Single-stranded small interfering RNA are more immunostimulatory than their double-stranded counterparts: A central role for 2′-hydroxyl uridines in immune responses


  • Mouldy Sioud Dr.

    Corresponding author
    1. The Norwegian Radium Hospital, Department of Immunology, Molecular Medicine Group Montebello, Oslo, Norway
    • The Norwegian Radium Hospital, Department of Immunology, Molecular Medicine Group Montebello, N-0310 Oslo, Norway, Fax: +47-22-500-730
    Search for more papers by this author


It has recently become apparent that certain small interfering RNA (siRNA) sequences stimulate the innate immunity through endosomal Toll-like receptors (TLR), particularly TLR7 and TLR8. However, it remains unclear whether siRNA duplexes act as specific ligands for these receptors. To address this question and to overcome the problem of immune activation by siRNA, several RNA sequences were chemically synthesized and their effects were investigated. Results indicate that human peripheral blood mononuclear cells (PBMC) recognize and respond to a large number of sense or antisense single-stranded (ss) siRNA. In most cases immunostimulatory RNA motifs are more effectively recognized by innate immunity in the context of ss siRNA as compared to siRNA duplexes. Novel immunostimulatory RNA motifs were identified and their replacement with adenosines abrogated immune activation. Most notably, replacement of the 2′-hydroxyl uridines with either 2′-fluoro, 2′-deoxy or 2′-O-methyl uridines abrogated immune activation. Thus, immune recognition of RNA by TLR can be evaded by 2′-ribose modifications of only uridines. Collectively, the data should facilitate the development of siRNA therapeutics and expand the understanding of how RNA is sensed by innate immunity.


N-[1–2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate






plasmacytoid dendritic cells


protein kinase R


RNA interference


short hairpin RNA


small interfering RNA




RNA interference (RNAi) is an evolutionarily conserved of post-transcriptional gene-silencing process, resulting in sequence-specific mRNA degradation in several organisms 1. In this process, double-stranded (ds) RNA is cleaved by Dicer into 21–25-nucleotide (nt)-small interfering RNA (siRNA) with 2-nt overhangs at the 3′ ends. Processed siRNA are incorporated into a protein complex, termed RNA-induced silencing complex (RISC) 2. Subsequently, the two strands of siRNA are separated by a helicase, and the retained RNA strand guides the activated RISC to identify complementary mRNA sequences. Identified mRNA are targeted for cleavage by an Argonaute protein, which has been shown to adopt an RNase H-like structure 3.

In contrast to plants, insects and worms 1, 4, mammalian cells react to the presence of long dsRNA, a product generated during intermediate stages of viral replication, by activating cytoplasmic receptors such as the dsRNA-dependent protein kinase R (PKR) and the recently identified retinoic acid inducible gene-1 (RIG-I) 5, 6. PKR is activated by autophosphorylation following binding to dsRNA. Once activated, it phosphorylates the eukaryotic translation initiation factor (EIF-2)–α, resulting in global inhibition of protein synthesis 5. PKR can also activate NF-κB and, therefore, connects to signal transduction pathways leading to type I IFN production. A family of 2′-5′-oligoadenylate synthetases (OAS) is also activated by dsRNA. This leads to the activation of RNase L, which degrades cellular mRNA. Although some studies have shown that even short dsRNA (11–17 bp) could activate PKR 7, it is believed that its full activation will not occur with dsRNA shorter than 30 bp. Because of this size discrimination, small siRNA produced by Dicer were found to induce sequence-specific gene regulation without triggering the host type I IFN pathway 8.

To date, the natural function of siRNA-mediated gene regulation in mammalian cells is not known. However, the conservation of the protein machinery involved in RNAi reflects the importance of a second class of small regulatory RNA, known as microRNA (miRNA) 9. In contrast to siRNA, miRNA are genome encoded as primary transcripts, which are processed in the nucleus by Drosha into primary transcripts of 60–80 nucleotides in length. Subsequent to transport to the cytoplasm and processing by Dicer, mature miRNA (22 nucleotides) are incorporated into a ribonucleoprotein complex that can either direct mRNA cleavage (in the case of perfect match) or suppress translation of mRNA (in the case of imperfect match). Because of this natural property, each siRNA sequence is likely to function as miRNA, leading to the inhibition of genes closely related to the target gene. In this respect, some studies have shown that even 11-bp matches between a siRNA and an unintended target is enough to impede that gene 10. In general, the magnitude of this unwanted activity known as an “off-target effect” was lower than that of the on-target gene, and can be avoided by careful selection of siRNA sequences and working with the lowest effective siRNA concentrations 11.

Although siRNA were initially thought to be small enough to avoid the activation of the IFN pathway 8, recent studies indicated that they could induce several unwanted effects, including IFN production. Initially, the IFN response was induced by siRNA derived from short hairpin RNA (shRNA) expressed from plasmids 12. Subsequently, it has been shown that IFN-stimulated genes (ISG) such as ISG20 and 2′-5′oligoadenylate synthetase 1 can be induced by chemically synthesized siRNA 13. These nonspecific effects were concentration dependent. In addition, the IFN-mediated JAK-STAT pathways and PKR are activated by siRNA 14. The observed effects were sequence independent and occurred at low siRNA concentrations. A recent study demonstrated that siRNA and shRNA activate TLR3 leading to the production of cytokines and type I IFN 15. Again, the effects were sequence independent and do not occur with the sense or antisense 21-bp ssRNA used to prepare the siRNA duplexes 15. Thus, how does a sequence independent receptor/protein avoid recognition of self-RNA? In general, siRNA enzymatically synthesized using T7 polymerase induced type I IFN response. However, only the 5′-triphosphate, a feature of T7-synthesized RNA, has been identified as the trigger of type I IFN production and subsequent activation of IFN-inducible genes 16. Taken together, these studies indicate that antiviral mechanisms associated with the introduction of dsRNA in mammalian cells are widespread and it is necessary to understand them in order to advance the therapeutic applications of siRNA.

During our studies with siRNA, initially we have noted that only certain chemically synthesized sequences were able to activate adherent PBMC, an enriched monocyte population, to produce inflammatory cytokines 17, 18. The observed effects were cell type dependent, thus suggesting the involvement of cellular factors that are differentially expressed by mammalian cells. In this respect, we 19 and others 20, 21 have recently demonstrated that PKR and TLR3 do not represent the major mechanisms by which chemically synthesized siRNA activate innate immunity. Moreover, we found that single-stranded (ss) siRNA (sense or antisense strands) and ds siRNA separately could activate innate immunity genes. In both cases, internalization of RNA and endosomal maturation is required for immune stimulation, suggesting the involvement of endosomal TLR, in particular TLR7 and TLR8. Indeed, inhibitors of endosomal maturation, such as bafilomycin, abrogated immune activation 19, 21. The data reported by Hornung et al. 20 and Judge et al. 21 suggest that immunostimulatory motifs are equally recognized within the context of either ss siRNA or siRNA duplexes because the magnitude of cytokines and IFN responses was comparable for both molecules. In contrast, we have found that ss siRNA were more potent than siRNA duplexes 19. A possible explanation for the differences could be related to variation in RNA sequences and secondary structures. These RNA-specific features may differ in triggering TLR signaling, thus resulting in different cytokine profiles.

In this study, we first investigated whether the cellular mechanisms, in particular TLR7 and 8, responsible for RNA recognition do effectively recognize immunostimulatory motifs in the context of siRNA duplexes. Then, we evaluated the effects of 2′-uridine modifications on RNA immunostimulatory activity. The data indicate that immunostimulatory RNA motifs are more effective in the context of ss siRNA as compared to siRNA duplexes and 2′-ribose modification of only uridines is sufficient to block immune activation by either ssRNA or siRNA duplexes.

Results and discussion

Is ds siRNA an ideal ligand for TLR7 and TLR8?

Since the discovery of RNAi in somatic mammalian cells 8, several groups have demonstrated that RNAi can be used to silence any gene of interest 2, 18, 2224. Indeed, exogenously provided siRNA or endogenously expressed hairpin RNA can induce the degradation of specific target mRNA in vitro and in mice, suggesting the potential for siRNA as therapeutics. However, recent studies indicate that there are nonspecific effects associated with the use of siRNA. In addition to off-target effects due to cross-hybridization with cellular mRNA, siRNA can activate innate immunity, leading to the production of inflammatory cytokines and type I IFN. Whole PBMC, monocytes, monocyte–derived DC (moDC) and plasmacytoid DC (pDC) are activated by siRNA to produce cytokines and type I IFN through the activation of TLR7 and/or TLR8 signaling pathways 1921. TLR are transmembrane proteins that are critical for the recognition of pathogen-specific molecular patterns (PAMP) derived from both bacterial and viral species 25. They are localized either on the plasma membrane or in intracellular vesicles such as endosomes. Among ten reported human TLR, TLR3, TLR7, TLR8 and TLR9 are involved in the recognition of microbial nucleic acids 25.

Judge et al. 21 identified one single RNA motif (5′-UGUGU-3′) and its immunostimulatory effect seems to be dependent on the GU-content as suggested by Heil et al.26. However, Hornung et al. 20 identified a second RNA motif (5′-GUCCUUCAA-3′) that is effectively recognized by TLR7 in the context of ds siRNA and the induction is unrelated to GU-content. Regardless of the GU-content, the data suggest that TLR7 and/or TLR8, recently discovered to recognize ssRNA 23, 27, can also recognize dsRNA with similar efficacy. Indeed, Hornung et al.20 argued that the mechanism responsible for recognizing ss siRNA is able to effectively recognize the immunostimulatory motifs within the siRNA duplexes. Thus, a first fundamental question is whether ds siRNA is an ideal ligand for TLR7 and TLR8 expressed by human blood cells.

Previous reports have shown that human monocytes express mRNA of TLR8, but not TLR3, or TLR9, suggesting that the TLR8 protein might be expressed by monocytes 19, 2830. Intracellular staining demonstrated that TLR8 and to less extent TLR7 are expressed by freshly isolated human monocytes (Fig. 1A). Notably, around 90% of the isolated cells are CD14 positive (Fig. 1B). To test the functionality of the expressed TLR, cells were incubated with the specific TLR7 and TLR8 ligand R-848 for 6 h and then TNF-α levels were measured in the supernatant. Consistent with the expression data, adherent PBMC responded to R-848 as shown by production of TNF-α. (Fig. 1C). It is worth noting that despite the absence of TLR9 expression, monocytes also responded to both shRNA and control plasmid DNA (data not shown). Thus, analysis of the stimulatory effects of shRNA in immune cells will be difficult to assess with accuracy. Given that freshly isolated monocytes did not express TLR9 and responded to plasmid DNA, it is more likely that a TLR9-independent pathway is responsible for this activation. Also, it is noteworthy that electroporation of in vitro transcribed shRNA lacking the 5' triphosphate and nimicking the structure of the in vivo transcribed shRNA did not induce an interferon response in adherent PBMC, suggesting that cytoplasmic shRNA and siRNA are not immunostimulatory.

Figure 1.

Expression of TLR7 and TLR8 by freshly isolated human monocytes. (A) Adherent mononuclear cells were stained for the intracellular expression of TLR7 and TLR8. (B) Surface expression of CD14. Cells were also stained with fluorescence-conjugated isotype-matched control antibodies. (C) The TLR7 and TLR8 agonist R-848 induced TNF-α production. Adherent PBMC cultured at 2 × 105 cells/mL were stimulated with R-848 (5 μM) for 6 h. Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. The data are representative of three separate experiments.

To address the question whether short RNA duplexes are an ideal ligand for TLR7 and TLR8, first the immunostimulatory effects of ss siRNA and siRNA duplexes of five target sequences were compared. Adherent PBMC were incubated overnight (∼18 h) with the test molecules and subsequently TNF-α levels in culture supernatants were measured by an ELISA (Fig. 2). Adherent PBMC do respond strongly to ss siRNA when compared to siRNA duplexes. These findings suggest that unlike their ds counterparts, ss siRNA are effective ligands for TLR7 and TLR8. Dose-response experiments confirm the potent immunostimulatory activity of ssRNA when compared to siRNA duplexes (Fig. 3). Several studies have established that ssRNA is a natural ligand for human TLR7 and TLR8 26, 27. In addition, it is worth noting that TLR8 is functional in humans but not in mice 26. This observation is in accordance with our early observation that a siRNA sequence activated TNF-α production in human monocytes, but not in mice monocytes, which mainly express TLR8 17.

Figure 2.

Effects of ss and ds siRNA on TNF-α production in adherent PBMC. Cells were transfected overnight with 50 nM ds siRNA (ds), sense siRNA strands (s) or antisense siRNA strands (as). Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. The data are representative of five independent experiments ±SD.

Figure 3.

Dose responses to ss and ds siRNA. Adherent PBMC were transfected overnight with the test molecules at various concentrations. Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. DOTAP concentrations were 3, 6, 9 and 12 μg/mL, lanes 1 to 4, respectively. The data are representative of three independent experiments.

In the view of the fact that ss siRNA are potent activators of immune cells and that sense and antisense strands should be perfectly annealed prior to use in RNAi experiments, we examined whether the addition of a small amount of ss siRNA to the siRNA duplexes would enhance the immune response. Addition of ss siRNA to the siRNA duplexes enhanced immune stimulation (Fig. 4). Thus, we conclude that ss siRNA are a potent inducer of innate immunity in vertebrates, requiring as little as a few intracellular molecules to invoke a strong response.

Figure 4.

Addition of small amounts of the antisense strand to the siRNA duplexes enhanced immune stimulation. Adherent PBMC were transfected overnight with ds siRNA-5 (ds) without or with the addition of free antisense strand (as). Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. The data are representative of three independent experiments.

To date, there are only two siRNA motifs known to effectively activate innate immunity in the context of ds siRNA to a level comparable to that obtained with free ssRNA 20, 21. These findings prompted us to perform some comparison experiments. Consistent with the reported data, sense strands but not antisense strands stimulated PBMC to produce significant amounts of TNF-α (Fig. 5A). However, under our experimental conditions, the magnitude of the response to the sense strands was higher than that of siRNA duplexes (Fig. 5A). Similarly, the sense strands of siRNA-J and siRNA-H induced more IFN-α than the siRNA duplexes (Fig. 5B).

Figure 5.

Responses of whole PBMC to ss and ds siRNA duplexes. (A) Whole PBMC were transfected overnight with the test molecules at 50 nM. Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. (B) IFN-α levels after overnight transfection were measured. SiRNA-H and siRNA-J correspond to the immunostimulatory sequences described by Hornung et al. 20 and Judge et al. 21, respectively. The data are representative of six independent experiments.

In addition to the sequence-dependent activation of innate immunity by siRNA, there is also qualitative difference in cytokine production by individual RNA sequences. Indeed, some sequences preferentially induced TNF-α more than INF-α, whereas others induced IFN-α more than TNF-α. This could be due to a functional difference between TLR7 and TLR8. Certain RNA sequences might bind more effectively to TLR7 than TLR8 and vice versa. In adherent PBMC, TNF-α is mainly produced by CD14+ monocytes. Virtually all nucleated cells can produce type I IFN. However, the major source of IFN-α production in whole blood is pDC, known as “professional” type I IFN-producing cells. They produce up to 100-fold more than most other cell types and respond to immunostimulatory siRNA 20. In addition to their crucial role in activating immunity against pathogens, inappropriate activation of pDC by immunostimulatory RNA, such as viral and bacterial RNA can enhance the reactivity of autoreactive B cells to self-antigens leading to IFN-dependent autoimmunity.

Based on the data discussed above, it would seem that siRNA duplexes are less effective than ss siRNA in activating innate immunity. This conclusion is first supported by the fact that several ss siRNA, but not their siRNA duplexes are more potent in stimulating innate immunity genes. Secondly, some siRNA duplexes failed to induce cytokine production, however, their ss siRNA counterparts did. The latter observation provides compelling evidence that the intracellular receptors for ss siRNA, in particular TLR8 and TLR7, do not effectively recognize most immunostimulatory RNA motifs in the context of siRNA duplexes.

The finding that certain siRNA duplexes can activate innate immunity, albeit less effectively than ssRNA, is not surprising because during endocytosis and endosomal maturation, siRNA duplexes may dissociate at low pH and then expose the ssRNA motifs 19. In this case, one may expect to see difference in immune stimulation between ss siRNA and siRNA duplexes. Alternatively, siRNA duplexes may bind directly to endosomal TLR7 and TLR8, but with low affinity as compared to ss siRNA. It is of note that the linkage of both RNA strands, as done by Hornung et al. 21, will not totally prevent the unwinding of the RNA duplexes and allowing signaling. Indeed, hairpin antisense oligonucleotides and molecular beacons recognize their intracellular targets (e.g. mRNA) as single-stranded molecules 31. Alternatively, siRNA duplexes may bind to endosomal TLR3. Further studies are required to identify the true endosomal receptors for siRNA duplexes. For example, in vitro binding experiments with purified recombinant TLR or the use of appropriate cell lines overexpressing specific TLR may shed light on the recognition of short RNA duplexes by TLR. However, it is worth noting that Judge et al.21 failed to detect signaling by siRNA duplexes in human embryonic kidney cells (HEK) 293 transfected with either TLR7 or TLR8.

Although the exact role of cationic liposome upon the immunostimulatory potency of ss siRNA when compared to siRNA duplexes is not known, coated ss siRNA may mimic viral particles that are actively recognized by the immune system. However, the data obtained with the 2′-modified uridine ss siRNA argue against any potential role of N-[1–2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate (DOTAP) on ss siRNA immunostimulatory potency. Indeed, 2′-fluoro modified immunostimulatory ssRNA formulated with the same DOTAP in a similar fashion as the unmodified sequences did not induce cytokine production (see below). Despite this strong argument, the effect of DOTAP on ss siRNA activity was further investigated. In these experiments ss siRNA were complexed separately with DOTAP, mixed in equimolar concentrations and then incubated with adherent PBMC. The immunostimulatory potency of these preparations was significantly lower than ss siRNA, but higher than the respective ds siRNA preparations. Although the data are consistent with the notion that siRNA duplexes are less immunostimulatory when compared to their ss siRNA counterparts, it should be interpreted with caution considering that a large amount of DOTAP-trapped ss siRNA might not form siRNA duplexes.

Identification of novel immunostimulatory RNA motifs

Because our primary goal was to identify therapeutically immunostimulatory RNA motifs, which can be used as vaccine adjuvants, we have analyzed the siRNA sequences for potential improved RNA motifs. Although ssRNA containing GU-motifs (e.g. 5′-GUAGUGU-3′, 5′-GUGAUUGU-3′, 5′-GUCUACUUU-3′, 5′-GUUCUU-3′, 5′-GUUGGU-3′) exhibited higher immunostimulatory activity compared to sequences lacking these motifs, the overall analysis indicated that the immunostimulation might depend on more than just the single-GU bases. Indeed, it was difficult to predict the levels of immunostimulation based on the primary RNA sequences. Moreover, certain ss siRNA with very low GU-content (e.g. 5′-UGCUAUUGGUGAUUGCCUCTT-3′) or without GU residues (e.g. 5′-GACUUGAGCGAGCGCUUUUTT-3′) also activated the immune system (see Fig. 1). Thus, it is likely that other characteristics such as RNA structure, base position and base composition of flanking sequences may contribute to immune recognition and signaling by RNA.

During our studies, however, we noted that a substantial number of immunostimulatory ss siRNA contained a high level of uridines, which may provide an additional molecular feature for recognition by immune cells. This observation prompted us to investigate their immunostimulatory potential. Replacement of uridines within the immunostimulatory RNA motifs with adenosines significantly reduced immune activation (Fig. 6). Thus, stretches of uridines should also be avoided during siRNA design in order to minimize immune activation. This observation is important because current guidelines for the selection of effective siRNA recommend U- or A-rich 3′-end of the sense strand 32.

Figure 6.

Replacement of uridines with adenosines inhibited the immunostimulatory activity of ss siRNA. Adherent PBMC were transfected overnight with 50 nM synthetic test molecules. Subsequently, culture supernatants were collected and TNF-α levels were measured by ELISA. The data are representative of three independent experiments.

Knowledge about immunostimulatory motifs will help in the design of therapeutic siRNA with minimal immune activation. If multiple RNA motifs are involved in RNA recognition by the immune system, it will be interesting to identify when one motif is favored over the other. Examination of the tested ss siRNA sequences indicates the presence of several RNA motifs (e.g. 5′-UGGUGAUUG-3′, 5′-GUGAUUG-3′) that strongly activated blood cells in the context of ss siRNA, but less effectively in the context of siRNA duplexes. To determine the contribution of these motifs in stimulating human blood cells, base substitution was performed. As illustrated in Fig. 7A, adenosine replacement abrogated immunological activity of the ssRNA.

Figure 7.

Effects of base substitutions and chemical modifications on immunostimulatory activity of ss siRNA. (A) Global substitutions of the putative immunostimulatory motifs with adenosines. Adherent PBMC were transfected overnight with the test molecules (50 nM). Subsequently, TNF-α levels were measured by ELISA. (B) Effects of 2′-fluoro uridine modification on the immunostimulatory potential of three ss siRNA: the antisense strand of siRNA-5, the sense strand of siRNA-3 and the sense strand of siRNA-J. Experimental conditions are as in (A). (C) Effects of 2′-deoxy or 2′-O- methyl uridine modification on the immunostimulatory potential of the sense strand of siRNA-3. Experimental conditions are as in (A). The data are representative of four independent experiments.

2′-ribose modification of uridines abrogates immune activation

Considering the simplicity of the identified RNA motifs and their frequent occurrence in mRNA sequences, it is more likely that a large number of ssRNA and siRNA duplexes will activate innate immunity. It is therefore important to identify additional strategies to selectively evade immune recognition of siRNA. At least two basic approaches can be used to interfere with immune activation. One is to use delivery agents that avoid the delivery and/or retention of siRNA within the endosomes. Another is to apply chemical modifications. The first approach has been illustrated by the recent use of protamine-antibody-mediated delivery of synthetic siRNA, which did not result in immune activation 33. The second approach involves chemical modifications. In this respect, locked nucleic acid (LNA) that contains a methylene linkage between the 2′ oxygen and the 4′ carbon of the ribose ring reduced partially the immunoactivity of siRNA 21. Also 2′-fluoro pyrimidine-modified siRNA exhibited no immune activity in vitro and in vivo (unpublished data).

In the past, modified oligonucleotides have been essential for the development of therapeutic antisense oligonucleotides and ribozymes. However, the chemical modifications that maximize ribozymes or siRNA stability and pharmacokinetic properties must be chosen carefully so as not to inhibit their silencing activity. Thus, finding the appropriate chemical modifications for blocking siRNA immune activation will be important for exploring their therapeutic applications. To date, several 2′-ribose and backbone chemical modifications have been incorporated into siRNA without reducing their ability to block gene expression in vitro and in vivo. In particular, siRNA modified with 2′-fluoro pyrimidines showed enhanced activity in cell culture as compared to unmodified siRNA 34. In addition, 2′-fluoro- pyridimidine (C and U)-modified siRNA retained their activity in mice when delivered through the hydrodynamic transfection method 35. Since most of the activity of the tested ss siRNA seems to correlate with their uridine content, we sought to investigate whether 2′-ribose modification of uridines evade immune activation. As shown in Fig. 7B, replacement of uridines with their 2′-fluoro modified counterparts abrogated immune activation for the three molecules tested (siRNA-3, siRNA-5 and siRNA-J). No TNF-α secretion was seen with the corresponding siRNA duplexes (data not shown).

To further support the notion that the 2′-hydroxyl of uridines was required for immune recognition and signaling, ss siRNA with either 2′-deoxy (H) or 2′-O-methyl uridines was tested. Interestingly, both modifications abrogated immune recognition (Fig. 7C). Thus, immune recognition can be evaded by several 2′-modifications of uridines. This finding offers the possibility of choosing the appropriate modifications that evade immune recognition without reducing siRNA silencing potency. Similarly, to 2′-fluoro modifications, 2′-deoxy uridines can selectively replace unmodified uridines with minimal reduction in silencing activity 34.

Regarding immune recognition of RNA, the present data indicate that the cellular components of the innate immune system can distinguish between modified and unmodified RNA. In addition, the data indicate that high concentration of self (sense strand) or non-self RNA (antisense strand) within the endosomes does not necessarily activate TLR signaling. Thus, the intracellular localization of TLR allows them to perfectly distinguish between immunostimulatory and non-immunostimulatory RNA. Whereas several TLR are expressed in the cell surface, TLR7 and TLR8 traffic between the endoplasmic reticulum and intracellular compartments such as the endosomes and the lysosomes. To date, the exact meaning of these trafficking pathways is not known. However, an obvious function is to scan for viral or bacterial RNA and to avoid encounter cytoplasmic cellular RNA. Notably, a number of viruses have been shown to traffic through the endosomal compartments during cellular entry or assembly of viral particles. However, we cannot rule out the involvement of subcellular compartments in TLR dimerization and optimal signaling. Another question is how immune cells will deal with cytoplasmic RNA and nuclear DNA during cell division where free nucleic acid diffusion takes place. Whatever the exact role of intracellular compartmentalization of TLR in self and non-self discrimination of nucleic acids by innate immunity, it is important to understand the mechanism(s) by which TLR sense 2′-ribose modifications. In theory, these modifications can affect the interaction of RNA with TLR or other accessory factors in a number of ways. They could alter or modify the correct RNA structures required for strong binding to TLR and/or adequate downstream signaling. Alternatively, modified RNA might bind to cellular proteins, then blocking their interactions with TLR. The exact mechanisms underlying RNA recognition by the immune system clearly deserve further studies.

Concluding remarks

Although certain siRNA duplexes activate innate immunity, they are less effective than their ssRNA counterparts are. In addition to GU-containing immunostimulatory RNA motifs, uridine-rich ss siRNA are immunostimulatory. Replacement of 2′-hydroxyl uridines with 2′-fluoro, 2′-deoxy, or 2′O-methyl modified counterparts abrogated TNF-α production (Fig. 7B and C), indicating that modified RNA are not effectively recognized by endosomal TLR despite being localized into the endosomes through cationic lipids. The use of either 2′-fluoro or 2′-deoxy modified uridines should facilitate the therapeutic applications of siRNA in humans because siRNA with these modifications retained comparable activity in vitro and in vivo as the unmodified siRNA (34, 35 data not shown). It is also clear that immunostimulatory ss siRNA or siRNA duplexes can be suitable and safe adjuvants for incorporation into cancer and viral vaccines. Indeed, they activated immune cells to produce cytokines and type 1 IFN, which are required for a Th-1 type response [19].

Materials and methods


The unmodified siRNA used in this study were chemically synthesized by Ambion, dissolved in RNase-free water and annealed in transfection buffer containing 20 mM Hepes, 150 mM NaCl, and pH 7.4. The 2′-fluoro uridine modified siRNA was made by Ambion and Eurogentec. The 2′-O-methyl uridine modified siRNA and 2′-deoxy modified siRNA were made by Eurogentec. The compound resiquimod R-848 (Pharm Tech) was dissolved in water at 10 mM and stored at 4oC.

Isolation of blood mononuclear cells

PBMC were isolated by density gradient centrifugation (Lymphoprep, Nycomed Pharm, Oslo, Norway) from buffy coats obtained from healthy adult donors. Enriched monocyte populations were isolated by plastic adherence as previously described 19. Briefly, after 2–3-h incubation in humidified 5% CO2 incubator at 37oC, cells were washed with warm medium and then adherent cells were detached, washed, and analyzed by flow cytometry for the expression of CD14 marker. In most cases, more than 85% of the obtained cells by this technique were CD14+ cells. Cells were cultured in RPMI-1640 medium supplemented with 10% heated inactivated FCS and antibiotics.

Transfection and ELISA

Freshly isolated whole PBMC were cultured at 2 × 106 cells/mL in 0.200 mL in 96-well, flat-bottom plates and then transfected overnight with siRNA/DOTAP complexes. Secreted TNF-α and IFN-α were measured by sandwich ELISA according to the manufacturer's protocols (R&D PharMingen and PBL Biomedical Laboratories). Samples were run in triplicate. Transfection was performed by using DOTAP at 10 μg/mL according to the manufacturer's instructions. Adherent PBMC were cultured at 2 × 106 cells/mL in 200 μL in 96-well, flat-bottom plates and then processed as whole PBMC.

Flow cytometric analysis

Adherent PBMC were stained with PE-labeled CD14 or isotype-matched control for 30 min in ice. Subsequently, cells were washed three times with the staining buffer (PBS containing 0.5% FCS and 0.1% azide) and then analyzed by flow cytometry. For intracellular staining, cells were washed three times with PBS, resuspended in 1% paraformaldehyde and incubated for 15 min in ice. After washing, cells were resuspended in 150 μL pure methanol and incubated overnight at –20oC. Then, permeabilized and fixed cells were stained with FITC-conjugated anti-TLR7, anti-TLR8 (ImmunoKontact) or isotype-matched control Ab. Stained cells were washed three times with the staining buffer, resuspended in PBS and then analyzed by flow cytometry. CellQest software was used for data analysis.


I thank Lise Forfang and Trang Hynh for technical assistance and Dr. Anne Dybwad for critical reading of the manuscript. This work was supported in part by a grant for the Norwegian Cancer Society (A01052/005).


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5


  6. 6


  7. 7