RNA interference and chemically modified small
`interfering RNAs
`Muthiah Manoharan
`
`RNA interference (RNAi) is a powerful biological process for
`specific silencing of gene expression in diversified eukaryotic
`cells and has tremendous potential for functional genomics,
`drug discovery through in vivo target validation, and
`development of novel gene-specific medicine. The future
`success of this technology relies on identifying appropriate
`chemical modifications to improve stability, potency and
`in vivo cellular delivery. The present review summarizes
`the role of the chemist’s toolbox in this
`emerging technology.
`
`Addresses
`Alnylam Pharmaceuticals, Inc., 300 Third Street, Cambridge,
`Massachusetts 02142, USA
`e-mail: mmanoharan@alnylam.com
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`This review comes from a themed issue on
`Biopolymers
`Edited by Christian Leumann and Philip Dawson
`
`Available online 22nd October 2004
`
`1367-5931/$ – see front matter
`# 2004 Elsevier Ltd. All rights reserved.
`
`DOI 10.1016/j.cbpa.2004.10.007
`
`Abbreviations
`ENA
`ethylene-bridge nucleic acids
`LNA
`locked nucleic acid
`0
`MOE
`-O-methoxyethyl
`2
`RISC
`RNA-induced silencing complex
`RNAi
`RNA interference
`siRNA small interfering RNA
`Tm
`melting temperature
`
`Introduction
`RNA interference (RNAi) is an evolutionarily conserved
`process for the control of gene expression. During gene
`silencing by RNAi, double-stranded RNA (dsRNA)
`guides the degradation of mRNAs that are homologous
`in sequence to the dsRNA. Gene silencing by long 299
`base-paired dsRNAs was first observed in Caenorhabditis
`elegans by Fire and Mello [1] and silencing by small RNAs
`in plants was observed by Hamilton and Baulcombe
`[2]. The pioneering work on the structure and origin of
`the small interfering RNAs (siRNAs) came from bio-
`chemical analysis in Drosophila melanogaster carried out
`by Zamore et al. and Elbashir et al. [3,4]. These studies
`clearly established that dsRNA is processed to 21 or 22
`nucleotide siRNAs that in turn mediate gene silencing.
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`Subsequently, synthetic RNAi has been shown to silence
`genes in vitro in cultured human cells [5]. After the RNAi
`assembly into a protein–RNA complex (described below
`in detail), one of the strands of the siRNA duplex, the
`antisense strand or guide strand, hybridizes to the mRNA
`through formation of the specific Watson–Crick base
`pairs. The RNAi pathway activated by the hybridization
`of the guide strand results in degradation of the targeted
`RNA. Thus, a disease-causing protein, for example, is not
`produced [6–9]. This review attempts to summarize the
`mechanism of RNAi and describes some of the biological
`challenges associated with using synthetic siRNAs to
`effect RNAi, different chemical modifications being used
`to overcome these challenges, and future prospects.
`
`Mechanism of RNA interference
`The underlying mechanism of RNAi is probably evolved
`from a natural defense mechanism against foreign DNA
`that is transcribed in cells into long dsRNA; dsRNA is
`also a key intermediate in the life cycle of many viruses.
`Unlike the cells of simpler organisms, such as plants and
`insects, mammalian cells react to the presence of long
`strands of dsRNA by triggering cellular suicide. Long
`dsRNA (independent of its sequence) is known to induce
`the interferon response that can lead to cell suicide.
`Concomitant with the interferon response, it is sometimes
`possible to detect a sequence-specific RNAi response.
`Researchers have shown that the RNAi pathway is inde-
`pendent of the interferon pathway and is activated in
`mammalian cells by short synthetic siRNA duplexes of
`21–25 base pairs [5]. Analysis of the siRNAs isolated from
`Drosophila melanogaster showed that approximately 50%
`of the siRNAs were exactly 21 nucleotides in length. The
`0
`0
`monophosphates and 3
`hydroxyl groups
`siRNAs had 5
`0
`0
`and 3
`and no sequence bias was observed for the 5
`nucleotides at the cleavage sites. When short duplexes
`were used, gene silencing was more efficient when the
`0
`fragments had two nucleotide 3
`overhangs at each end
`than when the duplex was blunt ended [4].
`
`The mechanism of RNAi is shown in Figure 1. Within
`cells, the RNAi process comprises at least four sequential
`assembly steps. First is the ATP-dependent processing of
`double-stranded RNA by the action of ‘Dicer’, an RNase
`III-family enzyme, into siRNAs. In the natural process
`(Figure 1a), long dsRNAs are processed to siRNAs. In the
`second step, these naturally derived siRNAs, or synthetic
`siRNAs introduced into the cell (Figure 1b), are incor-
`porated into an inactive ribonucleoprotein complex. In
`the third step, ATP-dependent unwinding of the siRNA
`
`www.sciencedirect.com
`
`Alnylam Exh. 1046
`
`

`

`RNA interference and chemically modified siRNAs Manoharan 571
`
`Figure 1
`
`(b)
`
`Synthetic siRNA
`
`(a)
`
`Dicer
`RNase III
`
`ATP
`ADP + Pi
`
`dsRNA trigger
`
`Transgene,
`transposon, or virus
`
`siRNAs
`
`ATP
`ADP + Pi
`
`RISC
`
`RISC
`
`Selective gene silencing
`
`Argonaute RNase H-like cleavage
`
`RISC
`
`mRNA
`
`RISC
`
`Cleaved mRNA
`
`Current Opinion in Chemical Biology
`
`The mechanism of RNAi. RNAi has been observed in plants, fungi, mammals, worms, and flies and offers significant therapeutic potential.
`(a) RNAi as a natural process. (b) RNAi using synthetic siRNAs.
`
`duplex generates an active RNA-induced silencing com-
`0
`phosphates on
`plex (RISC). ATP is used to maintain 5
`siRNAs. Finally, in an ATP-independent reaction, the
`RNA target is cleaved. RISC is a sequence-specific endo-
`nuclease complex that contains Argonaute proteins and
`the single-stranded guide siRNA strand. RISC can cleave
`target RNAs with multiple turnovers without additional
`ATP [10,14,15] (and reviewed in [11–13]). The high
`specificity of the process results from the Watson–Crick
`base pairing of the siRNA antisense oligonucleotide with
`the target mRNA. The mRNA cleavage products are
`presumably degraded due to the lack of either a poly
`0
`-cap [16]. Very recently it was demonstrated
`(A) tail or 5
`that the endonuclease, or ‘Slicer’, activity responsible for
`mRNA cleavage is exclusively associated with a particular
`member of the Argonaute family, Ago2 [17,18]. Recent
`crystal structures indicate that
`the Argonaute PIWI
`domain responsible for the ‘Slicer’ activity is very similar
`to known RNase H structures [19].
`
`In a 21-mer siRNA duplex with mRNA, the cleavage
`position on the mRNA is located between nucleotides
`0
`-end of the antisense
`10 and 11 when counting from the 5
`strand upstream from the target nucleotide paired to
`0
`the 5
`terminal of the antisense strand [16]. Recently
`the Tuschl and Zamore groups [20,21] independently
`
`established the mechanism of mRNA cleavage by RISC.
`The Tuschl group used affinity-purified minimal RISC
`from human cells and showed that cleavage proceeds
`0
`-hydroxyl and a
`via hydrolysis and the release of a 3
`0
`-phosphate. Zamore’s group confirmed the cleavage
`5
`products and found that the RISC-mediated cleavage
`requires Mg2+ and that Ca2+ will not substitute [21].
`Moreover, a single phosphorothioate in place of the
`scissile phosphate blocked cleavage; the phosphorothio-
`ate effect was rescued by the thiophilic cation Mn2+, but
`not by Ca2+ or Mg2+. These results suggest that during
`catalysis, Mg2+ is bound to the RNA substrate through a
`non-bridging oxygen of the scissile phosphate. In this
`regard, the mechanism is similar to that of RNase H and
`other RNase III superfamily enzymes.
`
`Design of siRNA
`Sequence considerations for design of siRNA
`Generally, each strand of the siRNA duplex is 21 nucleo-
`0
`tides in length, designed with two nucleotide 3
`over-
`hangs on each strand (the double overhang structure). It is
`the natural structure derived from the RNAi mechanism
`described above. Sequence constraints mostly derive
`from synthetic considerations, either to simplify chemical
`synthesis or to make use of pol III expression vectors. As
`with primer/probe design, a G/C content of around 50% is
`
`www.sciencedirect.com
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`

`

`572 Biopolymers
`
`usually selected and oligonucleotides with three or more
`of any purine nucleotide in a row are avoided. A BLAST
`search is carried out to ensure that the chosen sequence
`does not target unwanted genes.
`
`The sites of active duplexes are not predictable, but
`several groups have systematically studied numerous
`siRNAs targeting particular messages in attempts to
`correlate duplex features with function. In the first com-
`prehensive siRNA ‘gene walk’ study published, Harborth
`et al. evaluated 25 duplexes targeting the open reading
`0
`-untranslated region of lamin A/C [22]. The
`frame and 3
`G/C content of the oligonucleotides varied from 32% to
`0
`dTdT overhangs. All siRNAs substan-
`74% and had 3
`tially reduced message levels, with 23 of the 25 reducing
`expression by >75%. There was no bias based on position
`within the message. It has been observed that siRNA
`efficiency varies with small positional shifts [22,23]. The
`positional effects could be due to many factors. Possibi-
`lities include competition for binding with protein factors,
`thermodynamic stability, or sequence-dependence of the
`RISC complex. Naturally, the structure and accessibility
`of the target message will also play an important role [24].
`
`Recently, prediction algorithms have begun to offer some
`guidance for siRNA sequence selection [25]. Khvorova
`et al. [26] and Schwarz et al. [27] have shown that the most
`effective siRNAs tend to have weaker base pairing at the
`0
`0
`antisense side of the duplex relative to the 3
`antisense
`5
`side of the duplex. Reynolds et al. [28] have proposed
`several characteristics that correlate with efficacy from the
`dataset of Khvorova et al. [26]. Reynolds et al. [28] used
`their criteria, including base composition, strength of base
`pairing at particular positions in the duplex, overall ther-
`modynamic stability, and base identities, to select siRNA
`duplexes against five genes. The duplexes selected using
`their criteria had a mean percent silencing of 76% while
`those selected randomly gave only 39% mean silencing.
`Whereas introduction of mismatches in the central region
`of the antisense strand abolished silencing activity [16],
`the sense strand can tolerate limited mismatches [29].
`Schwarz et al. have intentionally incorporated mismatches
`0
`end
`in the terminal four base wing region between the 5
`0
`end of the sense strand
`of the antisense strand and the 3
`to improve activity [27]. These so called ‘forked’ siRNAs
`gave a mean silencing of 87% whereas non-mismatched
`duplexes had a mean silencing of 75%.
`
`In Drosophila, silencing was more efficient when the
`0
`overhangs at each end than
`dsRNA had two nucleotide 3
`when the duplex was blunt ended [3]. Most researchers
`0
`-TT overhangs (the
`still construct siRNA duplexes with 3
`‘Tuschl Design’) on both strands. Alternative designs are
`0
`possible: for example, siRNAs without 3
`overhangs were
`0
`-
`active in silencing in mammalian cells [30], and single 3
`overhang structures in the antisense strand were also
`active [31]. Several groups have also shown that dsRNA
`
`joined by a hairpin loop can effect target RNA cleavage
`through the RNAi pathway [22,32–34]. As single-stranded
`RNA is much more susceptible to nuclease degradation
`than double-stranded, and the recruitment by the RISC
`seems to require siRNA duplexes, most antisense-only
`single-stranded siRNAs have significantly less activity
`than the corresponding duplexes [35].
`
`Chemically modified siRNAs: need for
`chemical modifications
`Although the short double-stranded RNA molecules used
`for silencing have a relatively high stability and activity in
`cell culture, there are important reasons why chemical
`modification of one or both of the strands may be desired
`[36–38]. First, further enhancement of thermodynamic
`and nuclease stability are important for therapeutic appli-
`cations. Second, modifications will be required to increase
`the half-life of the siRNA duplexes in circulation in vivo.
`Third, chemical modifications are needed to improve
`the biodistribution and pharmacokinetic properties of
`siRNAs and to target siRNA to certain cell types. Fourth,
`one can envision improving the potency of siRNA with
`appropriate chemical modifications in terms of target
`binding affinity, favored conformational features of the
`modification such as RNA-like sugar pucker of the
`0
`nucleoside (C3
`-endo) and A-helix geometry, altering
`on-rates and off-rates of hybridization, and enhancing
`the product release. However, in designing chemical
`modifications, the safety issues of the chemical modifica-
`tions and their cost must also be addressed.
`
`Chemical modification of the termini and
`conjugate groups
`0
`A 5
`-phosphate group on the antisense strand, but not on
`the sense strand, is required for cleavage of target duplex
`0
`-hydroxyl group on
`[10,15]. Synthetic RNA bearing a 5
`the antisense strand can mediate RNAi function, as the
`RNA is phosphorylated by an endogenous kinase. When
`0
`-hydroxyl termini of the antisense strands were
`the 5
`0
`-OMe or other functionalities (Figure 2),
`blocked as 5
`0
`silencing was abolished [10,15,35]; 5
`modification of the
`sense strand had no effect [30,39]. As the antisense siRNA
`0
`phosphate
`with a fluorescent label conjugated to the 5
`group via a 6-carbon linker had the same activity as the
`unmodified oligonucleotide [22] and an siRNA with the
`0
`end by a 6-amino-
`antisense strand modified at the 5
`hexyl phosphodiester was effective [15], it appears that
`0
`phosphodiester linkage are
`modifications that retain the 5
`tolerated.
`
`0
`
`Structural studies [19,40] indicate that the 2-nucleotide 3
`overhangs used for most siRNA duplexes appear to be
`specifically recognized by the PAZ protein domain, found
`in Dicer RNase III as well as in Argonaute proteins.
`Interactions with this domain must contribute to the
`formation and possibly also the stability or binding of
`the guide (antisense) siRNA strand in RISC. Synthetic
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`www.sciencedirect.com
`
`

`

`RNA interference and chemically modified siRNAs Manoharan 573
`
`Figure 2
`
`O
`
`NH
`
`N
`
`O
`
`OH
`
`O
`
`MeO
`
`NH2
`N
`
`N
`
`O
`
`O
`
`O
`
`O
`
`O
`
`O
`5′-Methoxy modification
`
`3′-ddC
`
`O
`
`NH
`
`N
`
`O
`
`OH
`O
`
`PO
`
`O
`O
`
`NH3
`3′-Aminomodifier
`
`HN
`H
`
`O
`
`S
`
`NH
`H
`
`O
`
`HN
`
`O
`
`NH
`
`O
`
`B
`
`O
`
`OH
`P O
`
`O
`O
`
`H
`
`O
`ON
`H
`Cholesterol-aminohexanol
`
`OH
`
`NH
`
`O
`
`H
`
`HO
`
`H H
`
`HO
`
`H
`Lithocholicacid-aminohexanol
`
`OH
`
`NH
`
`N
`
`O
`
`O
`
`Laurin
`
`C32
`
`Current Opinion in Chemical Biology
`
`HO
`
`HO
`
`Illustrative terminal modifications evaluated.
`
`0
`
`siRNAs usually have two deoxy-thymidines at the 3
`termini of both strands. However, the available data
`0
`terminus of the antisense
`regarding modification of the 3
`strand of the siRNA duplex is not conclusive. In mam-
`0
`terminus were
`malian cells, unpaired nucleotides at the 3
`0
`end of the antisense strand
`not required [30]. When the 3
`and both termini of the sense strand were modified with
`either an inverted deoxy abasic residue or an amino group
`attached through a 6-carbon linker no reduction in activ-
`0
`0
`ity was observed [30]. 3
`-Puromycin and 3
`-biotin mod-
`ifications of the antisense strand had little or no effect on
`
`activity [39]. When the antisense strand was modified at
`0
`-end with ddC and an aminopropyl group was
`the 3
`0
`end via the phosphate group, the siRNA
`attached at the 3
`was active [15]. By contrast, siRNA duplexes lost all
`0
`-end of the antisense strand was mod-
`activity when the 3
`0
`0
`-O,4
`-C-
`ified with either 2-hydroxyethylphosphate or 2
`ethylene thymidine [29]. 2-hydroxyethylphosphate was
`tolerated in the sense strand [29].
`
`Certain terminal conjugates have been reported to
`improve cellular uptake (Figure 2). For example, siRNAs
`
`www.sciencedirect.com
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`

`

`574 Biopolymers
`
`0
`
`-end of the sense strand with choles-
`conjugated at the 5
`terol, lithocholic acid, lauric acid or long alkyl branched
`chains (‘C32’) improved in vitro cell permeation in liver
`cells [31]. Certain fluorescent chromophores did not
`0
`perturb gene silencing when conjugated to the 5
`-end
`0
`0
`-end of the sense siRNA strand or the 5
`-end of the
`or 3
`0
`-end of
`antisense siRNA strand, but conjugation to the 3
`the antisense siRNA abolished gene silencing [22].
`
`siRNA conjugated to TAT peptide or TAT-derived
`0
`oligocarbamate at the 3
`-end of antisense strand had
`efficient RNAi activity and perinuclear localization in
`HeLa cells similar to activity and localization of siRNA
`duplexes delivered by a transfection agent, but distinctly
`different from free TAT peptide nucleolar localization.
`The silencing activity was observed without the need for
`transfection agent, although a higher concentration was
`needed compared with normal transfection [41].
`
`Chemical modifications within the
`oligonucleotides
`Backbone modifications
`In general, backbone modifications cause a small loss in
`binding affinity, but offer nuclease resistance and may
`alter pharmacokinetic behavior of oligonucleotides. Har-
`borth et al. [22] targeted endogenous lamin A/C mRNA in
`human HeLa or mouse SW3T3 cells, with siRNA
`duplexes that contained phosphorothioate (P=S) back-
`bone modifications (Figure 3). This modification did not
`significantly affect silencing efficiency [22,42], although
`cytotoxic effects were observed when every second phos-
`phate (P=O) of an siRNA duplex was replaced by phos-
`phorothioate. Complete replacement of P=O by P=S also
`caused cytotoxicity [42]. Braasch et al. showed that phos-
`phorothioate linkages did not significantly enhance
`siRNA nuclease stability and reduced the melting tem-
`peratures of the duplexes as compared with unmodified
`RNA [43]. In another study, P=S modification was shown
`to reduce siRNA activity [44].
`
`Figure 3
`
`RNAs with a boranophosphonate backbone (P=B, Figure 3)
`have been synthesized using T7 RNA polymerase from
`nucleotide triphosphates with specific Rp linkages [45].
`The resultant compounds have Sp linkages and are
`chirally pure. They show improved nuclease resistance
`relative to racemic phosphorothioates and obviously have
`more nuclease resistance than unmodified RNAs. The
`0
`-modified compounds were resistant to diges-
`chimeric 5
`0
`-exonuclease bovine spleen phosphodies-
`tion with the 5
`terase. The oligonucleotides are lipophilic and maintain
`negative charge, which may contribute to important
`pharmacokinetic and pharmacodynamic properties. The
`P=B linkages were better tolerated in the sense strand
`than in the antisense strand of the siRNA duplex and
`exhibited good inhibition of green fluorescence protein
`activity [45]. While modifications in the central core
`region were not as well tolerated, siRNAs with many
`P=B modifications
`in the terminal
`regions of
`the
`oligonucleotides (equivalent to the ‘wings’ of antisense
`‘gapmers’) retained activity and silencing was observed
`for a longer time than with unmodified siRNAs.
`
`0
`-sugar modifications
`2
`Silencing by siRNA duplexes is compatible with some
`0
`types of 2
`-sugar modifications within the strands and
`effects depend on whether the sense or antisense strand is
`0
`0
`-H, 2
`-OMe,
`modified. Some of the sugar modifications (2
`0
`0
`0
`-O-MOE, locked nucleic acid (LNA) and
`2
`-F, 2
`-NH2, 2
`ethylene-bridge nucleic acids (ENA) evaluated are shown
`0
`-F, LNA and
`in Figure 4. Among these modifications, 2
`ENA provide significant increases in target binding affi-
`nities (with RNA hybridization melting temperature dif-
`ferences, DTm, of 2 to 48C per modification), others did
`0
`not have much effect, and 2
`-H cause a decrease in Tm
`(DTm = 0.58C per modification). Furthermore, but not
`
`surprisingly, the duplex must retain conformationally
`RNA-like A-type helical characteristics for effective gene
`silencing. However, conformational features alone do not
`decide the activity: even in A-helix forming RNA–DNA
`
`O
`
`O
`
`B
`
`O
`
`O
`
`B
`
`O
`
`O
`
`B
`
`B
`
`OH
`
`OH
`O
`
`O
`
`O
`H3B
`
`PO
`
`O
`
`B
`
`OH
`
`OH
`O
`
`O
`
`O
`S
`
`PO
`
`O
`
`Phosphorothioate RNA
`
`Boranophosphonate RNA
`
`Current Opinion in Chemical Biology
`
`B
`
`OH
`
`OH
`O
`
`O
`
`O
`O
`
`PO
`
`O
`
` RNA
`
`Backbone modifications evaluated.
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`www.sciencedirect.com
`
`

`

`RNA interference and chemically modified siRNAs Manoharan 575
`
`B
`
`O
`O
`
`O
`
`O
`
`B
`
`O
`
`O
`
`O
`X
`
`PO
`
`B
`
`O
`O
`
`CH3
`
`B
`
`O
`
`O CH3
`
`O
`2′-O-Methyl
`X = O, S
`
`O
`
`O
`
`O
`2′-O-MOE: 2′-O-(2-Methoxyethyl)
`X = O, S
`
`O
`
`S
`
`B
`
`O
`
`O
`
`O
`X
`
`PO
`
`B
`
`OH
`O
`
`S
`
`O
`O
`
`PO
`
`OH
`
`O
`4′-Thio-RNA
`
`Figure 4
`
`B
`
`O
`
`B
`
`O
`
`O
`
`O
`
`O
`X
`
`PO
`
`O
`
`DNA
`X = O, S
`B = Base
`
`B
`
`O
`
`O
`
`B
`
`T
`
`O
`
`OH
`
`O
`
`O
`O
`
`PO
`
`O
`
`F
`O
`
`O
`
`O
`
`B
`
`F
`
`O
`O
`
`PO
`
`2′-F: 2′-Deoxy-2′-Fluoro RNA
`
`2-Hydroxyethylphosphate DNA
`
`O
`
`O
`
`B
`
`NH2
`O
`
`O
`
`O
`O
`
`PO
`
`NH2
`
`O
`2′-NH2: 2′-Deoxy-2′-Amino RNA
`
`B
`
`O
`O
`
`O
`
`O
`ENA
`
`B
`
`O
`
`O
`
`O
`
`O
`S
`
`PO
`
`O
`
`O
`
`B
`
`B
`
`O
`
`O O
`
`O
`
`O O
`O
`
`PO
`
`LNA
`
`Current Opinion in Chemical Biology
`
`Sugar modifications evaluated.
`
`heteroduplexes, the activity was almost completely abol-
`0
`0
`ished when either of the strands was 2
`-deoxy, but 2
`-
`deoxy residues were tolerated in the overhangs and at the
`terminal ends [16,46]. Almost complete activity was
`0
`region of the antisense strand was free
`retained if the 5
`of modification [44]. On the other hand, several labora-
`0
`-fluoro modification with a
`tories have shown that the 2
`0
`high preference for C3
`-endo sugar pucker, was well
`tolerated with or without a phosphorothioate backbone
`0
`-F
`[22,43,44]. These results support the idea that the 2
`modification is well accommodated during the RNAi
`pathway. Capodici et al. treated HIV-infected activated
`0
`-F modified siRNA. The mod-
`CD4(+) T cells with a 2
`ified siRNA was resistant to RNase A, and the authors’
`observed activity without transfection agent. Under these
`
`inhibition of HIV infection was observed
`conditions,
`while the unmodified siRNA was ineffective [47].
`0
`
`In another study, 2
`-F modified siRNAs enhanced serum
`stability in cell culture and showed efficacy in silencing
`luciferase expressed from a co-injected plasmid in mice
`[48]. However, even though the modified siRNAs have
`greatly increased resistance to nuclease degradation in
`plasma, this increase in stability did not translate into
`enhanced or prolonged inhibition of target gene expres-
`sion in mice following tail vein injection. It appears that
`0
`-F modified siRNAs are functional in vivo, but are
`2
`not more potent than unmodified siRNAs in animals.
`0
`The unfavorable, pharmacokinetic properties of the 2
`-
`F-modified compound are likely to be the reasons, and
`
`www.sciencedirect.com
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`

`

`576 Biopolymers
`
`they compromised the pharmacodynamic features of this
`modification.
`
`0
`0
`Even though it has a C3
`-O-
`-endo sugar pucker, the 2
`methyl modification (DTm, ca. 08C to 18C per modifica-
`tion) was not tolerated when oligonucleotides were fully
`modified [16,30,43,44]. Incorporation of
`two to four
`nucleotides in either strand was well tolerated. When
`0
`-O-methyl,
`every other nucleotide was modified with 2
`oligonucleotides were relatively resistant
`to serum-
`derived nucleases and significant enhancement in activity
`was observed [30]. In a different study, full replacement
`0
`0
`0
`-OH by alternating 2
`-OMe/2
`-F improved the activ-
`of 2
`ity of an siRNA duplex significantly with an increase in
`nuclease resistance [49].
`0
`
`0
`
`-O-allyl modification in the 3
`-unpaired nucleotides
`The 2
`0
`-O-methoxyethyl
`had no effect on activity [42]. The 2
`(MOE) terminal modification with P=S linkage has been
`used to improve nuclease resistance in certain local
`delivery applications [50]. In a study targeting the PTEN
`gene, the MOE modification was tolerated in the anti-
`sense strand of the siRNA duplex. It was shown that
`appropriate placement of the MOE modification, for
`example, an alternating MOE/OH format or several
`MOE at the termini (as in an antisense ‘gapmer’), led
`to improved activity [51].
`0
`
`-amino modification has not been widely studied.
`The 2
`One study reported some loss of activity when the mod-
`
`ification was incorporated in the sense strand, and severe
`loss when the modification was in the antisense strand
`[46]. As the study involved long dsRNA in C. elegans, the
`loss of activity could have arisen from lack of efficient
`cleavage of the transcript by Dicer or at the amplifica-
`tion step. Additional studies with this modification are
`warranted.
`
`0
`-Thio-RNA
`4
`0
`-Thio-RNA modifications showed nearly 18C per mod-
`4
`ification increase in Tm in a hybridization assay. When
`both strands were modified, the increase was even higher.
`However, the increase in Tm per modification was less
`0
`-OMe RNA for
`the same
`than that observed for 2
`sequence. In a nuclease assay using isolated enzymes
`0
`0
`-exonuclease, 3
`-exonuclease, endonuclease S1,
`having 5
`0
`-thio RNA showed very high
`or ribonuclease A activity, 4
`nuclease resistance in each case. For gene silencing, this
`modification was better tolerated in the sense strand than
`in the antisense strand. In the sense strand, the modifica-
`tion was most effective when it was located at the termini
`of the strand rather than in the center [52,53]. In another
`0
`-thio modification was tolerated at the term-
`study, the 4
`inal ends of antisense and showed enhanced nuclease
`resistance [54].
`
`LNA modification
`The very high binding affinity modification LNA (Figure 4)
`has been evaluated as a modification for siRNA. Duplexes
`0
`0
`-termini, the 3
`-termini, or
`modified with LNA at the 5
`
`Figure 5
`
`O
`
`NH
`
`N
`
`O
`
`OH
`
`Br
`
`O
`
`O
`
`O
`
`O
`
`NH
`
`N
`
`O
`
`OH
`
`I
`
`O
`
`O
`
`O
`
`S
`
`N
`
`NH
`
`O
`
`OH
`
`O
`
`O
`
`O
`
`O
`
`Me
`
`N
`
`N
`
`O
`
`OH
`
`O
`
`O
`
`O
`
`5-Bromouridine residue
`
`5-Iodoouridine residue
`
`4-Thiouridine residue
`
`N-3-Me-uridine residue
`
`NH2
`N
`
`N
`
`NH2
`
`N N
`
`OH
`
`O
`
`O
`
`O
`
`O
`
`N
`
`NH
`
`N N
`
`OH
`
`O
`
`O
`
`O
`
`H2N
`
`O
`
`O
`
`NH
`
`N
`
`O
`
`OH
`
`O
`
`O
`
`5-(3-aminoallyl)- uridine residue
`
`Inosine residue
`
`2,6-Diaminopurine residue
`
`Current Opinion in Chemical Biology
`
`Nucleobase modifications, evaluated.
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`www.sciencedirect.com
`
`

`

`RNA interference and chemically modified siRNAs Manoharan 577
`
`both in both strands, were able to block gene expression
`[43]. This is in contrast with data from Hamada et al. [29]
`0
`0
`-O,4
`-C-ethylene thy-
`who did similar experiments with 2
`midine, a component of ethylene-bridge nucleic acids
`(ENA, Figure 4), at the termini of siRNA duplexes and
`observed complete loss of activity with ENA. An siRNA
`duplex with certain wing region LNA modifications was
`able to block protein expression with efficiency equal to
`the unmodified duplex [43]. As this duplex had a melt-
`ing temperature of over 958C, the experiment demon-
`strated that the Tm of duplexes can be extremely high
`without affecting the function of the RISC. This is
`consistent with the assumption that ATP hydrolysis
`drives the separation of the siRNA strands [10]. Some
`duplexes with LNA modifications in the central core
`were inactive. Both number and placement of LNA
`modifications affect activity, and the inactive duplexes
`apparently do not bind to RISC as they did not inhibit
`silencing of a control gene.
`
`Nucleobase modifications
`The nucleobase modifications evaluated in siRNA
`duplexes are shown in Figure 5. In their work in C. elegans,
`Parrish et al. showed that while 4-thiouridine and 5-
`bromouridine residues were tolerated in both sense
`and antisense strands of siRNA duplexes [46], bulky 5-
`iodo or cationic 5-(3-aminoallyl) modifications were not
`well tolerated. Replacement of guanosine by inosine
`reduced the activity of siRNA duplexes [46]. As this
`analysis was carried out with in vitro transcribed long
`dsRNA, the lack of gene inhibition could be due to inhi-
`bition of Dicer processing rather than to the base mod-
`ification. In a different study in HeLa cells, N-Me-uridine
`in place of uridine abolished the siRNA activity [44], and
`as a general conclusion from this study, all base modifica-
`tions produced deleterious effects of various magnitudes
`compared with the wild-type RNA. Surprisingly, this
`was the case with even 2,6-diaminopurine in place of
`adenine [44].
`
`Perspectives
`RNAi holds great promise as a tool for understanding
`biological functions of genes and as a therapeutic. Che-
`mical modification of one or both strands of the siRNA
`duplex has allowed numerous groups to understand the
`mechanism of siRNA processing and target cleavage.
`The effects of chemical modifications are being evalu-
`ated and certain rules are emerging, although some
`mixed observations exist. This may be due to differ-
`ences among the gene targets employed or to the dif-
`ferent sequences employed. Repeating the experiments
`with well-characterized siRNA duplexes in multiple
`targets and with multiple sequences will yield valuable
`information for better design of chemical motifs for
`efficacious siRNAs. Crystal structures of enzymes in
`the RISC pathway are being solved and RNA docking
`has allowed further insight into the mechanistic path-
`
`can be
`[18,19,40]. Chemical modifications
`ways
`designed for efficient RNAi using the results from these
`studies. Apart
`from such ‘rational drug design’
`approaches, chemical modifications will also be needed
`to improve pharmacokinetic properties of siRNA in
`order for effective use in human therapeutic applica-
`tions. The issues associated with tissue and cell-specific
`in vivo delivery of siRNAs have not been rigorously
`analyzed. The siRNA duplexes have at least 40 negative
`charges, high molecular weight and are very hydrophilic,
`and chemical modification will be required to allow
`broad biodistribution and effective cell permeation.
`Because of interest in antisense RNA and ribozymes
`over the past decade, the efficiency of RNA synthesis
`and analysis have improved considerably, but to be
`practical and competitive more efficient synthetic pro-
`cesses, especially with better protecting groups,
`improved purification techniques, and faster analytical
`methods need to be developed.
`
`Acknowledgements
`I am grateful to Professors T Tuschl, P Zamore and F Eckstein, for
`their advice. I want to thank my colleagues at Alnylam, in particular
`Drs KG Rajeev, P Hadwiger, T de Fougerolles, R Meyers, H-P
`Vornlocher for their input and Ms M Duckman for Figure 1.
`
`References
`1.
`Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE,
`Mello CC: Potent and specific genetic interference by
`double-stranded RNA in Caenorhabditis elegans.
`Nature 1998, 391:806-811.
`
`2. Hamilton AJ, Baulcombe DC: A species of small antisense
`RNA in posttranscriptional gene silencing in plants.
`Science 1999, 286:950-952.
`
`3.
`
`4.
`
`5.
`
`6.
`
`Zamore PD, Tuschl T, Sharp PA, Bartel DP: RNAi: double-
`stranded RNA directs the ATP-dependent cleavage of mRNA
`at 21 to 23 nucleotide intervals. Cell 2000, 101:25-33.
`
`Elbashir SM, Lendeckel W, Tuschl T: RNA interference is
`mediated by 21- and 22-nucleotide RNAs. Genes Dev 2001,
`15:188-200.
`
`Elbashir SM, Harborth J, Lendeckel W, Yalcin A, Weber K,
`Tuschl T: Duplexes of 21-nucleotide RNAs mediate RNA
`interference in cultured mammalian cells. Nature 2001,
`411:494-498.
`
`Kent OA, MacMillan AM: RNAi: running interference for the cell.
`Org Biomol Chem 2004, 2:1957-1961.
`
`7. McManus MT, Sharp PA: Gene silencing in mammals by small
`interfering RNAs. Nat Rev Genet 2002, 3:737-747.
`
`8.
`
`Tuschl T, Borkhardt A: Small interfering RNAs: a revolutionary
`tool for the analysis of gene function and gene therapy.
`Mol Interv 2002, 2:158-167.
`
`9. Mello CC, Conte D: Revealing the world of RNA interference.
`Nature 2004, 431:338-342.
`
`10. Nykanen A, Haley B, Zamore PD: ATP requirements and small
`interfering RNA structure in the RNA interference pathway.
`Cell 2001, 107:309-321.
`
`11. Bartel DP: MicroRNAs: genomics, biogenesis, mechanism,
`and function. Cell 2004, 116:281-297.
`
`12. Meister G, Tuschl T: Mechanisms of gene silencing by
`double-stranded RNA. Nature 2004, 431:343-349.
`
`13. Hutvagner G, Zamore PD: RNAi: nature abhors a double strand.
`Curr Opin Genet Dev 2002, 12:225-232.
`
`www.sciencedirect.com
`
`Current Opinion in Chemical Biology 2004, 8:570–579
`
`

`

`578 Biopolymers
`
`14. Martinez J, Patkaniowska A, Urlaub H, Luhrmann R, Tuschl T:
`Single-stranded antisense siRNAs guide target RNA cleavage
`in RNAi. Cell 2002, 110:563-574.
`
`15. Schwarz DS, Hutvagner G, Haley B, Zamore PD: Evidence that
`siRNAs function as guides, not primers, in the Drosophila and
`human RNAi pathways. Mol Cell 2002, 10:537-548.
`
`16. Elbashir SM, Martinez J, Patkaniowska A, Lendeckel W,
`Tuschl T: Functional anatomy of siRNAs for mediating efficient
`RNAi in Drosophila melanogaster embryo lysate.
`EMBO J 2001, 20:6877-6888.
`
`17. Meister G, Landthaler M, Patkaniowska A, Dorsett Y, Teng G,
`Tuschl T: Human Argonaute2 mediates RNA cleavage targeted
`by miRNAs and siRNAs. Mol Cell 2004, 15:185-197.
`
`18. Liu J, Carmell MA, Rivas FV, Marsden CG, Thomson JM, Song JJ,
`Hammond SM, Joshua-Tor L, Hannon GJ: Argonaute2 is the
`catalytic engine of mammalian RNAi. Science 2004,
`305:1437-1441.
`
`19. Song JJ, Smith SK, Hannon GJ, Joshua-Tor L: Crystal structure
`of Argonaute and its implications for RISC slicer activity.
`Science 2004, 305:1434-1437.
`
`phosphomonoester-producing
`2