Chemical Modification of siRNA
`
`UNIT 16.3
`
`Glen F. Deleavey,1 Jonathan K. Watts,2 and Masad J. Damha1
`
`1McGill University, Montr´eal, Qu´ebec, Canada
`2University of Texas Southwestern Medical Center, Dallas, Texas
`
`ABSTRACT
`
`The ability to manipulate the RNA interference (RNAi) machinery to specifically silence the
`expression of target genes could be a powerful therapeutic strategy. Since the discovery that
`RNAi can be triggered in mammalian cells by short double-stranded RNAs (small interfering
`RNA, siRNA), there has been a tremendous push by researchers, from academia to big pharma, to
`move siRNAs into clinical application. The challenges facing siRNA therapeutics are significant.
`The inherent properties of siRNAs (polyanionic, vulnerable to nuclease cleavage) make clinical
`application difficult due to poor cellular uptake and rapid clearance. Side effects of siRNAs have
`also proven to be a further complication. Fortunately, numerous chemical modification strategies
`have been identified that allow many of these obstacles to be overcome. This unit will present an
`overview of (1) the chemical modifications available to the nucleic acid chemist for modifying
`siRNAs, (2) the application of chemical modifications to address specific therapeutic obstacles,
`and (3) the factors that must be considered when assessing the activity of modified siRNAs. Curr.
`Protoc. Nucleic Acid Chem. 39:16.3.1-16.3.22. C(cid:2) 2009 by John Wiley & Sons, Inc.
`Keywords: chemical modification r siRNA r oligonucleotide r immunostimulatory effects r
`off-target effects
`
`INTRODUCTION
`The first report that long double-stranded
`RNA could elicit specific gene silencing was
`published in 1998 (Fire et al., 1998). This dis-
`covery prompted a tremendous research ef-
`fort to understand and manipulate the gene-
`silencing pathway now referred to as RNA
`interference (RNAi). Three years later a syn-
`thetic trigger of the RNAi pathway in mam-
`malian cells was identified (Elbashir et al.,
`2001a). These short RNAi triggers are 21-nt
`double-stranded RNAs aligned such that each
`(cid:3)
`-overhang, and are referred
`strand has a 2-nt 3
`to as siRNAs. In recent years, siRNAs have
`generated tremendous interest in the thera-
`peutic field, and rightly so. siRNAs can be
`used to silence the expression of virtually any
`gene. The potential applications of siRNAs in
`drug development are immediately obvious:
`siRNAs can be used in cell culture to inhibit
`specific genes as a means of target validation,
`and/or siRNAs can themselves be drugs used
`to silence therapeutic targets.
`The transition of siRNAs from the labo-
`ratory to the clinic has met with some sig-
`nificant obstacles. The polyanionic phospho-
`diester backbone of siRNA hinders cellular
`uptake, and numerous cellular enzymes are
`capable of rapidly degrading RNA strands,
`making for poor pharmacokinetics. Further-
`
`more, it is becoming increasingly clear that
`siRNAs can elicit undesirable side effects.
`siRNAs can act like microRNAs (miRNAs)
`and affect the expression of unintended genes
`through partial complementarity. More re-
`cently, it has been discovered that siRNAs can
`be potent stimulants of the innate immune sys-
`tem (Robbins et al., 2009). Mammalian cells
`have receptors that recognize dsRNA (often
`a hallmark of viral infection), which can be
`triggered by siRNAs to produce inflamma-
`tory cytokines. Recent findings suggest that
`these immune responses to siRNAs may con-
`fuse gene-silencing outcomes, and may pro-
`duce unwanted and potentially dangerous side
`effects. All of these side effects are collectively
`termed “off-target” effects.
`Fortunately, a variety of chemical modifi-
`cations have been proposed to address these
`issues. This overview describes a wide range
`of chemical modifications that have been ap-
`plied to siRNAs, and provides examples of
`how these modifications can improve po-
`tency and serum stability, and reduce off-
`target and immunostimulatory side effects.
`The goal of this review is to provide the
`nucleic acid researcher with an overview of
`the state of the art, and a comprehensive
`guide for the design of chemically modified
`siRNAs.
`
`Current Protocols in Nucleic Acid Chemistry 16.3.1-16.3.22, December 2009
`Published online December 2009 in Wiley Interscience (www.interscience.wiley.com).
`DOI: 10.1002/0471142700.nc1603s39
`Copyright C(cid:2) 2009 John Wiley & Sons, Inc.
`
`RNA Silencing
`
`16.3.1
`
`Supplement 39
`
`Alnylam Exh. 1035
`
`

`

`siRNA
`jjjjjjjjjjjjjjjjjjj
`
`long dsRNA
`jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj
`
`phosphorylation and
`assembly into RISC
`
`\ jjjjjjjjjjjjjjjjjjjji : )
`I
`
`hairpin RNA
`
`Dicer cleavage into
`shorter dsRNA and
`assembly into RISC
`
`RISC Complex
`with bound siRNA
`
`cleavage and unwinding
`of sense strand to
`generate activated RISC
`complete with loaded
`guide strand.
`
`guide strand directs RISC
`to complementary mRNA
`
`I
`
`mRNA
`
`cleavage of mRNA
`target and
`regeneration of
`activated RISC
`
`Figure 16.3.1 The mechanism of RNAi in human cells triggered by exogenously introduced
`dsRNAs. The largest of the ellipses represents AGO2, the catalytic engine of RISC. Adapted from
`Watts et al. (2008), with permission from Elsevier.
`
`A detailed understanding of the mecha-
`nism of RNAi and the enzymes involved is
`invaluable when designing chemically mod-
`ified siRNAs, an overview of which can be
`found in UNIT 16.1. For the purposes of this
`overview, a brief description of the RNAi
`process is included in Figure 16.3.1. Briefly,
`when an exogenous 19- to 21-bp siRNA is
`
`(cid:3)
`-
`introduced into a mammalian cell, the 5
`end is phosphorylated by Clp-1, a cellular
`kinase (Weitzer and Martinez, 2007). The
`siRNA is then loaded into the RNA-induced
`silencing complex (RISC). This multiprotein
`complex includes Argonaute2 (AGO2), Dicer,
`TRBP (HIV-1 TAR RNA-binding protein),
`and PACT (a dsRNA binding protein), as
`
`Current Protocols in Nucleic Acid Chemistry
`
`Chemical
`Modification of
`siRNA
`
`16.3.2
`
`Supplement 39
`
`

`

`well as other proteins, some of which remain
`unknown (Rana, 2007). The structural and
`mechanistic aspects of both Dicer and Arg-
`onaute2 have been recently reviewed (Jinek
`and Doudna, 2009).
`siRNAs are double-stranded molecules,
`and the two strands play very different roles in
`RNAi. One strand, referred to as the guide or
`antisense strand, is loaded into the RISC com-
`plex and will function to direct RISC to com-
`plementary mRNAs for post-transcriptional
`gene silencing (PTGS). The other strand, re-
`ferred to as the passenger or sense strand,
`is simply cleaved and/or unwound from the
`guide strand upon RISC loading. It appears
`that RISC differentiates between the guide and
`passenger strands based on the thermodynam-
`ics of the siRNA duplex. It has been reported
`that the strand with lower binding affinity at
`(cid:3)
`-end preferentially becomes the guide
`its 5
`strand (Khvorova et al., 2003; Schwarz et al.,
`2003). The other strand is designated as the
`passenger strand. It is cleaved and unwound,
`generating a single RNA guide strand asso-
`ciated with AGO2, which is an endonuclease
`at the heart of RISC and promotes hybridiza-
`tion and cleavage of mRNAs complementary
`to the guide strand (Jinek and Doudna, 2009).
`When modifying an siRNA duplex, it is im-
`portant to recognize that different modifica-
`tion approaches are required for the sense and
`antisense strands, because of their very differ-
`ent roles (Parrish et al., 2000; Hamada et al.,
`2002).
`It is typically assumed that the mecha-
`nism of siRNA-mediated gene silencing is
`unchanged in chemically modified siRNAs.
`Studies using modified siRNA have confirmed
`this by showing that the cleavage of com-
`plementary mRNA occurs between bases 10
`(cid:3)
`-end of the guide
`and 11, counting from the 5
`strand, which is the same result as for un-
`modified duplexes (Soutschek et al., 2004;
`Kraynack and Baker, 2006; Ui-Tei et al., 2008;
`Judge et al., 2009). In principle, this should be
`verified for each new pattern of modification.
`
`METHODS FOR DESIGNING
`MODIFIED siRNA
`Choosing the Right Sequence
`Many factors must be considered in the de-
`sign of an effective and specific siRNA. Firstly,
`the chosen siRNA sequence must be unique
`to the intended mRNA target. Even incom-
`plete complementarity to unintended mRNAs
`can result in significant unwanted off-target
`effects (Jackson et al., 2006a). For exam-
`
`Current Protocols in Nucleic Acid Chemistry
`
`ple, the silencing ability of siRNAs doubly
`mismatched to mRNA targets has been thor-
`oughly analyzed. Results indicate that in some
`cases doubly mismatched siRNAs could still
`effect >50% gene silencing (Dahlgren et al.,
`2008). Because siRNAs can sometimes enter
`microRNA-type pathways, which silence gene
`expression based on complementarity to the
`miRNA “seed region” (the 6-7 nucleotide re-
`(cid:3)
`end of the guide strand), siRNAs
`gion at the 5
`should be designed to reduce seed-region
`complementarity to other genes as much as
`possible.
`Additional considerations include the avail-
`ability of the targeted region of the mRNA
`for RISC binding. Secondary structure can in-
`terfere with RISC-mediated mRNA cleavage,
`and so prediction of mRNA secondary struc-
`ture can be useful (Brown et al., 2005; Gredell
`et al., 2008; Rudnick et al., 2008). Experiments
`can also be used to identify accessible mRNA
`regions (Sokol et al., 1998; Lee et al., 2002).
`These factors can complicate siRNA de-
`sign. Fortunately, computer programs have
`been developed that take these and other fac-
`tors into account. Many of them are extremely
`user friendly, and can be invaluable when de-
`signing siRNAs. Some noteworthy programs
`include:
`(http://www.biopredsi.org):
`BIOPREDsi
`This site was developed by the Novartis Insti-
`tutes for BioMedical Research. It requires and
`provides only essential information. Input is a
`gene Accession number or gene sequence, and
`output is a user-defined number of optimized
`siRNA sequences (Huesken et al., 2005).
`Whitehead siRNA Selection Web Server
`(http://jura.wi.mit.edu/bioc/siRNAext): This
`Web site was developed and is hosted by the
`Whitehead Institute. It is somewhat more com-
`plex, giving a large number of possible du-
`plexes along with their properties, e.g., ther-
`modynamic properties. An off-target search
`can be done for each duplex, within the site,
`but in a separate step (Yuan et al., 2004).
`For an excellent guide to siRNA sequence
`selection, see Pei and Tuschl (2006). Other
`siRNA design algorithms are reviewed by
`Boese et al. (2005).
`It is useful to test several duplexes target-
`ing a gene, even after using a selection Web
`site. The most thorough way to identify highly
`active siRNA sequences is to test as many as
`100 to 200 sequences against a given mRNA
`(Choung et al., 2006; Manoharan and Rajeev,
`2007), looking for the duplexes with the best
`gene silencing activity and minimum off-target
`effects. Furthermore,
`individual sequences
`
`RNA Silencing
`
`16.3.3
`
`Supplement 39
`
`

`

`canonical
`
`blunt-ended
`
`XX
`
`XX
`universal overhangs
`(X = aromatic groups)
`
`- - - -
`25/27mer
`sisiRNA
`Dicer-substrate
`
`- - --8 C=====8
`
`dumbbell
`
`asymmetric siRNA
`(various designs are possible)
`
`antisense only
`
`hairpin
`(shRNA)
`
`(cid:3)
`direction,
`to 3
`Figure 16.3.2 Functional siRNA architectures. The sense strand is always shown on top, in the 5
`(cid:3)
`(cid:3)
`to 5
`direction. Note that three of the structures (25/27mer,
`and the antisense strand is on the bottom, in the 3
`hairpin, and dumbbell) require the activity of Dicer before incorporation into RISC. Adapted from Watts et al. (2008),
`with permission from Elsevier.
`
`(cid:3)
`
`respond differently to chemical modifications,
`and modifications beneficial for one siRNA se-
`quence may be detrimental to another. There-
`fore, a variety of chemical modifications
`should be included in the sequence screen from
`fairly early in the process.
`
`Choosing an siRNA Architecture
`The canonical siRNA architecture is a 21-nt
`(cid:3)
`antiparallel, double-stranded RNA with 2-nt 3
`overhangs (Elbashir et al., 2001a). The canoni-
`cal siRNA design mimics the product of Dicer
`processing of long dsRNAs (Zamore et al.,
`2000; Bernstein et al., 2001). In the naturally
`occurring cellular pathway, these Dicer cleav-
`age products are the effector molecules loaded
`into RISC to effect gene silencing. However,
`researchers have discovered that a much wider
`variety of duplexes can be incorporated into
`the RNAi pathway.
`Indeed, we now understand that siRNA
`overhangs are particularly amenable to modi-
`(cid:3)
`-deoxynucleotide units
`fications. The use of 2
`(cid:3)
`-overhangs has been common since the
`in the 3
`earliest siRNA experiments; it reduces the cost
`of oligonucleotide synthesis and may increase
`(cid:3)
`-exonucleases (Elbashir et al.,
`resistance to 3
`2001a). Of course, RNA units also work well
`(Holen et al., 2002). Perhaps more interest-
`ingly, most of the chemistries reviewed below
`are well tolerated in the overhangs as well (i.e.,
`accepted by the RNAi machinery without loss
`of potency).
`Blunt-ended siRNAs have also been used
`(Czauderna et al., 2003) because blunt-ended
`(cid:3)
`-exonucleases.
`duplexes are more resistant to 3
`In addition, one study reported a greater tol-
`erance to chemical modifications in combina-
`tion with blunt-ended duplexes (Kraynack and
`Baker, 2006). However, blunt-ended siRNAs
`can be more immunogenic than duplexes with
`(cid:3)
`-overhangs (Marques et al., 2006).
`3
`(cid:3)
`overhang is
`It has been shown that the 3
`bound within a hydrophobic pocket of the
`
`PAZ domain (Lingel et al., 2003; Song et al.,
`2003; Yan et al., 2003; Ma et al., 2004). So,
`(cid:3)
`nucleotide overhangs can be replaced
`the 3
`by aromatic groups (described as “universal
`overhangs” due to the lack of base pairing),
`improving nuclease stability without harming
`activity (Ueno et al., 2009).
`(cid:3)
`-end of
`Chemical phosphorylation of the 5
`the antisense strand can help ensure high po-
`(cid:3)
`-
`tency, and may be necessary when the 5
`end of the strand is modified (Fisher et al.,
`2007; Watts et al., 2007; Zhang et al., 2009).
`(cid:3)
`-phosphorylation of the sense
`Conversely, 5
`strand can be blocked without loss of activity
`(Nykanen et al., 2001; Czauderna et al., 2003).
`In fact, blocking sense strand phosphorylation
`can be beneficial, ensuring proper guide strand
`selection by RISC (Chen et al., 2008).
`The siRNA architecture itself can be mod-
`ified via chemical synthesis (Fig. 16.3.2).
`While most siRNA duplexes are made up of
`two strands, an siRNA made of three strands
`(an intact antisense strand and two 9- to 13-nt
`sense strands) can reduce off-target effects and
`increase potency; the resulting nicked duplex
`is termed small internally segmented interfer-
`ing RNA (sisiRNA; Bramsen et al., 2007).
`Functional siRNA can also be made from
`just one strand,
`in any of the following
`ways: Hairpin-type duplexes, made from a
`single strand, can be introduced exogenously
`(Siolas et al., 2005) or expressed within a
`cell (Brummelkamp et al., 2002; reviewed in
`Banan and Puri, 2004). Dumbbells or nanocir-
`cles, made by closing the open end of the
`hairpin, retain RNAi activity while provid-
`ing complete protection from exonucleases
`(Abe et al., 2007). Finally, a single-stranded
`antisense RNA (not folded into a duplex
`at all) has been shown to enter the RNAi
`pathway, with potency occasionally approach-
`ing that of duplex siRNA (Martinez et al.,
`2002; Amarzguioui et al., 2003; Holen et al.,
`2003).
`
`Current Protocols in Nucleic Acid Chemistry
`
`Chemical
`Modification of
`siRNA
`
`16.3.4
`
`Supplement 39
`
`

`

`Recent findings have also demonstrated
`that asymmetrically designed siRNA duplexes
`can be potent gene silencing agents, compa-
`rable to canonical siRNAs. Indeed, siRNA
`duplexes can vary significantly in the lengths
`of either the sense or antisense strands. Asym-
`metrical siRNAs can contain ssRNA regions
`and comparatively short dsRNA regions with-
`out impairing activity in many cases. Active
`asymmetrical designs include 15 (sense, S)
`+ 21 (antisense, AS) (Sun et al., 2008), 16
`(S) + 21 (AS), 18 (S) +18 (AS) (Chu and
`Rana, 2008), and 16 (S) + 19 (AS) (Chang
`et al., 2009). Asymmetrical siRNA designs
`are intriguing because several of these designs
`significantly decrease off-targeting, especially
`the sense strand mediated off-targeting that
`occurs when the wrong strand is loaded into
`RISC.
`Asymmetry in siRNA overhangs can also
`be beneficial in directing the proper guide
`strand into RISC. Asymmetric siRNAs de-
`signed such that the guide strand has a 2-nt
`(cid:3)
`-overhang and the passenger strand lacks a
`3
`(cid:3)
`-overhang (blunt ended) can improve siRNA
`3
`potency (Sano et al., 2008). This effect is use-
`ful for siRNAs with unfavorable thermody-
`namics of loading.
`The length of an siRNA duplex can be
`altered. Most synthetic duplexes are 19 to
`21 bp in length, designed to mimic the natural
`products of the Dicer enzyme, which cleaves
`long dsRNAs into substrates for RISC. On the
`other hand, using a longer siRNA duplex can
`make it a substrate for Dicer, and has been
`found to increase potency of gene silencing
`and stability to degradation (Kim et al., 2005;
`Amarzguioui et al., 2006; Kubo et al., 2007).
`Enhanced potency may result from a physi-
`cal transfer of newly dicer-produced siRNA
`to RISC. It is important to maintain a length
`of <30 nt, to avoid triggering the interferon
`response (Minks et al., 1979).
`
`INCORPORATING MODIFIED
`NUCLEOSIDES AND BACKBONE
`LINKAGES
`While careful sequence selection is essen-
`tial and varying the architecture of an siRNA
`duplex can provide certain advantages, this re-
`view will focus on changing the chemical na-
`ture of the oligos themselves. The rationale for
`chemical modification is clear, and the right
`chemical modifications can address nearly all
`the shortcomings of siRNA therapeutics:
`
`1. Nuclease stability can be significantly im-
`proved.
`
`Current Protocols in Nucleic Acid Chemistry
`
`2. Cellular uptake can be increased.
`3. Immunostimulation can be reduced (or in-
`creased, if desired).
`4. Off-targeting can be reduced.
`
`This section will describe a wide variety
`of chemical modifications compatible with
`siRNA.
`
`Sugar Modifications
`The ribose sugar moiety is perhaps the
`most attractive component for chemical modi-
`fication. Changes in the sugar chemistry can
`strongly influence sugar conformation, and
`thus overall siRNA duplex structure. Sugar
`modifications can improve endonuclease re-
`sistance. siRNAs containing RNA-like nu-
`cleoside analogues can pass under the immune
`receptor radar screen without impairing gene
`silencing activity. A wide variety of sugar
`modifications can be used to modify siRNAs,
`although often with limitations on position and
`quantity. Various sugar modifications com-
`patible with RNAi are described below (see
`Fig. 16.3.3), with reference to features and
`properties that are useful when designing mod-
`ified siRNAs.
`A number of successful siRNA modifica-
`(cid:3)
`tions have focused on modification of the 2
`position. One of the earliest studies on chem-
`ically modified siRNA showed that while A-
`(cid:3)
`-OH
`form duplex structure is important, the 2
`is not required for active siRNA (Chiu and
`Rana, 2003). DNA is perhaps the simplest
`modification accepted within siRNA duplexes.
`(cid:3)
`-overhangs of syn-
`DNA has been used in the 3
`thetic siRNAs since the beginning (Elbashir
`et al., 2001a), and can be tolerated in lim-
`ited numbers within the duplex region as well
`(Parrish et al., 2000; Elbashir et al., 2001b;
`Ui-Tei et al., 2008). An antisense strand made
`(cid:3)
`F-RNA pyrim-
`entirely of DNA purines and 2
`idines is functional (Chiu and Rana, 2003),
`(cid:3)
`(cid:3)
`F-RNA strongly favors a northern (C3
`-
`since 2
`endo) sugar pucker and A-form helical struc-
`ture, presumably directing the conformation of
`(cid:3)
`-deoxynucleotides towards
`the more flexible 2
`northern, RNA-like conformations. dsDNA
`(cid:3)
`-end
`can be used in the 8-bp region at the 5
`of the guide strand to produce active siRNAs
`with reduced off-target effects (Ui-Tei et al.,
`2008).
`Binding affinity and nuclease stability
`(cid:3)
`-O-methylation of RNA,
`are increased by 2
`(cid:3)
`-O-Me-RNA can be well-tolerated
`and 2
`throughout the siRNA duplex. This makes it
`a popular and versatile siRNA modification.
`Many groups have found that large numbers
`
`RNA Silencing
`
`16.3.5
`
`Supplement 39
`
`

`

`~
`
`O
`
`base
`
`O
`
`~
`
`O
`
`base
`
`O
`
`JV"'
`\
`O
`
`base
`
`O
`
`JV"'
`\
`O
`
`base
`
`O
`
`O
`
`OH
`
`~
`
`O
`
`OCH3
`
`~
`
`R R R
`R
`~ ~ ~ \
`
`JV"'
`O
`
`base
`
`O
`
`\ R
`
`O
`
`O
`
`~ 7_
`
`CN
`
`O
`
`O
`
`O
`
`O
`
`O
`
`CH3
`
`RNA
`
`2'-O-Me-RNA
`
`2'-O-MOE-RNA
`
`2'-O-allyl-RNA
`
`2'-O-cyanoethyl
`
`~
`
`O
`
`base
`
`O
`
`~
`
`O
`
`base
`
`O
`
`R R
`~ \ 0
`
`O
`
`O
`
`~
`
`NH2
`
`O
`
`O
`,.;.,-r
`
`NO2
`
`_,,-::;
`
`~
`
`O
`
`base
`
`O
`
`R
`
`O
`
`F
`
`~
`
`O
`
`F
`
`O
`
`µ
`
`O
`
`base
`
`O
`
`O
`
`base
`
`O
`
`2'-O-ethylamine
`
`2'-O-DNP-RNA
`
`NO2
`
`~
`
`2'F-RNA
`
`~
`
`2'F-ANA
`
`~
`
`DNA
`
`~ µ
`I ,w
`
`JVVV
`
`H
`
`O
`
`O
`
`base
`
`O
`
`I
`
`./VV'
`
`~
`
`O
`
`base
`
`F
`
`S
`
`µ
`
`O
`
`~
`
`base
`
`O
`
`O
`
`~ g
`
`OO
`
`~
`
`~
`
`O
`
`base
`
`S
`
`R
`
`O
`
`OH
`
`~
`
`4'S-RNA
`
`4'S-FANA
`
`LNA
`
`tc-DNA
`
`O
`
`base
`
`O
`
`O
`
`O
`
`O
`
`O
`
`base
`
`base
`
`H2N
`
`O
`
`OH
`
`O
`
`base
`O
`
`O
`
`O
`
`O
`\
`~
`
`base
`
`O
`
`N
`
`CeNA
`
`ANA
`
`HNA
`
`aminoisonucleosides
`
`morpholino
`
`(cid:3)
`-O-alkyl mod-
`Figure 16.3.3 Sugar units that have been successfully used to modify siRNA duplexes. Top row, 2
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`-O-Me, 2
`-O-MOE, 2
`-O-allyl, 2
`-O-cyanoethyl); second row, other 2
`-modifications (2
`-O-ethylamine,
`ifications (2
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`(cid:3)
`-O-DNP, 2
`F-RNA, 2
`F-ANA, and DNA; third row, 4
`-modifications (4
`S-RNA and 4
`S-2
`F-ANA) and bicyclic/tricyclic
`2
`modifications (LNA and tc-DNA, respectively); bottom row, six-membered ring modifications (CeNA, ANA, HNA) and
`dramatically redesigned sugar structures (aminoisonucleosides, morpholino). Adapted from Watts et al. (2008), with
`permission from Elsevier.
`(cid:3)
`-O-Me modifications (in either strand)
`of 2
`decrease siRNA activity (Elbashir et al.,
`2001b; Braasch et al., 2003; Chiu and Rana,
`2003; Czauderna et al., 2003), but others re-
`(cid:3)
`-O-Me sense strands
`port that fully modified 2
`are functional (Choung et al., 2006; Kraynack
`and Baker, 2006). The retention of functional-
`(cid:3)
`-O-Me modifications
`ity may be due to the 2
`being made
`in
`blunt-ended
`duplexes
`(Kraynack and Baker, 2006), but at
`least
`one other group reports that even in this
`context,
`activity is greatly reduced by
`(cid:3)
`-O-Me modification (Czauderna
`heavy 2
`(cid:3)
`(cid:3)
`-O-methoxyethyl (2
`-
`et al., 2003). The 2
`O-MOE) modification has been used in
`(cid:3)
`-overhangs of siRNA targeting the
`the 3
`pain-related cation-channel P2X3 and has
`resulted in successful gene targeting in vivo
`(Dorn et al., 2004). Another group found that
`
`(cid:3)
`-O-MOE modifications could be used in the
`2
`sense strand, especially at the termini, but not
`in the antisense strand (Prakash et al., 2005).
`(cid:3)
`-O-allyl-modifications
`reduce
`Similarly,
`2
`activity at most positions in the duplex, but
`can be used (Odadzic et al., 2008). These
`modifications do not reduce activity in the
`(cid:3)
`-overhangs (Amarzguioui et al., 2003).
`3
`(cid:3)
`position in-
`A cyanoethyl group at the 2
`creases binding affinity and nuclease resis-
`tance, and may find applications in siRNA
`(Saneyoshi et al., 2005; Bramsen et al., 2009).
`(cid:3)
`-O-acetalester-protecting groups are being
`2
`developed for production of “proRNA,” which
`remains protected until demasked by cellu-
`lar esterases (Martin et al., 2009). These
`molecules may find application as “pro-
`siRNA” (Martin et al., 2009). In a similar
`(cid:3)
`-esterified units (levulinates) may
`fashion, 2
`
`Chemical
`Modification of
`siRNA
`
`16.3.6
`
`Supplement 39
`
`Current Protocols in Nucleic Acid Chemistry
`
`

`

`be used to create protected siRNA molecules
`(Lackey et al., 2007).
`(cid:3)
`-O-alkylamines also appear to be active
`2
`in siRNA duplexes (Odadzic et al., 2008;
`Bramsen et al., 2009). Minimal modification
`(cid:3)
`(cid:3)
`-O-ethylamine units at the 3
`end of the
`with 2
`antisense strand appears to be well-tolerated
`(Odadzic et al., 2008). Further characteriza-
`tion will improve our understanding of their
`gene silencing potential.
`(cid:3)
`-modifications are
`In most cases, bulkier 2
`not well tolerated in siRNA duplexes, ex-
`cept for limited modifications on the termini.
`(cid:3)
`-
`Interestingly, siRNAs with 70% of the 2
`OH groups converted at random into 2,4-
`(cid:3)
`-O-DNP) show a va-
`dinitrophenyl ethers (2
`riety of improved properties, including higher
`binding affinity, nuclease resistance, and po-
`tency (Chen et al., 2004). Recently Bramsen
`et al. (2009) have reported a large-scale screen
`of siRNA activity elicited by several nucle-
`oside analogues, showing that a variety of
`(cid:3)
`modifications can be tolerated, particularly
`2
`(cid:3)
`-
`in the sense strand. In some cases, bulky 2
`modifications were tolerated, but with slightly
`reduced activity.
`can also contain
`Functional
`siRNAs
`(cid:3)
`-position. One of
`the
`fluorine at
`the 2
`best-known
`and most
`versatile
`siRNA
`(cid:3)
`(cid:3)
`F-RNA. Partial 2
`F-RNA
`modifications is 2
`modification is tolerated throughout sense and
`antisense strands (Braasch et al., 2003; Chiu
`and Rana, 2003; Harborth et al., 2003), and
`there are reports of active siRNAs that are fully
`(cid:3)
`(cid:3)
`F-RNA (Blidner et al., 2007). 2
`F-
`modified 2
`RNA-modified siRNA duplexes have signifi-
`cantly increased serum stability (Layzer et al.,
`2004) and increased binding affinity of the
`duplex (Kawasaki et al., 1993; Monia et al.,
`(cid:3)
`F-RNA modification has also al-
`2003). The 2
`(cid:3)
`-
`lowed the synthesis of siRNAs containing 5
`(cid:3)
`(cid:3)
`(cid:3)
`,5
`-dideoxy-2
`-fluororibonucleic acid
`amino-2
`units that confer improved hydrolytic stability
`(Ueno et al., 2008).
`Changing the stereochemistry of the flu-
`(cid:3)
`(cid:3)
`(cid:3)
`F-RNA results in 2
`-deoxy-2
`-
`orine of 2
`(cid:3)
`F-ANA). Con-
`fluoroarabinonucleic acids (2
`(cid:3)
`F-ANA was originally devel-
`sidering that 2
`oped as a DNA mimic (Rosenberg et al., 1993;
`Berger et al., 1998; Damha et al., 1998; Ikeda
`et al., 1998; Wilds and Damha, 2000), it is
`(cid:3)
`F-ANA is also
`somewhat surprising that 2
`well-tolerated in siRNA duplexes, including
`fully modified sense strands and partially mod-
`ified antisense strands (Dowler et al., 2006;
`(cid:3)
`(cid:3)
`F-RNA, 2
`F-ANA
`Watts et al., 2007). Like 2
`binds with high affinity and increases nucle-
`
`Current Protocols in Nucleic Acid Chemistry
`
`(cid:3)
`-fluorinated analogues are
`ase stability. Both 2
`highly resistant to acidic conditions (Watts
`et al., 2009).
`(cid:3)
`-position,
`In addition to the oxygen at the 2
`the ring oxygen can be modified for siRNA ap-
`(cid:3)
`S-RNA, a high-
`plications. One example is 4
`affinity modification that gives a significant
`advantage in nuclease stability. Incorporation
`(cid:3)
`S-RNA units does not reduce silencing
`of 4
`when placed near the termini of siRNA du-
`plexes (Hoshika et al., 2005, 2007; Dande
`et al., 2006). An siRNA architecture consist-
`(cid:3)
`-thioribonucleotides on each end
`ing of four 4
`(cid:3)
`-end of
`of the sense strand and four at the 3
`the antisense strand has consistently worked
`well for silencing two target genes in three
`cell lines (Hoshika et al., 2007). When the
`(cid:3)
`S-RNA,
`antisense strand was modified with 4
`especially at the center, potency was reduced
`(Hoshika et al., 2005; Dande et al., 2006).
`(cid:3)
`(cid:3)
`S-RNA with 2
`-O-Me and
`Combinations of 4
`(cid:3)
`-O-MOE modifications at the termini of both
`2
`strands have also shown excellent potency and
`serum stability (Dande et al., 2006).
`(cid:3)
`(cid:3)
`S-2
`F-ANA, featuring modifications at
`4
`(cid:3)
`(cid:3)
`and 4
`positions, has a northern,
`both the 2
`RNA-like conformation (Watts et al., 2007),
`and is compatible with siRNA activity at vari-
`(cid:3)
`(cid:3)
`S-2
`F-ANA in
`ous positions in both strands. 4
`(cid:3)
`F-
`the antisense strand shows synergy with 2
`ANA modifications in the sense strand (Watts
`et al., 2007). However, the low-binding affin-
`(cid:3)
`(cid:3)
`S-2
`F-ANA means that only a small
`ity of 4
`number of modifications should be used.
`Locked nucleic acid (LNA), a confor-
`mationally constrained nucleotide adopting
`a northern structure (Petersen and Wengel,
`2003), has been extensively studied in siRNA
`(Braasch et al., 2003; Elmen et al., 2005;
`Hornung et al., 2005; Bramsen et al., 2007;
`Mook et al., 2007). Its conformational rigidity
`causes significant increases in binding affinity.
`Careful placement of LNA in siRNA duplexes
`can lead to functional duplexes of various
`types. Common terminal modification sites
`are the termini of the sense strand (Hornung
`(cid:3)
`-overhang of the anti-
`et al., 2005) and the 3
`sense strand (Elmen et al., 2005; Mook et al.,
`2007). Minimal modification of most inter-
`nal positions of the antisense strand is also
`tolerated (Braasch et al., 2003; Elmen et al.,
`2005), but heavier modification of the anti-
`sense strand is tolerated only in combination
`with a segmented sense strand (sisiRNA, de-
`scribed above; see Bramsen et al., 2007).
`LNA is one example of a larger family
`of nucleoside analogues, the bicyclic nucleic
`
`RNA Silencing
`
`16.3.7
`
`Supplement 39
`
`

`

`acids (BNAs). Another BNA, ethylene-bridge
`nucleic acid (ENA), has been tested in the
`siRNA overhang and been found to impair
`siRNA activity (Hamada et al., 2002). A rigid
`tricyclic modification (tc-DNA) has been suc-
`cessfully used at either end of the sense strand,
`and even improved activity when placed in the
`overhangs (Ittig et al., 2008). The development
`of a variety of BNAs has been reviewed by
`Obika (2004), although much remains to be
`understood regarding their compatibilities to
`elicit siRNA activity. Toward this end, Bram-
`sen et al. (2009) have published a large-scale
`screen of siRNA activity elicited by several
`nucleoside analogues, including several mem-
`bers of the BNA family. For the siRNA se-
`quence tested, several BNAs appear to be tol-
`erated in siRNA, especially within the sense
`strand.
`Interestingly, siRNA potency can be re-
`tained, and in some cases improved, when the
`5-membered sugar moiety is replaced with a
`6-membered ring. Cyclohexenyl nucleic acids
`(CeNA), altritol nucleic acids (ANAs), and
`hexitol nucleic acids (HNAs) have all been
`successfully incorporated into siRNAs. Incor-
`poration of one or two CeNA units can pro-
`duce siRNAs with similar or enhanced activity
`compared to unmodified siRNA (Nauwelaerts
`et al., 2007). ANAs support A-form structures,
`and can improve potency and duration of ac-
`tivity when incorporated at certain positions
`(Fisher et al., 2007), but can also slightly de-
`crease potency in some cases (Fisher et al.,
`2009). HNA can improve siRNA potency, and
`is tolerated in the sense strand opposite the po-
`sition on the antisense strand at which mRNA
`cleavage occurs (Fisher et al., 2009).
`siRNAs can be minimally modified with
`aminoisonucleosides (see Fig. 16.3.3) without
`reducing activity (Li et al., 2007). Aminoi-
`sonucleosides were better tolerated by the
`RNAi machinery in the sense strand than in
`the antisense strand, and a single insert could
`(cid:3)
`(cid:3)
`or the 3
`region of
`be used at either the 5
`the sense strand without impairing gene si-
`lencing. However, modifications were less tol-
`erated internally in the sense strand, reduced
`activity when placed in the antisense strand,
`and may be detrimental to overall nuclease
`stability. Aminoisonucleoside modification of
`siRNA is awaiting optimization.
`Morpholino nucleoside analogues can also
`be used in siRNA. Morpholino analogues are
`tolerated in the sense strand and on the over-
`hangs, but reduce activity in the antisense
`strand (Zhang et al., 2009).
`
`Backbone Linkage Modifications
`Chemical modifications to the oligonu-
`cleotide backbone are also tolerated by the
`RNAi machinery. Indeed, several variations
`on the phosphodiester linkage can be used
`for gene silencing (Fig. 16.3.4). Phospho-
`rothioate (PS) linkages, a popular modifica-
`tion in antisense technology, can be used in
`siRNA, but they often reduce potency relative
`to native siRNA (Chiu and Rana, 2003; Hall
`et al., 2004). However, in some cases their
`potencies are comparable to native siRNAs
`(Amarzguioui et al., 2003; Harborth et al.,
`2003). Some observations indicate that PS
`linkages are not accepted at the center of the
`duplex, especially at the sense strand scissile
`phosphate (Leuschner et al., 2006). Some cy-
`totoxicity has been observed with extensive
`PS modification (Amarzguioui et al., 2003).
`In marked contrast to PS-DNA antisense con-
`structs (Geary et al., 2001), PS modifications
`do not appear to have a major effect on biodis-
`tribution of siRNA (Braasch et al., 2004).
`siRNAs with boronophosphate linkages are
`functional. They show increased potency rel-
`ative to PS-modified siRNAs, and often to na-
`tive siRNAs as well. To maximize potency,
`the center of the antisense strand should not
`be modified (Hall et al., 2004). Boronophos-
`phate linkages in siRNAs provide a significant
`increase in nuclease stability over native RNA
`(Hall et al., 2004).
`Two nonionic amide linkages are toler-
`(cid:3)
`overhangs of siRNA: a sim-
`ated in the 3