Downloaded from
`
`
`
`rnajournal.cshlp.org Cold Spring Harbor Laboratory Press on December 26, 2017 - Published by
`
`
`
`
`
`siRNA function in RNAi: A chemical modification analysis
`
`YA-LIN CHIU and TARIQ M. RANA
`Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School,
`Worcester, Massachusetts 01605, USA
`
`ABSTRACT
`Various chemical modifications were created in short-interfering RNAs (siRNAs) to determine the biochemical properties
`required for RNA interference (RNAi). Remarkably, modifications at the 2ⴕ-position of pentose sugars in siRNAs showed the
`2ⴕ-OHs were not required for RNAi, indicating that RNAi machinery does not require the 2ⴕ-OH for recognition of siRNAs and
`catalytic ribonuclease activity of RNA-induced silencing complexes (RISCs) does not involve the 2ⴕ-OH of guide antisense RNA.
`In addition, 2ⴕ modifications predicted to stabilize siRNA increased the persistence of RNAi as compared with wild-type siRNAs.
`RNAi was also induced with chemical modifications that stabilized interactions between A–U base pairs, demonstrating that
`these types of modifications may enhance mRNA targeting efficiency in allele-specific RNAi. Modifications altering the structure
`of the A-form major groove of antisense siRNA–mRNA duplexes abolished RNAi, suggesting that the major groove of these
`duplexes was required for recognition by activated RISC*. Comparative analysis of the stability and RNAi activities of chemically
`modified single-stranded antisense RNA and duplex siRNA suggested that some catalytic mechanism(s) other than siRNA
`stability were linked to RNAi efficiency. Modified or mismatched ribonucleotides incorporated at internal positions in the 5ⴕ or
`3ⴕ half of the siRNA duplex, as defined by the antisense strand, indicated that the integrity of the 5ⴕ and not the 3ⴕ half of the
`siRNA structure was important for RNAi, highlighting the asymmetric nature of siRNA recognition for initiation of unwinding.
`Collectively, this study defines the mechanisms of RNAi in human cells and provides new rules for designing effective and stable
`siRNAs for RNAi-mediated gene-silencing applications.
`Keywords: RNAi; siRNA; human; nucleotide modification; GFP
`
`INTRODUCTION
`
`The evolutionarily conserved phenomenon RNA interfer-
`ence (RNAi), the process by which specific mRNAs are
`targeted for degradation by complementary short-interfer-
`ing RNAs (siRNAs), has increasingly become a powerful
`tool for genetic analysis and is likely to become a potent
`therapeutic approach for gene silencing (for review, see
`Hammond et al. 2001; McManus and Sharp 2002). Conse-
`quently, understanding the mechanism of RNAi has be-
`come critical for developing the most effective RNAi meth-
`odologies for both laboratory and clinical applications. The
`general mechanism of RNAi
`involves the cleavage of
`double-stranded RNA (dsRNA) to short 21–23-nt siRNAs.
`This processing event is catalyzed by Dicer, a highly con-
`served, dsRNA-specific endonuclease that is a member of
`the RNase III family (Hammond et al. 2000; Zamore et al.
`
`Reprint requests to: Tariq M. Rana, Chemical Biology Program, De-
`partment of Biochemistry and Molecular Pharmacology, University of
`Massachusetts Medical School, 364 Plantation Street, Worcester, MA
`01605, USA; e-mail: tariq.rana@umassmed.edu; fax: (508) 856-6696.
`Article and publication are at http://www.rnajournal.org/cgi/doi/
`10.1261/rna.5103703.
`
`2000; Bernstein et al. 2001; Hamilton et al. 2002; Provost et
`al. 2002; Zhang et al. 2002). Processing by Dicer results in
`siRNA duplexes that have 5⬘-phosphate and 3⬘-hydroxyl
`termini, and subsequently, these siRNAs are recognized by
`the RNA-induced silencing complex (RISC; Hammond et
`al. 2000). Active RISC complexes (RISC*) promote the un-
`winding of the siRNA through an ATP-dependent process,
`and the unwound antisense strand guides RISC* to the
`complementary mRNA (Nykanen et al. 2001). The targeted
`mRNA is then cleaved by RISC* at a single site that is
`defined with regard to where the 5⬘-end of the antisense
`strand is bound to the mRNA target sequence (Hammond
`et al. 2000; Elbashir et al. 2001b). For RNAi-mediated
`mRNA cleavage and degradation to be successful, 5⬘-phos-
`phorylation of the antisense strand must occur, and the
`double helix of the antisense-target mRNA duplex must be
`in the A form (Chiu and Rana 2002).
`One highlighted difference between mammalian RNAi
`and RNAi in other eukaryotes is the lack of an amplification
`system for long-term persistence of RNAi in mammalian
`cells. For example, in Drosophila, ∼35 molecules of dsRNA
`can silence ∼1000 copies of the targeted mRNA per cell and
`can persist over the course of many generations (Kennerdell
`
`1034
`
`RNA (2003), 9:1034–1048. Published by Cold Spring Harbor Laboratory Press. Copyright © 2003 RNA Society.
`
`Alnylam Exh. 1044
`
`

`

`
`
`Downloaded from rnajournal.cshlp.org on December 26, 2017 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`and Carthew 1998; Zamore 2001). In mammalian cells,
`RNAi only persists effectively for an average of ∼66 h before
`the siRNA is likely diluted out over the course of several cell
`divisions (Chiu and Rana 2002). The amplification that is
`seen in flies and other lower eukaryotes can potentially be
`attributed to three factors. One is that the conversion of
`long trigger dsRNA to smaller 21–23-nt siRNAs by Dicer
`adds a degree of RNAi amplification, whereas in mamma-
`lian cells long trigger dsRNA invokes the interferon re-
`sponse that activates the protein kinase PKR (Stark et al.
`1998). This suggests that only siRNA transfections success-
`fully trigger RNAi in mammalian cells without other side
`effects, and thus, no amplification would take place through
`the processing of longer RNAs. A second factor in amplifi-
`cation is the presence of RNA-dependent RNA polymerase
`(RdRP), which has been found in plants, worms, fungi, and
`flies (Cogoni and Macino 1999; Dalmay et al. 2000; Sijen et
`al. 2001). RdRP has been postulated to amplify target
`mRNA,
`through a random, degradative PCR model
`(Lipardi et al. 2001; Nishikura 2001; Sijen et al. 2001), into
`dsRNA, which can be targeted by Dicer. However, no RdRP
`homologs have been found in mammalian cells, and the
`3⬘-OH that is required for RdRP-dependent degradative
`PCR is not required for RNAi in mammalian cells (Chiu
`and Rana 2002; Schwarz et al. 2002; Stein et al. 2003),
`indicating that PCR-based amplification likely does not oc-
`cur in mammals. A third factor in amplification may be the
`high enzymatic turnover rate of RISC* during the targeting
`and cleavage of mRNA (Hutvagner and Zamore 2002),
`which may add a degree of amplification to RNAi induction
`in all eukaryotes, including mammals. However, as the per-
`sistence of RNAi occurs for only a short period of time,
`finding methods for increasing the longevity of siRNAs in
`human cells will be fundamental for applying RNAi to labo-
`ratory and therapeutic applications.
`To address this issue of siRNA stability for prolonging the
`duration of dsRNA-mediated gene silencing and to further
`dissect the mechanism of RNAi in human cells, various
`chemically modified nucleotides predicted to affect siRNA
`stability were incorporated into siRNAs to study whether
`specific modifications increased or decreased the efficacy
`and persistence of RNAi in vivo. The most important of
`these modifications was to the 2⬘-OH of the ribonucleotide
`that distinguishes RNA from DNA and is required for the
`nucleophilic attack occurring during the hydrolysis of the
`RNA backbone,
`the reaction catalyzed by degradative
`RNases. Our results showed that the 2⬘-OH was not re-
`quired for RNAi,
`indicating that structural rather than
`chemical properties of siRNA–mRNA duplexes were the key
`to inducing RNAi and that RISC* did not require the 2⬘-
`OH for ribonuclease activity. 2⬘-modified siRNAs also in-
`creased the persistence of RNAi in human cells. Modifica-
`tions that stabilized base-pairing interactions were also in-
`corporated into the antisense strand of siRNAs and were
`able to initiate RNAi, signifying that this class of chemical
`
`Chemical modification of siRNAs
`
`modifications could be used to increase the targeting effi-
`ciency of siRNAs for mRNA target sequences and for allele-
`specific inhibition of gene expression.
`Other chemical modifications affected the formation of
`the major groove of the A-form helix of the antisense-
`siRNA–target-mRNA duplex, and potentially disrupted H-
`bonds or sterically hindered protein contacts, most prob-
`ably preventing the RISC* complex from stably interacting
`with the dsRNA duplex. These modifications completely
`abolished RNAi, demonstrating that an intact major groove
`in the A-form helix and stable RNA–protein interactions
`were required for RNAi in human cells. Finally, previous
`observations of psorelan cross-linked siRNAs implied that
`unwinding of siRNA occurred from the 5⬘-end of the an-
`tisense strand and that complete unwinding may not be
`necessary for effective RNAi (Chiu and Rana 2002). By us-
`ing mismatched or chemically modified nucleotides on ei-
`ther the 3⬘ or 5⬘ half of the antisense strand within the
`siRNA duplex, we have shown here that RNAi depended on
`the integrity of the 5⬘, and not the 3⬘, half of the siRNA
`duplex, as defined by the antisense strand. Altogether, these
`results gave insight into the essential biochemical properties
`of functional siRNAs and how specific changes in the siRNA
`structure can affect the efficacy of RNAi. Furthermore, these
`studies present new methodologies for improving the sta-
`bility and utility of siRNAs for future RNAi applications.
`
`RESULTS
`
`2ⴕ-OH is not required for siRNA to enter the
`RNAi pathway
`
`Previous results showed that RNAi effects typically peaked
`between 42 and 54 h posttransfection, and targeted gene
`expression started to be restored by 66 h posttransfection
`(Chiu and Rana 2002). To determine if the duration of
`RNAi could be prolonged by increasing the half-life of
`siRNAs, various chemical modifications were made to
`nucleotides that affected siRNA stability. These modified
`siRNAs were then tested in an improved dual fluorescence
`reporter assay based on the original assay developed previ-
`ously (Chiu and Rana 2002). Briefly, GFP and RFP were
`constitutively expressed from pEGFP-C1 and pDsRed2-N1,
`respectively. EGFP mRNA was targeted for RNAi using a
`21-nt siRNA targeted to nucleotides 238–258 of the EGFP
`mRNA (Fig. 1A). The fluorescence intensity ratio of target
`(GFP) to control (RFP) fluorophores was determined in the
`presence of siRNA duplexes and normalized to that ob-
`served in the mock-treated cells. The sequence of EGFP
`siRNA and EGFP mRNA, the specific mRNA cleavage site,
`plus the structures of the chemically modified nucleotides
`are diagrammed in Figure 1. As outlined previously, the
`cleavage site was defined precisely 11 nt downstream of the
`target position complementary to the 3⬘-most nucleotide of
`
`www.rnajournal.org
`
`1035
`
`

`

`Downloaded from
`
`Chiu and Rana
`
`
`
`rnajournal.cshlp.org Cold Spring Harbor Laboratory Press on December 26, 2017 - Published by
`
`
`
`
`
`Sense strand (SSI
`
`siRNA
`
`21
`5· GCAGCACGACUUCUUCAAGdTdT
`
`Antisense strand (ASI dTdTCGUCGUGCUGAAGAAGUUC 5·
`21
`1
`
`s· m7GpppG,
`t
`Poly A
`/
`,,,.,.-- AAGCAGCACGACUUCUUCAAG---....
`258 ~
`'-----"""
`238
`I Cleavage on mRNA I
`
`A
`
`B
`
`EGFP mRNA
`
`1-~ -
`
`y o
`
`H
`
`2·-0eoxy
`
`2'-0Me
`
`2· -Fluoro-uridine = 2'FU
`
`2· -Fluoro-cytidine = 2'FC
`
`~
`
`o
`
`Y O
`
`OH
`
`5-Bromo-uridine = U[5Br]
`
`5-lodo-uridine = U[51]
`
`!-\:1
`I \:::1
`
`?
`
`O=P- S
`
`OH
`
`Y O
`
`OH
`
`Thioate linkage= P-S
`
`N3-Methyl Uridine = 3MU
`
`2,6-Dlamlnopurine = DAP
`
`and 2⬘-fluoro-cytidine (2⬘-FC), respec-
`tively, which have a fluoro group at the
`2⬘-position in place of the 2⬘-OH (Fig.
`1B). Where these modified 2⬘-FU, 2⬘-FC
`nucleotides reside in the siRNA se-
`quence are highlighted in red in Figure
`2A. Addition of a 2⬘-fluoro group
`should increase the stability of
`the
`siRNA by making the siRNAs less rec-
`ognizable to RNases, thereby providing
`siRNAs protection from degradation
`(see below). When measured in the dual
`fluorescence assay, 2⬘-FU, 2⬘-FC siR-
`NAs, modified only in the sense strand,
`only in the antisense strand, or in both
`strands, all
`showed decreased EGFP
`fluorescence when normalized to non-
`targeted RFP fluorescence that was com-
`parable to the normalized decrease seen
`with wild-type siRNAs (Fig. 2; Table 1,
`rows 1–4). These results indicated that
`the 2⬘-OH was not required for RNAi
`and that nucleotides modified with 2⬘-
`fluoro groups could be used in siRNA
`constructs to successfully induce RNAi-
`mediated gene silencing.
`To support the conclusion that the
`2⬘-OH was not required for RNAi, ad-
`enine and guanine deoxynucleotides
`that inherently have 2⬘-H in place of the
`2⬘-OH (Fig. 1B) were incorporated into
`the sense, antisense, or both strands of
`2⬘-FU, 2⬘-FC-modified EGFP siRNAs to
`determine their effect on RNAi (Fig. 2A;
`green nucleotides). When 2⬘-FU, 2⬘-FC
`nucleotides were incorporated into the
`EGFP siRNA antisense strand with gua-
`nine and adenine deoxynucleotides at
`positions 9, 10, and 13, which base pair
`with nucleotides lining the cleavage site
`(Fig. 2A), EGFP RNAi effects were al-
`most indistinguishable from wild-type
`levels (Fig. 2B; Table 1, row 5). This
`same antisense construct base-paired to
`2⬘-FU, 2⬘-FC-modified sense strands also showed consider-
`able EGFP silencing at ∼64% (Table 1, row 6). In addition,
`siRNAs that had the entire antisense strand replaced with
`2⬘-FU, 2⬘-FC, dATP, and dGTP nucleotides still showed
`moderate levels of RNAi activity at ∼42%, or ∼44% if the
`sense strand was also modified with 2⬘-FU, 2⬘-FC (Table 1,
`rows 7,8). All together, these results demonstrated that a
`2⬘-OH group was not required for RNAi-mediated degra-
`dation and, even more specifically, was not required for
`nucleotides base-paired with nucleotides lining the mRNA
`cleavage site. There was, however, a limit on the extent to
`
`FIGURE 1. Structures of EGFP siRNA and chemical modifications. (A) Graphical represen-
`tation of dsRNAs used for targeting EGFP mRNA. EGFP was encoded by the pEGFP-C1
`reporter plasmid. siRNAs were synthesized with 2-nt deoxythymidine overhangs at the 3⬘-end.
`The position of the first nucleotide of the mRNA target site is indicated relative to the start
`codon of EGFP mRNA. The sequence of the antisense strand of siRNA is exactly complemen-
`tary to the mRNA target site. (B) Structure and nomenclature of chemical modifications.
`
`the antisense guide siRNA (Elbashir et al. 2001a). The spe-
`cific chemical modifications, the particular siRNA strand(s)
`where modifications were made, and the effect of the
`chemically modified siRNA on RNAi activity are summa-
`rized in Table 1. The RNAi activity of siRNAs was evaluated
`with eight different siRNA concentrations (ranging from 1
`to 200 nM). Each experiment was completed in duplicate
`and repeated twice.
`The effects of modifying the 2⬘-OH of nucleotides on
`RNAi were studied by replacing uridine and cytidine in the
`antisense strand of siRNA with 2⬘-fluoro-uridine (2⬘-FU)
`
`1036
`
`RNA, Vol. 9, No. 9
`
`

`

`
`
`Downloaded from rnajournal.cshlp.org on December 26, 2017 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`Chemical modification of siRNAs
`
`TABLE 1. RNA interference mediated by chemically modified siRNAs
`
`Row no.
`
`EGFP siRNA
`
`Sense strand
`
`Antisense strand
`
`RNAi activity
`(%)
`
`RNAi activity
`(+ or −)
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`29
`30
`31
`32
`
`Unmodified Unmodified
`DS (WT)
`2⬘-FU, FC
`SS/AS-2⬘-FU, FC
`Unmodified
`2⬘-FU, FC
`SS-2⬘-FU, FC/AS
`Unmodified
`2⬘-FU, FC
`2⬘-FU, FC
`DS-2⬘-FU, FC
`2⬘-FU, FC + (9, 10, 13) dA, dG
`SS/AS-2⬘-FU, FC + (9, 10, 13) dA, dG
`Unmodified
`SS-2⬘-FU, FC/AS-2⬘-FU, FC + (9, 10, 13) dA, dG 2⬘-FU, FC
`2⬘-FU, FC + (9, 10, 13) dA, dG
`SS/AS-2⬘-FU, FC + dA, dG
`2⬘-FU, FC + dA, dG
`Unmodified
`SS-2⬘-FU, FC/AS-2⬘-FU, FC + dA, dG
`2⬘-FU, FC
`2⬘-FU, FC + dA, dG
`SS/AS-Deoxy
`Unmodified Deoxy
`SS-Deoxy/AS
`Deoxy
`Unmodified
`DS-Deoxy
`Deoxy
`Deoxy
`SS/AS-2⬘-OMe
`2⬘-OMe
`Unmodified
`SS-2⬘-OMe/AS
`2⬘-OMe
`Unmodified
`DS-2⬘-OMe
`2⬘-OMe
`2⬘-OMe
`SS/AS-P-S
`Unmodified
`P-S
`SS-P-S/AS
`P-S
`Unmodified
`DS-P-S
`P-S
`P-S
`SS/AS-2⬘-FU, FC + P-S
`2⬘-FU, FC + P-S
`Unmodified
`SS/AS-U[5Br]
`Unmodified U[5Br]
`SS/AS-U[5I]
`Unmodified U[5I]
`SS/AS-DAP
`Unmodified DAP
`SS-2⬘-FU, FC/AS-U[5Br]
`2⬘-FU, FC
`U[5Br]
`SS-2⬘-FU, FC, FC/AS-U[5I]
`2⬘-FU, FC
`U[5I]
`SS-2⬘-FU, FC/AS-DAP
`2⬘-FU, FC
`DAP
`SS/AS-3MU
`Unmodified
`3MU
`SS/AS-(11) 3MU
`Unmodified
`(11) 3MU
`SS/AS-(1, 2) mm
`Unmodified
`(1, 2) mm
`SS/AS-(18, 19) mm
`Unmodified
`(18, 19) mm
`SS/AS-2⬘-FU, FC + (1-13) dA, dG
`2⬘-FU, FC + (1-13) dA, dG
`Unmodified
`SS-2⬘-FU, FC/AS-2⬘-FU, FC + (1-13) dA, dG
`2⬘-FU, FC
`2⬘-FU, FC + (1-13) dA, dG
`SS/AS-2⬘-FU, FC + (9-19) dA, dG
`2⬘-FU, FC + (9-19) dA, dG
`Unmodified
`SS-2⬘-FU, FC/AS-2⬘-FU, FC + (9-19) dA, dG
`2⬘-FU, FC
`2⬘-FU, FC + (9-19) dA, dG
`
`93 ± 0.70
`83 ± 3.48
`92 ± 0.98
`83 ± 0.01
`85 ± 2.10
`64 ± 2.89
`42 ± 1.66
`44 ± 0.60
`0 ± 5.97
`38 ± 2.95
`0 ± 0.01
`16 ± 4.41
`25 ± 1.75
`0 ± 0.01
`42 ± 6.03
`62 ± 0.07
`47 ± 0.03
`22 ± 0.03
`70 ± 1.88
`59 ± 11.2
`51 ± 0.57
`31 ± 1.88
`42 ± 5.02
`35 ± 7.69
`0 ± 6.65
`0 ± 1.71
`35 ± 5.69
`77 ± 2.00
`43 ± 0.09
`45 ± 2.23
`91 ± 0.36
`64 ± 0.42
`
`++++
`++++
`++++
`++++
`++++
`+++
`++
`++
`−
`+
`−
`−
`+
`−
`++
`+++
`++
`+
`+++
`+++
`++
`+
`++
`+
`−
`−
`+
`+++
`++
`++
`++++
`+++
`
`Summary of the specific chemical modifications analyzed, the particular siRNA strand(s) modified, and the effect of the chemically modified
`siRNA on RNAi activity in HeLa cells. RNAi activity was quantified by the dual fluorescence assay and is presented as the inhibition efficiency
`of target gene (EGFP) expression when cells were treated with 50 nM modified siRNAs. For comparison, the RNAi activity of unmodified, or
`wild-type, duplex siRNA (DS) normalized to 93% was designated (++++). Modified siRNAs assigned (++++) showed >80% RNAi activity, (+++)
`showed 60%–80%, (++) showed 40%–60%, and (+) showed 20%–40%. Modified siRNAs showing <20% RNAi activity were considered
`nonfunctional (−) in the RNAi pathway. Each experiment measuring RNAi activity of siRNAs was completed in duplicate and repeated twice.
`
`which deoxynucleotides could substitute for ribonucleo-
`tides because replacing the entire siRNA sense strand with
`deoxynucleotides decreased EGFP gene silencing to ∼38%
`inhibition, and replacing either the antisense strand or both
`strands entirely with deoxynucleotides completely abolished
`EGFP RNAi (Fig. 2B; Table 1, rows 9–11). Nonetheless,
`these results collectively showed that nucleotides with either
`2⬘-F or 2⬘-H groups can selectively replace ribonucleotides
`within the siRNA sequence to effectively induce RNAi.
`An interesting result was seen by modifying the 2⬘-OH to
`a bulky methyl group to create 2⬘-OMe nucleotides that
`were incorporated into sense, antisense, or both strands of
`EGFP siRNAs (Fig. 1B). This modification was hypoth-
`esized to improve RNAi efficacy because 2⬘-OMe groups are
`thought to increase RNA stability by inducing an altered
`RNA conformation that is more resistant to nucleases
`(Cummins et al. 1995). This modification is also thought to
`
`increase RNA affinity for RNA targets and improve hybrid-
`ization kinetics (Majlessi et al. 1998). Despite these poten-
`tial benefits, 2⬘-OMe nucleotides incorporated into either
`the sense or antisense strand greatly diminished EGFP gene
`silencing to ∼25% or ∼16%, respectively, whereas double-
`stranded 2⬘-OMe-modified siRNAs completely abolished
`RNAi (Table 1, rows 12–14). These results indicated that
`the methyl group, as a bulky group, may severely limit the
`interactions between siRNAs, target mRNAs, and the RNAi
`machinery required for successfully mediating RNAi. It is
`worth noting that because the bulkiness of the methyl group
`would likely be the cause of decreased RNAi activity rather
`than the actual lack of the 2⬘-OH specifically, these studies
`still supported the conclusion that the 2⬘-OH was not re-
`quired for RNAi.
`In a final analysis of modifications that may potentially
`increase siRNA stability without disrupting RNAi potency,
`
`www.rnajournal.org
`
`1037
`
`

`

`
`
`rnajournal.cshlp.org Cold Spring Harbor Laboratory Press on December 26, 2017 - Published by
`
`
`
`
`
`Downloaded from
`
`Chiu and Rana
`
`A
`
`S' GCAGCACGACUUCUUCAA TdT
`c CGUCGUGCUGAAGAAGUUC s·
`
`s· GCAGCACGACUUCUUCAAG TdT
`dTd CGUCGUGCUGAAGAAGUUC s·
`
`WTDS siRNA
`
`SS/AS-2'FU , FC
`
`2' structure
`(cid:127) -OH
`(cid:127) -ti
`(cid:127) -F
`
`s· GCAGCACGACUUCUUCAAGHdT
`dTdTCGUCGUG GAAGAAGUUC s·
`
`SS/AS -Deoxy siRNA
`
`s· GCAGCACGACUUCUUCAAG JTdT
`Sense stra nd (SS)
`slRNA Antlsense strand (AS) dT TCGUCGUGCUG GAAGUUC 5'
`
`m7GpppG~ AAGCAGCACGACUUCUUCAAG--.....__,,, Poly A
`
`SS/AS -2'FU , FC+ (9, 10, 13) dA, dG
`
`EGFP mRNA
`
`Cleavage on mRNA
`
`B
`
`1.40 . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,
`
`1.20
`
`~
`~ 1.00
`0:::
`a:
`~ ffi 0.80
`
`"C
`Cl)
`N
`: : 0,60
`CV
`
`e 0
`
`Z 0.40
`
`~-"
`
`1.
`
`❖
`❖
`
`2
`
`3
`
`N
`
`4
`.,,
`
`5
`6
`~ SI
`
`7
`
`8
`
`N
`
`0.20
`
`0.00
`
`.,.
`8
`:IE
`
`EGFP
`siRNA
`(nM)
`
`10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
`9
`It) ~ !<l
`SI
`It) ~ !<l SI
`It) ~ !<l SI
`
`0
`0
`~ !<l
`
`N
`
`0
`0
`~ !<l
`
`N
`
`0
`0
`~ !<l
`
`OS
`
`SS/AS-Oeoxy
`
`SS/AS-2'FU , FC
`
`SS/AS -2'FU , FC +
`(9, 10, 13) dA, dG
`
`FIGURE 2. siRNA 2⬘-OH is not required to guide mRNA cleavage. (A) Sequence and structure of siRNA duplexes with modification at the
`2⬘-position of the sugar unit. Nucleotides with 2⬘-hydroxyl groups (-OH) are black. Nucleotides with 2⬘-deoxy groups (-H) are cyan. Nucleotides
`with 2⬘-fluoro groups (-F) are red. The cleavage site on the target mRNA is also shown (red arrow). (B) Ratios of normalized GFP to RFP
`fluorescence intensity in lysates from modified siRNA-treated HeLa cells. The fluorescence intensity ratio of target (GFP) to control (RFP)
`fluorophores was determined in the presence of EGFP siRNA duplexes with modifications at the 2⬘-position of the sugar unit. Normalized ratios
`at <1.0 indicate specific RNA interference effects. For comparison, results from unmodified duplex siRNA-treated cells are included.
`
`a thioate linkage (P–S) was integrated into the backbone of
`the EGFP siRNA strand(s). P–S linkages were previously
`used in antisense methodology for increasing resistance to
`ribonucleases (for review, see Stein 1996) and, therefore,
`were postulated to enhance the stability of siRNAs. Incor-
`porating the P–S linkages into the double-stranded siRNA
`sense strand led to moderate levels of RNAi activity (62%
`inhibition), whereas P–S linkages in either the antisense or
`both strands of the siRNAs led to just less than ∼50% RNAi-
`
`induced inhibition (Table 1, rows 15–17). These results im-
`plied that the P–S modifications did not prohibit RNAi-
`mediated degradation and only moderately affected the ef-
`ficiency of RNAi. Interestingly, incorporating 2⬘-FU, 2⬘-FC
`modifications into the antisense strand in addition to the
`added P–S linkages showed lower levels of EGFP gene si-
`lencing (Table 1, row 18), indicating that there was a syn-
`ergistic effect that decreased but did not inhibit RNAi-me-
`diated degradation when both the 2⬘-F groups and the
`
`1038
`
`RNA, Vol. 9, No. 9
`
`

`

`Downloaded from
`
`
`
`rnajournal.cshlp.org Cold Spring Harbor Laboratory Press on December 26, 2017 - Published by
`
`
`
`
`
`Chemical modification of siRNAs
`
`P–S linkages were incorporated into
`siRNAs.
`
`A
`
`(a)
`
`- - AS OS
`
`-- AS-2'FU, FC
`-- SSIAS-2'FU, FC
`
`DS-2'FU, FC
`
`10
`
`20
`
`"'
`"'
`Tlme(mln)
`
`50
`
`60
`
`- - AS
`- - OS
`_.,_ AS-P-S
`_.,_ SS/AS-P-S
`_.,_ 05-P-S
`
`100.0%
`
`.,_ ..
`
`80.0%
`
`< z
`a,:
`·;;;
`0
`~ 40.0%
`.5
`
`20.0%
`
`100.0%
`
`80.0"fo
`
`< z 60.0%
`a,:
`·;;
`0
`~ 40.0%
`.5
`
`20.0%
`
`(b)
`
`Stability of modified siRNAs and the
`persistence of their RNAi activity in
`vitro and in vivo
`
`As the above experiments showed that
`siRNAs modified with stabilizing 2⬘-FU,
`2⬘-FC groups could effectively mediate
`RNAi to levels comparable to wild type,
`it was necessary to show that
`these
`modifications did in fact enhance siRNA
`stability. To measure the stability of
`siRNA in cell extracts, unmodified or
`2⬘-FU, 2⬘-FC-modified EGFP antisense
`5⬘-labeled
`strand
`siRNAs
`with
`[␥-32P]ATP were annealed with sense
`strand siRNAs to form duplex siRNAs,
`which were then incubated in HeLa cell
`extracts. At various time points, siRNAs
`were extracted, analyzed on a 20% poly-
`acrylamide gel containing 7 M urea, and
`visualized by phosphorimager analysis.
`Smaller siRNA degradation products
`were visualized in this analysis (data not
`shown), indicating that the lost of intact
`siRNA observed during these experi-
`ments was not caused by dephosphory-
`lation of siRNAs. The top panel (a) of
`Figure 3A shows the stability of the vari-
`ous 2⬘-FU, 2⬘-FC-modified siRNAs as
`compared with wild-type siRNAs over
`time. Wild-type
`double-stranded
`siRNAs showed a steady loss of intact
`siRNAs over the course of the experi-
`ment, with only ∼7% of the original
`concentration of intact siRNAs remain-
`ing after 1 h inextract (Fig. 3A[a], dark
`blue line). Intact modified or unmodi-
`fied single-stranded antisense siRNAs
`were quickly lost over the time course
`and were virtually undetectable by 30
`min in extract (Fig. 3A[a], black and red
`lines). This result showed that single-
`stranded modified siRNA was as suscep-
`tible to degradation as wild-type siRNA,
`indicating that single-stranded siRNAs,
`modified or unmodified, are inherently
`less stable than duplex siRNA. Double-
`stranded siRNAs with 2⬘-FU, 2⬘-FC
`modifications in either the antisense
`strand or both strands remained pre-
`dominantly intact over the course of the
`experiment with ∼68% or ∼81%, respec-
`
`10
`
`20
`
`"'
`30
`Time(min)
`
`B
`
`1 . 20
`
`a. 1.00
`LL.
`
`f 0 .80
`
`LL.
`C,
`w
`'Cl 0 .60
`G.I
`N
`ii 0 .40
`E
`0 z 0 . 20
`
`EGFP siRNA
`aMOCK
`EIWT OS
`D DS-2'FU, FC
`
`o.oo
`
`6
`
`1 8
`
`30
`
`42
`
`54
`66
`Time {h)
`
`78
`
`90
`
`120
`
`144
`
`FIGURE 3. Extending the half-life of siRNA duplexes prolongs the persistence of RNA inter-
`ference in vivo. (A) Comparing the stability of unmodified siRNAs with siRNAs containing
`2⬘-fluoro-uridine and 2⬘-fluoro-cytidine (2⬘-FU, 2⬘-FC) modifications (a) and thioate linkage
`(P–S) modifications (b). Unmodified or modified EGFP antisense strand siRNAs (AS) were
`5⬘-labeled with [␥-32P]ATP by T4 polynucleotide kinases. Duplex siRNAs were formed by
`annealing equal molar ratios of sense-strand (SS) siRNAs with the 5⬘-32P-labeled antisense
`strand. To analyze siRNA stability in HeLa cell extract, 50 pmole of siRNA was incubated with
`500 µg of HeLa cell extract in 50 µL of reaction mixture containing 20 mM HEPES (pH 7.9),
`100 mM KCl, 10 mM NaCl, 2 mM MgCl2, and 10% glycerol. At various time points, siRNAs
`were extracted and analyzed on 20% polyacrylamide gels containing 7 M urea followed by
`phosphorimage analysis (Fugi). (B) Kinetics of RNAi effects of duplex siRNA with 2⬘-fluoro-
`uridine and 2⬘-fluoro-cytidine modification in HeLa cells over a 144-h time course. The
`fluorescence intensity ratio of target (GFP) to control (RFP) protein was determined in the
`presence of unmodified dsRNA (blue bars) and duplex siRNA with 2⬘-fluoro-uridine and
`-cytidine modifications (DS-2⬘-FU, 2⬘-FC, cyan bar) and normalized to the ratio observed in
`the presence of mock-treated cells (red bars). Normalized ratios at <1.0 indicated specific RNA
`interference.
`
`www.rnajournal.org
`
`1039
`
`

`

`
`
`Downloaded from rnajournal.cshlp.org on December 26, 2017 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`Chiu and Rana
`
`tively, of the original siRNA population remaining intact
`throughout the duration of the experiment (Fig. 3A[a],
`green and light blue lines). These results indicated that the
`2⬘-FU, 2⬘-FC modifications did, indeed, increase the stabil-
`ity of the siRNAs upon exposure to extract and that having
`these modifications in both strands provided the siRNAs
`with the most stability.
`In a similar experiment, the stability of P–S-modified
`EGFP siRNAs was evaluated. Unmodified, double-stranded
`antisense siRNAs showed about the same rate of siRNA loss
`as described in the above experiment (Fig. 3A[b], dark blue
`lines). However, P–S-modified single-stranded antisense
`siRNAs demonstrated a markedly increased rate of stability
`over the course of the experiment, showing ∼63% of the
`original siRNAs remaining intact after 1 h inextract as
`compared with 0% intact for single-stranded unmodified
`antisense siRNAs (Fig. 3A[b], black and red lines). The
`stability of double-stranded siRNAs with P–S modifications
`in both strands was comparable to the stability seen with the
`modified single-stranded antisense strand, with ∼63% of the
`original siRNA population remaining intact after 1 h (Fig.
`3A[b], light blue lines). Double-stranded siRNAs with P–S
`modifications in only the antisense strand showed weaker
`but still significant stability with ∼42% of the original
`siRNA population remaining intact through 1 h in extract
`(Fig. 3A[b], green lines). These results showed that the P–S
`modifications increased the stability of the siRNAs and,
`most notably, increased the stability of both single- and
`double-stranded siRNAs.
`These in vitro results indicated that siRNA stability is
`prolonged by these different modifications; however, it is
`important to note that these experiments address the gen-
`eral stability of siRNA in the context of endonucleases
`present in whole-cell extracts. Therefore, these experiments
`cannot distinguish whether the endonucleases affecting
`siRNAs in the in vitro assay would necessarily affect the
`stability of
`these various siRNAs in vivo. To address
`whether increased stability seen with modified siRNAs pro-
`longed the duration of RNAi in vivo, RNAi, induced by
`unmodified and 2⬘-FU, 2⬘-FC-modified double-stranded
`EGFP siRNAs, was assayed in the dual fluorescence reporter
`assay over a period of 144 h. To visualize RNAi effects over
`an even longer period of time, HeLa cells were transfected
`with modified or unmodified siRNA and, 36 h later, trans-
`fected with dual fluorescence reporter plasmids; RNAi ac-
`tivity persisted but was tapering by 168 h (data not shown).
`Also, growth of cells containing modified siRNAs was com-
`parable to cells containing wild-type siRNA, indicating that
`modified siRNAs were not affecting cell division (data not
`shown). Although 2⬘-FU, 2⬘-FC-modified EGFP siRNAs
`were slower to show RNAi effects by 6–18 h, maximal RNAi
`effects occurred by 42 h posttransfection for both modified
`and unmodified siRNAs. The maximal activity for both
`siRNAs was also in the same range, with both showing
`∼85%–90% inhibition of GFP expression. However, the
`
`1040
`
`RNA, Vol. 9, No. 9
`
`RNAi effects observed over the period of 66–120 h revealed
`that the effect of modified siRNAs was much more persis-
`tent than that of unmodified siRNAs. By 120 h posttrans-
`fection, the effect of modified siRNAs still remained at
`∼80% inhibition of GFP expression but the effect of un-
`modified siRNAs had dropped to less than ∼40% inhibition.
`Similarly, prolonged RNAi activity was observed with
`2⬘-FU, 2⬘-FC-modified siRNAs targeting endogenous hu-
`man Cyclin T1 mRNA when compared with wild-type siR-
`NAs targeting Cyclin T1 (see Discussion; Y.L. Chiu and
`T.M. Rana, unpubl.). Altogether, these results strongly in-
`dicated that there was a direct link between the duration of
`the RNAi effects and siRNA stability in human cells. Fur-
`thermore, these results showed conclusively that siRNAs
`stabilized by chemical modifications, like the 2⬘-FU, 2⬘-FC
`modifications, can be used to effectively induce and signifi-
`cantly prolong RNAi-mediated gene silencing in vivo.
`
`Modified siRNAs that stabilize A–U base-pair
`interactions can induce RNAi
`
`In addition to incorporating modifications that affected the
`stability of siRNAs, nucleotides chemically modified to
`strengthen the base-pair interactions between two comple-
`mentary bases were analyzed. In theory, increasing the sta-
`bility of base-pair interactions may increase the targeting
`efficiency of siRNAs to target mRNA sequences. Increasing
`targeting efficiency may then induce more robust RNAi
`effects with siRNAs that are weaker at binding to their target
`sequence or have mismatched sequences, and thus, are not
`showing a high degree of RNAi. This type of approach may
`also be used to significantly inhibit expression of one allele
`over another when both alleles are present in the same cell.
`To bolster base-pairing interactions, 5-bromo-uridine
`(U[5Br]), 5-iodo-uridine (U[5I]), or 2,6-diaminopurine
`(DAP; Fig. 1B), which are modified nucleotides known to
`increase the association constant between A–U base pairs
`(Saenger 1984), were incorporated into siRNAs and tested
`in the dual
`fluorescence report assay. Double-stranded
`siRNAs having U[5Br], U[5I], or DAP modifications incor-
`porated into the antisense strand were capable of inducing
`RNAi activity at levels of ∼70% for U[5Br], ∼5