Chemistry & Biology
`
`Review
`
`Designing Chemically Modified Oligonucleotides
`for Targeted Gene Silencing
`
`Glen F. Deleavey1,* and Masad J. Damha1,*
`1Department of Chemistry, McGill University, 801 Sherbrooke Street West, Montre´ al, QC H3A 0B8, Canada
`*Correspondence: glen.deleavey@mail.mcgill.ca (G.F.D.), masad.damha@mcgill.ca (M.J.D.)
`http://dx.doi.org/10.1016/j.chembiol.2012.07.011
`
`Oligonucleotides (ONs), and their chemically modified mimics, are now routinely used in the laboratory as
`a means to control the expression of fundamentally interesting or therapeutically relevant genes. ONs are
`also under active investigation in the clinic, with many expressing cautious optimism that at least some
`ON-based therapies will succeed in the coming years. In this review, we will discuss several classes of
`ONs used for controlling gene expression, with an emphasis on antisense ONs (AONs), small interfering
`RNAs (siRNAs), and microRNA-targeting ONs (anti-miRNAs). This review provides a current and detailed
`account of ON chemical modification strategies for the optimization of biological activity and therapeutic
`application, while clarifying the biological pathways, chemical properties, benefits, and limitations of oligo-
`nucleotide analogs used in nucleic acids research.
`
`The concept of using synthetic oligonucleotides (ONs) to control
`the expression of specific genes dates back to the late 1970s,
`when Zamecnik and Stephenson first demonstrated targeted
`gene silencing using a short synthetic oligonucleotide (Zamecnik
`and Stephenson, 1978; Stephenson and Zamecnik, 1978). From
`these seminal studies emerged the ‘‘antisense’’ approach for tar-
`geted gene silencing. Here, an exogenous synthetic ON (termed
`an antisense ON, or AON) complementary to a target mRNA is
`introduced into cells with the intent to block gene expression
`through either translational inhibition or enzymatic cleavage of
`the mRNA target. AON-based therapeutics have been under
`clinical
`investigation for more than 30 years, achieving one
`approved drug (Fomivirsen/Vitravene). The gradual progress of
`AON therapeutic development has led to some hesitation
`regarding the viability of the platform. Subsequent to Zamecnik’s
`discoveries, elucidation of the RNA interference (RNAi) pathway
`for modulation of gene expression, and the role of small inter-
`fering RNAs (siRNAs) in the process, changed our understanding
`of posttranscriptional gene expression control, and renewed
`excitement in the nucleic acid therapeutics community (Fire
`et al., 1998; Elbashir et al., 2001).
`It is well established that properly designed AONs or siRNAs
`can cause cleavage of specific messenger RNA (mRNA) strands,
`taking advantage of endogenous cellular pathways to potently
`silence the expression of specific genes. These synthetic ON-
`directed approaches target mRNAs directly, before translation,
`eliminating the need for protein/enzyme inhibition using small
`molecules. Although siRNAs and AONs are perhaps the most
`commonly discussed ON-based agents under development,
`they are certainly not the only promising short synthetic ONs.
`Other relevant ON agents include microRNA-targeting ONs (anti-
`miRNAs) and antagomirs (discussed below), aptamers (Keefe
`et al., 2010), DNA/RNAzymes (Chan and Khachigian, 2009;
`Mulhbacher et al., 2010), exon-skipping and splice-switching
`compounds (Kole et al., 2012; Saleh et al., 2012), and immunos-
`timulatory nucleic acids (Barchet et al., 2008).
`The most significant of obstacles impeding the path to ON
`therapeutics include (1) their poor extracellular and intracellular
`
`stability, (2) low efficiency of intracellular delivery to targets cells
`or tissues, and (3) the potential for ‘‘off-target’’ gene silenc-
`ing, immunostimulation, and other side effects. Fortunately, in
`attempts to overcome the therapeutically limiting features of
`DNA and RNA, a vast array of ON chemical modifications
`has been developed. These nucleic acid analogs are often ratio-
`nally designed, allowing specific alterations to many of the in-
`herent properties of ONs affecting their biological application
`and potency (e.g., target binding affinity, nucleoside/nucleo-
`tide/duplex conformation, hydrophobicity, enzyme interaction,
`nuclease resistance, and immunostimulatory properties).
`Keeping with the aim of this review, biological pathways
`relevant to AON, siRNA, and anti-miRNA ON agents are briefly
`described, providing a foundation for subsequent discussions
`of therapeutic ON analogs. The underlying chemistries driving
`innovative strategies to optimize ON therapeutics are the focus
`of this review, and are best considered in relation to the biolog-
`ical pathways in which these ONs function. Concepts essential
`to the design of AONs, siRNAs, and anti-miRNAs are presented,
`along with perspectives on the current and future state of the art.
`Combining advances in biological understanding with chemical
`know-how is key to advancing this promising field.
`
`Biological Pathways
`Although specific details involved in some of the biological path-
`ways described remain to be elucidated, the fundamental events
`of these processes are now understood to the point that chem-
`ical modifications can be utilized in an effort to enhance the
`potency and therapeutic potential of ONs. Here, AON-mediated
`gene silencing, miRNA- and siRNA-induced gene knockdown
`through RNAi, and gene expression modulation by anti-miRNAs
`are introduced. Focus is given to mechanism and nucleic
`acid-enzyme interactions where possible, such that subsequent
`discussions of oligonucleotide modifications and pathway com-
`patibility can be made.
`AON-Mediated Gene Silencing
`AONs are perhaps the oldest and most studied class of gene-tar-
`geting ONs. Indeed, AONs have been reported on and studied
`
`Chemistry & Biology 19, August 24, 2012 ª2012 Elsevier Ltd All rights reserved 937
`
`Alnylam Exh. 1080
`
`

`

`Chemistry & Biology
`
`Review
`
`Figure 1. Fundamentals of mRNA-
`Targeting by AONs
`Exogenously introduced AONs can recognize and
`bind to target mRNA sequences. This can result in
`either mRNA cleavage via recruitment of RNase
`H1 (in humans), or translational arrest. In both
`situations, the result is downregulation of gene
`expression. Although shown in the cytoplasm,
`RNase H may also function in the nucleus.
`
`the cellular target of the AON is a region
`of the mRNA from the gene of interest.
`Initially, the concept of gene silencing
`AONs that do not induce mRNA cleav-
`age as part of their biological function
`appears to be quite straightforward:
`AONs are chemically modified to have
`strong binding affinity to their mRNA
`targets, and function by binding RNA
`tightly, preventing ribosomal assembly
`(Bennett and Swayze, 2010). Trans-
`lational
`inhibition by these AONs is
`difficult to detect and measure (Bennett
`and Swayze, 2010), but can indeed be
`detected and assayed (e.g., via polysome
`profiling experiments) (Baker et al., 1997).
`However,
`the situation can become
`considerably more complex.
`In some
`cases, AONs can be designed to bind
`mRNA regions that prevent ribosomal
`0
`assembly at the 5
`cap (Bennett and
`Swayze, 2010), prevent polyadenylation
`during mRNA maturation (Bennett and
`Swayze, 2010), or even affect splicing
`events (Watts and Corey, 2012; Kole
`et al., 2012; Saleh et al., 2012).
`AONs that achieve gene silencing
`by directing mRNA cleavage events are
`widely studied and used in both research applications and in
`therapeutic development, with several candidates explored in
`the clinic. In almost all cases, AONs stimulate mRNA cleavage
`through the recruitment of an endogenous endonuclease known
`as RNase H. RNase H enzymes, which are involved in essential
`steps of the DNA replication process, are known to cleave the
`RNA strand of a DNA-RNA duplex (Stein and Hausen, 1969;
`Lima et al., 2007; Cerritelli and Crouch, 2009; Bennett and
`Swayze, 2010). In humans, the specific enzyme recruited by
`AON-mRNA duplexes is RNase H1 (Wu et al., 2004).
`The cleavage mechanism through which RNase H mediates
`hydrolysis of an internucleotide phosphate linkage in the RNA
`strand of a DNA/RNA hybrid duplex is under active investigation,
`and relies on divalent metal ion(s) (Mg2+) for catalysis (Ho et al.,
`2010; Hollis and Shaban, 2011). A widely accepted mechanism
`for RNase H phosphodiester hydrolysis involves a model
`in
`which two highly coordinated metal cations (often described
`0
`as ‘‘metal A’’ oriented 3
`to the scissile phosphate, and ‘‘metal
`0
`) are present in the active site, coordinated
`B’’ oriented to the 5
`by acidic amino acid residues, nonbriding oxygen in the ON
`
`for nearly two decades before the initial reports on the RNAi
`pathway appeared. AON-based therapeutics have seen periods
`of both optimism and pessimism, perhaps partly owing to the
`longer than anticipated development time (Watts and Corey,
`2012). Nonetheless, significant progress has been made in
`recent years, with one approved drug on the market (Fomi-
`versen) and more than 20 candidates in early to late-stage clin-
`ical trials (Sanghvi, 2011; Bennett and Swayze, 2010; Watts
`and Corey, 2012).
`A wide variety of RNA-targeting ONs can be considered as
`AONs, although they may act through different mechanisms.
`Specific and strong AON recognition and binding to the
`mRNA target
`is accomplished through Watson-Crick base
`pairing, which may or may not be augmented by chemical
`modifications to the AON internucleotide phosphate linkages,
`backbone sugars, or nucleobases. The majority of AONs can
`be divided into one of two groups: those that direct cleavage
`of the target mRNA, and those that alter mRNA translation
`without causing mRNA cleavage. The basic concepts of these
`two modes of action are outlined in Figure 1. In both cases,
`
`938 Chemistry & Biology 19, August 24, 2012 ª2012 Elsevier Ltd All rights reserved
`
`

`

`Chemistry & Biology
`
`Review
`
`Figure 2. mRNA-Targeting by miRNA and
`siRNA
`Naturally produced miRNAs, loaded into RISC,
`are capable of recognizing and binding partially
`complementary mRNAs, especially those with
`sequence complementarity to the miRNA ‘‘seed
`region.’’ The result is a decrease in gene ex-
`pression, adding an additional
`layer of post-
`transcriptional control over gene expression. Anti-
`miRNAs can target and inhibit these miRNAs.
`As well, synthetic siRNAs, mimicking natural
`miRNAs, can be introduced into cells to utilize the
`cellular RISC machinery for targeted mRNA gene
`silencing.
`
`mise RNase-H function (Mangos et al.,
`2003 and references therein).
`RNAi-, miRNA-, and siRNA-
`Mediated Gene Silencing
`The RNAi pathway is an important cellular
`process triggered by endogenous nucleic
`acids, as a means through which cells
`can achieve posttranscriptional gene ex-
`pression control
`through the use of
`miRNA expression (Figure 2). RNAi was
`discovered following the observation
`that dsRNA sharing sequence with a
`cellular mRNA (Onc-22) silenced ex-
`pression of that gene in Caenorhabditis
`elegans (Fire et al., 1998). It was subse-
`quently found that RNAi could be trig-
`gered in mammalian cells using 21 nt-
`long dsRNA duplexes (Elbashir et al.,
`2001). The RNAi pathway in the mamma-
`lian system is the focus in this overview.
`A summary of the RNAi pathway is
`shown in Figure 2. The miRNA triggers
`of RNAi are typically produced from tran-
`scription of intergenic or intronic DNA
`regions (Duroux-Richard et al., 2011), to
`form primary miRNA (pri-miRNA) hairpins
`that are further processed into dsRNA
`duplexes featuring a hairpin loop with
`imperfect sequence complementarity, termed precursor miRNA
`(pre-miRNA). Pre-miRNAs are exported out of the nucleus and
`processed by an endoribonuclease called Dicer (usually with
`an associated dsRNA-binding protein, TRBP) to form mature
`miRNA (for reviews, see Filipowicz et al., 2008; Kawamata and
`Tomari, 2010; Lennox and Behlke, 2011).
`0
`miRNAs, which are 21 nt-long dsRNAs with 2 nt-long 3
`overhangs and imperfect complementarity, are recognized and
`loaded into a complex of enzymes and proteins known as
`the RNA-induced silencing complex (RISC). One of the two
`miRNA strands, called the ‘‘guide strand’’ or miRNA strand, is
`selected and used in the RISC complex. The other, called the
`‘‘passenger strand’’ or miRNA* strand, is discarded. The guide
`strand is ‘‘antisense,’’ or complementary, to the sequence of
`the targeted mRNA, whereas the passenger strand is ‘‘sense’’
`to the mRNA sequence. The RISC complex subsequently finds
`cellular mRNAs partially complementary to the loaded guide
`
`backbone, and water (Ho et al., 2010; Hollis and Shaban, 2011).
`Metal A may facilitate activation of water for in-line SN2-like
`nucleophilic attack on the internucleotide linkage, with metal B
`0
`-OH and
`in position to stabilize the transition state leading to 3
`0
`5
`-phosphate RNA cleavage products without cleaving the
`DNA strand. There is also evidence that a third Mg2+ cation
`may be involved in facilitating enzymatic activity, which can
`help explain the sensitivity of RNase H activity toward Mg2+
`concentration changes (Ho et al., 2010).
`Duplex conformation and flexibility in the AON/RNA hybrid is
`critical for the stimulation of RNase H activity, and could provide
`an explanation for the specificity of RNase H for DNA/RNA hybrid
`duplexes, but not dsDNA or dsRNA (Noy et al., 2004, 2005,
`2008). Structurally, dsDNA duplexes are B-form, and do not
`closely resemble DNA/RNA heteroduplexes. A-form dsRNA is
`more similar in conformation to DNA/RNA hybrid duplexes, but
`the rigidity of the two strands in dsRNA duplexes may compro-
`
`Chemistry & Biology 19, August 24, 2012 ª2012 Elsevier Ltd All rights reserved 939
`
`

`

`Chemistry & Biology
`
`Review
`
`During RISC loading with siRNA, one of the two siRNA
`strands is loaded into RISC, while the other is cleaved and
`unwound from the guide strand to be discarded (Matranga
`et al., 2005; Leuschner et al., 2006), although cleavage is not
`obligatory (Matranga et al., 2005). In flies, Dicer-2 and R2D2
`(an associated dsRNA binding protein)
`interact with siRNA,
`and facilitate this loading into RISC (Kawamata and Tomari,
`2010; Betancur and Tomari, 2012). In mammals, RISC loading
`is somewhat less understood, and at least in some cases, Dicer
`is not required for siRNA loading (Betancur and Tomari, 2012).
`The strand selection process in which RISC designates guide
`and passenger strands from duplex siRNAs and miRNAs is
`based on several factors, including duplex thermodynamics
`0
`(where the strand with the least tightly bound 5
`end is usually
`designated the guide strand; Khvorova et al., 2003; Schwarz
`0
`phosphate (that is required
`et al., 2003); the presence of a 5
`for binding of the guide strand within RISC; Schwarz et al.,
`0
`0
`2002); and the sequence at the 5
`of the strands (U and A 5
`nucleotides may be preferred over G and C for AGO2; Frank
`et al., 2010). After loading, siRNA directs gene silencing through
`mRNA cleavage and catalytic turnover mediated by a compo-
`nent of RISC, an endoribonuclease called Argonaute-2 (AGO2)
`(Figures 2 and 3).
`A particularly elegant model for siRNA guide strand selection
`and RISC loading has recently been proposed, in which dsRNA
`is processed in a catalytic region of Dicer to form an siRNA or
`miRNA product, which is subsequently released and reposi-
`tioned to make contacts with the Dicer helicase domain. Dicer
`interacts with the dsRNA on the less stable end of the duplex,
`and TRBP (or PACT) on the more stable end, spatially orienting
`the duplex for a subsequent AGO2 loading step in which TRBP
`(or PACT) may hand off the stable end of the duplex to the
`0
`-binding domain) of AGO2 (No-
`PAZ domain (the guide strand 3
`land et al., 2011). This proposed model suggests a mechanism
`through which guide strand selection can occur based on siRNA
`duplex thermodynamic asymmetry, corresponding to observa-
`0
`tions that the siRNA strand with the least tightly bound 5
`end
`frequently becomes the RISC guide (Khvorova et al., 2003;
`Schwarz et al., 2003).
`The fate of siRNA passenger strand during RISC loading has
`yet to be fully elucidated. Currently, there are three proposed
`mechanisms for dealing with the siRNA passenger strand during
`loading: (1) the siRNA passenger strand is cleaved by the hAGO2
`PIWI subdomain upon loading of the guide strand and released
`(Gaynor et al., 2010); (2) a bypass mechanism independent of
`passenger strand cleavage exists, in which an ATP-dependent
`helicase activity is invoked for passenger strand unwinding (Gay-
`nor et al., 2010); and (3) the passenger strand is ‘‘nicked’’ by
`hAGO2, triggering C3PO (component 3 promoter of RISC) to
`degrade the remaining passenger strand fragments (Ye et al.,
`2011). Interfering with passenger strand cleavage through intro-
`0
`modifications at the scissile nucleotide position can
`duction of 2
`significantly impair gene silencing, although not in all cases (Mar-
`tinez and Tuschl, 2004; Matranga et al., 2005; Muhonen et al.,
`0
`2007). RISC-mediated ON cleavage produces a 3
`-OH fragment
`0
`-phosphate fragment (Martinez and Tuschl, 2004),
`and a 5
`0
`0
`0
`suggesting a free 2
`hydroxyl group available to form the 2
`–3
`cyclic phosphate observed for some other nucleases is not
`necessary. These observations should be carefully considered
`
`Figure 3. hAGO2 with Loaded siRNA
`(A) A simplified diagram of AGO2 domains, and their interaction with siRNA.
`0
`The MID sub-domain binds the 5
`P and nucleotide of the guide strand. The
`PIWI sub-domain contains the RNase H-like fold with endonucleolytic
`cleavage activity, which cuts the siRNA passenger strand during loading, and
`the mRNA during gene silencing. The PAZ domain binds the guide strand
`0
`3
`OH.
`(B) Classical siRNA structure; 21 nt RNA duplex with 2 nt 3
`overhangs. The
`antisense strand (‘‘guide’’ strand) is complementary to target mRNA. The
`sense strand (‘‘passenger’’ strand) is complementary to the guide strand.
`0
`The guide strand contains the seed region at the 5
`.
`
`0
`
`strand sequence and prevents translation, either via translational
`arrest or mRNA cleavage. The extent of sequence complemen-
`tarity between the miRNA guide strand and the mRNA target is
`thought to determine whether translation arrest (partial comple-
`mentarity) or mRNA cleavage (near full complementarity) results
`from mRNA recognition by RISC (Humphreys et al., 2005).
`Most miRNAs are only partially complementary to their mRNA
`targets, thus translational arrest is more common. For the gene
`silencing event to occur at all, a short region composed of nucle-
`0
`otides 2–8 counting from the 5
`end of the guide strand (the
`‘‘seed region’’) must recognize a complementary sequence in
`0
`UTR (Humphreys et al., 2005).
`the mRNA, usually within the 3
`From a chemical perspective, miRNAs are essentially short
`dsRNA molecules, mimics of which can be readily synthesized.
`Small interfering RNAs, or siRNAs (Figure 3), are exogenously
`produced double-stranded RNAs, typically 21–24 nt in length
`0
`with 2 nt 3
`overhangs, designed to mimic miRNAs. siRNAs func-
`tion very similarly to miRNAs in gene silencing, and naming
`convention with siRNAs is the same, the loaded strand is the
`guide and the discarded strand is the passenger. Unlike
`miRNAs, siRNAs are designed to share full sequence comple-
`mentarity with a single target mRNA.
`
`940 Chemistry & Biology 19, August 24, 2012 ª2012 Elsevier Ltd All rights reserved
`
`

`

`Chemistry & Biology
`
`Review
`
`when designing chemically modified siRNAs, especially those
`modified near the hAGO2 cut site.
`Clearly, the RISC complex plays a central role in the RNAi
`pathway, performing essential
`functions in RNAi-mediated
`gene silencing. Optimizing interactions with RISC may be the
`key to improving the potency of siRNA therapeutics. At the
`core of the RISC complex lies an endoribonuclease belonging
`to the Argonaute (AGO) family, which both anchors the loaded
`guide strand from miRNA/siRNA, and performs the key endonu-
`clease function of the RISC complex via an RNase H-like fold (Liu
`et al., 2004). In humans, the AGO clade of the Argonaute family
`has four members: hAGO1, hAGO2, hAGO3, and hAGO4 (Gay-
`nor et al., 2010; Gagnon and Corey, 2012). Of these, only
`hAGO2 has Slicer-type activity (Meister et al., 2004), and thus
`hAGO2 is of particular interest as the core catalytic component
`of RNAi gene silencing.
`four principal domains:
`Human AGO2, hAGO2, has
`(see Figure 3)
`(Jinek and
`N-terminal, PAZ, MID, and PIWI
`Doudna, 2009; Gaynor et al., 2010; Gagnon and Corey, 2012).
`When an siRNA is loaded into RISC, the guide strand becomes
`0
`highly associated with hAGO2. The 3
`end of the guide strand
`0
`is recognized by the PAZ domain (Ma et al., 2004), and the 5
`0
`nucleotide (with an essential 5
`phosphate) is anchored in the
`0
`MID domain 5
`binding pocket (Ma et al., 2005; Frank et al.,
`0
`2010). The 5
`phosphate can be added synthetically or by an
`endogenous kinase (i.e., Clp1) (Weitzer and Martinez, 2007).
`0
`0
`nucleotide to the 5
`binding pocket bends
`The anchoring of the 5
`it out and away from the rest of the guide strand, making it
`unavailable for the target mRNA base-pairing interactions seen
`in the neighboring seed region nucleotides (Frank et al., 2010;
`Gaynor et al., 2010). The PIWI subdomain of hAGO2 contains
`an RNase H-like activity dependent on two Magnesium ions
`(as described above), and catalyzes the mRNA target cleavage
`(Liu et al., 2004; Gaynor et al., 2010). The cleavage of the target
`RNA, either the siRNA passenger strand, or the mRNA, occurs
`between the nucleotides paired to nucleotides 10 and 11 of the
`0
`hAGO2 guide strand (measured from the 5
`end) (Soutschek
`et al., 2004; Kraynack and Baker, 2006; Ui-Tei et al., 2008; Judge
`et al., 2009).
`Anti-miRNAs and Gene Regulation
`As our understanding of the complex roles played by miRNAs
`continues to grow, so does the view that a wide range of disease
`might be treatable through miRNA targeting (for a good com-
`mentary, see Ambros, 2008). We now recognize that miRNAs
`play strikingly important functions. Hundreds of miRNAs have
`been identified in the human genome, which may be capable
`of regulating up to 60% of protein-coding genes (Filipowicz
`et al., 2008; Friedman et al., 2009). Because seed-region com-
`plementarity is all that is necessary to produce mRNA downregu-
`lation, and many different mRNAs can all contain the same short
`sequences in their UTRs, a single miRNA sequence is capable
`of exerting control over multiple genes. Indeed, miRNAs are
`involved in regulating numerous biological processes, ranging
`from cell differentiation and development to apoptosis, which
`has motivated significant efforts to develop means for exerting
`control over miRNA activity both in vitro and in vivo. Malfunctions
`in the miRNA regulation system have been implicated in devel-
`opmental changes, metabolism, rheumatoid arthritis (Duroux-
`Richard et al., 2011), viral
`infection, and cancer (Sonenberg
`
`and Hinnebusch, 2009; Lennox and Behlke, 2011; Bhayani
`et al., 2012).
`The importance of miRNA gene regulation has attracted a
`substantial amount of research into the development of agents
`and techniques for exerting control over miRNA function. Fortu-
`nately, successes in RNAi and AON research can be readily
`applied to miRNA study. Elucidation of the RNAi pathway has
`revealed the importance of miRNAs, their structure, and their
`production in the cell, and advances in AON design have facili-
`tated the development of miRNA-targeting ONs. Termed anti-
`miRNA ONs (sometimes called AMOs)
`(Lennox and Behlke,
`2011), the majority of this class of ONs function via a steric block
`mechanism, in which miRNA function is inhibited by strong
`hybridization with exogenously introduced anti-miRNAs in order
`to block RISC loading (Figure 2). Anti-miRNAs may be designed
`to trigger miRNA cleavage by RNase H as well, but this has been
`less common thus far (Lennox and Behlke, 2011). A typical anti-
`miRNA is a perfect complement to the miRNA target (Lennox and
`Behlke, 2011), and typically features chemical modifications to
`enhance nuclease stability and target binding affinity.
`
`Challenges in the Field
`The development of ON therapeutics based on AON, siRNA, and
`anti-miRNA platforms have all proven highly potent and effective
`in in vitro cell assays. However, as previously mentioned, trans-
`lation from the bench to the clinic has been hampered by signif-
`icant challenges arising from: (1) the poor in vivo stability of
`nucleic acids, (2) ineffective uptake of nucleic acids to target
`cells, and (3) the potential for off target effects (OTEs) and immu-
`nostimulation. Chemical modifications to native DNA and RNA
`structures have gone a long way in abrogating many of these
`obstacles, and will be discussed in the subsequent section,
`following a brief description of the specific challenges facing
`ON therapeutics.
`Nuclease Stability
`Nucleic acids, RNAs in particular, are rapidly degraded in cells.
`AONs, siRNAs, and anti-miRNAs all suffer from these cellular
`degradation mechanisms,
`leading to shortened duration of
`activity and systemic delivery challenges. Duplex RNAs, such
`as siRNAs, are more nuclease resistant than RNA single strands,
`however unmodified siRNAs are also degraded quickly in serum
`(Turner et al., 2007). A human ortholog of ERI-1 may be a major
`contributor in siRNA degradation (Kennedy et al., 2004). In the
`case of siRNAs, it appears that cleavage frequently happens
`after pyrimidines (Turner et al., 2007), which could be useful
`to consider when designing chemically modified siRNAs. As
`discussed below, chemical modification strategies have been
`developed to improve the nuclease resistance of ONs, without
`altering the nucleic acid sequence.
`Effective and Targeted Delivery
`Primary approaches being explored for achieving efficient
`cellular delivery of therapeutic ONs include both the develop-
`ment of ON conjugates (therapeutic ONs are covalently attached
`to moieties facilitating cellular uptake), and the development of
`delivery vehicles designed to encapsulate and shield ONs, as
`well as facilitate and target their cellular uptake. Strategies may
`be designed for either local or systemic administration, and
`systemic delivery approaches are now being investigated in clin-
`ical trials with some success (Yuan et al., 2011). However, the
`
`Chemistry & Biology 19, August 24, 2012 ª2012 Elsevier Ltd All rights reserved 941
`
`

`

`Chemistry & Biology
`
`Review
`
`delivery challenge remains formidable, and new, improved de-
`livery technologies will be welcome additions to the field.
`Nucleic acids can benefit tremendously from conjugation
`with biologically relevant moieties. There are examples in which
`nucleic acid conjugates have demonstrated increased circula-
`tion time, improved cellular uptake and endosomal release,
`and targeted tissue uptake and cellular localization. For thorough
`reviews of ON conjugates, their biological activity, and the
`synthetic strategies employed in oligonucleotide conjugate
`chemistry, please refer to these excellent reviews: Manoharan
`(2002), Lo¨ nnberg (2009), and Juliano et al. (2012).
`The use of delivery vehicles to transport siRNA has proven
`a very promising mode of ON delivery, protecting siRNA from
`degradation and increasing cellular uptake. In fact, the first
`example of targeted systemic delivery of siRNAs in humans
`was recently reported, in which cyclodextrin-based nanopar-
`ticles were used to deliver siRNAs to patient tumors, effecting
`reduction of target gene mRNA and protein (Davis et al., 2010).
`0
`An RNAi mechanism was evidenced by 5
`-RACE PCR, which
`demonstrated that mRNA cleavage occurred precisely at the
`predicted AGO2 cleavage site in the mRNA (Davis et al., 2010).
`However, delivery vehicles often enter cells via endocytosis,
`which can lead to entrapment of siRNAs in endosomes and lyso-
`somes rather than achieving their efficient release into the cyto-
`plasm (Li and Huang, 2006). Additionally, delivery vehicles often
`expose siRNAs to endosomal
`immune receptors, potentially
`leading to undesirable immunostimulatory responses.(White-
`head et al., 2011) Again, chemical modification of the siRNAs
`themselves may be the only option for abrogating immune
`responses. Further discussion of current ON delivery strategies
`(including in vivo strategies) can be found in recent reviews
`(Yuan et al., 2011; Juliano et al., 2012; Rettig and Behlke, 2012).
`OTEs and Immunostimulation
`The knockdown of unintended genes (mRNAs) following ON
`treatment is a well-known side effect of both siRNA and AON-
`mediated gene silencing, although these effects are more
`frequently associated with siRNAs. These hybridization-depen-
`dent OTEs are generally observed as a result of partial comple-
`mentarity between the ON agent and mRNAs other than the
`intended target. In the case of siRNAs, OTEs can happen when
`the siRNA passenger strand is selected as the RISC guide, or
`when the loaded guide strand functions in miRNA fashion,
`recognizing targets mainly on the basis of seed region comple-
`mentarity, which is a short sequence more likely to be found in
`multiple mRNAs (Jackson et al., 2006b). In the case of siRNAs,
`two sets of OTEs are possible. Careful sequence selection
`when designing siRNAs and AONs, as well as chemical modifi-
`cations to reduce the likelihood of siRNA passenger strands
`acting as guide strands, can be beneficial for minimizing OTEs
`(Deleavey et al., 2009). In addition, chemical modifications of
`the siRNA seed region can be used to reduce OTEs (Jackson
`et al., 2006a). In the case of anti-miRNAs, sequence selection
`is limited due to the inherent small size of the miRNA target,
`which could be problematic when targeting one miRNA from
`a family with significant sequence homology. Microarray tech-
`nology can be used to monitor changes in cellular gene expres-
`sion following oligonucleotide treatment (Jackson et al., 2003).
`Immunostimulation following ON treatment is another poten-
`tial side effect that is of concern in the development of ON ther-
`
`apeutics (Marques and Williams, 2005; Watts et al., 2008a), and
`can confuse experiments designed to measure gene silencing
`potencies of AONs and siRNAs (Marques and Williams, 2005;
`Kleinman et al., 2008). For example, the nonspecific innate
`immune responses triggered by siRNAs and AONs can cause
`changes in cellular gene expression levels, affecting gene
`silencing data and antiviral activity measurements, and can
`lead to phenotypic changes such as reduced tumor angiogen-
`esis (Kleinman et al., 2008). Immune system responses to short
`nucleic acids such as siRNAs is a complex topic, and has
`been reviewed in detail elsewhere (Whitehead et al., 2011).
`For the purposes of this review, it should be noted that dif-
`ferent cellular immune receptors, positioned in different cellular
`locations, detect AONs and siRNAs, potentially leading to cyto-
`kine release and changes in gene expression. siRNA receptors
`include TLR3 (dsRNA, cell surface, and endosomal), TLR7
`(ssRNA, endosomal), TLR8 (ssRNA, endosomal), MDA5 (cyto-
`plasm), RIG-I (cytoplasm), and PKR (cytoplasm).(Takeda and
`Akira, 2005; Judge and MacLachlan, 2008; Kleinman et al.,
`2008; Watts et al., 2008a; Zamanian-Daryoush et al., 2008; Del-
`eavey et al., 2009; Whitehead et al., 2011) AONs can be immu-
`nostimulatory as well. For example, when DNA sequences
`0
`-CpG motif(s), they can
`such as AONs contain unmodified 5
`be immunostimulatory upon recognition by TLR9 (endosomal)
`(Hemmi et al., 2000; Krieg, 2012). Immunostimulatory sequence
`0
`-UGU
`motifs have also been identified for siRNAs, such as the 5
`0
`GUU (Judge et al., 2005) and 5
`-GUCCUUCAA motifs (Hornung
`et al., 2005). In addition, it should be noted that the type of
`delivery vehicle used will influence the mode of cellular uptake,
`which can determine the number and type of immune receptors
`to which the oligonucleotides can be exposed (Whitehead et al.,
`2011). Many of the chemical modifications described below can
`dramatically reduce the immunostimulatory properties of nucleic
`acids, providing an effective means for avoiding these potential
`side effects. It should be noted that abrogating immunostimula-
`tion is not always beneficial, as in the cases of isRNAs (Schlee
`et al., 2006) and oligonucleotide adjuvants (Klinman, 2004), but
`in siRNAs, AONs, and anti-miRNAs, minimizing immunostimula-
`tion is often desired. Discussion of immunostimulatory oligonu-
`cleotides can be found elsewhere (Barchet et al., 2008; White-
`head et al., 2011).
`
`Chemical Modifications of ONs
`The continued development of chemically modified nucleoside
`analogs has provided nucleic acid chemists with the tools
`necessary to exert a remarkable amount of control over many
`important nucleic acid properties, including: binding affinity for
`RNA targets, structural preferences, nuclease stability, and im-
`munostimulatory properties. An exhaustive list of all nucleic
`acid chemical modifications is beyond the scope of this review;
`however, this section will discuss many of the frequently utilized
`chemical modifications, as well as some interesting r