|Home | About | Journals | Submit | Contact Us | Français|
The biological activity of siRNA seems to be influenced by local characteristics of the target RNA, including local RNA folding. Here, we investigated quantitatively the relationship between local target accessibility and the extent of inhibition of the target gene by siRNA. Target accessibility was assessed by a computational approach that had been shown earlier to be consistent with experimental probing of target RNA. Two sites of ICAM-1 mRNA predicted to serve as accessible motifs and one site predicted to adopt an inaccessible structure were chosen to test siRNA constructs for suppression of ICAM-1 gene expression in ECV304 cells. The local target-dependent effectiveness of siRNA was compared with antisense oligonucleotides (asON). The concentration dependency of siRNA-mediated suppression indicates a >1000-fold difference between active siRNAs (IC50 ≈ 0.2–0.5 nM) versus an inactive siRNA (IC50 ≥ 1 µM) which is consistent with the activity pattern of asON when relating target suppression to predicted local target accessibility. The extremely high activity of the siRNA si2B (IC50 = 0.24 nM) indicates that not all siRNAs shown to be active at the usual concentrations of >10–100 nM belong to this highly active species. The observations described here suggest an option to assess target accessibility for siRNA and, thus, support the design of active siRNA constructs. This approach can be automated, work at high throughput and is open to include additional parameters relevant to the biological activity of siRNA.
The increasing use of siRNA to specifically suppress target gene expression in mammalian cells progressively reveals an influence of local target characteristics on the biological effectiveness (1–3). In major part, this is assigned to characteristics of the local folding of target RNA (1–3) and a recent study indicates to some extent a similarity between the site-dependent biological activity of phosphorothioate-derived antisense oligonucleotides (asON) and siRNA (3). When taken together, these findings on target structure-related biological activity of siRNA in mammalian cells, raise two increasingly relevant questions. First, what are the rules governing the structure–function relationship of siRNA and, secondly, can one find ways to distinguish effective and non-effective siRNA constructs on the basis of target structure analysis. The latter issue guides one to the systematic design of effective siRNA.
Since siRNA is regarded to be highly sequence-specific (4), it is reasonable to assume that a recognition step occurs between siRNA and its target sequence. This step might occur between the siRNA double strand and the target or subsequent to a possible preceding dissociation of the siRNA with its antisense strand. In any case it seems to be reasonable to assume that the local target RNA segment has to be accessible to intermolecular nucleotide–nucleotide interactions with siRNA or its antisense strand, respectively. There is a need to find a mammalian test system and experimental conditions to study the influence of target structure on the activity of siRNA.
Here, we used the human gene ICAM-1/CD54, which is one of the best studied targets for antisense nucleic acids (5,6), and siRNA (3) to investigate the extent of apparent gene expression in relation to local target accessibility. Target gene suppression in this system is reproducible and robust. Regarding RNA folding of the complete ICAM-1 mRNA, we had access to a highly systematic computational RNA secondary structure analysis (6; R.Kretschmer-Kazemi Far, J.Leppert and G.Sczakiel, manuscript in preparation). Even a less detailed analysis was earlier shown to give rise to results that are highly consistent with structural probing of RNA in the presence of cellular extracts (7), suggesting that the theoretical analysis of target accessibility as performed here and as described recently (6), represents a valid basis to study the relation between accessibility and efficacy of siRNA. However, as an additional measure of the reliability of this kind of computational analysis of target accessibility, we included asON in this study. Furthermore, asON were used here to compare their effectiveness with siRNA on a quantitative level.
All oligonucleotides were purchased from commercial suppliers. asON were synthesized as phosphorothioate-modified oligodeoxyribonucleotides. All single-stranded oligoribonucleotides constituting siRNA were synthesized with two 3′-deoxythymidine nucleotides. The following phosphorothioate-modified control oligonucleotides with scrambled sequences were used: 5′-GGTACGTGCTGAGGCCTG-3′; for siRNA we used the scrambled control: 5′-CGAACUCACUGGUCUGACCdtdt-3′ (sense strand), 5′-GGUCAGACCAGUGAGUUCGdtdt-3′ (antisense strand). The yield and purity of all oligonucleotides was determined by UV absorption spectroscopy and the integrity was controlled with 20% denaturing polyacrylamide gels followed by staining with Stains-All (Sigma-Aldrich, Deisenhofen, Germany). The sequences of all oligonucleotides are shown in Figure Figure11.
Complementary RNA strands at 20 µM were annealed in 50 mM potassium acetate, 1 mM magnesium acetate and 15 mM HEPES pH 7.4 by heating at 90°C for 2 min followed by incubation at 37°C for 1 h. The formation of double-stranded RNA was confirmed by non-denaturing (15%) polyacrylamide gel electrophoresis and staining with Stains-All.
To predict secondary structures of target sequences, we used the software mfold version 2.3 (8,9) which is included in the software Wisconsin Package Version 10.0, Genetics Computer Group. Characteristic parameters of oligonucleotide sequences were determined using the software Oligo version 3.4, which is based on the work of Rychlik and Rhoads (10).
The cell line ECV304 was described as an endothelial cell line derived from spontaneously immortalized human umbilical vein cells (11). However, the DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany) showed by DNA fingerprinting that this cell line is a derivative of the human urinary bladder carcinoma cell line T-24 (12), which expresses ICAM-1 in an inducible manner. ECV304 cells were maintained in medium 199 (Sigma-Aldrich) buffered with 25 mM HEPES pH 7.4 and supplemented with 0.68 mM l-glutamine and 10% FCS (Invitrogen, Karlsruhe, Germany). Cells were routinely split two to three times a week after trypsinization. For stimulation of ICAM-1 (CD54) 200 U/ml interleukin 1β (IL-1β; PromoCell, Heidelberg, Germany) was added to the medium and cells were incubated overnight (for 16–18 h).
ECV304 cells were seeded in 12-well culture plates at a density of 1.5 × 105 cells/well, 15 h prior to oligonucleotide treatment. Then, the cells were washed once with pre-warmed (37°C) serum-reduced Opti-MEM I medium (Invitrogen). To deliver asON or siRNA to cells we used Lipofectamine 2000 (Invitrogen). The transfection of asON or siRNA was performed with 0.4 ml of Opti-MEM I medium containing 0.1 µM asON or siRNA and 10 µg/ml Lipofectamine 2000 per well. The cells were incubated for 4 h at 37°C, 5% CO2. Subsequently, the transfection medium was replaced by medium 199 containing 10% FCS. After an incubation of 4 h at 37°C, 5% CO2, the medium was again substituted with medium 199 containing 10% FCS and 200 U/ml IL-1β to stimulate ICAM-1 expression. For immunofluorescence analysis, cells were removed from the culture dish by trypsinization with 100 µl of 0.25% trysin/0.02% EDTA in PBS for 5 min at 37°C, 16–18 h after stimulation with IL-1β. ICAM-1-specific gene expression of cells treated with IL-1β and control asON or siRNA (scrambled sequence for both) was set at 100%, while gene expression of cells treated with control asON or siRNA only was defined as the basal expression level.
Cells were washed with phosphate-buffered saline (PBS; pH 7.4) and incubated with a phycoerythrin (PE)-conjugated CD54 monoclonol antibody (clone LB-2) or an isotype-identical PE control (both, Becton Dickinson, Heidelberg, Germany) in 50 µl of PBS containing 1% BSA at 4°C for 30 min. Subsequently, cells were washed with PBS and suspended in PBS containing 1% paraformaldehyde (pH 7.4). The cells were analyzed using a Becton Dickinson FACSCalibur and CellQuest Pro software (Becton Dickinson). After gating out dead cells, the mean (PE) fluorescence intensity was determined and the inhibition of ICAM-1 gene expression was standardized to the IL-1β-induced state which was set at 100%. On average, basal gene expression levels were ~35%.
Total RNA was extracted from transfected cells using the RNeasy mini kit including treatment with RNase-free DNase I (Qiagen, Hilden, Germany). The yield and purity of RNA were determined by spectrophotometry. Synthesis of cDNA was carried out using random hexamer primers and Superscript II RNase H– reverse transcriptase according to the manufacturer’s specifications (Invitrogen). Quantitative PCR was performed using the GeneAmp 5700 sequence detection system (Applied Biosystems, Darmstadt, Germany) and SYBR green PCR core reagents (Eurogentec, Seraing, Belgium). ICAM-1 cDNA was amplified with the forward primer 5′-GCCACTTCTTCTGTAAGTCTGTGGG-3′ and reverse primer 5′-CTACCGGCCCTGGGACG-3′, resulting in a fragment of 300 bp. Samples were standardized using primers specific to cDNA encoding human GAPDH (forward primer 5′-AACAGCGACACCCACTCCTC-3′ and reverse primer 5′-GGAGGGGAGATTCAGTGTGGT-3′) resulting in a product of 258 bp. Standard curves were obtained after amplification of 2.5 × 103–2.5 × 107 copies of purified plasmid pP5, a derivative of pEGFP-C1 carrying the amplicon generated with the ICAM-1 primer set and plasmid pCR-GAPDH, a derivative of pCR 2.1 harboring the GAPDH amplicon sequences.
Many approaches have been developed in order to obtain insights into the relationship between local target accessibility and the effectiveness of antisense nucleic acids in living cells (reviewed in 13). Accessibility of a given RNA for intermolecular base-pairing is directly related to RNA structure which can be assessed experimentally (14–18) or theoretically (reviewed in 19). Based on the identified or presumed local accessibility, complementary nucleic acids can be selected. Among the theoretical ways to design effective asON, a specific systematic computational analyses of predicted RNA structures and nucleotide sequence characteristics has been shown to be significantly related to the biological activity of asON in living cells (6,20). This theoretical approach has also been shown to be closely related to the experimental probing of target RNA in the presence of cellular extracts indicating its significance (7). This methodology (6) was applied here to the sequence of the ICAM-1 mRNA in order to define local folding, i.e. structural accessibility. Window sizes ranging between 600 and 1400 nt were analyzed by mfold (see Materials and Methods), shifted along the target sequence at a step width of 100 nt, and the 10 lowest energy structures of each were recorded and used to select local targets that occurred in all predictions at high abundance (data not shown).
To investigate and to compare the possible role of local target characteristics for the efficacy of asON and siRNA, three local target sites along the ICAM-1 mRNA were chosen. The local sequence segment at position 1430–1490 termed T1 in this work is predicted to adopt a local RNA structure that is not regarded as being accessible to complementary strands and, thus, it should not support their binding and efficacy in living cells (Fig. (Fig.1,1, top right). Conversely, the local sequence segments in the vicinity of positions 1570 and 1640 are predicted to be accessible in structural terms and, thus, are regarded to favor the biological activity of complementary nucleic acids including asON and, possibly, siRNA (Fig. (Fig.1,1, middle and bottom right). In addition to structural characteristics of the target RNA, the design of asON sequences in this work was based on criteria related to their sequences as described recently (20), including the results of nucleotide sequence homology searches, the exclusion of specific motifs such as G-quartets and intramolecular folding, or a high potential of dimer formation. The sequences of siRNA constructs were chosen such that they are directed against the central portion of the target motifs shown in Figure Figure1.1. All siRNA constructs were tested for nucleotide sequence homology with unrelated human transcribed sequences and they all share the same GC content and length.
The target gene ICAM-1 is expressed at low basal levels and at increased levels in the presence of stimulatory cytokines such as IL-1β. Routinely, inhibitory effects in this system are monitored by the suppression of IL-1β-induced stimulation of ICAM-1 expression. Here, we also tested effects on the basal levels of ICAM-1 expression since this increases the range in which the effects of siRNA and asON can be tested. If one compares the effectiveness of different classes of biologically active molecules such as asON and siRNA, then it is reasonable to attempt to apply optimal experimental conditions for each. Experimental evidence suggests that the transfection step is critical for the extent of apparent target gene suppression by siRNA (21–23). In this work, we used a constant concentration of 10 µg/ml Lipofectamine 2000 in cell culture with ECV304 cells. At 100 nM asON this is equivalent to a ratio of Lipofectamine 2000 to asON of ~16:1 (w/w) and to siRNA of ~7:1 (w/w). Under these conditions we observed significantly stronger siRNA-mediated inhibition than with the use of the transfection reagent ‘lipofectin’. We performed further experiments using 3 µg/ml Lipofectamine 2000, equivalent to a ratio of Lipofectamine 2000/asON of 5:1 (w:w), which showed a decrease in asON-mediated inhibition. At 3 µg/ml Lipofectamine 2000, the transfection conditions were appropriate, in principle, although at the higher dose of Lipofectamine 2000, suppressive effects were stronger. For further experiments we used Lipofectamine 2000 at 10 µg/ml in cell culture because this favored the effectiveness of siRNA and, to a minor extent, also that of asON. This allowed a greater dynamic range for measuring suppression of target gene expression.
The biological activity of the set of asON schematically depicted in Figure Figure11 was monitored by suppression of IL-1β-induced expression of ICAM-1 in ECV304 cells at the protein level as described (6). Values for asON-mediated inhibition of ICAM-1 were related to control oligonucleotides with a scrambled order of nucleotides at a comparable base composition. The results clearly show strong inhibition of IL-1β-induced gene expression of ICAM-1 by asON directed against the favorable target sites T2 and T3 and even indicates a suppression of gene expression below the levels of constitutive gene expression (Fig. (Fig.2).2). The unfavorable target site T1 (Fig. (Fig.1A)1A) does not give rise to significant inhibitory effects by asON (Fig. (Fig.2A).2A). The observations described here are consistent with extensive studies on asON-mediated inhibition of gene expression where asON had been selected according to the protocol used in this work. This experiment shows a large dynamic range of suppression of ICAM-1 gene expression when comparing the unfavorable local target T1 with the favorable targets, T2 and T3, respectively.
A real antisense effect implies that the target gene is suppressed at the level of RNA which gives rise to reduced levels of the protein encoded by the target gene. Levels of target mRNA were measured by quantitative RT–PCR and standardized to the levels of GAPDH mRNA. The inhibition data summarized in Figure Figure2B2B show that this is the case for the antisense species used here. When considering the phosphorothioate chemistry of the asON it is important to note that no side effects were observed in the use of the scrambled controls (data not shown). Together, these findings are consistent with the view that the asON-mediated down-regulation of ICAM-1 gene expression observed in this study is specific and follows an antisense mechanism.
The effects of siRNA were monitored at the protein level by FACS analysis of transfected cells and standardized to IL-1β-induced expression of ICAM-1 in the presence of a scrambled siRNA control (Fig. (Fig.3A).3A). The most potent siRNA species were directed against the local targets T2 and T3 which gave rise to the highest inhibition levels in the use of asON. The siRNA directed against the local target T1 which was unfavorable for asON showed no significant influence of ICAM-1 gene expression. This differential behavior is highly consistent with the inhibition properties of the asON investigated in this work. The biologically active siRNA constructs suppress stimulation of the target completely and even reduce basal levels of ICAM-1 gene expression, indicating a significantly increased activity when compared with asON (Figs (Figs22 and and33).
Since the action of siRNA is presumed to occur post-transcriptionally and involves the degradation of target RNA, we also measured effects in this system at the level of ICAM-1 mRNA (Fig. (Fig.3B).3B). The results of this experiment are compatible with siRNA-mediated inhibition at the protein level. To further support the presumed siRNA mechanism, we tested whether the antisense strand of the siRNA construct si2B is more sensitive to modifications than the sense strand, as one would believe from the literature (24,25). We transfected DNA–RNA hybrid constructs in which either the RNA sense strand or the RNA antisense strand was substituted by a DNA strand (si2B-H1, si2B-H2; Fig. Fig.1).1). A substantial loss of activity was observed when the antisense strand was DNA rather than RNA but the activity remained high when the sense strand was replaced. Moreover, the antisense or sense strand of the active siRNA si2B did not show significant suppression of target gene expression at the same concentration (Fig. (Fig.3C).3C). These observations suggest that siRNA is the active molecule in the test system used here.
In order to quantitatively compare the biological activity of asON and siRNA and the influence of the choice of the local target, we determined apparent values of half maximal inhibition (IC50 values) and the maximal achievable extent of inhibition (Fig. (Fig.4).4). At the level of the ICAM-1 protein, the values of the apparent IC50 for siRNA against T2 and T3 were 0.24 and 0.58 nM, respectively (Fig. (Fig.4B4B and C), whereas si1 did not exert any significant inhibition at concentrations of up to 100 nM (Fig. (Fig.4A).4A). For antisense constructs, an IC50 value of 47 (as2B) and 22 nM (as3C) was measured, whereas the possible IC50 value for as1 against T1 cannot be derived from the experiments. It is reasonable to assume that this value is not <500 nM (Fig. (Fig.4A).4A). In the latter case, a substantial cytotoxicity was observed at concentrations >100 nM. The IC50 value of the most active siRNA construct si2B is ~100-fold lower than the value for one of the most effective known phosphorothioate-substituted asON, as3C. When assuming that as3C represents the maximal achievable level of target gene suppression in the use of phosphorothioate-substituted asON, and si2B represents the corresponding level in the use of siRNA, then this suggests that under the conditions used here, siRNA is ~100-fold more active in mammalian cells than asON. Furthermore, even though si3 (Fig. (Fig.4C;4C; IC50 = 0.58 nM) is ~40-fold more potent than the homologous asON as3C (Fig. (Fig.4C;4C; IC50 = 22 nM), the maximal extent of inhibition is greater with the use of the asON (~30% gene expression) than with the siRNA (~50% gene expression).
At a first look one might hypothesize that observed IC50 values monitor the biological activity of the siRNAs used here. When comparing the effectiveness of si1, si2B and si3 at 100 nM (Fig. (Fig.3A3A and B) with their dose-dependency, then it seems that the maximal achievable level of target inhibition rather than the IC50 value is more closely related to biological activity. This means that above certain concentrations of siRNA a further increase does not increase the extent of inhibition; this indicates a bottle neck for the effectiveness of siRNA in mammalian cells that cannot be compensated by increased doses. One might deduce that the mode of action of siRNA includes at least two critical parameters: a concentration-dependent and a concentration-independent parameter.
In this study, experimental evidence is provided for the view that structural local target characteristics are related to target accessibility and to the biological activity of siRNA. At a first glance this relationship seems to be similar to the presumed structure–function relationship of asON which is compatible between this work and detailed published work in the same biological system (6). The findings described here are based on a computational analysis of target accessibility. The significance of this theoretical RNA structure analysis is supported (i) by its consistency with further detailed studies on the structure–function relationship of asON (7,26,27), (ii) by its close relationship with an experimental approach to monitor target accessibility (7) and (iii) by the results in the use of asON as measured here.
Local RNA structures along the ICAM-1 target mRNA that are predicted to be accessible, serve as favorable targets for siRNA; this suggests the usefulness of the theoretical accessibility approach for the selection of siRNA constructs. This observation is consistent with a recent report in which the oligonucleotide scanning array technology has been used to experimentally monitor target accessibility (2). In principle, this work suggests on a quantitative level that target accessibility is correlated with biological activity of asON and siRNA. It remains to be tested, however, whether the significant effects of single additional or lacking nucleotides of siRNA with respect to complementary nucleotides of the target strand give rise to a heterogeneous pattern of effectiveness within a given local target as observed for asON (7,28,29). At a lower resolution this possibility is supported by a study with overlapping siRNAs shifted along a target sequence by 3 nt (1). Kawasaki et al. (30) provide experimental evidence for the view that target structure matters with regard to the biological activity of siRNA. Interestingly, this seems to be overcome with the use of siRNA that has been exposed to recombinant Dicer in vitro (30).
The biological activities of si2A and si2B are comparable (Fig. (Fig.3A).3A). If one speculates on the importance of the termini of siRNA or its complementary single strands, then one might conclude that there is no difference between the ends with respect to their effectiveness in mammalian cells. When considering the local target for both siRNAs (T2; Fig. Fig.1),1), then it does not seem to matter whether either terminus of the siRNA is directed against a sequence segment that is involved in intramolecular base-pairing or not. It seems to be reasonable, however, to assume that a closer look at the relationship between local target characteristics and the efficacy of siRNA will reveal new mechanistic insights that can be exploited for the design of highly effective siRNA species.
The extent of inhibition of ICAM-1 gene expression in the use of asON versus siRNA indicates an ~100-fold increase in the apparent IC50 value of siRNA (Fig. (Fig.4B).4B). This observation is consistent with a recent report that provides experimental evidence for a similar gap of activity reflected by an ~100-fold increased activity of siRNA versus asON (23). The authors further suggest that the activity of double-stranded RNA is dependent on ‘positional effects’ of the target RNA.
The ICAM-1 mRNA was used as a target RNA for a large set of asON and siRNA in a recent study (3). We compared the inhibition in this study with local target accessibility using the approach in our study. For a number of cases we also find a significant relationship between local target accessibility and efficacy. Conversely, most local target sites described in Vickers et al. (3) cannot be assessed by the theoretical approach used here and, to some extent, the inhibition data on the level of target RNA and target protein are not consistent. Thus, a systematic analysis of the data in Vickers et al. (3) using the concept in our work is not possible.
The work on asON-mediated effects made use of phosphorothioate-derived asON. It might well be that improved second generation chemistries significantly reduce the gap of biological activity between asON and siRNA. In this context, it is also interesting to look at the sequence specificity and side effects of siRNA and asON at the available different chemistries.
One might hypothesize that sequence-specific interactions between siRNA and its target RNA are crucial and are closely related to efficacy. As a consequence, insights into the biochemical properties of this step might help to understand the structure–function relationship of siRNA, thereby supporting the rational improvement of the efficacy. It is likely that the recognition process between double-stranded and single-stranded RNA inside mammalian cells is accompanied by cellular proteins and other endogenous compounds. It remains open, however, as to whether such additional substances influence the molecular mechanism of RNA interactions (e.g. by changing RNA folding and the mode of recognition) or whether they catalyze this reaction in an enzyme-like fashion in order to facilitate this process and increase the kinetics without affecting the mechanism [e.g. as observed in the case of facilitator-promoted annealing of complementary RNA which does not measurably influence the underlying structure–function relationship (31)].
We thank M. Lehmann and T. Sabelhaus for providing plasmid constructs. This work was supported by a grant from the Faculty of Medicine of the University of Lübeck (FSO) to R.K.K.F. and G.S.