that the 5`-triphosphate end of A/PR/8/MDCK-RNA complexed to DOTAP or Lipofectamine 2000 was important for inducing IFN-α in murine immune cells.

Dephosphorylation of the A/PR/8/MDCK-RNA reduced signaling, suggesting that RIG-I senses the A/PR/8/MDCK-RNA in a 5`-triphosphate dependent manner in Flt3-derived dendritic cells. For the RIG-I ligand 5`-3P RNA complexed to Lipofectamine 2000, we observed the same behaviour as for A/PR/8/MDCK-RNA; the IFN-α induction was 5`-triphosphate dependent.

However, for human PBMC cells we observed only a minor decrease in the IFN-α response upon dephosphorylation of A/PR/8/MDCK-RNA complexed to DOTAP or Lipofectamine 2000. In contrast, for the RIG-I ligand 5`-3P RNA complexed to Lipofectamine 2000, the IFN-α response was abrogated upon dephosphorylation. We hypothesize that human immune cells express more RIG-I proteins than murine immune cells, so that in the human system a higher number of putative binding sites between RIG-I and the A/PR/8/MDCK-RNA are available, and the 5`-triphosphate end is no longer important for interaction. For the RIG-I ligand 5`-3P RNA, the nucleic acid length from 100 bp is not sufficient in order to have interactions with RIG-I proteins in a 5`-triphosphate independent way. Another possibility is that the murine RIG-I molecule and the human RIG-I molecule are different in the dependency of the 5`-triphosphate end. But the results obtained with the RIG-I ligand 5`-3P RNA complexed to Lipfectamine 2000 argue against the idea that the murine and the human RIG-I molecules behave differently as regards the 5`-triphosphate ends.

In untransfected HEK293 cells, the immune response to A/PR/8/MDCK-RNA complexed to DOTAP or Lipofectamine 2000 was abrogated after dephosphorylation, whereas in RIG-I-transfected HEK293 cells dephosphorylation of A/PR/8/MDCK-RNA had no influence. The RIG-I ligand 5`-3P RNA complexed to Lipofectamine 2000 induced IFN-ß activation only in RIG-I-transfected HEK293 cells, and the IFN-ß response was diminished after dephosphorylation. It seems that the A/PR/8-infected MDCK-RNA shows more putative binding sites even for only lowly expressed RIG-I proteins in untransfected HEK293 cells than the RIG-I ligand 5`-3P RNA. As discussed for PBMCs, we hypothesize that immune cells expressing more RIG-I proteins contain a higher number of putative

binding sites for the ligands and that, in this case, the 5`-triphosphate end is not important for recognition of the RNA through RIG-I. But further experiments are required, like radioactive labeling of the RNAs to prove that the phosphate group was efficiently removed.

Influence of the ds character of RNA for RIG-I signaling

The "panhandle" conformation of the A/PR/8/MDCK-RNA represents a second independent target structure for RIG-I signaling. Schmidt et al. have found that RIG-I ligands require base-paired structures in conjugation with a free 5`-triphosphate to trigger a type-I interferon response (Schmidt et al. 2009). We could show that immunostimulatory RNA from virus-infected cells was sensitive to but not ss-specific RNases. The ds-specific RNase III abolished signaling of A/PR/8/MDCK-RNA complexed to DOTAP or Lipofectamine 2000 in human and murine immune cells, which indicates that the A/PR/8/MDCK-RNA contain base-paired regions.

In the past, most studies suggested that single-stranded 5`-triphosphate RNA is sufficient to bind to and activate RIG-I (Pichlmair et al. 2006). These studies used in vitro transcription for generation of 5`-triphosphate RNA without analyzing the purity of those RNA molecules. In fact, an unintended formation of dsRNA is the cause of RIG-I activity of the in vitro transcribed RNA (Schlee et al. 2009; Schmidt et al. 2009). Even for negative-strand RNA viruses known to activate RIG-I, it was shown by Schlee et al. that they contain 5` and 3` sequences that form a short ds with a perfect blunt end (panhandle) (Schlee et al. 2009). Interestingly, Myong et al. showed that RIG-I translocates on synthetic double-stranded RNA molecules and that this movement is enhanced in the presence of 5`-triphosphate (Myong et al. 2009). Some studies suggested that for ssRNA-viruses no dsRNA can be detected (Hornung et al. 2006; Pichlmair et al. 2006; Weber et al. 2006). However, the antibody used to demonstrate the absence of dsRNA (Weber et al. 2006) is limited to the detection of dsRNA longer than 40 bases (Bonin et al. 2000;

Schlee et al. 2009). In contrast, other studies demonstrated that the most widely postulated triggers for cytokine induction on influenza virus infection are dsRNA intermediates produced during replication (Jacobs and Langland 1996; Majde 2000). Long dsRNA molecules are absent from uninfected cells. Madje et al. showed that dsRNA is released from dying influenza virus-infected cells (Majde et al. 1998).

Comparing the ligands A/PR/8/MDCK-RNA and the RIG-I ligand 5`-3P RNA in human

following: For A/PR/8/MDCK-RNA complexed to DOTAP or Lipofectamine 2000, phosphatase treatment had no influence on the IFN-α-inducing ability, but cleavage with the ds-specific RNase III abrogated IFN-α secretion in human PBMC cells. This shows that, for the recognition of RNA from virus-infected cells, the ds character or at least a

"panhandle" conformation by pairing complementary 5` and 3` ends is important. In contrast, for the RIG-I ligand 5`-3P RNA complexed to DOTAP or Lipofectamine 2000, only the 5`-triphosphate seemd to be important, whereas treatment with the ds-specific RNase III had no influence for IFN-α secretion in human PBMC cells. A possible explanation is that for longer ligands the ds character is sufficient to induce IFN-α, and even after phosphatase treatment they remain immunostimulatory. In contrast, shorter ligands like the RIG-I ligand 5`-3P RNA need a 5`-triphosphate end for IFN-α secretion. At the moment, it is not clear whether dsRNA or hairpin dsRNA regions of the virus-infected RNA are responsible for inducing IFN-α. It is also possible that the RIG-I ligand 5`-3P RNA shows hairpin dsRNA regions, but that for inducing IFN-α a 5`-triphosphate end is necessary. Thus, we conclude that for the ligands A/PR/8/MDCK-RNA and RIG-I ligand 5`-3P RNA two important features for RIG-I signaling are necessary: the ds character and the 5`-triphosphate end. On the other hand, some studies indicate that RNase III cuts ssRNA specifically and seems to degrade dsRNA rather nonspecifically (Dunn 1976). In contrast, Sun et al. demonstrated that RNase III retains strict specificity for dsRNA (Sun et al. 2001). At the moment, it is not clear whether RNase III is a ds-specific RNase. Further analysis is required in order to prove that the ds character of the A/PR/8/MDCK-RNA is responsible for inducing IFN-α.