NUCLEIC ACID THERAPEUTICS
`Volume 21, Number 3, 2011
`ª Mary Ann Liebert, Inc.
`DOI: 10.1089/nat.2011.0286
`
`K-OTS Reviews
`
`Structural Diversity Repertoire
`of Gene Silencing Small Interfering RNAs
`
`Chan Il Chang,1,2* Helena Andrade Kim,3* Pooja Dua,1,4
`Soyoun Kim,4 Chiang J. Li,2 and Dong-ki Lee1
`
`Since the discovery of double-stranded (ds) RNA-mediated RNA interference (RNAi) phenomenon in Cae-
`norhabditis elegans, specific gene silencing based upon RNAi mechanism has become a novel biomedical tool that
`has extended our understanding of cell biology and opened the door to an innovative class of therapeutic agents.
`To silence genes in mammalian cells, short dsRNA referred to as small interfering RNA (siRNA) is used as an
`RNAi trigger to avoid nonspecific interferon responses induced by long dsRNAs. An early structure–activity
`relationship study performed in Drosophila melanogaster embryonic extract suggested the existence of strict
`siRNA structural design rules to achieve optimal gene silencing. These rules include the presence of a 30
`overhang, a fixed duplex length, and structural symmetry, which defined the structure of a classical siRNA.
`However, several recent studies performed in mammalian cells have hinted that the gene silencing siRNA
`structure could be much more flexible than that originally proposed. Moreover, many of the nonclassical siRNA
`structural variants reported improved features over the classical siRNAs, including increased potency, reduced
`nonspecific responses, and enhanced cellular delivery. In this review, we summarize the recent progress in the
`development of gene silencing siRNA structural variants and discuss these in light of the flexibility of the RNAi
`machinery in mammalian cells.
`
`Introduction
`
`RNA interference (RNAi) is an evolutionarily conserved
`
`mechanism of posttranscriptional gene silencing by
`double-stranded (ds) RNAs (Hannon, 2002). Originally dis-
`covered by Fire and Mello in Caenorhabditis elegans (Fire et al.,
`1998), long (typically 300–1000 bp) dsRNAs introduced into
`cells or organisms effectively trigger RNAi to specifically
`inhibit target gene expression in a wide range of organisms.
`The RNAi pathway is initiated upon cleavage of long dsRNA
`into *21-nucleotide (nt)-long small interfering RNA (siRNA)
`by a ribonuclease III enzyme called Dicer. This siRNA duplex
`subsequently gets assembled into an RNA-induced silencing
`complex (RISC), in which one strand (sense or passenger
`strand) is eliminated and the other (antisense or guide strand)
`recognizes and cleaves the complementary mRNA with the
`help of Argonaute-2 (Ago-2) and other auxiliary RISC proteins.
`Because of the outstanding potency and specificity compared
`with other loss-of-function technologies, RNAi-mediated gene
`
`silencing has rapidly become a fundamental tool for gene
`function studies (Fraser, 2004) and a promising therapeutic
`modality for a variety of diseases (Lares et al., 2010).
`However, in contrast to other organisms, the initial effort
`to use long dsRNAs to trigger RNAi in mammalian cells was
`largely unsuccessful, because of the strong induction of in-
`terferon and the activation of protein kinase R (PKR), pro-
`duced as a consequence of an antiviral response to the long
`dsRNA molecules. This undesired response results in the
`nonspecific degradation of mRNAs and inhibition of protein
`synthesis (Stark et al., 1998; Caplen et al., 2000; Ui-Tei et al.,
`2000). Successful silencing of specific genes via an RNAi
`mechanism in mammalian cells was first reported by the
`Tuschl group, who demonstrated that chemically synthe-
`sized siRNA, a structural mimic of the Dicer cleavage
`product of long dsRNA, could trigger efficient and specific
`target gene silencing in mammalian cells without generating
`undesired interferon responses (Elbashir et al., 2001a,
`2001b).
`
`1Global Research Laboratory for RNAi Medicine, Department of Chemistry, Sungkyunkwan University, Suwon, Korea.
`2Skip Ackerman Center for Molecular Therapeutics, Beth Israel Deconness Medical Center, Harvard Medical School, Boston,
`Massachusetts.
`3PCL, Inc., Seoul, Korea.
`4Department of Biomedical Engineering, Dongguk University, Seoul, Korea.
`*These two authors equally contributed to this work.
`
`125
`
`Alnylam Exh. 1090
`
`

`

`126
`
`CHANG ET AL.
`
`The same group also performed a structure–activity rela-
`tionship study to define the structural features of potent siR-
`NAs (Elbashir et al., 2001c). Using Drosophila melanogaster
`embryo extract as a model experimental system, they inves-
`tigated the gene silencing activity of various dsRNA struc-
`tures, ranging in length from 19 to 25 nt, with different
`overhang structures. From this experiment, they found that
`there was a strict limit to the siRNA duplex length for optimal
`gene silencing activity; 19-bp-long duplexes showed optimal
`gene silencing, whereas duplexes shorter or longer than 19 bp
`were significantly less potent or inactive. Their results also led
`them to emphasize the importance of overhang structures;
`duplexes without overhangs (blunt-ended) or with 50 over-
`hangs were less potent than duplexes with 2-nt-long 30 over-
`hangs. Therefore, they concluded that a 19-bp RNA duplex
`with 2-nt 30 overhangs at both ends, often referred to as the
`‘‘19 þ 2 structure,’’ is the most potent siRNA structure for gene
`silencing, and this structure was adopted as the standard in
`the RNAi field.
`Soon after the application of siRNAs in functional genomic
`studies and the development of therapeutics, it was found
`that siRNAs triggered several unintended nonspecific re-
`sponses when introduced into cells and animals (Tiemann
`and Rossi, 2009). These nonspecific responses included off-
`target gene silencing triggered by the incorporation of the
`sense strand into the RISC or incomplete base pairing of the
`siRNA antisense strand with nontarget mRNA, activation of
`nonspecific innate immune responses by pattern recognition
`receptors [eg, Toll-like receptors (TLR) and retinoic acid in-
`ducible gene I (RIG-I)-like cytoplasmic helicases], and RNAi
`machinery saturation by excess exogenous siRNA, which in-
`hibits endogenous microRNA processing. These nonspecific
`responses have limited the use of siRNA as a specific tool for
`therapeutics and loss-of-function studies.
`To circumvent these problems, chemical modifications
`have been introduced into the classical 19 þ 2 siRNA back-
`bone. However, chemical modification of siRNA is often as-
`sociated with unfavorable side effects such as toxicity and
`reduced silencing efficiency. Therefore, the design of novel
`siRNA backbone structural variants that can not only trigger
`efficient gene silencing but also ameliorate nonspecific re-
`sponses triggered by classical siRNA structures is an impor-
`tant area of current research.
`Several recent studies reported the design and analysis of
`novel RNAi-triggering structures distinct from the classical
`19 þ 2 siRNA structure (Fig. 1). These novel RNAi-triggering
`structures do not conform to the key features of classical
`siRNA in terms of overhang,
`length, or symmetry. Im-
`portantly, the newly designed siRNA structural variants
`show improved functionality over classical siRNAs such as
`more potent gene silencing activity, reduction of nonspecific
`responses, immune stimulation, or enhanced internalization.
`In this review, we summarize several recent findings that
`challenge the dogma surrounding the 19 þ 2 siRNA archi-
`tecture and discuss these findings in view of the flexibility of
`the RNAi machinery in mammalian cells.
`
`siRNA Without 30 Overhangs
`
`The initial structure–activity relationship study of gene si-
`lencing siRNAs using D. melanogaster embryo lysates con-
`cluded that the presence of 2-nt-long 30 overhangs was an
`
`FIG. 1. Structures of siRNA structural variants. Top strand:
`sense (passenger) strand; bottom strand: antisense (guide)
`strand; ellipsoid: RNA; square box: DNA. (a) Classical 19þ 2
`siRNA; (b) Dicer-substrate siRNA (R 25/27D form); (c) Di-
`cer-noncleavable long siRNA (25 bp); (d) 38-bp-long dual-
`target siRNA; (e, f) asymmetric short siRNAs; (e) aiRNA; (f)
`asiRNA; (g) sisiRNA; (h) immunostimulatory siRNA; (i)
`ssiRNA.
`
`essential feature of siRNAs (Elbashir et al., 2001c). Deviations
`in the overhang length, substitution with deoxyribonucleo-
`tides (Hohjoh, 2002), or the addition of chemically modified
`residues (Harborth et al., 2003) has been shown to reduce the
`silencing efficiency of siRNAs. All these 30 overhang varia-
`tions affect the recognition of siRNA by the PAZ domain of
`Ago-2 (Ma et al., 2004). In contrast, however, many studies
`have shown that modifications in the length and chemistry of
`the overhang residues do not affect silencing (Chiu and Rana,
`2002; Amarzguioui et al., 2003), raising questions about the
`proposed limitations in overhang design and the importance
`of the PAZ domain–siRNA interaction in the human RNAi
`machinery. Even dsRNAs without any 30 overhangs have
`been shown to be highly efficient in gene silencing in mam-
`malian cells (Czauderna et al., 2003). The dispensability of the
`30 overhang structure was confirmed in another study (Chang
`et al., 2007). These findings gave a hint that siRNA structural
`requirement in mammalian cells might be more flexible than
`those in D. melanogaster embryo lysates (Elbashir et al., 2001c)
`and stimulated several other studies (described later) that
`demonstrate the structural flexibility of siRNAs in mamma-
`lian cells.
`
`

`

`SIRNA STRUCTURAL VARIANTS
`
`Long siRNA Structural Variants
`
`Similar to the overhang structure, it was originally thought
`that the length of the siRNA duplex was critical for triggering
`efficient target gene silencing (Elbashir et al., 2001c). How-
`ever, several recent studies have demonstrated that a great
`deal of flexibility exists in the length of the siRNA duplex. The
`first example of efficient gene silencing by siRNA duplexes of
`different lengths was reported by Kim et al. (2005). This group
`tested several dsRNA molecules longer than 19 bp and found
`that 27-bp-long siRNAs could trigger efficient RNAi and were
`also 10–100-fold more potent at gene silencing than the 19-bp-
`long siRNAs, depending upon the target sequence. Notably,
`even though the 27-bp siRNAs were longer than 19-bp ones,
`they did not induce interferon or activate PKR. The authors of
`that study found that the 27-bp duplex was processed effi-
`ciently by Dicer and that the introduction of chemical modi-
`fications that compromised Dicer processing also affected
`gene silencing activity. Based on this observation, it was hy-
`pothesized that the increased potency of 27-bp siRNAs was
`because Dicer was provided with the substrate duplex rather
`than the product (19þ 2 siRNA), thus improving the effi-
`ciency of siRNA entry into the RISC. Thus, this 27-bp-long
`siRNA structure was termed ‘‘Dicer-substrate siRNA.’’
`In a follow-up study, functional polarity was introduced
`into the Dicer-substrate siRNA to define the specific Dicer-
`cleavage product generated from the 27-bp-long siRNA.
`Specifically, the ‘‘R 25D/27’’ structure was developed, which
`consists of a 25-nt sense strand with 2 nucleotides on the 50 end
`replaced with DNA, and a 27-nt antisense strand, resulting in
`a blunt-ended antisense 50 end and a 2-nt antisense 30 over-
`hang (Rose et al., 2005). Compared with other Dicer-substrate
`siRNA structures, the R 25D/27 structure produced defined
`Dicer cleavage product and showed higher gene silencing
`activity. The asymmetric overhang structure also resulted in
`preferential incorporation of the antisense strand into the
`RISC. A mechanism proposed to explain this observation is
`that Dicer preferentially associates with the 30 overhang and
`spatially orients the siRNA cleavage product in the correct
`position for association of the 50 end of the antisense strand
`with Dicer, which remains in the RISC.
`Another long siRNA structural variant was recently intro-
`duced by Salomon et al. (2010). In their report, they described
`blunt-ended, 25-bp-long, chemically modified siRNA du-
`plexes. One of the chemical modification patterns they eval-
`uated consists of 4 20-OMe modifications on both ends of the
`sense strand and an unmodified antisense strand, designated
`as the ‘‘4/4’’ pattern. The striking finding was that while this
`modification rendered the 25-bp siRNA completely resistant
`to Dicer-mediated cleavage, it still triggered efficient target
`gene silencing. Therefore, although structurally similar, the
`proposed mechanism of action of Dicer-noncleavable 25-bp
`siRNA is different
`from that of Dicer-substrate siRNA.
`However, no side-by-side comparison data of the gene si-
`lencing activity of Dicer-noncleavable siRNA vs. Dicer-sub-
`strate 27-bp siRNA duplex were presented in their study.
`Therefore, a direct parallel comparison awaits further studies.
`The finding that the Dicer-noncleaved 25-nt-long antisense
`strand could execute gene silencing via the RNAi machinery
`might seem puzzling at first, because the initial X-ray crys-
`tallography studies proposed a model in which the MID do-
`main and PAZ domain of Ago-2 bind the 50 phosphate and 30
`
`127
`
`hydroxyl ends of the antisense strand, respectively, with an
`optimal interval of 21 nt (Wang et al., 2008). However, a recent
`structural study revealed that in a ternary complex of cata-
`lytically inactive bacterial Ago-2/21 nt antisense DNA/target
`RNA, the antisense DNA base-paired with the target RNA
`only up to the 16th nucleotide from the 50 end, and the 30 end
`of the antisense DNA was released from the PAZ domain
`(Wang et al., 2009; Sashital and Doudna, 2010). This new
`structural finding is consistent with the observation that Di-
`cer-noncleaved 25-nt-long antisense strand can trigger effi-
`cient RNAi. Although further structural and biochemical
`studies are warranted to clearly understand how RNAi is
`triggered by long antisense strand RNA, these results em-
`phasize the structural flexibility of the RNAi machinery.
`Although the Dicer-substrate siRNA did not induce cy-
`tokines by triggering the innate immune response in HeLa
`cells (Kim et al., 2005), a later study by Marques et al. (2006)
`showed that in cell lines that maintain the dsRNA immune
`response such as T98G, blunt-ended, 23–27-bp-long siRNAs
`could trigger the innate immune response, and the level of
`the immune response increased as the duplex length in-
`creased. This innate immune response appears to be sensed
`by RIG-I, a cytoplasmic RNA-sensing receptor, and can be
`attenuated by adding 30 overhang structures. Therefore, al-
`though the 30 overhang structure might be dispensable for
`RNAi activity, it is still important for triggering specific
`RNAi without inducing nonspecific immune responses. In
`view of this, it is notable that the R 25D/27 structure, which
`contains a 30 overhang and a blunt-end with DNA modifi-
`cation, not only showed optimal gene silencing activity, but
`also significantly reduced innate immune stimulation. Thus,
`the R 25D/27 structure appears to be an optimal structure
`for Dicer-substrate siRNA.
`A key conclusion from an earlier study that defined the
`optimal siRNA structure for mammalian gene silencing was
`that dsRNA with a duplex length of 30 bp or longer triggered a
`potent interferon response, which was avoided by using 19-bp-
`long siRNA (Elbashir et al., 2001a). Although longer than the
`classical 19-bp siRNA structure, both Dicer-substrate siRNA
`and Dicer-noncleavable siRNA are still below the 30 bp length
`limit. However, Chang et al. (2009a) recently demonstrated
`that 34–38-bp-long synthetic dsRNA can specifically trigger
`RNAi-mediated gene silencing without induction of the inter-
`feron response or PKR activation. Therefore, the original hy-
`pothesis that RNA duplexes longer than 30 bp could not trigger
`specific RNAi because of the induction of potent antiviral re-
`sponses needs to be revised. This is an important finding, be-
`cause the ability to design long dsRNAs for gene silencing will
`allow researchers to design a variety of complex, multifunc-
`tional siRNA structures. Further, using this longer duplex
`structure, they generated siRNA structural variants that can
`simultaneously knock down the expression of 2 target genes
`(termed as dual-target siRNAs or dsiRNAs). Once again, these
`results demonstrate the structural flexibility of gene silencing
`siRNAs and propose a simple and useful strategy to develop a
`multitarget gene silencing strategy against diseases with mul-
`tiple gene alterations, such as viral infections and cancer.
`
`Gene Silencing by Short siRNAs
`
`Several recent studies also reported successful gene si-
`lencing with siRNA structural variants shorter than 19 bp.
`
`

`

`128
`
`CHANG ET AL.
`
`Chu and Rana (2008) tested the activity of dsRNAs shorter
`than 19 bp in mammalian cells. They claimed that, in human
`cell lines, 16-bp siRNA with 2-nt 30 overhangs triggered gene
`silencing of target mRNAs with greater potency than 19-bp
`siRNA and thus suggested that the minimal requirement for
`siRNA molecule is a *42 A˚ A-form helix with *1.5 helical
`turns (Chu and Rana, 2008). In another study, Li et al. (2009)
`developed a forward genetic approach to identify nontoxic,
`highly potent siRNAs from bacterially delivered RNAi li-
`braries. Interestingly, some of the highly potent siRNAs se-
`lected against MVP mRNA were only 16 bp in length, and
`increasing the length of the native sequences dramatically
`reduced RNAi potency. These results indicate that depending
`upon sequence features, siRNAs shorter than 19 bp can effi-
`ciently trigger target gene silencing in mammalian cells and
`sometimes even better than the classical 19-bp siRNAs.
`Although these studies stated that shorter siRNAs showed
`greater gene silencing potency than 19-bp siRNAs, some of
`the claims should be interpreted with caution. For example,
`Chu and Rana (2008) concluded that 16-bp siRNAs were more
`effective at gene silencing than the corresponding 19-bp siR-
`NAs. However, a recent study demonstrated that when the
`short siRNA was made by trimming the 50 end of siRNA
`duplex with respect to the polarity of the antisense strand, the
`seed sequence that dictated the target site specificity changed,
`thus preventing direct comparison of the activities of the 19-
`bp siRNA and its shorter version. In fact, when short siRNA
`was made by trimming the 30 end of duplex with respect to the
`antisense strand (thus keeping the seed sequence identical),
`the shorter siRNAs were less active than their 19-bp coun-
`terparts.
`It is possible that although dsRNA shorter than 19 bp could
`be recognized by and incorporated into the RNAi machinery,
`the length of the antisense strand must be 19 nt or longer to
`trigger efficient gene silencing. Based upon this idea, several
`different asymmetric, shorter duplex siRNA structural vari-
`ants have been proposed by 3 independent groups (Sano et al.,
`2008; Sun et al., 2008; Chang et al., 2009b).
`In the study by Sano et al. (2008), they reported that the
`siRNA terminus is a crucial factor in strand selection and
`RNAi activity. In agreement with the conclusion drawn by
`Rose et al. (2005), they demonstrated that siRNAs with an
`asymmetric 3’ overhang structure, that is, a 2-nt overhang
`present only at the 30 end of the antisense strand, showed
`better antisense strand selection and enhanced efficacy than
`the symmetric counterparts. Further, they showed that as
`long as the 30 overhang at the end of the antisense strand was
`maintained, partial deletion or DNA substitution of the sense
`strand did not affect gene silencing activity, demonstrating
`that asymmetric siRNA with a shortened sense strand can
`trigger efficient gene silencing. Other groups also reported
`asymmetric shorter duplex siRNA structural designs. Sun
`et al. (2008) showed efficient mammalian gene silencing using
`asymmetric RNA duplexes termed aiRNAs, which are com-
`posed of a 15-nt-long sense strand and a 21-nt-long antisense
`strand, resulting in both 30 and 50 antisense overhangs. Chang
`et al. (2009b) also reported asymmetric shorter duplex siRNA
`structures, termed asiRNA, composed of a 16-nt-long sense
`strand and 19–21-nt-long antisense strand, resulting in a 50
`blunt end and a 30 overhang with respect to the polarity of the
`antisense strand. asiRNA duplexes were incorporated into the
`RISC and mediated sequence-specific cleavage of
`target
`
`mRNA. These results further support the idea that neither
`structural symmetry nor a duplex length of 19 bp needs to be
`maintained for efficient recruitment of the RISC and gene si-
`lencing by siRNAs. Nonetheless, it should be noted that
`asymmetric short siRNA duplexes shorter than 14 bp showed
`severely compromised gene silencing activity (Chu and Rana,
`2008; Chang et al., 2009b), suggesting that a certain minimal
`RNA duplex length is essential for optimal recognition of
`siRNA by the RNAi machinery.
`
`Reduction of Nonspecific Responses
`by Asymmetric Shorter Duplex siRNAs
`
`Sense strand-mediated off-target silencing occurs because
`of the incorporation of the sense strand into RISC. This can
`often lead to unexpected gene silencing ( Jackson et al., 2003;
`Clark et al., 2008). Although it has been shown that siRNA
`strands with a thermodynamically unstable 50 end are pref-
`erentially incorporated into the active RISC (Khvorova et al.,
`2003; Schwarz et al., 2003), many siRNAs show sense strand
`incorporation into the RISC irrespective of their sequence
`features. Because sense strand incorporation into the RISC is
`due to the symmetric nature of classical siRNAs, siRNA
`structural variants with asymmetric designs have been shown
`to successfully ameliorate this nonspecific response. All 3
`asymmetric shorter duplex siRNAs described earlier showed
`reduced sense strand off-target silencing (Sano et al., 2008;
`Sun et al., 2008; Chang et al., 2009b). It is likely that both an
`asymmetric terminal structure and a shortened sense strand
`contribute to the preferential usage of the antisense strand for
`RNAi.
`Besides asymmetric siRNAs, a triple-stranded RNA variant
`of siRNA called small internally segmented interfering RNA
`(sisiRNA) has also been also shown to reduce sense strand off-
`target silencing (Bramsen et al., 2007). sisiRNA is composed of
`an intact antisense strand base-paired with 2 shorter 9–13-nt
`sense strands, debilitating the sense strand for subsequent
`RISC association and target gene silencing. Because these
`segmented sense strands were too short to stably maintain the
`RNA duplex, locked nucleic acid modifications were intro-
`duced to increase the thermal stability of the duplex structure.
`Chang et al. (2009b) also showed that asiRNAs could alle-
`viate the saturation of the endogenous RNAi machinery that
`was observed with the corresponding 19 þ 2 siRNA struc-
`tures. Interestingly, the reduction in the saturation of the
`RNAi machinery was dependent on the duplex length of the
`asiRNAs. Because RNAi machinery saturation by exoge-
`nously added siRNAs could disturb cellular miRNA biogen-
`esis and potentially lead to nonspecific gene expression
`changes (Khan et al., 2009) or even cell death (Grimm et al.,
`2006), the ability of asymmetric shorter duplex siRNAs to
`alleviate RNAi machinery saturation indicates that they have
`a great advantage over classical siRNAs with regard to the
`development of improved functional genomics tools and
`therapeutic modalities.
`The first-in-class siRNA drug to enter clinical trial as a ther-
`apeutic molecule was designed to silence vascular endothelial
`growth factor A (VEGF-A), a gene believed to be largely re-
`sponsible for vision loss in wet age-related macular degenera-
`tion (AMD). Another siRNA targeting VEGF for AMD
`therapeutics was also under phase II trial. However, while the
`trials were being perused, a report by Kleinman et al. (2008)
`
`

`

`SIRNA STRUCTURAL VARIANTS
`
`demonstrated that intraocular injection of classical 19þ 2 siR-
`NA structures induced sequence-independent angiogenesis
`suppression via nonspecific activation of TLR3. In their study,
`nontargeted siRNA suppressed dermal neovascularization in
`mice as effectively as VEGF-A siRNA. This study raised serious
`concerns about the specificity of siRNA drugs and resulted in
`the termination of phase III clinical trials for bevasiranib.
`A computer modeling study suggested a potential interac-
`tion between 21-nt siRNA and TLR3, leading to TLR3 dimer
`stabilization and receptor activation, which explained the re-
`markable length-based discrimination of dsRNAs by TLR3
`(Kleinman et al., 2008). According to this study, dsRNAs shorter
`than 17 bp did not activate TLR3, as they interacted with TLR3
`with a free energy of binding below the threshold for receptor
`activation. This indicates that specific therapeutics targeting
`diseases such as AMD can be developed using asymmetric
`short siRNAs without nonspecific TLR3 activation. Taken to-
`gether, these findings suggest that asymmetric short siRNA
`structures are a superior alternative to classical siRNAs for more
`specific gene silencing with reduced nonspecific effects.
`
`Immunostimulatory siRNA Structural Variants
`
`Although it is generally considered that immune stimula-
`tion by siRNAs should be avoided to achieve specific gene
`silencing, it might be therapeutically beneficial to design im-
`munostimulatory siRNAs that could provoke both an im-
`mune response and gene silencing for antiviral and antitumor
`therapy (Schlee et al., 2006). Indeed, a proof-of-principle study
`of a bifunctional immunostimulatory siRNA approach to
`achieve enhanced antitumor activity was recently published
`(Poeck et al., 2008). In this study, siRNA targeting anti-
`apoptotic BCL2 mRNA with a 50 triphosphate (3p) modifica-
`tion, which is known to induce the immune response by
`activating the RIG-I pathway, was used. The BCL2-targeting
`3p-siRNA induced stronger apoptosis than corresponding
`siRNA without 3p modification in B16 melanoma cell lines in
`vitro and also in a lung metastasis model in vivo.
`Gantier et al. (2010) recently presented a rational design for
`immunostimulatory siRNA backbone structures without any
`chemical modification. They introduced a 4-nt uridine bulge
`in the middle of the Dicer-substrate siRNA structure. This
`structure triggered not only efficient gene silencing but also
`activation of TLR8, which resulted in the induction of various
`proinflammatory cytokines. However, although the addition
`of 4-nt uridine bulge into the 21-nt siRNA structure also po-
`tentiated immunostimulation, it reduced target gene silenc-
`ing, presumably because of the instability of the duplex
`structure. As this uridine-bulge siRNA structure does not
`require a special chemical synthesis steps, it could be poten-
`tially used for large-scale industrial production of
`im-
`munostimulatory siRNAs.
`
`Multimeric siRNA Structural Variants
`with Enhanced Intracellular Delivery
`
`One of the main hurdles in the development of RNAi
`therapeutics is the efficient delivery of siRNAs to target cells/
`tissues (Li et al., 2006). A number of lipid- and polymer-based
`carriers have been developed to facilitate intracellular deliv-
`ery of nucleic acid drugs, including siRNAs, both in vitro and
`in vivo (Aigner, 2007; Whitehead et al., 2009). However, be-
`
`129
`
`cause siRNAs are less charged than gene-encoding plasmids,
`their interactions with cationic liposomes or polymers are
`relatively weaker. Thus, conventional cationic polymer-com-
`plexed siRNAs are easily expelled by negatively charged cell
`surface proteins, compromising the delivery efficiency of
`siRNAs. Although long dsRNA structures might interact
`more strongly with cationic delivery reagents than classical
`dsRNA structures, they could also potentially induce antiviral
`responses (Stark et al., 1998).
`To overcome this limitation, Bolcato-Bellemin et al. (2007)
`designed gene-like siRNAs with short complementary A6-8/
`T6-8 overhangs to form transient concatemers, termed ssiR-
`NAs. When complexed with polyethylenimine (PEI), a cat-
`ionic polymer carrier widely used for gene delivery, the
`concatemer structure was stabilized and formed a strong
`complex with PEI. This design increased siRNA-PEI complex
`stability and protected the siRNA from nuclease attack, re-
`sulting in enhanced delivery efficiency (up to 10-fold). Simi-
`larly, Lee et al. (2010) and Mok et al. (2010) constructed
`multimeric siRNA polymers with reducible disulfide bond
`linkages between the siRNAs; these multimeric siRNA poly-
`mers are cleaved into 19-bp siRNA molecules upon internal-
`ization. Like ssiRNAs, these multimeric siRNA structures also
`showed increased silencing efficiency because of the en-
`hanced delivery and protection from nuclease attack afforded
`by complexation with PEI. These results indicate that multi-
`meric siRNA structures are potential gene silencing thera-
`peutics that have increased delivery efficiency compared with
`monomeric siRNA structures.
`
`Conclusion and Perspectives
`
`In this review, we have summarized the recent develop-
`ments in siRNA structural variants and discussed the ad-
`vantages of these variants compared with classical 19 þ 2
`siRNA structures. The added benefits of using siRNA
`structural variants, such as improved gene silencing effi-
`ciency, alleviation of off-target gene silencing, reduction in
`RNAi machinery saturation, reduction in innate immune
`responses, and enhanced delivery, justifies the efforts made
`to identify novel gene-silencing siRNA structures. The in-
`creasing diversity of siRNA structural variants reported
`clearly indicates that a great deal of mechanistic flexibility
`exists in the mammalian RNAi machinery. Further bio-
`chemical, structural, and molecular studies into the mecha-
`nisms of action of these siRNA structural variants will shed
`light on the detailed mechanistic features of the RNAi ma-
`chinery and allow the design of more efficient siRNAs. Ad-
`vancements in the design and application of siRNA
`structural variants will not only inform RNAi biology but
`also provide new tools to combat various diseases with im-
`proved safety and efficacy.
`
`Acknowledgment
`
`This work was supported by a Global Research Laboratory
`grant from the Ministry of Education, Science, and Technol-
`ogy (MEST) of Korea (No. 2008-00582).
`
`Author Disclosure Statement
`
`No competing financial interests exist.
`
`

`

`130
`
`References
`
`AIGNER, A. (2007). Nonviral in vivo delivery of therapeutic
`small interfering RNAs. Curr. Opin. Mol. Ther. 9, 345–352.
`AMARZGUIOUI, M., HOLEN, T., BABAIE, E., and PRYDZ, H.
`(2003). Tolerance for mutations and chemical modifications in
`a siRNA. Nucleic Acids Res. 31, 589–595.
`BOLCATO-BELLEMIN, A.L., BONNET, M.E., CREUSAT, G.,
`ERBACHER, P., and BEHR, J.P. (2007). Sticky overhangs en-
`hance siRNA-mediated gene silencing. Proc. Natl. Acad. Sci.
`U. S. A. 104, 16050–16055.
`BRAMSEN, J.B., LAURSEN, M.B., DAMGAARD, C.K., LENA,
`S.W., BABU, B.R., WENGEL, J., and KJEMS, J. (2007). Im-
`proved silencing properties using small internally segmented
`interfering RNAs. Nucleic Acids Res. 35, 5886–5897.
`CAPLEN, N.J., FLEENOR, J., FIRE, A., and MORGAN, R.A.
`(2000). dsRNA-mediated gene silencing in cultured Droso-
`phila cells: a tissue culture model for the analysis of RNA
`interference. Gene 252, 95–105.
`CHANG, C.I., HONG, S.W., KIM, S., and LEE, D.K. (2007). A
`structure-activity relationship study of siRNAs with structural
`variations. Biochem. Biophys. Res. Commun. 359, 997–1003.
`CHANG, C.I., KANG, H.S., BAN, C., KIM, S., and LEE, D.K.
`(2009a). Dual-target gene silencing by using long, synthetic
`siRNA duplexes without triggering antiviral responses. Mol.
`Cells 27, 689–695.
`CHANG, C.I., YOO, J.W., HONG, S.W., LEE, S.E., KANG, H.S.,
`SUN, X., ROGOFF, H.A., BAN, C., KIM, S., LI, C.J., and LEE,
`D.K. (2009b). Asymmetric shorter-duplex siRNA structures
`trigger efficient gene silencing with reduced nonspecific ef-
`fects. Mol. Ther. 17, 725–732.
`CHIU, Y.L., and RANA, T.M. (2002). RNAi in human cells: basic
`structural and functional features of small interfering RNA.
`Mol. Cell. 10, 549–561.
`CHU, C.Y., and RANA, T.M. (2008). Potent RNAi by short RNA
`triggers. RNA 14, 1714–1719.
`CLARK, P.R., POBER, J.S., and KLUGER, M.S. (2008). Knock-
`down of TNFR1 by the sense strand of an ICAM-1 siRNA:
`dissection of an off-target effect. Nucleic Acids Res. 36, 1081–
`1097.
`CZAUDERNA, F., FECHTNER, M., DAMES, S., AYGUN, H.,
`KLIPPEL, A., PRONK, G.J., GIESE, K., and KAUFMANN, J.
`(2003). Structural variations and stabilising modifications of
`synthetic siRNAs in mammalian cells. Nucleic Acids Res. 31,
`2705–2716.
`ELBASHIR, S.M., HARBORTH, J., LENDECKEL, W., YALCIN,
`A., WEBER, K., and TUSCHL, T. (2001a). Duplexes of 21-nu-
`cleotide RNAs mediate RNA interference in cultured mam-
`malian cells. Nature 411, 494–498.
`ELBASHIR, S.M., LENDECKEL, W., and TUSCHL, T. (2001b).
`RNA interference is mediated by 21- and 22-nucleotide RNAs.
`Genes Dev. 15, 188–200.
`ELBASHIR, S.M., MARTINEZ, J., PATKANIOWSKA, A., LEN-
`DECKEL, W., and TUSCHL, T. (2001c). Functional anatomy of
`siRNAs for mediating efficient RNAi in Drosophila melano-
`gaster embryo l