REVIEWS
`
`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`Chemical modification of siRNA plays an essential role in moving siRNA toward the clinic.
`
`JONATHAN K. WATTS
`
`Jonathan K. Watts received his BSc in chemistry from
`Dalhousie University in Halifax, Canada. He has just
`finished his PhD in the group of Masad Damha at
`McGill University. He has been awarded a Natural
`Sciences and Engineering Research Council of Canada
`(NSERC) postgraduate fellowship, the McGill Tom-
`linson Fellowship and a postdoctoral fellowship from
`FQRNT Quebec. His primary research interest is the
`interface of chemistry and biology in the field of small
`RNAs.
`
`GLEN F. DELEAVEY
`
`Glen F. Deleavey received his BSc in biology–
`chemistry from the University of New Brunswick in
`Fredericton, Canada. He is currently a PhD candidate
`in the group of Masad Damha at McGill University. He
`has been awarded a postgraduate fellowship from
`NSERC and a 2007 SCI Merit Award (Canadian
`section). His research interests lie in the field of
`chemical biology, with a focus on chemically modified
`siRNAs.
`
`MASAD J. DAMHA
`
`Masad J. Damha received his BSc and PhD degrees
`from McGill University, the latter in the group of Prof.
`Kelvin Ogilvie, on synthesis and conformational ana-
`lysis of RNA and its analogues. After beginning his
`academic career at the University of Toronto, he
`returned to McGill in 1992, where he is currently
`James McGill Professor of Chemistry. His research
`interests include synthesis of RNA (including novel
`RNA structures) and the application of oligonucleo-
`tide derivatives as therapeutics. He was awarded the
`2007 Bernard Belleau Award from the Canadian
`Society for Chemistry, honoring significant contri-
`butions to the field of medicinal chemistry, for the
`0
`development of 2
`F-ANA.
`
`Chemically modified siRNA:
`tools and applications
`
`Jonathan K. Watts, Glen F. Deleavey and Masad J. Damha
`
`Department of Chemistry, McGill University, 801 Sherbrooke Street West, Montreal, QC H3A 2K6, Canada
`
`Chemical modification provides solutions to many of the challenges
`facing siRNA therapeutics. This review examines the various siRNA
`modifications available, including every aspect of the RNA structure and
`siRNA duplex architecture. The applications of chemically modified siRNA
`are then examined, with a focus on specificity (elimination of immune
`effects and hybridization-dependent off-target effects) and delivery. We
`also discuss improvement of nuclease stability and potency.
`
`Introduction
`It has been ten years since the publication of the seminal paper demonstrating the high potency
`of long double-stranded RNA in gene knockdown [1]. Shortly thereafter, it was discovered that the
`same effect could be produced in mammalian cells using synthetic short RNA duplexes [2]. The
`relatively few years since then have seen an explosion of research into therapeutic applications of
`RNAi. Several companies have been formed to pursue the technology, and transactions involving
`these companies have recently been measured in the billions of dollars.
`The reason for the excitement is that RNAi allows potent knockdown of virtually any gene. This
`in turn allows rapid progression from target selection to preclinical trials. siRNA has become the
`most common tool in functional genomics, and therefore can often also help at the target
`identification stage. Furthermore, some targets that are not druggable by traditional methods can
`be targeted by gene knockdown.
`In spite of the immense attractiveness of gene knockdown as a therapeutic strategy, siRNA
`duplexes are not optimal drug-like molecules. RNA is highly vulnerable to serum exo- and endo-
`nucleases, leading to a short half-life in serum. siRNA duplexes are composed of two strands that
`can drift apart in a dilute environment-like serum. Because oligonucleotides are polyanions they
`do not easily cross cell membranes and, because this charge density leads to extensive hydration,
`they do not easily interact with albumin and other serum proteins, leading to rapid elimination.
`Unmodified oligonucleotides have limited tissue distribution. And finally, oligonucleotides can
`have off-target effects, either by stimulating the immune system or by entering other endogenous
`gene regulation pathways.
`A wide variety of chemical modifications have been proposed to address these issues. In this
`review, we examine the principles of chemical modification of siRNA duplexes. We will briefly
`look into the toolbox; that is, summarize the possible ways that siRNA duplexes can be modified.
`
`Reviews 
`
`KEYNOTE REVIEW
`
`Corresponding authors: Watts, J.K. (jonathan.watts@mail.mcgill.ca), Damha, M.J. (masad.damha@mcgill.ca)
`
`Alnylam Exh. (cid:20)(cid:19)(cid:21)(cid:26)
`
`842 www.drugdiscoverytoday.com
`
`1359-6446/06/$(cid:1) -(cid:1) see(cid:1) front(cid:1) matter(cid:1)ß(cid:1) 2008(cid:1) Elsevier(cid:1) Ltd.(cid:1) All(cid:1) rights(cid:1) reserved.(cid:1) doi:10.1016/j.drudis.2008.05.007
`
`

`

`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`REVIEWS
`
`ReviewsKEYNOTEREVIEW
`
`duplex is then assembled into the RNA-induced silencing complex
`(RISC), a multiprotein complex including Argonaute2 (AGO2),
`Dicer, TRBP (HIV-1 TAR RNA-binding protein) and PACT (a
`dsRNA-binding protein), as well as other proteins, some of which
`are yet unknown [3]. The strand with the lower binding affinity at
`0
`its 5
`-end becomes the antisense (guide) strand [6,7], and the other
`strand (known as the sense or passenger strand) is cleaved and
`unwound, to leave a single-stranded RNA associated with Argo-
`naute2 (AGO2), an endonuclease at the heart of RISC that pro-
`motes location of complementary mRNA, hybridization and
`cleavage of the mRNA target. When modifying an siRNA duplex
`it
`is
`important
`to remember
`that different modification
`approaches are required for the sense and antisense strands,
`because of these very different roles [8,9].
`In most cases it is simply assumed that the RNAi mechanism
`is unaffected by chemical modification of siRNA duplexes. A few
`studies using modified siRNA have confirmed this by showing
`that the cleavage of complementary mRNA occurs between
`0
`bases 10 and 11, counting from the 5
`-end of the guide strand,
`as is the case for unmodified duplexes [10–12]. However,
`in principle, this should be verified for each new pattern of
`modification.
`
`Toolbox
`siRNA duplexes have been chemically modified in a wide variety of
`ways. Some of the results in the literature, however, seem to
`contradict one another, or to work on one system but not another.
`This field is still very young, and it will take time for the more
`robust and universal modifications to be recognized as such. In the
`meantime, it is useful to have many options so that at least one of
`the chemistries can be used to modify an siRNA without compro-
`mising its potency.
`In this section, we will briefly review the most significant siRNA
`modifications in the literature, drawing attention to those that
`have proven most useful and robust up to now. Our goal in this
`section is not to explore the advantages of each modification in
`detail but simply to present all the known possibilities in a
`straightforward way. In the second half of this review we will
`explore how these modifications can help move siRNA toward the
`clinic. We hope that by examining the ‘toolbox’ of chemical
`modifications separately from the ‘task list’ of required properties
`we will inspire creative new combinations and applications.
`
`Sugar modifications
`The most widely used siRNA modifications are on the sugar moiety
`(Fig. 2). One of the earliest studies on chemically modified siRNA
`0
`-
`showed that, while A-form duplex structure is important, the 2
`0
`OH is not required for active siRNA [13]. Therefore, the 2
`position
`has been extensively modified.
`0
`-O-Methylation of RNA increases binding affinity and nucle-
`2
`0
`-O-Me-RNA can be well-tolerated
`ase stability, and the resulting 2
`throughout the duplex, making it one of the most popular and
`versatile siRNA modifications. Many groups have found that large
`0
`-O-Me modifications (in either strand) decrease
`numbers of 2
`siRNA activity [13–16], but others have found that fully modified
`0
`-O-Me sense strands are functional [11,17]. Kraynack and Baker
`2
`0
`-O-Me modifica-
`attribute these differences to their finding that 2
`tions work best in blunt-ended duplexes [11], but at least one
`
`www.drugdiscoverytoday.com 843
`
`GLOSSARY
`
`RNA interference (RNAi) An evolutionarily conserved
`cellular mechanism for gene knockdown found in fungi,
`plants, and animals, in which double-stranded RNA (dsRNA)
`triggers the specific cleavage of complementary mRNA
`molecules via endogenous cellular machinery.
`Short interfering RNAs (siRNAs) These triggers of RNAi are
`0
`-
`dsRNAs that typically contain 19–21 bp and 2-nt 3
`overhangs. siRNAs are naturally produced by Dicer-mediated
`cleavage of larger dsRNAs, but may also be introduced into
`cells exogenously.
`microRNAs (miRNAs) Endogenous small non-coding RNAs
`that play an important role in the regulation of many genes.
`Precursor miRNAs (pri- and pre-miRNAs) contain hairpin
`structures, often with bulged regions, which are processed
`into duplexes resembling siRNAs.
`Short hairpin RNAs (shRNAs) ssRNA molecules that fold into
`hairpin-shaped structures containing dsRNA stems and a
`ssRNA loop. While shRNAs resemble naturally occurring pre-
`miRNAs, shRNA typically refers to exogenous RNA molecules
`introduced into cells, or produced within cells by introducing
`exogenous DNA.
`Antisense (guide) strand The strand of an siRNA duplex that
`is loaded into the RISC complex and which guides RISC to
`complementary mRNA.
`Sense (passenger) strand The strand of an siRNA duplex
`that is not loaded into the RISC complex. It should have the
`same sequence as the target mRNA.
`0
`end of the antisense
`Seed region A 6–7-nt region at the 5
`RNA strand (from nucleotides 2–7 or 2–8). The seed region is
`especially important for mRNA target recognition, and
`complementarity to the seed region is often sufficient to
`reduce gene expression through a miRNA-type mechanism.
`Innate Immune Response A nonspecific cellular response to
`foreign material. The innate immune response is often
`associated with the production of cytokines.
`Cytokines A class of signaling proteins and glycoproteins
`including interferons (INF), interleukins (IL) and tumor
`necrosis factors (TNF). They are stimulated as part of an
`innate immune response, and activate further immune
`system responses.
`Toll-like receptors (TLRs) A class of pattern-recognition
`receptors that recognize molecular structures generally
`associated with pathogens. Human TLR3 can recognize
`dsRNA, and human TLR7 and TLR8, primarily considered
`ssRNA receptors, can also be stimulated by siRNA. Stimulation
`of TLRs leads to activation of innate immune responses.
`Photocaging Temporarily blocking the activity of a drug by
`appending a photolabile group to it. The group can later be
`cleaved by treatment with light, activating the drug.
`
`Following this, we will review the ways these tools have been
`applied to move siRNA toward the clinic, including the use of
`chemical modifications to improve potency, serum stability, spe-
`cificity and delivery. We will point out the most useful and
`universal modifications as well as some of the most creative
`modifications and applications, which stretch our paradigms
`and open new avenues of research into RNAi-based drugs.
`Much excellent work has led to significant growth in under-
`standing the mechanism of RNAi [3–5] (Fig. 1, also see Glossary of
`specialist terms). When an exogenous 19–21 bp siRNA is intro-
`0
`-end is phosphorylated. The
`duced into a mammalian cell the 5
`
`

`

`REVIEWS
`
`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`group found that, even in this context, activity is greatly reduced
`0
`-O-Me modification [16].
`by heavy 2
`0
`-modifications are not well-tolerated in
`In general, bulkier 2
`siRNA duplexes, except small numbers of modifications on the
`0
`termini. Davis et al. demonstrated that bulky 2
`-substituents cor-
`relate with poor activity as siRNA sense strands and, for the same
`reasons, excellent activity as ‘anti-miRNA oligonucleotides’ (to
`block miRNAs) [18].
`0
`0
`-O-MOE) modification has been
`-O-methoxyethyl (2
`Thus, the 2
`0
`-overhangs of an siRNA targeting the pain-related
`used in the 3
`cation-channel P2X3, and resulted in successful gene targeting in
`0
`-O-MOE modifications could
`vivo [19]. Another group found that 2
`be included in the sense strand, especially at the termini, but not in
`0
`-O-allyl-modifications
`the antisense strand [20]. Similarly, 2
`caused a reduction in activity at most positions in the duplex,
`0
`-overhangs [21].
`but can be used in the 3
`0
`-OH groups in both strands are
`siRNAs in which 70% of the 2
`0
`-O-DNP)
`converted at random into 2,4-dinitrophenyl ethers (2
`show a variety of improved properties, including higher binding
`affinity, nuclease resistance and potency [22].
`Instead of a hydroxyl, alkoxy or aryloxy substituent, functional
`0
`0
`-position. 2
`F-RNA is one of
`siRNAs can contain fluorine at the 2
`0
`F-RNA mod-
`the best-known siRNA modifications, and partial 2
`ification is tolerated throughout the sense and antisense strands
`0
`F-RNA siRNAs are also active
`[13,14,23], and some fully modified 2
`0
`F-RNA-modified siRNA duplexes have
`significantly
`[24]. 2
`0
`F-RNA also increases the binding
`increased serum stability [25]. 2
`affinity of the duplex.
`0
`F-RNA gives
`Changing the stereochemistry of the fluorine of 2
`F-ANA, which was originally developed as a DNA mimic [26,27].
`2
`0
`F-ANA is also
`Considering this fact, it is somewhat surprising that 2
`well-tolerated in siRNA duplexes, including fully modified sense
`strands and partial modification of the antisense strand [28,29].
`0
`F-RNA, it binds with high affinity and increases
`Like its epimer 2
`nuclease stability.
`DNA itself, a ‘modification’ containing no electronegative sub-
`0
`, can also be accepted within siRNA duplexes. For
`stituent at 2
`0
`-overhangs has been well-known
`example, use of DNA in the 3
`since the earliest days of synthetic siRNA research [2] and DNA can
`also be tolerated, in limited numbers, within the base-paired
`region of an siRNA duplex [9,12,15]. An antisense strand made
`0
`F-RNA pyrimidines is functional
`entirely of DNA purines and 2
`0
`F-RNA strongly favors a northern sugar pucker and A-
`[13], since 2
`form helical structure, and presumably directs the conformation of
`0
`-deoxynucleotides. Substitution with dsDNA in
`the more flexible 2
`0
`-end of the guide strand gives active
`the 8-bp region at the 5
`duplexes with reduced off-target effects [12].
`S-RNA is a high-
`The ring oxygen has also been modified: 4
`affinity modification that gives a significant advantage in nuclease
`0
`S-RNA is very well tolerated near the termini of siRNA
`stability. 4
`duplexes [30–32]. In the antisense strand some loss of potency was
`0
`-O-Me-RNA at the same
`observed, but not as much as with 2
`0
`-end of the antisense strand could be mod-
`positions [30]. The 5
`0
`S-RNA inserts without significant loss of potency
`ified with a few 4
`
`0
`
`0
`
`844 www.drugdiscoverytoday.com
`
`FIGURE 1
`The mechanism of RNAi in human cells. The largest of the ellipses signifies
`AGO2, the catalytic engine of RISC.
`
`Reviews  KEYNOTE REVIEW
`
`

`

`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`REVIEWS
`
`ReviewsKEYNOTEREVIEW
`
`FIGURE 2
`0
`Sugar units that have been successfully used to modify siRNA duplexes. Top row: 2
`0
`0
`0
`-modifications (2
`F-RNA, 2
`F-ANA, DNA), 4
`Bottom row, from left to right: other 2
`modification (LNA).
`
`0
`0
`0
`0
`-O-DNP) modifications.
`-O-aryl (2
`-O-allyl) and 2
`-O-MOE, 2
`-O-Me, 2
`-O-alkyl (2
`0
`0
`0
`-modifications (4
`S-RNA, 4
`S-FANA) and a conformationally constrained
`
`0
`
`[31]. The center of the antisense strand cannot be modified with
`0
`S-RNA without significant loss of potency [30,31]. A strand
`4
`0
`-thioribonucleotides on each
`architecture consisting of four 4
`0
`-end of the antisense
`end of the sense strand and four at the 3
`strand worked consistently well against two target genes in three
`0
`0
`0
`-O-Me and 2
`-O-
`S-RNA with 2
`cell lines [32]. Combinations of 4
`MOE modifications at the termini of both strands showed excel-
`lent potency and serum stability [31].
`0
`0
`positions, has a
`and 4
`S-FANA, with modifications at the 2
`4
`northern, RNA-like conformation [29]. It has also shown siRNA
`0
`0
`S–2
`F-ANA is
`activity at various positions in both strands. When 4
`0
`F-ANA sense strand
`in the antisense strand it shows synergy with 2
`modifications [29]. However, its low-binding affinity makes it
`suitable only for a small number of modifications in a duplex.
`The conformationally constrained nucleotide LNA [33] has also
`been included in siRNA [14,34–37]. Its conformational rigidity
`
`0
`
`leads to significant increases in binding affinity. Careful placement
`of LNA in siRNA duplexes has led to functional duplexes of various
`types. The most common sites of modification are the termini of
`0
`-overhangs of the antisense strand
`the sense strand [37] and the 3
`[34,36]. Minimal modification of most internal positions of the
`antisense strand is also tolerated [14,34], but heavier modification
`of the antisense strand is tolerated only in combination with a
`segmented sense strand (described in more detail below) [35].
`
`Phosphate linkage modifications
`Several variations on the phosphodiester linkage are also accepted
`by the RNAi machinery (Fig. 3). Phosphorothioate (PS) linkages
`can be used, with comparable [21,23] or lower potency [13,38] to
`that of native siRNA. Some groups have found that PS linkages are
`not accepted at the center of the duplex, especially at the scissile
`phosphate [39]. However, the ability to accept fully modified PS
`
`FIGURE 3
`Internucleotide linkages used in siRNA. The phosphodiester linkage can be modified as a phosphorothioate or boranophosphate, which retain the negative
`0
`0
`,5
`-linked DNA (X=H) or RNA (X=OH) can be used in the sense strand.
`charge of a phosphate, or a neutral amide-linked RNA can be used at select positions. Either 2
`
`www.drugdiscoverytoday.com 845
`
`

`

`REVIEWS
`
`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`F-ANA
`
`0
`conjunction with sugar modifications such as DNA, 2
`and LNA.
`Surprisingly, some atypical base structures have been used in
`siRNA. A difluorotoluyl base, which has the same shape as thymine
`but cannot form hydrogen bonds, can replace uracil at single
`positions throughout an siRNA duplex, without significantly
`decreasing the potency or changing the mechanism of cleavage
`[45]. A nonaromatic base, dihydrouracil, can also be used, but
`because it cannot contribute to base stacking it lowers the binding
`0
`-end of the duplex,
`affinity of the duplex and is best placed at the 5
`as defined by the antisense strand [43]. Bulky or cationic base
`modifications have not been well-tolerated [9].
`
`Modifications to the overhangs and termini
`Early studies showed that the ideal siRNA consisted of 21-nt in
`0
`-overhangs [15]. These overhangs
`both strands, including 2-nt 3
`can be modified in various ways: from the beginning deoxy units
`0
`overhangs to reduce costs and
`have often been used in the 3
`0
`-exonucleases [2]. However, RNA
`possibly increase resistance to 3
`units also work well [46], most of the chemistries reviewed above
`work well in the overhangs, and blunt-ended siRNAs have also
`0
`-
`been used [16]. Blunt-ended duplexes are more resistant to 3
`exonucleases, and one study reported a greater tolerance to che-
`mical modifications in combination with blunt-ended duplexes
`0
`-
`[11]. However, they are more immunogenic than duplexes with 3
`overhangs [47].
`The termini of the strands can also be modified. Chemical
`0
`-end of the antisense strand helps ensure
`phosphorylation of the 5
`high potency, and is often necessary when the strand is modified.
`0
`-phosphorylation of the sense strand can be blocked
`By contrast, 5
`without loss of activity [16,48], and can in fact be beneficial [49].
`Furthermore, various groups can be conjugated to the ends of
`siRNA duplexes, especially the termini of the sense strand. These
`groups can include an inverted abasic end cap [50,51], which helps
`with exonuclease stability. Conjugation of fluorescent dyes [23] or
`biotin [52] has allowed important biophysical/biochemical stu-
`dies. Finally, conjugation of membrane-penetrating peptides [53]
`and lipophilic groups including steroids and lipids [10,54] has
`helped with delivery (see the section Delivering siRNA). An intri-
`guing recent study showed that including 5–8 dA and dT units on
`0
`-ends of the strands can lead to reversible concatemerization
`the 3
`through these sticky ends, which in turn leads to higher efficiency
`delivery in complex with PEI [55].
`0
`-end of the antisense strand
`The general consensus is that the 5
`is most sensitive to modifications [16,52,56] and does not tolerate
`most of the above-mentioned modifications. However, one group
`observed that fluorescein could be conjugated to any of the
`0
`-end of the antisense strand [23]. As long
`termini except the 3
`0
`-phosphate is present, attaching a group to it does
`as an antisense 5
`not necessarily eliminate RNAi activity [57].
`
`Modifications to the duplex architecture
`The duplex architecture itself can be modified through chemical
`synthesis (Fig. 5). While most siRNA duplexes are made up of two
`strands, it has been shown that an siRNA made of three strands (an
`intact antisense strand with two 9–13 nt sense strands) can reduce
`off-target effects and increase potency; the resulting duplex is
`termed small internally segmented interfering RNA (sisiRNA)
`
`strands may depend on strand architecture [11]. Some cytotoxicity
`has been observed with extensive PS modification [21]. PS mod-
`ifications do not appear to have a major effect on biodistribution of
`siRNA [40].
`siRNAs with boranophosphate linkages are functional and show
`increased potency relative to PS-modified siRNAs and often to
`native siRNAs as well but, to maximize potency, the center of the
`antisense strand should be unmodified [38]. Boranophosphate
`siRNAs provide a significant increase in nuclease stability over
`native RNA [38].
`0
`0
`-RNA) can substitute
`,5
`-DNA or 2
`,5
`,5
`-linkage (either 2
`A 2
`0
`0
`,5
`linkage, but only in the sense strand of the
`for the native 3
`duplex and with some reduction in potency [41]. A nonionic
`0
`-overhangs of siRNA
`amide linkage has been used in the 3
`duplexes [42].
`
`0
`
`0
`
`0
`
`0
`
`Base modifications
`Use of modified bases in siRNA has been somewhat more limited,
`but there have been several examples (Fig. 4). Modified bases that
`stabilize A-U base pairs (5-Br-Ura and 5-I-Ura instead of uracil, and
`diaminopurine instead of adenine) were tolerated in siRNA
`duplexes, although their activity was somewhat reduced [13]. 4-
`Thiouracil has also been used [9]. 2-Thiouracil [43,44] and the C-
`linked base pseudouracil [43,44] increase binding affinity and can
`be used to increase potency and specificity if placed appropriately
`within the duplex (see the sections: Increasing potency and Mod-
`ulating immunostimulatory activity). 5-Methylation of pyrimidines
`(i.e. use of T and 5-Me-C instead of U and C) is common in
`
`FIGURE 4
`Nucleobase modifications used in siRNA. R represents the ribose phosphate
`backbone. Top two rows: 5-halouracils, 2-thiouracil, pseudouracil and
`diaminopurine increase the strength of A-U base pairs. Bottom row: highly
`atypical modified bases including a base that cannot form hydrogen bonds
`(difluorotoluene) and a non-aromatic base (dihydrouracil).
`
`846 www.drugdiscoverytoday.com
`
`Reviews  KEYNOTE REVIEW
`
`

`

`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`REVIEWS
`
`ReviewsKEYNOTEREVIEW
`
`FIGURE 5
`0
`0
`direction, and the antisense strand is on the bottom, in the 3
`to 3
`Functional siRNA architectures. The sense strand is always shown on top, in the 5
`Note that three of the structures (25/27mer, hairpin and dumbbell) require the activity of Dicer before incorporation into RISC.
`
`0
`
`0
`
`to 5
`
`direction.
`
`[35]. Functional siRNA can also be made from just one strand, in
`one of the various ways. Hairpin-type duplexes, made from a single
`strand, can be introduced exogenously [58] or expressed within a
`cell [59,60]. Closing the other end of the hairpin results in a
`dumbbell or nanocircle which retains RNAi activity while provid-
`ing complete protection from exonucleases [61]. And finally, a
`single-stranded antisense RNA (which does not fold into a duplex
`at all) has been shown to enter the RNAi pathway, with potency
`approaching that of the duplex siRNA in some cases [21,56,62].
`The length of an siRNA duplex can also be changed. Most
`synthetic duplexes are 19–21 bp in length, mimicking the natural
`products of the Dicer enzyme. However, increasing the length of
`an siRNA duplex makes it a substrate for Dicer and has been found
`to increase its potency [63]. It is important to keep the length
`below 30 nt, to avoid triggering the interferon response [64].
`
`Applications
`Improving serum stability
`Unprotected RNA is very quickly degraded in cells. The fact that
`siRNA is double-stranded provides it with some degree of protec-
`tion, but not enough for in vivo use. A nuclease called eri-1 has been
`found to play a key part in the degradation of siRNA [65], and
`expression levels of eri-1 inversely correlate with duration of siRNA
`activity [66]. This and other data suggest that increasing the
`nuclease resistance of siRNAs can prolong their activity. Chemical
`modification is the principal strategy used to improve the nuclease
`resistance of siRNAs.
`Essentially all of the modifications in the toolbox can be used to
`increase the serum half-life of siRNAs. Within the therapeutic
`siRNA community, however,
`two schools of
`thought have
`emerged regarding the best paradigm for protecting siRNAs against
`nucleases. The first strategy favors extensive or entire chemical
`modification. This paradigm is exemplified by the research of Sirna
`Therapeutics (http://www.sirna.com/), who have published work
`on heavily modified siRNA duplexes. For example, a fully modified
`siRNA with significantly increased potency in a hepatitis B virus
`0
`F-RNA
`(HBV) mouse model consisted of a sense strand made of 2
`0
`0
`and 3
`inverted abasic end caps.
`pyrimidines, DNA purines, and 5
`0
`0
`-O-Me
`The antisense strand was made of 2
`F-RNA pyrimidines, 2
`0
`-terminus
`purines and a single phosphorothioate linkage at the 3
`[51]. This fully modified duplex had a half-life in serum of two to
`three days, as compared with 3–5 min for the unmodified duplex
`[51]. This improved stability got translated into higher efficacy in
`vivo. Higher potency was later obtained by including one to three
`0
`-end of the antisense strand, and this heavily
`RNA inserts at the 5
`
`modified siRNA still had a serum half-life nearly 30 times longer
`than that of unmodified siRNA [50].
`A few other examples of fully modified duplexes have been
`0
`0
`-O-
`F-RNA and 2
`reported. An siRNA made entirely of alternating 2
`Me units was found, unsurprisingly, to have greatly increased
`stability in serum [67]. This architecture also maintains or
`improves potency, as discussed below. A functional duplex made
`0
`-O-Me-RNA sense strand and a PS-RNA
`with DNA overhangs, a 2
`antisense strand were nearly all intact after 48 h in serum [17], and
`0
`F-RNA siRNA has also shown excel-
`a functional fully modified 2
`lent nuclease resistance [24].
`Such a large degree of modification, however, may not always be
`necessary. The second paradigm for creating stabilized siRNAs
`involves minimal, selective modification. It is exemplified in
`the research of, among others, Alnylam Pharmaceuticals (http://
`www.alnylam.com/). Because endonuclease degradation is a
`major mechanism of degradation of siRNAs [16], the endonuclease
`cleavage pattern of a given siRNA duplex is first characterized (this
`is often dominated by cleavage after a pyrimidine nucleotide, and
`can be readily characterized by mass spectrometry [68]). The
`vulnerable positions are then selectively modified, usually with
`0
`0
`-O-Me or 2
`F-RNA nucleotides, which considerably increases the
`2
`stability of the siRNA with minimal modification [69,70].
`Besides these empirically determined internal positions, key
`positions for modification include the termini of the strands,
`0
`0
`-termini, protecting the duplex from 3
`-exonu-
`especially the 3
`clease degradation [17].
`
`Increasing potency
`The RNAi pathway is very efficient and unmodified siRNA is a very
`potent gene silencing agent, although potency does depend on
`cell type, target and siRNA sequence. In general,
`increasing
`potency is not considered the primary objective of chemical
`modification: it is sufficient to maintain the potency of unmodi-
`fied siRNA while increasing its serum half-life and its specificity.
`However, as the requirements for effective RNAi are increasingly
`well understood, we can foresee an increase in the use of chemical
`modifications to optimize potency as well, through features such
`as target-binding affinity (enhancing hybridization on-rates and
`off-rates), conformational preorganization (A-form helical struc-
`ture) and flexibility.
`It is very rare to find patterns of chemical modification that
`universally increase potency. By contrast, there are several known
`modifications that increase potency, sometimes very significantly,
`for particular sequences or systems. One of the most dramatic
`
`www.drugdiscoverytoday.com 847
`
`

`

`REVIEWS
`
`Drug Discovery Today  Volume 13, Numbers 19/20  October 2008
`
`kinase R (PKR) and helicases such as RIG-I and MDA5. Recognition
`by these receptors can lead to a number of cellular responses
`including release of cytokines and changes in gene expression
`[72]. The types of immunoreceptors involved, and the level of
`immune activation, depend on a number of factors including the
`length, sequence and cellular delivery method of the RNA, as well
`as the type of immune cell involved. These receptors exist on cell
`surfaces (TLR3) [77], in endosomes (TLR3/7/8) [78–80] and in the
`cytoplasm (RIG-I, MDA5, PKR) [79].
`The sequence dependence of immune responses is not fully
`understood. Judge et al. have identified immunostimulatory
`0
`motifs in mice and in vitro in human blood, most notably 5
`-
`0
`0
`regions [74]. Hornung et al. demonstrated that 5
`UGUGU-3
`-
`0
`is also a potent immunostimulatory motif
`GUCCUUCAA-3
`[37]. In fact, any U-rich RNA sequence may be sufficient for
`0
`-triphosphate or
`recognition by TLR7 [81]. The presence of a 5
`blunt end can lead to RIG-I-mediated immunostimulation
`[44,47,75]. Finally, duplexes between 23 and 30 bp may still
`induce the interferon response in a length-dependent manner,
`depending on cell type [76].
`Chemical modifications can be used to reduce the immunos-
`timulatory properties of siRNAs [79]. For example, siRNA
`0
`0
`duplexes >90% modified with 2
`-O-Me and DNA resi-
`F-RNA, 2
`dues were shown to cause no detectable effect on interferon levels
`or cytokines, while unmodified siRNA duplexes caused significant
`activation [50]. Because of immunostimulation, mice treated
`with unmodified siRNA showed increased levels of serum transa-
`minases and signs of systemic toxicity such as decreased body
`weight, transient lymphopenia and thrombocytopenia, and
`piloerection, but these adverse effects were not evident in those
`mice treated with the modified siRNA [50]. More recently, it has
`0
`-O-Me modification is
`been shown that a far smaller degree of 2
`sufficient to eliminate immunostimulatory activity [82]. Almost
`any degree of modification was sufficient, with the exception that
`0
`-O-methylation of cytidine residues was ineffective in reducing
`2
`0
`0
`-O-Me mod-
`immunostimulatory activity [82]. 2
`F-RNA and/or 2
`ification of uridine resulted in the elimination of immune
`off-target effects [44], including TLR-dependent and TLR-inde-
`pendent immune effects [83]. Strikingly, even the presence of
`0
`-O-Me-modified RNAs on separate, noncomplementary strands
`2
`abrogates the TLR7-dependent immunostimulatory activity of
`0
`-O-Me-RNA itself is a potent
`unmodified siRNAs, indicating that 2
`antagonist of TLR7 [84].
`0
`0
`-O-Me and 2
`-F modifications are not alone in reducing immu-
`2
`nostimulation. LNA has been used to modify the ends of an siRNA
`sense strand containing an immunostimulatory motif, abrogating
`the IFN-a immunostimulatory activity of the duplex without
`affecting the silencing activity [37]. Base modifications, too, can
`help reduce immune activation: modification with pseudouracil
`or 2-thiouracil prevented the RIG-I-mediated immunostimulation
`0
`due to a 5
`-triphosphate [44], and various base modifications
`abrogated immune effects mediated by TLR3, TLR7 and TLR8 [85].
`Indeed, different types of chemical modifications can reduce
`the immune effects from different receptors. For example, of
`the base modifications useful for siRNA, one study found that
`only 2-thiouridine was able to reduce TLR3-mediated immunos-
`timulation, but several others were able to reduce TLR7- or
`TLR8-mediated effects [85]. In another recent study, the