Activation of the mammalian immune system by siRNAs

NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 11 NOVEMBER 20051399Activation of the mammalian immune system by siRNAsJoao T Marques & Bryan R G WilliamsInhibition of gene expression through RNA interference (RNAi) is emerging as a powerful experimental tool for gene function and target validation studies. The potential uses of this technology seem unlimited, extending to the prevention and therapy of human diseases. However, recent work demonstrating that there are unanticipated, different nonspecific effects associated with the use of small interfering RNAs in mammals has raised concerns about the safe use of RNAi in vivo . These nonspecific effects include activation of the immune system, potentially harming the individual. The application of screening assays for nonspecific activation of both innate and acquired immunity will be necessary for further development of RNAi as a therapeutic tool.RNA interference (RNAi) was first described in Caenorhabditis ele-gans 1 in 1998, where long double-stranded RNA (dsRNA) injected into adult animals silenced the expression of genes with sequence homol-ogy to the dsRNA 2. Subsequently, similar phenomena were described in Drosophila melanogaster plants and undifferentiated mammalian cells 2–4. When dsRNA is introduced into the cytoplasm, it is processed by an RNase III enzyme, Dicer, into 21-nucleotide (nt) dsRNAs with 2-nt 3′ overhangs, termed small interfering RNAs (siRNAs) (see Fig. 1 and Box 1 for glossary)5,6. SiRNAs are incorporated into the RNA-induced silencing complex (RISC), which mediates the degradation of complementary mRNAs, silencing gene expression 7. Initially, it was believed that nonspecific activation of the antiviral response by dsRNA longer than 30 nt would preclude the use of RNAi as an experimental or therapeutic tool in mammals 3,8,9. However, in 2001, Tuschl and col-leagues 8 and Caplen et al.9 described the delivery to the cytoplasm of mammalian cells of siRNAs that efficiently mediated gene silencing. Small interfering RNAs could be easily chemically synthesized allowing exploration of their use to regulate gene expression. Importantly, this was different from previous strategies such as antisense and ribozymes, in that RNAi exploited an endogenous mechanism 10. However, dur-ing attempts to develop techniques to efficiently deliver siRNAs and trigger RNAi in vivo, it was noted that siRNAs could activate cells of the immune system and induce the production of cytokines both in vivo and in vitro 11,12. Together with earlier reports of activation of the interferon (IFN) system in cell culture by siRNA 13–16, this has raised questions about the specificity of RNAi. This review describes what is known about nonspecific activation of the immune system by siRNAs and suggests future directions that may allow use of RNAi in vivo by bypassing or controlling these unwanted effects.Activation of the immune system by siRNAsThe pathways responsible for the recognition of siRNAs by immune cells were initially not clear 11,12. Mammalian immune cells express a family of toll-like receptors (TLRs), which recognize pathogen-associated molec-ular patterns including unmethylated CpG DNA and viral dsRNA 17. TLR3, the receptor for dsRNA, was a logical candidate for the recogni-tion of the siRNAs in the context of immunostimulation. Indeed, TLR3 overexpressed in cultured human embryonic kidney (HEK) 293 cells was capable of recognizing siRNAs 12. However, the activation of immune cells by siRNAs is sequence dependent and sense or antisense strands separately can induce cytokine production as efficiently as duplex siR-NAs 18–20. Therefore, as TLR3 neither recognizes single-stranded RNA (ssRNA) nor requires sequence specificity 17, it is unlikely involved in activation of the immune system by siRNAs. It has now become clear that TLR7 and TLR8 mediate the recognition of siRNAs in a sequence-dependent manner; using the appropriate knockout mice, it has been shown that TLR7 is absolutely required for induction of cytokines in murine immune cells in response to siRNAs 18–20.TLR7 and TLR8 were initially shown to mediate the recognition of RNA viruses 21,22 and small synthetic antiviral compounds referred to as imidazoquinolines, such as imiquimod and its derivative resiquimod (R-848)23. Interestingly, not all ssRNA molecules are capable of activating TLR7 and TLR8 but rather U- and G-rich ssRNAs seem to be preferen-tially recognized 21,22. Despite these prior observations, recent reports of the activation of the immune system by siRNAs in a sequence-dependent manner have surprised the field 18–20.Studies of HEK 293 cells overexpressing TLRs have indicated that ssRNAs are recognized by mouse TLR7 and human TLR8 but not by mouse TLR8 or human TLR7 (ref. 21). However, siRNAs can be recog-nized by human plasmacytoid dendritic cells (pDCs) through TLR7 (refs. 18,19,24) and human monocytes likely via TLR8 (refs. 24,25). The vari-ability in the immunostimulatory activity of different siRNA sequences observed in immune cells from human and mice may result from siRNA recognition by TLR7 or TLR8 homo or heterodimers as determined byDepartment of Cancer Biology, Lerner Research Institute, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, Ohio, 44195, USA. Correspondence should be addressed to B.R.G.W. (williab@)Published online 4 November 2005; doi:10.1038/nbt1161R E V I E W©2005 N a t u r e P u b l i s h i n g G r o u p h t t p ://w w w .n a t u r e .c o m /n a t u r e b i o t e c h n o l o g ytheir expression patterns in different cell types. It is also possible that siRNA recognition by TLR7 and/or TLR8 requires accessory proteins. Although such cofactors have not been identified for TLR7 or 8, this pos-sibility is supported by the finding that autologous plasma enhances the induction of cytokines by siRNA in peripheral blood mononuclear cells (PBMC 19) (Fig. 1). Furthermore, precedent for a cofactor requirement exists since recognition of bacterial lipopolysaccharide by TLR4 requires the CD14 and MD-2 proteins 17. There has also been speculation about possible coreceptors for TLR9, which recognize unmethylated CpG DNA (Box 2) and could modulate different responses 26.TLR7, TLR8 and TLR9 all recognize nucleic acids, are expressed in T o further complicate the issue of TLR7, TLR8 and TLR9 expression in immune cell subsets, the data differ between mice and humans 24,25,28. Therefore, it remains unclear which TLRs are expressed in each cell type of the immune system in humans and mice.All vertebrate species analyzed possess a single copy of the TLR7 and TLR8 genes present in tandem on the same chromosome 36. This indi-cates that TLR7 and TLR8 likely originated through duplication of an ancestral gene before the evolution of vertebrates and that there was selective pressure for the maintenance of both receptors 36.Therefore, despite the fact that TLR7 and TLR8 are closely related, it is likely that they fulfill nonoverlapping functions. Indeed, synthetic agonists that R E V I E Wn o l o g yNATURE BIOTECHNOLOGY VOLUME 23 NUMBER 11 NOVEMBER 2005 1401humans that complicate functional analy-sis of siRNA signaling using mouse models. Furthermore, a recent report that a poly(G) oli-godeoxynucleotide was stimulatory to human CD4+ T cells in a TLR8-dependent, TLR7- and TLR9-independent manner further illustrates the functional differences between TLR7 and TLR8 in humans 38.Endosomal recognition of siRNAsAs mentioned above, TLR7, TLR8 and TLR9 are expressed in endosomes and require endo-somal maturation for efficient signaling 27. Given the similarities among TLR7, TLR8 andTLR9, it is possible to apply the large body of knowledge about signaling by CpG DNA through TLR9 (refs. 26,27,39,40) (Box 2) and speculate about TLR7- and TLR8-dependent signaling. Endosomal recognition by TLR9 is not bypassed by combining the nucleic acid with cationic lipids or polymers and, in fact, this has been shown to extend the time that CpG oligodeoxynucleotides (ODNs) migrate through the endo-somes 41 (Box 2). Similarly, several studies have demonstrated that the method of delivery is critical for the immunostimulatory activity of siRNAs 18–20. Injection of siRNAs in complex with cationic lipids into mice induces a potent cytokine response that is debilitating to the organ-ism 18,19. On the other hand, injection of naked, unmodified siRNAs or siRNAs conjugated to cholesterol has no significant effect on immune system activation 42,43.There are two possible explanations for the absence of immunos-timulation by naked siRNAs. First, unmodified siRNAs have a short half-life in serum and may be degraded before being recognized by specific receptors. Second, siRNA in complex with lipids, perhaps by resembling a viral particle, might be more readily recognized by immune cells than would naked siRNA (Figs. 1 and 2). In favor of the second option, Sioud 20 reported that naked nuclease-resistant siRNAs contain-ing a phosphothioate backbone were not immunostimulatory. However, these results should be interpreted carefully considering that naked CpG ODNs can be immunostimulatory independent of backbone modifica-tions 26 (Box 2).Recognition of siRNAs by TLRs takes place in the endosome, before the siRNAs enter the cytoplasm (Fig. 1). Therefore, it would be expected that if siRNAs could enter the cytoplasm avoiding the endosome, they should bypass the activation of immune cells but still mediate gene silencing. Although electroporation should provide direct cytoplasmic delivery of siRNAs without their migrating through the endosome, con-flicting results have been obtained using this technique. Sioud 20 did not observe cytokine production after siRNA delivery through electropora-tion of monocytes, but Hornung et al.18 showed that a naked siRNA still induced IFN-α when pDCs were electroporated. Unfortunately, these studies did not assess the subcellular localization of the siRNAs or their silencing efficiency in either monocytes or pDCs. The discrep-ancies between Sioud 20 and Hornung et al .18 could reflect differential trafficking of the siRNAs in the two cell types. For example, the IFN induction observed by Hornung et al.18 after siRNAs entered the cell by electroporation might be due to trafficking of small amounts of siRNA through the endosome in pDCs.Notably, when siRNAs are coated with poly-L -lysine, significantly more IL-6 and tumor necrosis factor are induced in PBMCs than when they are in complex with cationic lipids or polyethylenimine 19. In con-trast, more IFN-α is induced by siRNA in complex with cationic lipids or polyethylenimine compared with the same sequence coated with poly-L -lysine 19. It is possible that different siRNA complexes could either modify the fate of the siRNA in a given cell type or change which cell types are preferentially stimulated. In accordance with the latter hypothesis, TNF induction by siRNAs is abrogated in PBMCs depleted of monocytes, whereas IFN-α induction is abrogated only in pDC-depleted PBMCs 19. Of note, there is a difference in the cytokine profile induced in myeloid dendritic cells (mDCs) and pDCs by the TLR7 and TLR8 ligand, R-848, with IL-12 being induced only in mDCs and IFN-α being induced only in pDCs 28. Similar differential responsiveness to different sequences or complexes also occurs with the immune responses to CpG ODNs medi-ated by TLR9 (ref. 26) (Box 2).The importance of sequence, size and structureA complete understanding of the sequence requirements for siRNA-mediated immune activation and gene silencing is necessary before attempts can be made to engineer siRNAs that shut down gene expres-sion without inducing deleterious side effects. U- and G-rich regions seem to be a common feature among the immunostimulatory motifs;R E V I E W©2005 N a t u r e P u b l i s h i n g G r o u p h t t p ://w w w .n a t u r e .c o m /n a t u r e b i o t e c h n o l o g y1402 VOLUME 23 NUMBER 11 NOVEMBER 2005 NATURE BIOTECHNOLOGYGUCCUUCAA and UGUGU are present in the immunostimulatory ssRNAs 18,19 (Table 1). However, as mentioned above, several immuno-stimulatory siRNAs do not contain either of these motifs, whereas oth-ers are not stimulatory despite being U- and G-rich 20. Therefore, other characteristics, such as the position or proportion of UG dinucleotides within the molecule or the composition of flanking sequences, may play a role in determining the nature and potency of the cytokine response induced by ssRNA.In addition to sequence composition, the size of ssRNAs is also important for activation of the immune system. Hornung et al .18 observed that 12-nt ssRNAs containing the immunostimulatory motif (GUCCUUCAA) were poor inducers of IFN-α in pDCs but that increas-ing their size to 16 nt or 19 nt completely restored cytokine induction. It remains unclear if the recognition of immunostimulatory motifs in the siRNA can occur in the context of the duplex. Sioud 20 has observed that the single strands separately tend to be stronger stimuli than the duplex, indicating that the presence of single strands is critical. An excess of one of the strands during annealing can lead to the presence of single strands in a siRNA preparation, but in most cases denaturation of the duplex in the endosome will release the single strands. This was likely the case where a short hairpin RNA containing the immunostimulatory motif was still capable of activating the immune response 18.There might be qualitative differences among siRNA sequences that are recognized by TLR7 and TLR8 as has been demonstrated for CpG ODNs. The latter can be divided into groups based upon sequences flanking the CpG motif that determine the cytokine profile they induce 26 (Box 1). In humans, substitution of U with A in the siRNA immunos-timulatory sequence abrogates TNF and IL-6 induction in PBMCs, but the substitution of G with A abrogates only induction of IFN-α in pDCs without affecting the induction of TNF, IL-6 and IL12 in PBMCs 21. Moreover, siRNAs with the same UGUGU motif embedded in differ-ent sequence contexts induce qualitatively different cytokine profiles in human PBMCs 19. This suggests that there are important qualitative differences in responses to different ssRNAs, especially in humans, where both TLR7 and TLR8 are functional. It remains to be determined if het-erodimerization between TLR7 and/or TLR8 and possibly other acces-sory proteins determines sensitivity to different immunostimulatorysequences. Interestingly, a ssRNA containing a CpG motif and a 6 nt poly-(G) run at the 3′ end has been shown to activate monocytes, but the receptors responsible for recognition of this molecule remain unknown(Table 1). When TLR3, TLR7, TLR8 and TLR9 are overexpressed sepa-rately in HEK 293 cells, they are not able to provide responsiveness to this sequence 44. It is important to note that for recognition of CpG ODNs by TLR9, the sequences flanking the CpG motif that induce the strongest immunostimulatory activity vary between mice and humans 26. Therefore, human and mouse TLR7s likely have different preferences in terms of immunostimulatory sequences. Purified recombinant TLR9 binds CpG ODNs directly in a sequence- and pH-dependent manner, implying that all of the in vivo requirements for recognition of unmeth-ylated CpG DNA are provided by the receptor itself 40. The crystal struc-ture of TLR3 also suggests a mechanism for direct binding of dsRNA 45. Despite these precedents, it is not yet known whether TLR7 and TLR8 can bind siRNAs directly.Because the sequence and chemical nature of siRNAs can determine not only their gene silencing potential, but also their ability to induce immune responses, interest has been intense in identifying chemical modifications that might preserve silencing while bypassing immune system activation 18,46. Locked nucleic acid (LNA) modifications of the 3′ or 5′ termini of the sense strand of siRNAs can abrogate their immunostimulatory activity 18. LNA nucleotides contain a methylene linkage between the 2´ oxygen and the 4´ carbon of the ribose ring in the ribonucleotides 47,48. However, depending on the extent and loca-tion of the LNA modifications, they can block not only the inductionR E V I E W©2005 N a t u r e P u b l i s h i n g G r o u p h t t p ://w w w .n a t u r e .c o m /n a t u r e b i o t e c h n o l o g yNATURE BIOTECHNOLOGY VOLUME 23 NUMBER 11 NOVEMBER 2005 1403of IFN but also the gene silencing activity of the siRNA 18. Different modifications of the siRNA backbone have also been tested and several identified that do not affect silencing activity, at least in cell culture systems 49. The inclusion of DNA bases in siRNAs is well tolerated and may even increase silencing activity 50. In vivo , these DNA-RNA hybrids could potentially activate not only TLR7 and TLR8, but also TLR9 if they contain CpG motifs 51.In contrast to unmodified molecules, siRNAs containing such back-bone modifications, such as 2´ F, 2´ O-methyl and 2´ H, do not induce cytokines when injected into mice while maintaining silencing activ-ity 46. More studies are necessary to determine the optimal modifications that abrogate immunostimulatory activity without loss of silencing. The effects of these modifications, however, have to be determined empiri-cally on a case-by-case basis because their immunostimulatory proper-ties are not completely understood and careful studies will be required before modified siRNA can be safely used in therapeutic applications.acids are known 53cytosine is usually methylated bacterial DNA has a much higher frequency of CpG dinucleotides and they are not methyl-ated 54for the lack of immunostimulatory activity of mRNAs 56. However, an in vitro –transcribed mRNA encoding green fluo-rescent protein (GFP) has been shown to trigger production of IFN-α by pDCs, regardless of the presence of either a short poly(A) tail or the 7-methyl-guanosine cap characteristic of eukaryotic mRNAs 22. These data do not rule out the possibility that a long poly(A) tail might act as a decoy preventing recognition of immunostimulatory motifs. Notably, human mRNAs, when in complex with protamine 57 or a cationic lipid 58, are also immunostimulatory to dendritic cells. Therefore, as mentioned above for CpG DNA, multiple levels of control are likely necessary to modulate recognition of ssRNAs. Compartmentalization of cellular RNAs may account at least in part for their failure to be recognized as immunostimulatory under normal conditions 21. Coated siRNA may resemble a viral particle and recognition of ssRNAs in such a context may be required for activation of the immune system, in addition to the presence of the immunostimulatory motifs. The notion that, as for CpG ODNs, particular sequence motifs in ssRNAs influence immunostimula-tion 18,19,21 is intriguing and indicates that to avoid immune recognition protein; OAS, 2´5´-oligoadenylate synthetase.R E V I E W©2005 N a t u r e P u b l i s h i n g G r o u p h t t p ://w w w .n a t u r e .c o m /n a t u r e b i o t e c h n o l o g y1404 VOLUME 23 NUMBER 11 NOVEMBER 2005 NATURE BIOTECHNOLOGYSequence-independent recognition of siRNAsNonimmune cells and immune cells can also recognize siRNAs indepen-dent of the sequence leading to the induction of type I IFNs 13–16. The signaling pathways that mediate this recognition are complex and involve a number of dsRNA-sensing receptors, both intracellular and membrane bound (Figs. 1 and 2). The intracellular pathways include the dsRNA-activated protein kinase (PKR) and the 2´5´-oligoadenylate synthetases that can differentially induce general protein synthesis inhibition 61, and signaling pathways that activate transcription factors including NF-κB and IFN regulatory factor (IRF)-3, leading to induction of genes, such as those encoding the IFNs 62 (Fig. 2). TLRs are primarily expressed in cells of the immune system, but can also be present in other cell types, includ-ing endothelial cells, fibroblasts and hepatocytes 30,35,63–65. In cultured cells, PKR and TLR3 have been shown to mediate the induction of IFN by siRNAs 12,14 (Figs. 1 and 2). Recognition of siRNAs by these pathways could potentially synergize with TLR7 and TLR8 in the activation of the immune system. It is known, for example, that dsRNA and CpG ODN can synergize in the activation of innate immune responses leading to enhanced tumor clearance 66. In addition, viruses can induce high levels of IFN in pDCs in the absence of both TLR7 and PKR signaling, presum-ably through production of intracellular dsRNA that is recognized by other RNA sensors 67. Such molecules might include caspase recruitment domain (CARD)-containing helicases, such as retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated gene-5 (MDA-5) proteins 68–70. Although their activation is capable of triggering IFN production, these helicases have not been implicated in recognition of siRNAs. However, experiments with TLR7-knockout mice indicate that TLR3, PKR and CARD-containing helicases all have at least minor roles in siRNA recognition in vivo 18. Additional experiments in TLR3- and PKR-knockout mice are needed to define the role of these proteins in response to siRNAs in vivo . It is also important to keep in mind that TLR8 is functional in humans but not in mice and until the differences between TLR7- and TLR8-mediated recognition of siRNAs are eluci-dated, studies in mice must be interpreted with caution.Implications for siRNA-mediated therapyThe immunostimulatory ‘side effects’ of siRNAs must be taken into account when considering possible applications of siRNAs for in vivo therapy. The potentially deleterious results of these side effects are illustrated by experiments in which mice injected with immunostim-ulatory siRNAs present signs of toxicity, including elevated levels of serum alanine and aspartate aminotransferases and reduced numbers of lymphocytes and platelets 19,52. As detailed above, there may be ways to experimentally bypass siRNA-induced side effects on the immune system while preserving gene-silencing activity. These might include modification of the siRNA backbone 18,52 or avoidance of immunos-timulatory sequence motifs 19. Furthermore, naked siRNAs conjugated to cholesterol or in complex with atelocollagen have been successfully used to promote gene silencing in vivo 42,71 and as discussed previously, the delivery of naked siRNAs might bypass activation of the immune system 43. Notably, endogenous dsRNAs that work through the RNAi pathway, called micro RNAs 72, likely present base modifications (e.g., methylation) or secondary structures that preclude the activation of the immune system, which could be incorporated into synthetic siRNAs to avoid nonspecific effects.M oreover, in some scenarios, activation of the immune system together with siRNA-mediated gene silencing might actually be desired, leading to enhanced therapeutic effects. Therefore, in the treatment of viral infections and tumors, siRNAs containing potent immunostimu-latory motifs could prove useful. However, it will be important to first understand whether there are distinct immunostimulatory sequencesthat determine qualitative differences in cytokine induction as well as the effect of different delivery methods on immune responses to siRNAs.Although the use of siRNA-mediated therapy in vivo is more complex than originally expected because of nonspecific effects on the immune system, there remains great potential for RNAi as a tool for in vitro experimentation as well as in vivo disease treatment.ACKNOWLEDGMENTSWe would like to thank Patricia Stanhope-Baker, Anthony Sadler, Mark Whitmore and Michelle Holko for helpful discussion and suggestions. Work in the Williams laboratory is supported by National Institutes of Health grants RO1 AI34039 and PO1 CA 62220.COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.Published online at /naturebiotechnology/Reprints and permissions information is available online at /reprintsandpermissions/1. Fire, A. et al. Potent and specific genetic interference by double-stranded RNA inCaenorhabditis elegans. Nature 391, 806–811 (1998).2. Mello, C.C. & Conte, D. Jr. Revealing the world of RNA interference. Nature 431,338–342 (2004).3. Yang, S., Tutton, S., Pierce, E. & Yoon, K. Specific double-stranded RNA interferencein undifferentiated mouse embryonic stem cells. Mol. Cell. Biol. 21, 7807–7816 (2001).4. Baulcombe, D. RNA silencing in plants. Nature 431, 356–363 (2004).5. Meister, G. & Tuschl, T. Mechanisms of gene silencing by double-stranded RNA. Nature431, 343–349 (2004).6. Elbashir, S.M., Lendeckel, W. & Tuschl, T. RNA interference is mediated by 21- and22-nucleotide RNAs. Genes Dev. 15, 188–200 (2001).7. Tomari, Y. & Zamore, P .D. Perspective: machines for RNAi. Genes Dev. 19, 517–529(2005).8. Elbashir, S.M. et al. Duplexes of 21-nucleotide RNAs mediate RNA interference incultured mammalian cells. Nature 411, 494–498 (2001).9. Caplen, N.J., Parrish, S., Imani, F ., Fire, A. & Morgan, R.A. Specific inhibition of geneexpression by small double-stranded RNAs in invertebrate and vertebrate systems. Proc. Natl. Acad. Sci. USA 98, 9742–9747 (2001).10. Robinson, R. RNAi therapeutics: how likely, how soon? PLoS Biol. 2, E28 (2004).11. Sioud, M. & Sorensen, D.R. Cationic liposome-mediated delivery of siRNAs in adultmice. Biochem. Biophys. Res. Commun. 312, 1220–1225 (2003).12. Kariko, K., Bhuyan, P ., Capodici, J. & Weissman, D. Small interfering RNAs mediatesequence-independent gene suppression and induce immune activation by signaling through toll-like receptor 3. J. Immunol. 172, 6545–6549 (2004).13. Sledz, C.A. & Williams, B.R. RNA interference in biology and disease. Blood 106,787–794 (2005).14. Sledz, C.A., Holko, M., de Veer, M.J., Silverman, R.H. & Williams, B.R. Activationof the interferon system by short-interfering RNAs. Nat. Cell Bio l. 5, 834–839 (2003).15. Kim, D.H. et al. Interferon induction by siRNAs and ssRNAs synthesized by phagepolymerase. Nat. Biotechnol. 22, 321–325 (2004).16. Persengiev, S.P., Zhu, X. & Green, M.R. Nonspecific, concentration-dependentstimulation and repression of mammalian gene expression by small interfering RNAs (siRNAs). RNA 10, 12–18 (2004).17. Takeda, K. & Akira, S. Toll-like receptors in innate immunity. Int. Immunol. 17, 1–14(2005).18. Hornung, V. et al. Sequence-specific potent induction of IFN-alpha by short interferingRNA in plasmacytoid dendritic cells through TLR7. Nat. Med. 11, 263–270 (2005).19. Judge, A.D. et al. Sequence-dependent stimulation of the mammalian innate immuneresponse by synthetic siRNA. Nat. Biotechnol. 23, 457–462 (2005).20. Sioud, M. Induction of inflammatory cytokines and interferon responses by double-stranded and single-stranded siRNAs is sequence-dependent and requires endosomal localization. J. Mol. Biol. 348, 1079–1090 (2005).21. Heil, F . et al. Species-specific recognition of single-stranded RNA via toll-like receptor7 and 8. Science 303, 1526–1529 (2004).22. Diebold, S.S., Kaisho, T., Hemmi, H., Akira, S. & Reis e Sousa, C. Innate antiviralresponses by means of TLR7-mediated recognition of single-stranded RNA. Science 303, 1529–1531 (2004).23. Hemmi, H. et al. Small anti-viral compounds activate immune cells via the TLR7MyD88-dependent signaling pathway. Nat. Immunol. 3, 196–200 (2002).24. Iwasaki, A. & Medzhitov, R. Toll-like receptor control of the adaptive immuneresponses. Nat. Immunol. 5, 987–995 (2004).25. Liu, Y.J. IPC: professional type 1 interferon-producing cells and plasmacytoid dendriticcell precursors. Annu. Rev. Immunol. 23, 275–306 (2005).26. Verthelyi, D. & Zeuner, R.A. Differential signaling by CpG DNA in DCs and B cells: notjust TLR9. Trends Immunol. 24, 519–522 (2003).27. Heil, F . et al. The Toll-like receptor 7 (TLR7)-specific stimulus loxoribine uncoversa strong relationship within the TLR7, 8 and 9 subfamily. Eur. J. Immuno l. 33, 2987–2997 (2003).R E V I E W©2005 N a t u r e P u b l i s h i n g G r o u p h t t p ://w w w .n a t u r e .c o m /n a t u r e b i o t e c h n o l o g y。