REVIEW ARTICLE
`published: 20 August 2012
`doi: 10.3389/fgene.2012.00154
`
`Development of therapeutic-grade small interfering RNAs
`by chemical engineering
`
`Jesper B. Bramsen* and Jørgen Kjems
`
`Interdisciplinary Nanoscience Center, Department of Molecular Biology and Genetics, Aarhus University, Aarhus C, Denmark
`
`Edited by:
`Kumiko Ui-Tei, University of Tokyo,
`Japan
`Reviewed by:
`Terrence Chi-Kong Lau, City University
`of Hong Kong, Hong Kong
`Heh-In Im, Korea Institute of Science
`and Technology, South Korea
`*Correspondence:
`Jesper B. Bramsen, Interdisciplinary
`Nanoscience Center, Department of
`Molecular Biology and Genetics,
`Aarhus University, C.F. Mollersalle
`Building 1130, 8000 Aarhus C,
`Denmark.
`e-mail: jebb@mb.au.dk
`
`Recent successes in clinical trials have provided important proof of concept that small inter-
`fering RNAs (siRNAs) indeed constitute a new promising class of therapeutics. Although
`great efforts are still needed to ensure efficient means of delivery in vivo, the siRNA mol-
`ecule itself has been successfully engineered by chemical modification to meet initial
`challenges regarding specificity, stability, and immunogenicity. To date, a great wealth of
`siRNA architectures and types of chemical modification are available for promoting safe
`siRNA-mediated gene silencing in vivo and, consequently, the choice of design and modifi-
`cation types can be challenging to individual experimenters. Here we review the literature
`and devise how to improve siRNA performance by structural design and specific chemical
`modification to ensure potent and specific gene silencing without unwarranted side-effects
`and hereby complement the ongoing efforts to improve cell targeting and delivery by other
`carrier molecules.
`
`Keywords: RNAi, siRNA, chemical modification, immunogenicity, off-target effect, LNA, OMe, siRNA therapeutic
`
`SILENCING GENES USING NUCLEIC ACID
`NUCLEIC ACID-BASED THERAPEUTICS
`Nucleic acid-based therapeutics promise to overcome the major
`limitation of existing medicine, which can currently only tar-
`get a limited number of proteins involved in disease pathways
`(Melnikova, 2007). Such promise rely on the high predictabil-
`ity of nucleic acid base-pairing which provides an ideal frame-
`work for gene silencing technologies (GSTs) by offering unpar-
`alleled specificity, rapidity of development and renders, at least
`in principle, all human genes “druggable”(Krieg, 2011). Pioneer-
`ing work in the 1970–1980s established the nucleic acid anti-
`sense technology as an universal GST by developing synthetic
`antisense oligonucleotides (ASOs) and ribozymes, which base
`pair to and inhibit the function of any desired messenger RNA
`(mRNA; Zamecnik and Stephenson, 1978; Potera, 2007). Today,
`two ASO-based drugs have been commercialized and several
`modern antisense design variants (Monia et al., 1993; Highley-
`man, 1998; Elmen et al., 2008; Gupta et al., 2010; Cirak et al.,
`2011) may be on the verge of success with >50 RNA or RNA-
`derived therapeutics reaching clinical testing (Sanghvi, 2011; Bur-
`nett and Rossi, 2012). This journey has, however, been far from
`straightforward and tedious efforts have been invested to engi-
`neer poorly performing drug candidates such as first generation
`ASO designs by chemical modification (Stein and Krieg, 1994)
`to meet therapeutic standards of potency and safety (Potera,
`2007).
`
`EXPLOITING RNAi PATHWAYS FOR THERAPEUTICS
`The discovery of RNAi interference (RNAi), gene silencing by
`double-stranded RNA (dsRNA), in the nematode worm C. ele-
`gans in 1998 (Fire et al., 1998) and the observation in 2001 that
`synthetic 21-mer dsRNA, named small interfering RNA (siRNA),
`triggered potent and specific gene silencing in mammalian cells
`
`(Elbashir et al., 2001a) provided researchers with an unprece-
`dentedly powerful gene silencing tool. The obvious therapeutic
`potential of siRNAs immediately renewed the scientific and com-
`mercial interest in developing nucleic acid drugs capable of low-
`dose, non-toxic targeting of mRNAs to treat human diseases. As
`compared to other nucleic acid-based technologies, siRNA bene-
`fits from harnessing endogenous RNAi pathways to effectuate gene
`silencing (Figure 1); upon introduction of synthetic siRNAs into
`the cell cytoplasm they are incorporated into an RNA-induced
`silencing complex (RISC; Hammond et al., 2001) by a RISC load-
`ing complex (RLC; Maniataki and Mourelatos, 2005) containing
`the RNase III enzyme Dicer (Bernstein et al., 2001). By sensing
`the thermodynamic asymmetry of siRNA duplex ends (Khvorova
`et al., 2003; Schwarz et al., 2003), RLC loads the siRNA guiding
`antisense strand into a cleavage-competent RISC containing Arg-
`onaute 2 (Ago2; Martinez and Tuschl, 2004), whereas the passenger
`sense strand (SS) is cleaved and released (Matranga et al., 2005;
`Leuschner et al., 2006). Subsequently, Ago2-RISC will efficiently
`guide and effectuate multiple rounds of target RNA cleavage result-
`ing in gene “knockdown” (KD; Hutvagner and Zamore, 2002).
`Furthermore, the structural similarity of endogenous microRNAs
`(miRNAs) and artificial siRNA triggers may be expected to render
`these undetectable to cellular sensors of (foreign) dsRNA thereby
`preventing induction of innate immune-responses. In effect, har-
`nessing siRNA to effectively enter the endogenous RNAi pathway
`translates into high silencing efficiencies, predictability, and reli-
`ability (Bertrand et al., 2002) but concurrently hold the potential
`to disturb endogenous gene regulation by the native inhabitants
`of the RNAi pathway, the miRNAs.
`
`siRNA AS A THERAPEUTIC PLATFORM
`Small interfering RNAs have gained increasing popularity in vivo
`(Behlke, 2006, 2008; Higuchi et al., 2010; Lares et al., 2010)
`
`www.frontiersin.org
`
`August 2012 | Volume 3 | Article 154 | 1
`Alnylam Exh. 1036
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`FIGURE 1 | The benefits and limitations of synthetic siRNA application.
`The most widely used siRNA type is the “canonical” synthetic 21-mer siRNA
`composed of two 21 nt RNA strands annealed to form a 19-bp dsRNA duplex
`stem and 2 nt 3(cid:2)-overhangs at both ends (the passenger strand is shown in
`black and the guide strand is shown in red). Also synthetic, dicer-substrate
`27-mer siRNAs (DsiRNA) has provided a popular alternative. Both design
`types can be delivered in vivo either unformulated or upon formulation of
`
`various types of delivery agents into the cell cytoplasm (light gray circle)
`where siRNAs are taken up by a RISC loading complex (RLC), which upon a
`dicer cleavage event (27-mer designs only, 21-mer siRNAs are
`dicer-independent) is structurally rearranged into a pre-RISC. Here the siRNA
`passenger strand is cleaved leading to the establishment of an active RISC
`that assists and ensures efficient degradation of RNA target sharing perfect
`(Continued)
`
`Frontiers in Genetics | Non-Coding RNA
`
`August 2012 | Volume 3 | Article 154 | 2
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`FIGURE 1 | Continued
`sequence complementarity to the siRNA guide stands. A number of
`bottleneck in siRNA applications are currently being resolved by
`chemical modification strategies (red circles). Unformulated siRNAs
`are sensitive to nuclease degradation in extracellular environment
`(cid:2) and, although degradation rates are much lower in the
`cytoplasm, siRNAs stabilization by modification is suggested to also
`enhance intracellular availability and silencing persistence (cid:3). Also,
`extracellular siRNAs can be rapidly cleared from the body, e.g., by
`renal filtration (cid:4) and can induce innate immune responses via TLR3
`in certain cells (cid:5). Delivery across the target cell membrane (cid:6) and
`endosomal release of endocytosed (cid:7) are currently the main
`bottlenecks in siRNA applications in vivo and siRNAs may induce
`
`TLR7/8-mediated immune-responses upon endosome retention in
`immune cells (cid:8). Also, all cells can respond to foreign cytoplasmic
`RNA via the PRRs, PKR, RIG-I, and Mda5 (cid:9). siRNA may disturb
`natural miRNA pathways, that processes nuclear pri-miRNA
`transcripts (dark gray circle) via a pre-miRNA intermediate and
`miRNA duplex into a single-stranded miRNA in RISC, by direct
`competition for RISC loading (cid:10) or by clotting the pathway due to
`slow siRNA processing and turnover (cid:11). Finally, all siRNA will trigger
`miRNA-like off-targeting effects on unintended targets upon
`base-pairing of the guide strand seed region and positions within
`mRNA 3(cid:2) UTRs leading to transcript destabilization and/or
`translational repression
`. Please refer to main text for more
`detail.
`
`and the number of RNAi-based preclinical and clinical tri-
`als have increased rapidly over recent years with ∼22 differ-
`ent siRNA or short hairpin RNA (shRNA) therapeutics reach-
`ing clinical testing for the treatment of at least 16 diseases
`(http://maps.google.com/maps/ms?ie=UTF8&source=embed&oe
`=UTF8&msa=0&msid=117696484602143675789.000476c449bf3
`97da6d6c; DeVincenzo et al., 2008; Davis et al., 2010; Leachman
`et al., 2010; Burnett et al., 2011; Davidson and McCray, 2011;
`Burnett and Rossi, 2012). This is truly an amazing achievement
`for a such a fledgling technology considering that conventional
`development of small-molecule drugs takes at least 5–7 years for
`preparing a single drug candidate for human clinical trials (Krieg,
`2011). For comparison, the first clinical trial of a siRNA-based
`drug was initiated in 2004 only 3 years after the initial application
`of siRNA in mammalian cell cultures (Shukla et al., 2010) and
`successful siRNA designs may easily be adaptable to other target.
`
`siRNA NEED CHEMICAL ENGINEERING TO SUCCEED AS THERAPEUTIC
`PLATFORM
`Building superior siRNAs is a combination of choosing an opti-
`mal siRNA-target sequence, optimal type of siRNA design, and,
`importantly, to introduce the proper combination of chemical
`modifications to suit the particular application. As a scientific
`tool in mammalian cell cultures, the potency and specificity of
`unmodified synthetic siRNA may be considered sufficient, yet
`chemical engineering of siRNAs is a prerequisite to transform
`them into a novel class of safe therapeutics, a natural progression
`similar to the development of second generation ASO (Potera,
`2007). Recent concerns regarding siRNA delivery and safety have
`dampened initial excitement and Big Pharma have recently down
`sized their investments in RNAi R&D (Ledford, 2010; Krieg, 2011;
`Schmidt, 2011). In particular, the size, lability, and negative charge
`of siRNAs severely complicate efficient intracellular delivery in vivo
`(Meade and Dowdy, 2009) and siRNA may trigger innate immune-
`response and lead to unintended deregulation of endogenous gene
`expression in several ways, as described in the following sections.
`Encouragingly, these concerns may be fully addressable by care-
`ful chemical modification of the synthetic siRNA molecule, an
`ongoing task that have already gone a long way; the first KD of
`an endogenous gene, apolipoprotein B (ApoB), was observed in
`mouse livers after low-pressure intravenous injections of a chemi-
`cally modified, but naked (non-formulated) siRNA already in 2004
`(Soutschek et al., 2004). Also, the first successful KD via RNAi of a
`cancer target gene in a human, the M2 subunit of ribonucleotide
`
`reductase (RRM2), was achieved in 2010 upon systemic siRNA
`delivery, a holy grail in siRNA therapeutics, using siRNA nanopar-
`ticles in a clinical phase-I trial in tumors from melanoma patients
`(Davis et al., 2010).
`
`STRUCTURAL siRNA DESIGNS
`Today, a variety of siRNA design types are available for gene silenc-
`ing each offering benefits and disadvantages (Figure 2): The by far
`most popular siRNA design mimics natural Dicer cleavage prod-
`ucts and comprises a 21 nucleotide (nt) guiding strand antisense
`to a given RNA target and a complementary passenger strand
`annealed to form a siRNA duplex with a 19-bp dsRNA stem
`(cid:2)
`overhangs at both ends (here referred to as canoni-
`and 2 nt 3
`cal 21-mer siRNAs; Elbashir et al., 2001a,b). Longer design types,
`collectively referred to as Dicer-substrate siRNAs (DsiRNAs) struc-
`turally mimic various Dicer substrates to enhance incorporation
`into RNAi pathways and potentially siRNA potency (Kim et al.,
`2005; Rose et al., 2005; Siolas et al., 2005; Amarzguioui et al., 2006;
`Collingwood et al., 2008; Hefner et al., 2008; Tanudji et al., 2009).
`Also shorter or truncated siRNA designs are gaining popularity
`such as 16-mer siRNA (Chu and Rana, 2008), shRNAs with RNA
`stems≤19 bp (Ge et al., 2009a,b), blunt 19-bp siRNAs (Czauderna
`et al., 2003; Prakash et al., 2005; Hogrefe et al., 2006; Ghosh et al.,
`2009), asymmetrical siRNAs (aiRNA; Sun et al., 2008), and asym-
`metric shorter-duplex siRNA (asiRNA; Chang et al., 2009). Finally,
`fork siRNAs (Hohjoh, 2004; Petrova Kruglova et al., 2010), single-
`stranded siRNAs (ss-siRNAs; Martinez et al., 2002; Holen et al.,
`2003; Hall et al., 2006), Dumbbell-shaped circular siRNAs (Abe
`et al., 2011), bulge-siRNA (Dua et al., 2011), and sisiRNAs (Bram-
`sen et al., 2007) have been successfully utilized, but may require
`more testing to qualify as a therapeutic siRNA platform. Recently,
`siRNAs have also been incorporated in larger nucleic acid struc-
`tures (Afonin et al., 2011; Grabow et al., 2011) with the prospect
`of enhancing delivery and bio-availability in vivo.
`
`TOLERANCES FOR CHEMICAL MODIFICATION OF siRNAs
`The chemical synthesis of siRNAs allows the position-specific
`incorporation of chemically modified nucleotides in the siRNA to
`modulate, e.g., thermostability, nuclease resistance, duplex struc-
`ture, and base-paring properties. For a decade, the compatibility
`of a diversity of chemical modifications with siRNA function has
`been mapped by empirical testing. Early siRNA chemical modifi-
`cation schemes quite naturally focused on modification types pre-
`viously used to potentiate and stabilize ASOs (reviewed in Kurreck,
`
`www.frontiersin.org
`
`August 2012 | Volume 3 | Article 154 | 3
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`FIGURE 2 | Popular siRNA design types. The canonical 21-mer siRNA is the
`most popular siRNA design today. Dicer-substrate siRNAs such as 27-mer
`siRNA, shRNA, pre-miRNA mimics, or fork siRNA may enhance siRNA
`potencies. Asymmetrical siRNAs (aiRNA), asymmetric shorter-duplex siRNA
`(asiRNA), bulge-siRNAs and sisiRNA may enhance silencing specificity.
`
`Blunt-end siRNA are reported to be more nuclease resistant. Single-stranded
`siRNAs (ss-siRNAs) and 16 mer are functional but may required higher siRNA
`concentrations. Dumbbell-shaped circular siRNAs may have longer silencing
`duration. Passenger strands are shown in black and guide strands in red.
`Please refer to main text for more detail.
`
`2003; Wilson and Keefe, 2006) with hopes of similar improvements
`in siRNA performance. The toolbox of chemical modification
`types seems ever expanding and current efforts should determine
`how and which chemical modifications types are best combined
`in single siRNAs to simultaneously reduce siRNA immunogenicity
`(Sledz et al., 2003), miRNA-like off-targeting (Jackson et al., 2003),
`to enhance nuclease resistance/bio-availability in vivo (Soutschek
`et al., 2004; Gao et al., 2009; Merkel et al., 2009), and silencing
`duration while preserving siRNA potency.
`Although effects are naturally rather chemistry-specific, the
`positional tolerance for chemical modification of siRNAis fairly
`established (Elbashir et al., 2001b; Chiu and Rana, 2002, 2003;
`Hamada et al., 2002; Amarzguioui et al., 2003; Braasch et al., 2003;
`Czauderna et al., 2003; Grunweller et al., 2003; Harborth et al.,
`2003; Prakash et al., 2005; Choung et al., 2006; Bramsen et al.,
`2009). As a general trend, the entire passenger strand as well as
`(cid:2)
`(cid:2)
`-proximal part and 3
`overhang of the guiding stand is most
`the 3
`tolerant to chemical modification, which agrees with the observa-
`tion that only position 2–16 of the guide strand base pairs with
`(cid:2)
`its target prior to cleavage (Wang et al., 2009b). The siRNA 3
`overhang of the guide strand, which is bound by the Ago PAZ
`domain during loading, is conveniently tolerant to chemical mod-
`ification. This reflects a limited role of PAZ binding during target
`(cid:2)
`overhang is released from
`cleavage (Ma et al., 2004) where the 3
`the PAZ domain (Tomari and Zamore, 2005; Wang et al., 2009b).
`(cid:2)
`overhangs relatively safe to modify, even
`This renders siRNA 3
`with bulky modifications incompatible with the size of the PAZ
`binding pocket.
`(cid:2)
`(cid:2)
`-proximal part, and central
`phosphate, the 5
`In contrast, the 5
`positions of the guide strand are sensitive, especially to multi-
`ple, thermo-modulating, or bulky modifications that influence the
`properties of minor groove. These tolerances agree nicely with the
`(cid:2)
`phos-
`structure of the Ago2-guide strand complexes; Here the 5
`phate of the guide strand is consistently found in the Ago MID
`domain (Wang et al., 2008), an essential interaction for strand
`loading into RISC (Nykanen et al., 2001; Lima et al., 2009). Once
`bound by Ago2, the initial interactions between the guide strand
`
`(cid:2)
`proximal siRNA seed
`and target RNA is mediated only by the 5
`region (positions 2–8 of the RISC-associated strand) selectively
`exposed to the solvent (Wang et al., 2008) and subsequent Ago2-
`mediated cleavage of target RNA requires forming of an RNA-like
`A-type helix structure between the guide strand and the target
`spanning both the seed region and around the cleavage site (oppo-
`site of guide strand position 10/11; Meister et al., 2004) hereby
`explaining the sensitivity of these AS positions to modification.
`
`TOOLS FOR CHEMICAL MODIFICATION OF siRNAs
`Mainly four classes of chemical modifications is utilized to modify
`ASOs and now siRNAs: (i) modification of the negatively charged
`phosphodiester backbone is primarily utilized to enhance siRNA
`nuclease resistance or affect RNA biodistribution and cellular
`(cid:2)
`-OH group is widely
`uptake; (ii) modifications at the ribose 2
`used to modulate most aspects of siRNA behavior including mod-
`ulating siRNA nuclease resistance, potency, specificity of silencing
`and to reduce siRNA immunogenicity; (iii) modifications of the
`ribose ring and nucleoside base is utilized to modulate siRNA sta-
`bility and base-pairing properties; (iv) dual modifications harbor
`two modified functionalities in a single nucleotide and especially
`(cid:2)
`the combination of backbone and ribose modifications with 2
`OH substitutions are currently gaining momentum (for a non-
`exhaustive selection of popular chemical modification types in
`siRNA design see Figure 3).
`
`siRNA BACKBONE MODIFICATION
`A classic and popular phosphate backbone alteration is the phos-
`phoromonothioate (PS) modification where one of the non-
`bridging phosphate oxygens is replaced with sulfur (Braasch et al.,
`2004). Also phosphorodithioate (PS2) substitutions, where both
`non-bridging oxygen atoms are replaced, were recently tested in
`siRNA designs and slightly increased siRNA potencies and nucle-
`ase resistance as compared to PS and unmodified siRNA (Yang
`et al., 2012). Moderately PS-modified siRNAs support efficient
`RNAi, yet effects are very position-dependent (Amarzguioui et al.,
`2003; Braasch et al., 2003; Chiu and Rana, 2003; Grunweller et al.,
`
`Frontiers in Genetics | Non-Coding RNA
`
`August 2012 | Volume 3 | Article 154 | 4
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`FIGURE 3 | Popular chemical modification types in siRNA design. RNA,
`ribonucleic acid; PS, phosphothioate; PS2, phosphodithioate; EA,
`2(cid:2)-aminoethyl; DNA, deoxyribonucleic acid; 2(cid:2)-F, 2(cid:2)-fluoro; 2(cid:2)-OMe 2(cid:2)-O-methyl;
`2(cid:2)-MOE, 2(cid:2)-O-methoxyethyl; F-ANA, 2(cid:2)-deoxy-2(cid:2)-fluoro-β-d-arabinonucleic acid;
`HM, 4(cid:2)-C-hydroxymethyl-DNA; LNA, locked nucleic acid; carboxylic LNA,
`
`2(cid:2),4(cid:2)-carbocyclic-LNA-locked nucleic acid; OXE, oxetane-LNA; UNA, unlocked
`nucleic acid; 4(cid:2)-S, 4(cid:2)-thioribonucleis acid; F-SRNA,
`2(cid:2)-deoxy-2(cid:2)-fluoro-4(cid:2)-thioribonucleic acid; ME-SRNA,
`2(cid:2)-O-Me-4(cid:2)-thioribonucleic acid; 4(cid:2)-S-F-ANA, 2(cid:2)-fluoro-4(cid:2)-thioarabinonucleic
`acid; ANA, altritol nucleic acid; HNA, hexitol nucleic acid; B, base.
`
`2003; Harborth et al., 2003; Choung et al., 2006) but extensive
`PS modification reduces silencing (Chiu and Rana, 2003; Hall
`et al., 2004) and has toxic side-effects (likely due to a tendency to
`bind non-specifically to cellular membrane proteins; Amarzguioui
`et al., 2003; Harborth et al., 2003). Still, moderate PS modifi-
`cation levels have been widely and successfully combined with
`(cid:2)
`-OH modifications in both ASO and siRNA designs to
`ribose 2
`stabilize naked RNA for systemic application in mice (Soutschek
`et al., 2004; Elmen et al., 2008). Both PS and PS2 will slightly
`
`thermo-destabilize siRNA duplexes (∼0.5˚C per modification;
`Eckstein, 2002; Amarzguioui et al., 2003; Harborth et al., 2003;
`Yang et al., 2012). Substitution of the phosphodiester linkage with
`a boranophosphate linkage have been explored in gene silencing
`using a canonical siRNA (Hall et al., 2004) or single-stranded siR-
`NAs (Hall et al., 2006). Similarly to PS and PS2, boranophosphate
`linkages decrease the T m of RNA duplexes by ∼0.5–0.8˚C per
`modification (Li et al., 2007) and enhances nuclease resistance as
`compared to unmodified siRNAs, yet is better-tolerated in siRNA
`
`www.frontiersin.org
`
`August 2012 | Volume 3 | Article 154 | 5
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`(cid:2)
`
`design than PS (Hall et al., 2004). Still this modification type has
`not been widely used.
`A phosphonoacetate (PACE) modification consists of an acetate
`group in place of a non-bridging oxygen in an internucleotide
`phosphate linkage whereas a thioPACE modification is an anal-
`ogous substitution in a phosphorothioate linkage. Both mod-
`ifications are completely resistant to degradation and PACE is
`electrochemically neutral if esterified with, e.g., methyl groups
`(Sheehan et al., 2003), which allows modified oligonucleotides to
`be taken up by cells in the absence of delivery reagents (Yamada
`(cid:2)
`et al., 2007). Also dual modification types combining 2
`-OMe
`and PACT/ThioPACT have recently been tested in siRNA design
`(cid:2)
`(cid:2)
`(see below). Finally, amide linkages (Iwase et al., 2007) and 2
`,5
`-
`linkages (Prakash et al., 2006) have been found to enhance nuclease
`resistance of siRNAs, yet are little used so far.
`RIBOSE 2(cid:2)-OH SUBSTITUTIONS
`-OH is the most popular approach
`Modifications of the ribose 2
`(cid:2)
`in siRNA design in which the 2
`-OH is either substituted with
`(cid:2)
`other chemical groups or the 2
`-oxygen is “locked” via intramol-
`ecular linkages as in bridged nucleic acids (BNAs). So far, the
`(cid:2)
`(cid:2)
`(cid:2)
`small electronegative 2
`substituents such as 2
`-fluoro (2
`-F), DNA
`(cid:2)
`(cid:2)
`(cid:2)
`(2
`-H), 2
`-O-methyl (2
`-OMe) are most widely used as they are
`well-tolerated, will generally enhance siRNA nuclease resistance
`and not dramatically affect siRNA thermostability, dsRNA duplex
`conformation nor siRNA activity.
`(cid:2)
`Fluorine substitution (2
`-F) slightly stabilizes dsRNA duplexes
`(∼1˚C increase in T m per modification; Freier and Altmann, 1997;
`Allerson et al., 2005), is among the best tolerated modification
`types and therefore allows the creation of highly modified, active
`(cid:2)
`siRNAs: both strands tolerate 2
`-F modification at most positions
`(Braasch et al., 2003; Chiu and Rana, 2003; Harborth et al., 2003;
`Prakash et al., 2005; Choung et al., 2006; Manoharan et al., 2011)
`and substitution of all siRNA pyrimidines was reported to greatly
`enhance serum stability and to support effective silencing in vitro
`and in vivo (Capodici et al., 2002; Layzer et al., 2004; Morrissey
`(cid:2)
`et al., 2005a). Also, 2
`-F has proven superior during an applica-
`tion in vivo using a mouse model for silencing of the factor VII
`(cid:2)
`gene as directly compared to other 2
`OH modifications types such
`(cid:2)
`as the popular 2
`-OMe (Manoharan et al., 2011). Even fully sub-
`(cid:2)
`stituted siRNA containing alternating modifications of 2
`-F and
`(cid:2)
`DNA (Blidner et al., 2007) or 2
`-OMe (Allerson et al., 2005) can
`be both highly potent and nuclease resistant.
`(cid:2)
`(cid:2)
`Similarly to 2
`-OMe modification
`-F, the naturally occurring 2
`(cid:2)
`was among the first and most extensively tested 2
`-substitutions
`(Amarzguioui et al., 2003; Chiu and Rana, 2003; Czauderna et al.,
`2003; Grunweller et al., 2003; Prakash et al., 2005; Choung et al.,
`(cid:2)
`2006; Jackson et al., 2006a); 2
`-OMe slightly enhances the binding
`affinity toward RNA (T m increase of 0.5–0.7˚C per modification)
`and is well-tolerated in most duplex positions (Freier and Alt-
`mann, 1997; Kraynack and Baker, 2006), although less so than
`(cid:2)
`2
`-F in the guide strand. More extensive or full modification, par-
`ticularly of the guide strand seed region, can reduce siRNA potency
`(Elbashir et al., 2001a; Braasch et al., 2003; Chiu and Rana, 2003;
`Czauderna et al., 2003) although conflicting results suggest that
`tolerances are siRNA sequence-dependent (Choung et al., 2006;
`(cid:2)
`Kraynack and Baker, 2006). Traditionally, 2
`-OMe modification
`
`has been extensively used to increase siRNA nuclease resistance
`(cid:2)
`-
`and proven particular successful in combination with other 2
`(cid:2)
`modifications, e.g., 2
`-F, to generate fully substituted, nuclease
`resistant, yet functional siRNAs (Allerson et al., 2005). Also, par-
`(cid:2)
`-OMe/PS-modified siRNA conjugated with cholesterol was
`tially 2
`the first chemical design to successfully silence an endogenous gene
`in mice using a systemic delivery strategy suitable for therapeu-
`(cid:2)
`tics (Soutschek et al., 2004). More recently 2
`-OMe modification is
`also gaining popularity for reducing siRNA immunogenicity and
`off-targeting (see Reducing siRNA Immunogenicity by Chemical
`Modification and Reducing siRNA Off-Target Effects by Chemical
`Modification).
`DNA modification, typically dTdT, has long been the industry
`(cid:2)
`standard for modifying siRNA 3
`overhangs to reduce cost, while
`conferring nuclease resistance (Elbashir et al., 2001a) and preserv-
`ing siRNA potency (Braasch et al., 2003; Chiu and Rana, 2003).
`(cid:2)
`(cid:2)
`As for 2
`-F and 2
`-OMe, DNA is well-tolerated in the passenger
`strand with little effect on siRNA potency (Hogrefe et al., 2006;
`(cid:2)
`Pirollo et al., 2007). Similarly to 2
`-OMe, only partially DNA sub-
`stituted guide strands are functional (Parrish et al., 2000; Chiu
`and Rana, 2003) whereas alternating the DNA modification with
`(cid:2)
`2
`-F has created fully substituted, active guide strands (Chiu and
`Rana, 2003). Notably, full DNA substitution of the guide strand
`seed region can reduce off-target effects (see Reducing siRNA Off-
`Target Effects by Chemical Modification) albeit its influence on
`siRNA potencies may be somewhat sequence-dependent (Ui-Tei
`et al., 2008) and dTdT overhangs seem to reduce silencing duration
`(see Strapps et al., 2010; Enhancing Silencing Duration).
`(cid:2)
`(cid:2)
`(cid:2)
`More bulky 2
`-modifications such as 2
`-O-MOE, 2
`-O-allyl and
`others are only tolerated at certain positions within the siRNA
`duplex, likely owing to their distortion of RNA helix structure
`essential to Ago2 cleavage (Amarzguioui et al., 2003; Prakash
`et al., 2005; Odadzic et al., 2008; Bramsen et al., 2009). This
`somewhat limits their use for general siRNA design as similar
`enhancements in, e.g., nuclease resistance may be obtained by
`(cid:2)
`the better-tolerated, smaller 2
`substitutions described above. Still,
`(cid:2)
`(cid:2)
`bulky 2
`-modifications are best tolerated at the 3
`-ends of siRNA
`strands (Odadzic et al., 2008), where they may be used to mod-
`ulate thermodynamic stability and duplex asymmetry to enhance
`(cid:2)
`siRNA potency: e.g., an 2
`-aminoethyl modification in the passen-
`(cid:2)
`ger strand 3
`-end can enhance siRNA potency, likely by affecting
`strand selection during RISC loading (Bramsen et al., 2009). Still,
`(cid:2)
`bulky 2
`-modifications are not widely used other than in siRNA
`overhangs to enhance nuclease resistance (Amarzguioui et al.,
`2003; Prakash et al., 2005; Odadzic et al., 2008; Bramsen et al.,
`2009).
`
`BRIDGED NUCLEIC ACIDS
`(cid:2)
`Another class of 2
`-modification is the BNA in which the ribose
`(cid:2)
`(cid:2)
`2
`-oxygen is linked to the 4
`-carbon via a methylene bridge as in
`Locked Nucleic Acid (LNA; Wengel et al., 2001) and carbocyclic-
`LNA (Srivastava et al., 2007; Bramsen et al., 2009), via an ethylene
`bridge as in ethylene-bridged nucleic acid (ENA; Hamada et al.,
`2002) and carbocyclic-ENA (Srivastava et al., 2007; Bramsen et al.,
`(cid:2)
`2009), or to the 1
`-carbon as in oxetane (OXE; Pradeepkumar et al.,
`2003; Bramsen et al., 2009). This radical modification type gen-
`erates nucleotides with interesting properties to siRNA design;
`
`Frontiers in Genetics | Non-Coding RNA
`
`August 2012 | Volume 3 | Article 154 | 6
`
`

`

`Bramsen and Kjems
`
`Chemical siRNA
`
`The methylene bridge in the very popular LNA-modification
`(cid:2)
`locks the sugar moiety in the RNA-helical C3
`-endo conformation,
`which additively increase RNA duplex thermostability by 2–10˚C
`per LNA (Petersen and Wengel, 2003). Although this dramatic
`thermo-stabilization limits siRNA modification levels by LNA
`(Braasch et al., 2003; Grunweller et al., 2003; Elmén et al., 2005),
`it provides unique opportunities to modulate the local thermody-
`namic profile within the siRNA duplex to optimize strand selection
`and thereby enhance specificity (Elmén et al., 2005; Bramsen et al.,
`2007, 2009), to enhance nuclease resistance in vitro (Braasch et al.,
`2003; Bramsen et al., 2009) and in vivo (Mook et al., 2007; Glud
`et al., 2009) and reduce siRNA immunogenicity (Hornung et al.,
`2005) as described in the following sections. Also, the thermo-
`stabilizing LNA may be used to ensure the integrity of siRNA
`designs relying on only short regions of oligo base-pairing such
`as the three stranded sisiRNA design (Bramsen et al., 2007). As a
`note of caution, LNA-modified ASO scan induce profound hepa-
`totoxicity in mice (Swayze et al., 2007) albeit this is not observed
`in other studies, e.g., in primates (Elmen et al., 2008).
`
`ALTERATION OF THE RIBOSE MOIETY
`Modification types based on sugar moieties other than ribose have
`been successfully used in siRNA designs upon incorporation of,
`(cid:2)
`e.g., altritol nucleic acid (ANA), hexitol nucleic acid (HNA), 2
`-
`(cid:2)
`(cid:2)
`deoxy-2
`-fluoroarabinonucleic acids (2
`-F-ANA), and cyclohex-
`enyl nucleic acid (CeNA) nucleotides, which are based on altritol,
`hexitol, arabinose, and cyclohexenyl sugars, respectively (Dowler
`et al., 2006; Fisher et al., 2007, 2009; Nauwelaerts et al., 2007; Watts
`et al., 2007; Bramsen et al., 2009).
`(cid:2)
`The 2
`-F-ANA modification is structurally similar to DNA
`(cid:2)
`(C2
`-endo conformation) and increases the T m of the siRNA
`duplex by ∼0.5–0.8˚C per modification and will, due its structural
`similarity to DNA, structurally distort siRNA duplexes making it
`little suited to modify the seed regions of the guide strand. Full
`modification of the SS can lead to significant enhancements in
`(cid:2)
`potency and nuclease resistance and 2
`-F-ANA-modification may
`(cid:2)
`be particular useful in the guide strand to create high-affinity 3
`overhangs, similarly to DNA (Dowler et al., 2006; Watts et al.,
`2007). Moreover, heavily modified siRNAs that contain combi-
`(cid:2)
`(cid:2)
`nations of 2
`-F-ANA and 2
`-F or LNA show superior properties
`(Deleavey et al., 2010).
`The ANA modification will slightly enhance siRNA thermosta-
`bility, are stable against enzymatic degradation and can moderately
`enhance siRNA activity and silencing duration upon incorpo-
`(cid:2)
`ration into of both duplex 3
`-ends (Fisher et al., 2007) whereas
`overhang modification slightly decrease RISC affinity (Maiti et al.,
`(cid:2)
`2011). Similarly, the incorporation of HNA at both strand 3
`-ends
`enhanced silencing potency, serum stability, and silencing dura-
`tion of a siRNA against B-Raf, even more so than observed for
`ANA-modifications upon direct comparison (Fisher et al