| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
REVIEW |
1 Faculdade de Ciências Farmacêuticas de Ribeirão Preto, Universidade de São Paulo, Ribeirão Preto, SP, Brazil
2 Department of Pharmaceutical Sciences, University of Tennessee Health Science Center, Memphis, Tennessee 38163, USA
 |
ABSTRACT |
|---|
Keywords: RNA interference; small interfering RNA; bioconjugation; chemical modification; complex formation
|
INTRODUCTION |
|---|
). The process of RNAi consists of an initiator step, in which long double-stranded RNA (dsRNA) is cleaved into siRNA fragments, and an effector step, in which these fragments are incorporated into a protein complex, dissociated, and used as a guiding sequence to recognize homologous mRNA that is subsequently cleaved. RNAi is considered a self-defense mechanism of eukaryotic cells to combat infection by RNA viruses and transposons. It is also assumed to tightly regulate protein levels in response to various environmental stimuli (Meister and Tuschl 2004).
The revolutionary finding of RNAi resulted from the work of Andrew Fire and coworkers at the Carnegie Institute in Washington, D.C., who demonstrated in 1998 that injection of dsRNA into Caenorhabditis elegans leads to efficient sequence-specific gene silencing (Fire et al. 1998). At that time, the state of the art in gene silencing was the use of antisense oligodeoxynucleotide (ODNs), which comprise single strands of short DNA or RNA complementary sequences that hybridize with the targeted mRNA (Mahato et al. 2005). However, the dsRNA seemed to induce silencing through a pathway distinct from classical antisense therapies due to the catalytic nature of RNAi, in which one siRNA can be used over and over to guide the cleavage of many mRNA molecules (Dykxhoorn et al. 2003). Bertrand et al. (2002) have compared the effects of antisense ODNs and siRNAs targeting green fluorescent protein (GFP) in vitro and in vivo. siRNA was quantitatively more efficient, and its effect lasted for a longer time in cell culture. In mice, siRNAs were able to silence gene expression, whereas no effect was observed in the presence of antisense ODNs.
The first evidence that siRNAs can mediate sequence-specific gene silencing in mammalian cells was provided in 2001 when the conversion of dsRNA into short RNA fragments was shown to be bypassed by the transfection of siRNA molecules into cells (Elbashir et al. 2001a). After that, various in vivo effects of siRNA and short hairpin RNA (shRNA) have been reported (Lewis et al. 2002; McCaffrey et al. 2002; Xia et al. 2002; Song et al. 2003b). For example, McCaffrey et al. (2002) showed that siRNA and shRNA reduce luciferase (Luc) expression in the liver in a sequence-specific manner. Song et al. (2003b) found that siRNA targeted to the Fas receptor protects mice from liver fibrosis.
Human clinical trials of RNAi-based drugs are currently under way by Acuity Pharmaceuticals and Sirna Therapeutics. Both companies are working on intravitreal administration of siRNA targeting vascular endothelial growth factor (VEGF), whose overexpression is the primary cause of age-related macular degeneration (AMD). The first results are encouraging in terms of tolerability of siRNA compounds. Other clinical trials involving siRNA for treating chronic myeloid leukemia and respiratory syncytial virus infection are also being carried out by the Hadassah Medical Organization and Alnylam Pharmaceutics, Inc., respectively.
Despite early excitement, siRNAs have shown to activate immune response in a sequence- and concentration-dependent manner, leading to nonspecific gene silencing (Jackson et al. 2003; Sioud 2005). In this respect, the introduction of chemical modifications and generation of designed siRNAs have become essential for achieving high gene silencing with a low degree of undesired effects. Chemical modifications of siRNAs also appear to stabilize these molecules in serum and show enhanced gene silencing (Braasch et al. 2003; Chiu and Rana 2003; Layzer et al. 2004). A breakthrough in this area has recently been achieved by linking cholesterol (Chol) to siRNAs (Soutschek et al. 2004), and by encapsulating them into stable nucleic acid lipid particles (Zimmermann et al. 2006). Undoubtedly, effective in vivo delivery of siRNAs will be a key factor in turning siRNA into a new class of therapeutics. In this review, we discuss several strategies to abrogate undesired effects and improve the efficiency of siRNA, including chemical modifications, siRNA sequence design, bioconjugation, complex formation, and encapsulation into lipid particles.
|
MECHANISM OF RNA INTERFERENCE |
|---|
; Zamore and Haley 2005). Processed siRNA is incorporated into a protein complex, termed RNA-induced silencing complex (RISC). Dicer is also involved in the early steps of RISC formation and may be required for siRNA entry into RISC (Lee et al. 2004). In mammalian cells, the protein Argonaute 2 (Ago2) is the catalytic component of RISC that cleaves target mRNAs (Liu et al. 2004b; Meister et al. 2004). In the RISC assembly process, siRNAs are initially loaded into Ago2 as duplexes, and then Ago2 cleaves the passenger strand, thereby liberating the guide strand from the siRNA duplex and producing active RISC capable of cleaving target mRNAs (Matranga et al. 2005; Leuschner et al. 2006). The guide strand serves as a template for the recognition of homologous mRNA, which upon binding to RISC is cleaved between bases 10 and 11 relative to the 5' base of the guide siRNA strand by the catalytic activity of Ago2 (Elbashir et al. 2001c). The template siRNA is not affected by this reaction, so the RISC can undergo numerous cycles of mRNA cleavage that comprise the high efficiency of RNAi (Elbashir et al. 2001b; Meister and Tuschl 2004). The mechanism by which passenger strands are cleaved by Ago2 follows the same rules established for the siRNA-guided cleavage of a target mRNA (Leuschner et al. 2006).
|
). The mature miRNA is subsequently incorporated into the silencing complex (RISC), where it mediates translational repression. During RISC assembly, the Ago2-mediated cleavage of the passenger strand is blocked by mismatches between the guide and passenger strands of miRNAs. In this case, a slower "backup" pathway dissociates and destroys the passenger strand, liberating active RISC (Matranga et al. 2005; Leuschner et al. 2006). In contrast to siRNA, miRNAs mediate RNAi through a translational inhibition mechanism involving imperfect complementarity to target mRNAs (Jackson et al. 2003; Bartel 2004).
One major difference between mammalian RNAi and RNAi in other eukaryotes is the lack of an amplification system for long-term persistence of RNAi in mammalian cells. For example, in Drosophila, 
35 molecules of dsRNA can silence 1000 copies of a target mRNA per cell and can persist over the course of many generations (Nykanen et al. 2001). In contrast, in mammalian cells RNAi persists effectively only for an average of 66 h due to its dilution during cell divisions (Chiu and Rana 2002).
|
METHODS OF siRNA PRODUCTION |
|---|
|
). The chemical synthesis of siRNA requires several steps including the generation of two homologous strands, annealing of the strands, addition of chemical entities to increase stability, and ensuring that 2-nt overhangs are present. The siRNA duplex requires a 3'-hydroxyl group and a 5'-phosphate group for functional activity (Dykxhoorn et al. 2003). Unlike DNA, RNA possesses an additional hydroxyl group at the 2' position of each ribose building block, which destabilizes RNA under the basic conditions generally present in DNA synthesis reactions. Hence, the most difficult step in RNA synthesis is the simultaneous protection of the 5'- and 2'-hydroxyl groups during solid-phase chemistry.
One drawback of using chemically synthesized siRNA is that the most effective target sequence is unpredictable since gene silencing efficiency may vary depending on the segments of the transcripts that are targeted. For example, Holen et al. (2002) observed that only a few siRNAs resulted in a significant reduction of human tissue factor (HTF) expression after targeting its mRNA with several siRNAs synthesized against different sites of the same mRNA. The results suggested that accessible siRNA target sites may be rare in some human mRNAs. However, more effective gene silencing can be achieved by targeting different segments of the same transcript simultaneously. Significant enhancement in gene silencing was achieved by targeting the HIV-1 coreceptor CXCR4 and the apoptosis-inducing Fas ligand with two or more siRNAs against different sites of the same mRNA (Ji et al. 2003).
In vitro transcription and enzymatic digestion
Since chemically synthesized siRNAs are expensive, attempts are being made to produce siRNAs by in vitro transcription using T7 RNA polymerase. Synthetic DNA templates containing the T7 RNA polymerase promoter region followed by desired RNA sequence can be produced using an automated DNA synthesizer and then amplified using PCR. The T7 polymerase binds to the promoter sequence, initiates transcription, and then moves along the template strand toward its 5' end, elongating the RNA transcript as it goes. Although a termination region on the DNA may be used to halt transcription, the runoff transcription, which stops when the 5' end of the DNA template strand is reached, is also commonly used. Transcription of the PCR fragment by the T7 RNA polymerase produces both sense and antisense RNAs, which spontaneously anneal, forming full-length dsRNA (Milligan and Uhlenbeck 1989; Donze and Picard 2002).
The in vitro transcription approach is limited by specific sequence requirements related to T7 polymerase. The last guanosine of the T7 promoter is invariably the first ribonucleotide that is incorporated into the RNA by the T7 polymerase during transcription. Therefore, all siRNAs produced by this method start with a 5'-G residue and require a C-3' residue at position 19 (i.e., 5'-G-N17-C-3') to allow annealing with the complementary RNA, which also has to start with a 5'-G residue. Furthermore, since the efficacy of siRNAs targeted to different regions of a gene varies dramatically, the number of sequences that can be targeted using siRNAs generated by this method is limited (Donze and Picard 2002). Kim et al. (2004) demonstrated that siRNAs synthesized from the T7 polymerase system can trigger a potent induction of interferon 
and 
in a number of cell lines. The induction of these interferons was also seen by short single-stranded RNAs transcribed with T3, T7, and Sp6 RNA polymerases. These investigators further demonstrated that the presence of triphosphates on the 5' end of T7 transcripts is required for interferon induction, since the treatment of T7 transcripts with phosphatase could abrogate this effect.
Sohail et al. (2003) reported an alternative approach for producing a desired siRNA sequence using T7 polymerase in vitro transcription followed by transcript digestion by deoxyribozymes, which are known as "DNAzymes" or catalytic DNA. The cleavage of RNA by DNAzymes occurs between two specific nucleotides, and the requirement for this dinucleotide sequence is different for different DNAzyme groups, making them flexible tools for digesting a variety of sequences (Breaker and Joyce 1994; Feldman and Sen 2001). The siRNAs produced by this method caused dose-dependent inhibition of type 1 insulin-like growth factor receptor in human breast cancer cells comparable to that induced by chemically synthesized siRNAs of the same sequence (Sohail et al. 2003).
Both recombinant human Dicer (re-hDicer) and Escherichia coli RNAse III are used to cleave long dsRNA produced by in vitro transcription. Treatment with E. coli RNase III or re-hDicer randomly cleaves the RNA into a population of siRNA molecules that are effective mediators of gene silencing in a manner similar to that observed when using synthetic siRNA. Usually, the cleavage products of E. coli enzyme RNase III range from 12 to 15 base pairs (bp) in length with termini identical to those produced by Dicer (Amarasinghe et al. 2001). Although these short dsRNAs are not long enough to trigger an RNAi in mammalian cells, the average product length generated by RNase III digestion can be increased by altering digestion conditions. Yang et al. (2002) have shown that 20–25-bp siRNAs produced by RNase III digestion efficiently inhibited various endogenous genes in different mammalian cell lines without nonspecific effects. In addition, RNase III can digest dsRNA faster than re-hDicer and can be easily overexpressed and purified (Amarasinghe et al. 2001; Bernstein et al. 2001). The siRNAs produced by re-hDicer are 21–23 bp in length and contain 2-nt 3' overhangs with 5'-phosphate and 3'-hydroxyl termini, which are essential for RNAi activity. Dicer-generated siRNAs have been shown to be effective in silencing exogenous and endogenous gene expression in mammalian cells (Kawasaki et al. 2003; Myers et al. 2003).
The enzymatic synthesis of siRNA may provide more effective gene silencing than chemically synthesized siRNA, as the enzymatically generated siRNA can correspond to sequences overlapping the entire gene. However, complex siRNA populations may have a higher probability of targeting other genes, and therefore promoting nonspecific effects, than do discrete siRNAs. Additionally, enzymatic synthesis of siRNA requires separation of siRNA from uncleaved RNA duplexes and residual nucleic acids (Yang et al. 2002; Myers et al. 2003).
Endogenous expression
The application of synthetic siRNAs is restricted by both low-to-moderate transfection efficiency and the short-term persistence of transient gene expression. A single transfection of siRNA may not provide a sufficient window of functional depletion for proteins with long half-lives. Another potential problem inherent in chemically synthesized siRNA is variability in transfection efficiency, especially in difficult-to-transfect cells. To circumvent these limitations, expression vectors currently in use employ siRNA or shRNA expression cassettes that resemble pre-miRNAs and undergo processing by Dicer. Like synthetic siRNAs, they are designed to pair perfectly with the target mRNA to induce RNAi. These shRNAs are designed for either transient or long-term gene silencing and can be produced from plasmid or viral expression vectors (Amarzguioui et al. 2005).
Plasmid vectors
shRNA, siRNA, and miRNA can be produced from plasmid vectors containing promoters that are dependent on either RNA polymerase (Pol) II or Pol III. Among them, Pol III promoters are used most frequently because it is possible to express small RNAs that carry the structural feature of siRNA. Figure 2 is a schematic representation of different strategies used to create expression cassettes using RNA polymerase promoters for generation of siRNA, shRNA, and miRNA. In the first strategy, the sense and antisense strands are expressed as two independent transcripts that hybridize within the cells to form functional siRNA duplexes (Fig. 2A). In the second strategy, the sense and antisense strands are expressed as a single transcript separated by a short loop of 4–10 nt of sequence. The transcript forms a hairpin structure that can be processed by Dicer into functional siRNA (Fig. 2B). In the third strategy, miRNAs that are complementary to the target gene are expressed using the Pol II promoter (Fig. 2C; Dykxhoorn et al. 2003; Sioud 2004).
|
; Paddison et al. 2002). In addition to these two promoters, other Pol III promoters such as 5S, 7SK, and tRNA promoters have also been used for expressing siRNAs endogenously (Paul et al. 2002; Czauderna et al. 2003b; Kawasaki and Taira 2003). The activities of these promoters vary from cell type to cell type. To optimize shRNA expression, it is beneficial to create expression vectors with at least two different promoters and transfect them into the cells being targeted for gene knockdown. Construction of shRNA expression vectors poses two serious technical challenges. First, it is difficult to sequence constructs that contain a hairpin region, probably because of the tight palindromic structure. Second, 20%–40% of constructs get mutated within the hairpin region (Brummelkamp et al. 2002; Paddison et al. 2002; Paul et al. 2002).
Viral vectors
The introduction of siRNA expression plasmids into cells often requires electroporation, microinjection, or complex formation with synthetic carriers (lipids, polymers, or peptides). While most rapidly dividing cell lines are easily transfected using shRNA expression plasmids, these plasmid vectors are not easily transfected into primary cells, stem cells, and nondividing cells. In the absence of cell division, the siRNA expression plasmids cannot be introduced into the nucleus, where the DNA is transcribed. To overcome this limitation, different viral vectors encoding shRNA including retroviral, adenoviral, and adeno-associated viral (AAV) are being developed. Typically, these vectors use a Pol III promoter, such as U6, H1, or transfer RNA promoters (Dykxhoorn et al. 2003).
Retroviral vectors have been reported to mediate an efficient and stable siRNA expression (Rubinson et al. 2003; Liu et al. 2004a). Unlike Moloney murine leukemia virus (MoMLV), lentiviral vectors efficiently integrate into the genome of nondividing cells, such as pancreatic islets, hematopoietic stem cells, or terminally differentiated cells. A lentiviral vector encoding shRNA has been shown to effectively silence GFP, BCL-2, and Interleukin (IL) 12 receptor (CD25) genes (Tiscornia et al. 2003; Schomber et al. 2004; Wong et al. 2005). Lentiviral vectors encoding shRNA have also been shown to inhibit HIV-1 infection in hematopoietic stem cells and human CD4+ T-cells (Li et al. 2003; Nishitsuji et al. 2006). However, lentiviral vectors are associated with infrequent insertional mutagenesis. Consequently, the use of adenoviral vectors is being explored since these vectors do not integrate into the host genome and efficiently transduce both dividing and nondividing cells (Wu et al. 2003). Adenoviruses contain a linear double-stranded DNA genome that remains episomal after infection. Recombinant adenoviral vectors containing expression cassettes of interest are readily generated and can be purified to very high titers (up to 1013 infection units/mL) (Huang and Kochanek 2005).
A variety of properties make AAV vectors an interesting tool for organ-directed shRNA expression, including the lack of pathogenicity, and ease of vector production. The AAV-2 type is highly prevalent in the human population and frequently neutralized by antibodies. Therefore, it is important to evaluate different AAV serotypes for organ-directed shRNA expression. AAV-8 vectors expressing shRNA are reported to transduce almost 100% of hepatocytes after intravenous injection into mice (Grimm and Kay 2006). However, overexpression of shRNA caused length- and dose-dependent liver injury and ultimate death. Morbidity was associated with the down-regulation of liver-derived miRNA, indicating possible competition between miRNA and shRNA for nuclear Krayopherin exportin 5 (Grimm et al. 2006). Therefore, it is important to optimize siRNA dose, length, and sequence to avoid oversaturation of endogenous small RNA pathways.
Inducible expression vectors
Plasmid- and viral-vector-based constitutive expressions of shRNAs by RNA Pol III U6 and H1 promoters often fail to adjust the levels of gene silencing necessary for cell survival and growth. Besides, gene silencing for longer periods may result in nonphysiological responses. This problem can be circumvented by generating inducible regulation of RNAi (Czauderna et al. 2003b; Wiznerowicz and Trono 2003; Gupta et al. 2004). Ohkawa and Taira (2000) described the successful regulation of gene silencing by the integration of the bacterial tetracycline operon sequence (tetO) into the U6 promoter. Matsukura et al. (2003) have applied this tetracycline-inducible U6 promoter for in vivo transcription of shRNA, which enables stable transfection followed by conditional expression of shRNA. The main drawback of the tetracycline-inducible system is a relatively high background expression in the uninduced state in certain cell lines (No et al. 1996; Van Craenenbroeck et al. 2001). In contrast, the ecdysone-inducible system is tightly regulated, with no expression in the uninduced state and a rapid inductive response (Gupta et al. 2004).
Owing to the time-consuming process of cloning siRNA into plasmid constructs and the need for verification of the cloned sequence, an easier approach of screening sequences involving the production of siRNA expression cassettes (SECs) was developed (Castanotto et al. 2002). SECs are PCR-derived siRNA expression templates introduced directly into the cells. SECs consist of a Pol III promoter, a sequence encoding an shRNA, and an RNA polymerase III termination site. The final result is a PCR product that contains a Pol III promoter, a DNA sequence that, once transcribed, forms shRNA, and a terminator sequence. After transcription, the shRNA is intracellularly cleaved into siRNA by the RNAi machinery. By incorporating restriction sites at their ends, SECs effectively elicit gene silencing and can be cloned into a plasmid to create an siRNA expression vector. The expediency and low cost of this procedure lends itself to mass screening of siRNA libraries as well as identification of siRNA target sites (Chang 2004; Wooddell et al. 2005).
|
BARRIERS TO RNAi-BASED THERAPIES |
|---|
Immune response activation
Figure 3 is a schematic representation of the immune response activation by siRNAs. Non-immune and immune cells get activated in the presence of long dsRNA, leading to the activation of cytoplasmic receptors such as the dsRNA-dependent protein kinase R (PKR) and the retinoic acid-inducible gene-I (RIG-I) (Saunders and Barber 2003; Yoneyama et al. 2004). PKR is activated by autophosphorylation following binding to dsRNA. Once activated, it phosphorylates the eukaryotic translation initiation factor (EIF-2)-, leading to the global suppression of protein biosynthesis and subsequent programmed cell death. PKR can also activate nuclear factor 
B (NF-B) with consequent induction of type-I IFN production. A family of 2'-5'-oligoadenylate synthetases (2'5'-OAS) is also activated by dsRNA. This leads to the activation of RNase L, which eventually triggers the nonspecific degradation of mRNA (Marques and Williams 2005).
|
; Marques and Williams 2005). The helicase RIG-I is necessary for recognizing blunt-ended siRNAs, leading to the activation of dsRNA signaling. RIG-I unwinds siRNAs containing blunt ends more efficiently than siRNAs containing 3' overhangs. The efficiency in duplex unwinding is then translated into signaling downstream to interferon regulatory factor 3 (IRF-3) and NF-B activation (Marques et al. 2006).
Mammalian immune cells express a family of toll-like receptors (TLRs), which recognize pathogen-associated molecular patterns including unmethylated cytosine-guanine motifs (commonly known as CpG motifs) and viral dsRNA. TLR3, the receptor for dsRNA, was a logical candidate for recognizing siRNA in the context of immunostimulation. Indeed, TLR3 overexpressed in cultured human embryonic kidney (HEK) 293 cells was capable of recognizing siRNAs (Marques and Williams 2005). However, the activation of immune cells by siRNAs is sequence dependent, and either sense or antisense strands individually can induce inflammatory cytokine production as efficiently as duplex siRNAs (Judge et al. 2005; Sioud 2005, 2006). Thus, as TLR3 neither recognizes single-stranded RNA (ssRNA) nor requires sequence specificity, it is unlikely involved in activation of the immune system by siRNAs (Marques and Williams 2005; Sioud 2005).
TLR7 and TLR8 were initially shown to mediate the recognition of RNA viruses and small synthetic antiviral compounds referred to as imidazoquinolines. TLR7, TLR8, and TLR9 are expressed in endosomes and require endosomal maturation for efficient signaling (Marques and Williams 2005). siRNA recognition by TLR7, TLR8, and TLR9 results in activation of NF-B and IRFs, which induce inflammatory cytokines and IFNs, respectively (Fig. 3; Marques and Williams 2005). Interestingly, not all ssRNA molecules are capable of activating TLR7 and TLR8, but, rather, U- and G-rich ssRNAs seem to be preferentially recognized (Sioud 2006). It has now become clear that TLR7 and TLR8 mediate the recognition of siRNAs in a sequence-dependent manner (Heil et al. 2004; Hornung et al. 2005; Judge et al. 2005).
Recognition of siRNAs by TLRs takes place in the endosome, before the siRNAs enter the cytoplasm. Therefore, it is expected that if siRNAs could enter the cytoplasm avoiding the endosome, they should bypass the activation of immune systems but still mediate gene silencing. Robbins et al. (2006) tested the immunostimulatory effects of lipid-delivered siRNAs versus Pol III promoter-expressed shRNAs in primary CD34+ progenitor-derived hematopoietic cells. They observed that in this system, cationic lipid/siRNA complexes are potent inducers of IFN- and type-I IFN gene expression, whereas the same sequences when expressed endogenously are nonimmunostimulatory.
The method of siRNA delivery also has a great influence on the immunostimulatory activity of siRNAs (Fig. 3). Injection of siRNAs after complex formation with cationic liposomes into mice induces the release of inflammatory cytokines, including IL-6, tumor necrosis factor- (TNF-), and IFN- (Judge et al. 2005; Sioud 2005). In contrast, injection of naked, unmodified siRNAs or siRNAs conjugated to cholesterol has no significant effect on immunostimulation (Heidel et al. 2004; Marques and Williams 2005). There are two possible explanations for the absence of immunostimulation by naked siRNAs. First, unmodified siRNAs have a short half-life in serum and may be degraded before being recognized by specific receptors. Second, cationic lipid/siRNA complexes are more readily recognized by immune cells than are naked siRNAs. In fact, Sioud (2005) reported that naked siRNAs containing a phosphorothioate (PS) backbone were not immunostimulatory. However, these results should be interpreted carefully considering that naked CpG ODNs are known to be immunostimulatory independent of backbone modifications (Marques and Williams 2005).
Off-target effects and nonspecific gene silencing
For its therapeutic applications, siRNA must not cause any unintended effect other than sequence-specific gene silencing. However, recent studies indicate that there are off-target effects associated with the use of siRNA (Jackson et al. 2003). Off-target effects consist of any gene silencing effect caused by siRNAs in nontarget mRNAs through the RNAi mechanism. Generally, it occurs due to lack of complementarity between siRNAs and target mRNAs. An explanation for the fact that siRNAs can induce silencing of nontarget genes can be found in the RNAi machinery. Although the actual substrate specificity of individual siRNAs appears to be very high, RNAi machinery can tolerate single mutations located in the center of the siRNA molecule without losing the ability to induce gene silencing. It makes some siRNAs able to silencing other than those genes related to their homologous mRNA. However, complete inactivation of the RNAi mechanism can occur if four or more mutations are introduced in the guide strand, making the active RISC unable to recognize the target mRNA (Jackson et al. 2003; Persengiev et al. 2004; Sioud 2004).
siRNAs and miRNAs were found to be functionally interchangeable. siRNAs can act as miRNAs depending on the degree of complementarity with the target mRNA. If synthetic siRNAs bear a sufficiently low degree of complementary bases, target mRNA translation occurs instead of mRNA degradation, whereas miRNAs will lead to mRNA degradation if perfect complementarity with target mRNA exists. This process is one of the reasons for the off-target effects of siRNAs (Jackson et al. 2003; Bartel 2004). Another reason for the occurrence of off-target effects is that not only the antisense but also the sense strand of an siRNA can direct gene silencing of nontarget genes, and it has been documented to occur when as few as 15 bp of complementarity exists between the siRNA and mRNA (Jackson et al. 2003). Birmingham et al. (2006) also demonstrated that the presence of one or more perfect matches between the hexamer or heptamer seed (positions 2–7 or 2–8 of the antisense strand) of an siRNA and the 3'-untranslated region (UTR), but not the 5'-UTR or open reading frame, is associated with off-targeting. A high proportion of "off-target" transcripts silenced by siRNAs have been shown to have 3'-UTR sequence complementarity to the seed region of the siRNA, since the base mismatches within the siRNA seed region reduced the set of original off-target transcripts. Since there is no algorithm that can eliminate significant numbers of 7–8-base matches of siRNAs to the transcriptome, it will be difficult to achieve perfect specificity (Jackson et al. 2006). These findings suggest a strong mechanistic parallel between siRNA off-targeting and miRNA-mediated gene regulation.
Nonspecific gene silencing is those effects caused by siRNAs through any pathway other than RNAi. Semizarov et al. (2003) investigated the overall cellular effects of siRNAs on transcription levels. In a concentration-dependent manner, siRNAs nonspecifically interfered in the expression of a significant number of genes, many of which are known to be involved in apoptosis and stress response. Persengiev et al. (2004) reported that treatment with siRNA may result in diverse cellular activities, such as cell signaling, cytoskeletal organization, gene expression, metabolism, and cell adhesion. Jackson et al. (2003) attributed the nonspecific effects to cross-hybridization of transcripts containing regions of partial homology with the siRNA sequence. However, the nonspecific effects observed by Persengiev et al. (2004) cannot be explained by off-target regulation because the siRNAs tested lacked significant sequence similarity to any human gene. Moreover, for all of the genes tested, expression was affected by two siRNAs with sequences that are completely unrelated. Bridge et al. (2003) also reported that shRNAs can affect the expression of many genes, including several IFN targets. The activation of sequence-independent inhibition of gene expression by siRNA or shRNA-expression vectors seems to be, at least in part, due to the induction of type-I IFN, PKR, and signaling through TLR3 (Kariko et al. 2004; Marques and Williams 2005).
|
BIODISTRIBUTION OF siRNA |
|---|
siRNA duplexes are much more stable than ssRNAs. However, since they are similar to single-stranded antisense oligonucleotides, they still tend to degrade on incubation with serum, contributing to their short half-lives in vivo. 125I-labeled siRNAs are widely distributed in the body after intravenous injection into mice, with preferential accumulation in the liver and kidney (Braasch et al. 2004). Radiolabeled siRNA is also distributed to the heart, spleen, and lung. However, very little siRNA is detected in the brain (Braasch et al. 2004; Soutschek et al. 2004; Santel et al. 2006). Intravenous administration of naked siRNA in rats showed a short half-life of 6 min and a clearance of 17.6 mL/min. The poor pharmacokinetic properties of siRNAs are related to their low in vivo stability, since they can be degraded by endogenous RNases. Another reason is their fast elimination by kidney filtration because of their small molecular mass (7 kDa) (Soutschek et al. 2004). Consequently, improving the pharmacokinetic properties of siRNA duplexes is an important goal for developing siRNA-mediated gene silencing.
|
STRATEGIES FOR IMPROVING RNAi-BASED THERAPIES |
|---|
). Backbone modifications and bioconjugation with lipids and peptides are known to improve the stability and cellular uptake of siRNAs (Chiu and Rana 2003; Lorenz et al. 2004). Several reagents have also been employed for systemic administration of siRNAs as an alternative to the covalent modifications of the siRNA molecule (Simeoni et al. 2003; Morrissey et al. 2005b; Urban-Klein et al. 2005; Santel et al. 2006). Cationic liposomes are routinely used for transfection of nucleic acids into mammalian cells. Positively charged carriers, such as protamine-antibody fusion proteins, polyethyleneimine (PEI), and cyclodextrin-containing polycation, have been tested for complex formation with siRNAs. To achieve cell-specific delivery of siRNAs, several targeting ligands, including cell membrane receptors and antibodies, have been explored (Song et al. 2005).
Design of siRNAs
A thorough understanding of the sequence, size, and structural requirements of siRNAs is essential to effectively mediate RNAi. Table 2 summarizes the most important siRNA design rules to improve gene silencing and avoid undesired effects.
|
; Paddison et al. 2002; Reynolds et al. 2004). Several sequence motifs are consistent with effective siRNAs, including AAN19TT, NAN19NN, NARN17YNN, and NANN17YNN (where N is any nucleotide, R is a purine, and Y is a pyrimidine). In principle, any region of the mRNA can be targeted. However, it should be pointed out that complementarity is often found in the 3'-UTR of the off-target sequences. Furthermore, sequence complementarity between the 5' end of the guide strand and the mRNA is known to be the key to off-target silencing. Additionally, intronic sequences must be avoided since RNAi is a cytoplasmic process (Dykxhoorn et al. 2003).
Ui-Tei et al. (2004) proposed that highly effective RNAi can be achieved if siRNA satisfying the following four sequence conditions at the same time is used: (1) AU richness in the 5'-terminal, 7-bp-long region of the antisense strand; (2) G/C at the 5' end of the sense strand; and (3) the absence of any long GC stretch of >9 bp in length. Even though Elbashir et al. (2002) reported that the target region should be at least 50–100 nt downstream from the start codon, Dykxhoorn et al. (2003) suggest that there is a predisposition for effective siRNA-directed silencing toward the 3' portion of the gene. Moreover, sequences with even representation of all nucleotides on the antisense strand are favored, and those regions with stretches of a single nucleotide, especially G, should be avoided as G-rich oligonucleotides have a tendency to form quartets. Reynolds et al. (2004) analyzed 180 siRNAs targeting every other position of two 197-base regions of firefly luciferase and human cyclophilin B mRNA in a total of 90 siRNAs per gene. A two-base shift in the target position was sufficient to significantly alter siRNA activity, suggesting that its functionality is determined by the siRNA-specific properties. Most potent siRNA has a G/C content ranging from 36% to 52%. siRNA duplexes with a GC content >52% may have difficulty in dissociating into passenger and guide strands, while siRNAs with a GC content <30% may interact less well with the mRNA recognition site (Dykxhoorn and Lieberman 2006).
Low internal stability of the siRNA at the 5' antisense end is a prerequisite for effective silencing and probably important for duplex dissociation after assembly with RISC (Schwarz et al. 2003). The asymmetric RISC formation model proposed by Schwarz et al. (2003) predicts that duplex siRNA dissociation preferably occurs at an "easier" duplex end, possessing A:U, G:U, or unpaired bases at its 5'-end position and being thermodynamically less stable. Consequently, the end of the siRNA with the lower thermodynamic stability will favor selection of the antisense strand into RISC as the guide strand. The importance of thermodynamically unstable or flexible base pairs at or near the guide strand for siRNA duplex dissociation in HEK 293 cells has also been pointed out by Khvorova et al. (2003).
The immunostimulatory activity of an siRNA duplex is, at least in part, a function of its nucleotide sequence. The GU-rich sequence seems to be a common feature among the immunostimulatory motifs: UGUGU and GUCCUUCAA are present in the immunostimulatory ssRNAs. A single U-to-C base substitution in the 5'-UGUGU-3' siRNA sequence substantially reduces the siRNA-mediated induction of IFN- by human peripheral blood mononuclear cells. In contrast, the opposite substitution forming a 5'-UGU-3' motif renders the siRNA immunostimulatory activity. The sequence recognition mechanism is stringent enough that minimal base substitutions can have profound effects on the immunostimulatory capacity of the siRNAs. However, it is important to note that several immunostimulatory siRNAs do not contain either of these motifs, whereas others are not stimulatory despite being U- and G-rich (Heil et al. 2004; Judge et al. 2005).
Targeting the human TLR-9 gene with a pool of siRNAs in plasmacytoid dendritic cells (PDC) (also called IFN producing cells), Hornung et al. (2005) found that nine bases at the 3' end (5'-GUCCUUCAA-3') of the sense strand of siRNA are responsible for its immunostimulatory activity. The immunostimulation and gene silencing were shown to be two independent functional activities. Regardless of the GU content, it seems that TLR7 and/or TLR8 can also recognize dsRNA with similar efficacy (Heil et al. 2004). Indeed, Hornung et al. (2005) argued that the mechanism involved in single-strand siRNA recognition by TLRs is also responsible for recognizing the immunostimulatory motifs within the siRNA duplexes. Nonetheless, to date, there are only two siRNA motifs known to effectively activate innate immunity in the context of double-strand siRNA to a level comparable to that obtained with free single-strand RNA. siRNA duplexes seem to be less effective than siRNA in activating innate immunity. The intracellular receptors for single-strand siRNA, in particular TLR8 and TLR7, do not effectively recognize most immunostimulatory RNA motifs in the context of siRNA duplexes (Sioud 2006).
A number of academic and commercially affiliated Web-based softwares have been developed to assist researchers in the identification of efficient siRNA sequences (Naito et al. 2004; Yuan et al. 2004). Since the rules that govern efficient siRNA-directed silencing are still unknown, researchers seeking to silence gene expression should synthesize several siRNAs to a gene and validate empirically the efficiency of each one. To ensure that the chosen siRNA sequence targets a single gene, a search of the selected sequence should be carried out against sequence databases. This can be done using the Smith–Waterman algorithm or the Basic Local Alignment Search Tool (BLAST) located at the National Center of Biotechnology Information Web site (http://www.ncbi.nlm.gov). Sequences in these databases that share partial homology with siRNAs might be targeted for silencing by the siRNAs. Potential off-target effects of the siRNAs might be minimized by choosing an siRNA with maximal sequence divergence from the list of genes with partial sequence identity to the intended mRNA target.
siRNA size
Synthetic RNA duplexes of 25–30 nt in length (more specifically, 27 nt), which are Dicer substrates, have been shown to be up to 100-fold more potent than corresponding conventional 21-nt siRNAs (Kim et al. 2005a). The use of 27-nt dsRNAs, also called Dicer substrate dsRNA (disRNA), allows targeting of some sites within a given sequence that are refractory to suppression with traditional 21-nt siRNAs. The use of disRNAs to trigger RNAi can result in enhanced and prolonged gene silencing at lower concentrations than those required for conventional 21-nt siRNAs applications (Kim et al. 2005a). It is important to note that in vitro Dicer cleavage of 27-nt disRNAs before transfection does not significantly enhance gene silencing efficiency. Additionally, it has been shown that chemically synthesized shRNAs that are Dicer substrates are more potent inducers of RNAi than conventional siRNAs. Moreover, a two-base 3' overhang directs Dicer cleavage (Siolas et al. 2005). The enhanced potency of the longer duplexes is attributed to the fact that they are substrates of the Dicer endonuclease, directly linking the production of siRNAs for incorporation into the RISC (Kim et al. 2005a).
The size of siRNA also plays an important role in the activation of immune response in a sequence-independent manner. Although it was initially postulated that a minimum of 30 nt was necessary to activate dsRNA signaling through PKR activation, shorter siRNAs, as few as 21–23 nt, are capable of binding and activating PKR in vitro (Marques et al. 2006). In addition, Hornung et al. (2005) observed that 12-nt ssRNAs containing the immunostimulatory motif (GUCCUUCAA) were poor inducers of IFN- in PDCs. However, the investigators observed that the motif identified needs to be part of a longer ODN sequence, at least 19 bases, to induce IFN- activation. The approach reported by Kim et al. (2005a), which consists of Dicer substrate 27-nt dsRNA with increased RNAi potency relative to conventional 21-nt siRNAs, may facilitate the use of lower concentrations of duplex RNA. Therefore, since the undesired effects can occur in an siRNA dose-dependent manner, the use of Dicer substrate 27-nt dsRNA may be an alternative to overcome this problem without losing efficacy.
siRNA structure
According to Chiu and Rana (2002), the A-form helix of the guide strand-mRNA duplex is required for the mechanism of RNAi. Chemical modifications disrupting the functional groups of the major groove of the A-form helix formed by the guide strand and its mRNA specifically at the cleavage site inhibit RNAi. The major groove is also required for promoting RNA–RISC interactions that subsequently lead to mRNA cleavage (Chiu and Rana 2002, 2003).
The 2'-OH of the ribonucleotide of RNAs is required for the nucleophilic attack during the hydrolysis of the RNA backbone, the reaction catalyzed by degradative RNases (Chiu and Rana 2003). However, 2'-OH is not required for RNAi-mediated degradation, and, even more specifically, it is not required for nucleotide base-pairing with nucleotides lining the mRNA cleavage site (Chiu and Rana 2003). Additionally, Sioud (2006) has shown that the 2'-OH of uridines is required for immune recognition and signaling of siRNAs, being responsible, at least in part, for activation of the immune response.
New design approaches that improve the performance of long dsRNAs as Dicer substrates have been reported (Kim et al. 2005a). These new approaches include 25–30-nt asymmetric dsRNAs with a 5' blunt end and a 2-nt 3' overhang on the other end. The improved efficacy of disRNAs is postulated to result from their recognition and cleavage by Dicer, followed by their subsequent incorporation into the RISC complex. This interpretation is supported by observations that Drosophila Dicer is not only instrumental in handing over siRNA to nascent RISC, but it is itself a component of the latter (Lee et al. 2004). Providing the RNAi machinery with a Dicer substrate presumably results in more efficient incorporation of the active 21-nt siRNA strand into RISC. These modifications also direct the way Dicer processes the dsRNA substrate, resulting in a more homogeneous pool of siRNA products.
The extremities of the siRNA duplex are critical determinants of its capacity to trigger immune response activation. A blunt structure at the 3'end is the strongest terminal structure for promoting activation of dsRNA signaling through the PKR pathway, followed by a 5'overhang (Marques et al. 2006). In contrast, 3'overhangs, normally present in endogenous Dicer products such as miRNAs, allow RNAi to proceed without activation of dsRNA-dependent PKR (Marques et al. 2006). The presence of the 3' overhangs also precludes activation of dsRNA signaling by siRNAs and reveals an important basis for discrimination between self and nonself dsRNAs. Interestingly, this type of discrimination appears to take place after RIG-I binds to dsRNAs (Marques et al. 2006).
Chemical modifications
Chemical modifications to sugars, backbones, or bases of oligonucleotides can be used to control their pharmacokinetic profiles and reduce nonspecific effects without affecting their biological activity. The valuable know-how of backbone modifications of antisense ODNs can be readily adapted to develop new siRNA technologies (Marshall and Kaiser 2004). Although siRNA duplexes used for silencing are inherently more stable than ssRNAs, there are reasons to chemically modify one or both strands. Apart from improving stability and reducing off-target effects, chemical modifications can aid in broadly targeting siRNAs into cells and tissues, and certain conjugates can enhance uptake in specific cell types. Multiple chemical modifications can be introduced at various positions with the siRNA duplexes. Figure 4 shows the most common chemical modifications introduced in siRNAs.
|
; Mahato et al. 2005).
Braasch et al. (2003) reported that PS siRNAs containing 12–21 PS substitutions per strand were stable during extended incubation in fetal calf serum media, but their stability was not higher than unmodified siRNA duplexes. The investigators also monitored the inhibition of human caveolin-1 (hCav) expression by PS siRNAs. The modified siRNAs were able to silence the hCav gene, but less inhibition was observed by fully modified siRNAs. The nuclear uptake of PS siRNAs was greater than that of unmodified siRNAs, which can explain their reduced activity. Although backbone modifications did not reduce the silencing activity of siRNAs targeting the HTF gene in human keratinocytes, most extensively PS siRNAs proved to be cytotoxic, resulting in 70% cell death compared with mock-transfected cells (Amarzguioui et al. 2003). The introduction of PS linkages to siRNA yields mixed results for siRNA distribution to various organs. In another study, Braasch et al. (2004) observed an increase in the distribution of siRNAs to spleen, heart, and lung, while the distribution to kidney and liver decreased. However, the effects were modest, suggesting that the introduction of PS linkages may not play a major role in determining the distribution of siRNA.
An alternative backbone modification that confers increased biological stability to nucleic acids is the boranophosphonate linkage. In boranophosphonate ODNs, the nonbridging phosphodiester oxygen is replaced with an isoelectronic borane (–BH3) moiety. The charge distribution of boranophosphonate ODNs also differs from that of normal phosphate and phosphorothioate ODNs. Thus, it changes the polarity and increases the hydrophobicity of the molecule. While boranophosphonate modification has been less widely studied, it has shown more advantages than PSs. Targeting GFP in HeLa cells, Hall et al. (2004) observed that the activity of boranophosphonate siRNAs exceeds that of PS siRNAs, and it was often greater than that of unmodified siRNAs, irrespective of the base or strand modified. However, boranophosphonate modifications placed at the center of the antisense strand reduced RNAi activity. Boranophosphonate siRNAs modified at minimal levels also showed improved stability over unmodified siRNAs against nuclease degradation. Additionally, increasing the number of modifications yielded more stable siRNAs.
2'-Sugar modifications
Modifications of RNA at the 2'-position of the ribose ring have been shown to increase siRNA stability against endonucleases and reduce immune response activation. These modifications include 2'-O-methyl (2'-OMe), 2'-deoxy-2'-fluoro modifications, and locked nucleic acid.
Fluoro and methyl linkages
The siRNA motif consisting of 2'-OMe and 2'-fluoro nucleotides has enhanced plasma stability and increased in vivo potency. The 2'-OMe sugar modification retains the canonical right-handed A-form helical geometry, which is required for siRNA activity. This modification has also been shown to increase the nuclease resistance of ODNs and siRNA duplexes (Chiu and Rana 2003). The effect of 2'-OMe modification has been found to be dependent on both position and extent of incorporation. siRNAs with fully 2'-OMe-substituted sense strands were functional when the duplexes were of 20-bp blunt construction, but not for canonical duplexes of 19-bp constructs with 3' overhangs (Kraynack and Baker 2006). In contrast, siRNAs with alternating 2'-OMe and unmodified nucleotides (Czauderna et al. 2003a), or alternating 2'-OMe or 2'-O-fluoro nucleotides (Allerson et al. 2005), had activity similar to that of the unmodified duplexes. This suggests that minimal chemical modifications are compatible with siRNA function. Jackson et al. (2006) reported that 2'-OMe modifications to specific positions within the siRNA seed region reduce both the number of off-target transcripts and the magnitude of their regulation, without significantly affecting silencing of the intended targets. Current evidence sugg
TOP
ABSTRACT
INTRODUCTION
