Published online 16 January 2012
`
`4125–4136
`Nucleic Acids Research, 2012, Vol. 40, No. 9
`doi:10.1093/nar/gkr1301
`
`mRNA knockdown by single strand RNA is
`improved by chemical modifications
`
`Henry J. Haringsma, Jenny J. Li, Ferdie Soriano, Denise M. Kenski,
`W. Michael Flanagan and Aarron T. Willingham*
`
`Sirna Therapeutics, 1700 Owens Street, Fourth Floor, San Francisco, CA 94158, USA
`
`Received November 8, 2011; Revised December 16, 2011; Accepted December 19, 2011
`
`ABSTRACT
`
`While RNAi has traditionally relied on RNA duplexes,
`early evaluation of siRNAs demonstrated activity of
`the guide strand in the absence of the passenger
`strand. However, these single strands lacked the
`activity of duplex RNAs. Here, we report the system-
`atic use of chemical modifications to optimize
`single-strand
`RNA
`(ssRNA)-mediated mRNA
`0
`F ribose modifica-
`knockdown. We identify that 2
`0
`-end phosphorylation vastly
`tions coupled with 5
`improves ssRNA activity both in vitro and in vivo.
`The impact of specific chemical modifications on
`ssRNA activity implies an Ago-mediated mechanism
`but the hallmark mRNA cleavage sites were not
`observed which suggests ssRNA may operate
`through a mechanism beyond conventional Ago2
`slicer activity. While currently less potent
`than
`duplex siRNAs, with additional chemical optimiza-
`tion and alternative routes of delivery, chemically
`modified ssRNAs could represent a powerful RNAi
`platform.
`
`INTRODUCTION
`
`RNA interference is a process inherently driven by RNA
`duplexes. The double-strand endonuclease Dicer processes
`long double-stranded RNAs (dsRNA) into small 21–23 nt
`0
`phosphate (1–4). The re-
`RNA duplexes containing a 5
`sulting siRNAs and miRNAs load into the RNA-induced
`silencing complex (RISC) where the passenger strand
`(sense) is dissociated, leaving the active guide strand (anti-
`sense) loaded in Ago2 and available to base pair with
`target mRNAs and initiate degradation (5,6). Exogenous
`delivery of small interfering dsRNAs (siRNAs) to mam-
`malian cells was shown to be an effective strategy to
`engage this endogenous mechanism of RNA-mediated
`regulation (7,8). This
`seminal discovery resulted in
`
`for studying
`siRNA becoming an indispensible tool
`gene function and now represents a new therapeutic
`modality (9,10).
`Despite the central role of duplex RNAs in triggering
`RNAi, in the early days of the development of siRNAs
`there were reports of single-strand RNA (ssRNA) acting
`to initiate mRNA knockdown. Using extracts from
`lines, Martinez et al.
`human cell
`(11) reported that
`ssRNA could reconstitute RISC; however, compared to
`duplex siRNAs, 10- to 100-fold higher concentrations of
`ssRNA were required. They also transfected single strands
`and duplexes into cells, thus demonstrating that ssRNA
`knockdown was optimal with a size >17 nt and improved
`0
`by the addition of a 5
`phosphate. Again, duplex siRNAs
`remained more potent. Schwarz et al. (12) evaluated both
`miRNA and siRNA, demonstrating that ssRNAs corres-
`ponding to the let-7 miRNA or luciferase could trigger
`0
`target cleavage but required a 5
`phosphate. The observa-
`tion that phosphorylation improves ssRNA activity indir-
`ectly argues for Ago-mediated activity. Phosphorylation
`0
`of the 5
`-end of the guide strand plays an important role
`in guide strand binding within the MID domain of Ago2
`and the subsequent positioning of the guide strand for
`accurate cleavage (13,14).
`Single-strand-mediated knockdown was shown to be
`dose-dependent but was 8-fold less effective than the
`corresponding duplex. Another study found that ssRNA
`targeting the human blood clotting initiator tissue factor
`(TF) was as effective in human cells as the duplex siRNA
`at high concentrations of 100–200 nM (15). In contrast to
`other reports, exogenous phosphorylation of the ssRNA
`was not found to improve knockdown activity. In a
`follow-up study, a dose–response comparison of
`the
`same single strand and duplex siRNAs
`found the
`dsRNA was 5- to 6-fold more potent than the correspond-
`ing ssRNA and analysis by northern blot demonstrated
`that ssRNA produced mRNA cleavage fragments (16).
`Core components of RISC have been shown to bind to
`ssRNAs. In vitro binding studies with purified human
`Ago2 demonstrate significantly higher binding affinity
`
`*To whom correspondence should be addressed. Tel: +1 650 496 1284; Fax: +1 650 496 6556; Email: aarron.willingham@merck.com
`
`ß The Author(s) 2012. Published by Oxford University Press.
`This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
`by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
`
`Alnylam Exh. (cid:20)(cid:19)(cid:22)(cid:26)
`
`

`

`4126 Nucleic Acids Research, 2012, Vol. 40, No. 9
`
`for ssRNA which bound 70- to 100-fold tighter than
`dsRNA (17). Recombinant Dicer has also been shown
`to bind 21 nt ssRNAs in vitro with preferential affinity
`0
`for ssRNAs containing a 5
`phosphate (18). This led the
`authors to hypothesize that Dicer may play a role in
`ssRNA knockdown by either facilitating loading of
`ssRNA into Ago2 or by binding to ssRNAs and thus pro-
`tecting them from nucleases (18).
`Chemical modifications have been employed to improve
`the pharmacokinetic properties of siRNAs including
`increasing
`nuclease
`stability,
`improving
`potency,
`reducing off target effects and abrogating immune stimu-
`0
`0
`ribose modifications (2
`-
`lation (19–21). Generally, 2
`0
`methoxy and 2
`-fluoro in particular) are broadly tolerated
`within the guide strand of duplex siRNAs. Holen et al.
`reported the first evaluation of the impact of chemical
`modifications on ssRNA activity and compared them to
`corresponding duplex siRNAs
`(16). The positional
`chemical sensitivity seen for duplex siRNAs (15) was
`mirrored by the ssRNAs which the authors interpret as
`indicating ssRNA and dsRNA act through a common
`RNAi pathway (16). This hypothesis was supported by
`their observation that an excess of dsRNA can competi-
`tively block the activity of ssRNAs in vitro (16).
`Hall et al. (22) reported that ssRNA potency can be
`significantly improved by the replacement of some of
`the phosphodiester backbone with boranophosphate
`modfications, where a non-bridging oxygen is replaced
`with an isoelectronic borane. Inclusion of 5–8 borano-
`phosphate modifications allowed ssRNAs to achieve
`knockdown comparable to duplex siRNAs at 10–
`25 nM concentrations.
`Importantly, HeLa cells pre-
`treated with siRNAs targeting Ago2 had significantly
`reduced target knockdown when they were subsequently
`transfected with ssRNAs, suggesting that ssRNAs act
`through Ago2 (22). Despite their beneficial properties,
`boranophosphate modifications present a number of syn-
`thesis and chemistry challenges and have thus far not been
`widely adopted (21).
`Recently,
`in unpublished work presented at the 7th
`Annual Oligonucleotide Therapeutics Society meeting
`et al.
`(September 2011), Swayze
`(23) described a
`modified ssRNA that included phosphorothioate back-
`0
`-end vinyl phosphonate modifications that
`bone and 5
`were intended to improve the stability of the ssRNA
`molecule. This modification pattern was a refinement of
`0
`F and
`an earlier published pattern of evenly spaced 2
`0
`2
`OMe modifications reported to improve the activity of
`duplex siRNAs (24). Swayze et al. (23) also reported that
`ssRNAs were inactive in Ago2 knockout cells indicating
`Ago2 mediates ssRNA activity. Importantly, subcutane-
`ously administered unformulated ssRNAs were active in
`rodents and comparable to the potency of second gener-
`ation RNase H antisense oligonucleotides.
`As discussed above, the limited exploration of chemical
`modifications has shown promise for improving the
`activity of ssRNA. Therefore, we conducted a systematic
`0
`ribose modifications on the ac-
`analysis of the impact of 2
`tivity of both duplex and ssRNA-mediated knockdown—
`screening six different RNAs to ensure the generality of
`0
`our observations. Modifications of the 2
`ribose ring
`
`0
`0
`F), and deoxy
`OMe), fluoro (2
`included methoxy (2
`0
`H), which applied to either purines or pyrimidines
`(2
`present on the guide strand. Additionally, guide strands
`were synthesized with or without phosphorylation of the
`0
`-end. In vitro screening in tissue culture cells identified
`5
`0
`0
`that 2
`F content and 5
`phosphorylation are key for
`optimal ssRNA-mediated knockdown of target mRNAs.
`Optimized ssRNAs were tested in mice using a lipid
`nanoparticle delivery vehicle. Knockdown of
`target
`mRNA by ssRNA exceeded 90% at the highest dose
`tested; however, the duplex siRNAs in the same study
`had greater overall knockdown and much longer duration.
`0
`Modification of the guide strand 5
`-end with abasic nu-
`cleotides negatively impacted the activity of both duplex
`and ssRNAs, consistent with an earlier report that an
`abasic nucleotide reduced Ago2 binding affinity and
`0
`cleavage activity (17). Taken together with the role of 5
`phosphorylation,
`these
`observations
`offer
`indirect
`evidence that ssRNAs act
`through an Ago-mediated
`mechanism. However, analysis of cleavage products by
`0
`5
`RACE and sequencing failed to identify distinct
`cleavage sites for ssRNAs despite the identification of
`hallmark Ago2 cleavage for duplex RNAs. This suggests
`ssRNA may act through a mechanism beyond canonical
`Ago2 slicer activity.
`0
`0
`While 2
`phosphorylation were
`-fluoro content and 5
`identified as greatly improving ssRNA activity, our side
`by side comparison with duplex siRNAs demonstrates
`that duplex RNAs are consistently more potent than
`their corresponding single strands both in vitro and
`in vivo.
`Importantly, as highlighted by preliminary
`results from Swayze et al. (23), with further improvements
`in chemical optimization and evaluation of alternative
`dosing strategies, such as sub-cutaneous delivery, ssRNA
`could represent a powerful alternative to duplex siRNAs.
`
`MATERIALS AND METHODS
`
`RNA oligo sequence and synthesis
`0
`
`A systematic screen of 2
`ribose modifications on six dif-
`ferent target sites (Table 1) resulted in a total of 180
`siRNAs which were then compared as duplex and single
`strands (360 in total). Individual sequences and modifica-
`tion patterns are listed in Supplementary Tables S3
`0
`and S4. Inverted abasic modifications present at the 5
`0
`and 3
`of the passenger strand serve to block loading
`into Ago2 (25,26). Sequences for siRNAs were selected
`using a previously published algorithm developed to
`
`Table 1. Target gene, site, accession and sequence for siRNAs
`
`Target
`gene
`
`ApoB
`ApoB
`ApoB
`ApoB
`SSB
`SSB
`
`Target
`site
`
`6981
`8786
`9470
`10127
`386
`963
`
`Target
`accession
`
`NM_009693
`NM_009693
`NM_009693
`NM_009693
`NM_009278
`NM_009278
`
`Target sequence
`
`CACAATGCATTTAGATCAA
`CCAGTAAGGCTTCTCTTAA
`CTTTAACAATTCCTGAAAT
`TCATCACACTGAATACCAA
`TAACAACAGACTTTAATGT
`AAATCATGGTGAAATAAAA
`
`

`

`Nucleic Acids Research, 2012, Vol. 40, No. 9 4127
`
`predict maximal knockdown while minimizing off-target
`hybridizations (27). siRNAs were synthesized using previ-
`ously described methods (28,29). Individual oligonucleo-
`tide strands were synthesized separately using solid-phase
`methods, purified by ion-exchange
`chromatography
`and duplexed when appropriate. Knockdown studies
`were normalized using a non-targeting control siRNA
`0
`0
`0
`OH (r), 2
`F (f), 2
`OMe
`sequence which contains ribose 2
`0
`H (d) residues at the indicated positions as well
`(m) and 2
`as inverted abasic caps (iB) on the passenger strand: guide
`0
`0
`–3
`)
`(fC;fC;fU;mG;mA;mA;mG;mA;mG;mA;mG;fU;
`(5
`0
`0
`fU;mA;mA;mA;rA;rG;rA;mU;mU) and passenger (5
`–3
`)
`( iB; fU; fC;fU;fU;fU;fU;dA;dA;fC;fU;fC;fU;fC;fU;fU;fC;
`dA;dG;dG;dT;dT;iB).
`
`diafiltration as previously described (30,32). The cat-
`ionic lipid CLinDMA (2-{4-[(3b)-cholest-5-en-3-yloxy]-
`butoxy}-N,N-dimethyl-3-[(9Z,12Z)-octadeca-9,12-dien-1-
`yloxy]propan-1-amine),
`cholesterol
`and PEG-DMG
`(monomethoxy(polyethyleneglycol)-1,2-dimyristoylglycer-
`ol) were mixed together at a molar ratio of 50:44:6.
`PEG-DMG was purchased from NOF Corporation, chol-
`esterol
`from Northern Lipids, and CLinDMA was
`synthesized by Merck & Co., Inc. Particle size was
`measured by dynamic light scattering using a Wyatt
`DynaPro plate reader and percent encapsulation was
`determined using a SYBR Gold fluorescence assay
`(Invitrogen) and were within pre-established quality
`metrics (32).
`
`In vitro mRNA knockdown
`
`In vivo mRNA knockdown
`
`Mouse Hepa1–6 cells were cultured in Dulbecco’s-
`modified Eagle Medium supplemented with 10% fetal
`bovine
`serum, 1% penicillin–steptomycin and 1%
`sodium bicarbonate. These cells were plated in a 96-well
`culture plates at a density of 3000 cells/well 24 h prior to
`transfection. Transfections were
`performed
`using
`Opti-MEM I Reduced Serum Media and Lipofectamine
`RNAiMAX as previously described (30). Final siRNA
`concentrations range from 100 to 1 nM for in vitro
`cell-based screens with concentrations varying for
`ssRNA (100 nM, 10 nM) and dsRNA (100 nM, 10 nM,
`and 1 nM) (see Supplementary Table S3). Final siRNA
`concentrations for the dose–response curves range from
`40 to 0.002 nM along an eight-point, 4-fold titration
`curve. Twenty-four hours post-transfection cells were
`washed with phosphate-buffered saline and processed
`using
`the TaqMan Gene Expression Cells-to-CT
`(Invitrogen), per manufacturer’s instructions, to extract
`RNA, synthesize cDNA and perform RT–qPCR using
`an Ssb (Mm00447374_m1) or ApoB (Mm01545154_g1)
`specific Taqman primer/probe set on an ABI Prism
`7900HT Sequence Detector. Reverse transcription condi-
`
`tions were as follows: 60 min at 37
`C followed by 5 min
`
`at 95
`C. RT–qPCR conditions were as follows: 2 min at
`
`
`50
`C, 10 min at 95
`C, followed by 40 cycles of 15 s at
`
`
`95
`C, and 1 min at 60
`C. Gapdh mRNA levels were
`used for data normalization (Taqman part number
`4308313). Knockdown of Ssb/ApoB was calculated as
`the percent knockdown in Ssb/ApoB cDNA measured in
`experimentally treated cells relative to the Ssb/ApoB
`cDNA levels measured in non-targeting, control-treated
`cells. The
`comparative Ct
`calculation method for
`knockdown has previously been described (31). Briefly,
`mRNA CtGapdh
` Ct = CtTarget
` Ct =
`and
` Ct(siRNA) Ct(non-targeting control) and relative expres-
` Ct
`sion
`level = 2
`and % knockdown =
`100 (1 2
` Ct). Potency (EC50) was calculated using
`a four-parameter curve fit
`tool and Prism graphing
`software (GraphPad Software).
`
`Lipid nanoparticle formulation
`
`siRNA lipid nanoparticles (LNPs) were assembled by
`simultaneous mixing of a lipid mixture in ethanol
`with an aqueous
`solution of
`siRNA followed by
`
`All in vivo work was approved by an Institutional Animal
`Care and Use Committee and adhered to standards rec-
`ommended by the Association for Assessment and
`Accreditation of Laboratory Animal Care, International.
`C57/BL6 mice (Jackson Labs) were dosed by intravenous
`injection with siRNA formulated in lipid nanoparticle at 3
`or 6 mg/kg as previously described (30,33). Animals were
`euthanized by CO2 inhalation and immediately after eu-
`thanasia liver sections were excised, placed in RNA Later
`
`C until ready for analysis. Liver
`(Qiagen) and stored at 4
`tissue was homogenized in Qiazol using stainless steel
`beads and a TissueLyser (Qiagen). Following homogen-
`ization,
`chloroform was added and samples were
`centrifuged. The aqueous layer was combined with an
`equal volume of 70% ethanol and samples were purified
`using an RNeasy purification kit per manufacturer’s
`directions
`(Qiagen). The
`resulting RNA was
`then
`normalized, cDNA was synthesized, and RT-qPCR was
`performed using ApoB specific Taqman primer/probe sets
`on an ABI Prism 7900HT Sequence Detector. Reverse
`transcription conditions were as follows: 60 minutes at
`
`
`37
`C followed by 5 min at 95
`C. RT–qPCR conditions
`
`
`were as follows: 2 min at 50
`C, 10 min at 95
`C, followed
`
`
`by 40 cycles of 15 s at 95
`C and 1 min at 60
`C. Gapdh
`mRNA levels were used for data normalization.
`Knockdown of ApoB was calculated using the same
`method described for the in vitro experiments.
`0
`0
`
`-RLM-RACE was performed using GeneRacer kit
`5
`(Invitrogen) following the manufacturer’s instructions.
`Briefly, 100 ng untreated RNA was directly ligated to the
`RNA linker. After phenol/chloroform extraction and pre-
`cipitation, the first-strand cDNA was synthesized using
`0
`the Oligo-dT primer. For
`the first-round 5
`RACE
`reaction, 1 ml of the first-strand cDNA was amplified
`0
`using the GeneRacer 5
`primer and a target gene specific
`primer (Supplementary Table S5) with cycling as follow-
`
`
`ing: 1 cycle of 94
`C for 2 min, then 5 cycles of 94
`C for 30 s
`
`
`and 72
`C for 1 min, and then 5 cycles of 94
`C for 30 s, and
`
`
`70
`C for 1 min, followed by 20 cycles of 94
`C for 30 s,
`
`
`65
`C for 30 s and 68
`C for 1 min. After additional
`
`
`C, ml of
`C and cool down to 4
`10-min incubation at 68
`the first PCR reaction was used for the second round
`
`5
`
`-RACE and sequencing of cleavage products
`
`

`

`4128 Nucleic Acids Research, 2012, Vol. 40, No. 9
`
`0
`
`0
`Nested
`RACE (Nested PCR) using the GeneRacer 5
`5
`primer and gene-specific nested primer with the 1 cycle
`
`
`
`C for 2 min, and 25 cycles of 94
`C for 30 s, 65
`C
`at 94
`
`C for 1 min. From this reaction, 10 ml was
`for 30 s, and 68
`analyzed on a 2% agarose gel. See Supplementary Table
`S5 for sequence and other information for the primers
`used. To confirm the cleavage site of target mRNA, the
`gel bands corresponding to the size of the expected
`0
`RACE amplicon was
`excised and purified using
`5
`QIAquick Gel Extraction kit (QIAGEN). The purified
`PCR products were sequenced by GENEWIZ (South
`Plainfield, NJ, USA).
`
`RESULTS
`
`Selection of siRNAs and modification patterns
`
`ApoB has been a frequent target of siRNA knockdown
`because of its potential therapeutic value in lowering
`serum cholesterol. Furthermore, given its predominant
`expression in liver hepatocytes, ApoB transcripts can be
`readily targeted in vivo using currently available siRNA
`delivery technologies (34–36). The Ssb gene encodes the
`La antigen which is involved in transfer RNA maturation
`(37) and has been implicated in the autoimmune disease
`Sjogren’s syndrome (38). Ssb was selected primarily for its
`ubiquitous cellular expression pattern. Four siRNAs tar-
`geting ApoB and two siRNAs targeting SSB were chosen
`to evaluate the generality of our findings across multiple
`siRNAs (Table 1).
`0
`0
`0
`H),
`-deoxy (2
`OH), 2
`Various combinations of ribose (2
`0
`0
`0
`0
`-fluoro (2
`F), and 2
`-methoxy (2
`OMe) modifications
`2
`were assigned based on the pyrimidine (Y) or purine (R)
`identity of the siRNA sequence (Table 2). Methoxy modi-
`fication of position 14 of
`the guide strand was not
`permitted due to reported sensitivity of this position to
`modifications
`(17). These
`combinations were
`also
`0
`phosphorylation
`evaluated in the presence or absence of 5
`0
`PO4). Placement of abasic residues
`of the guide strand (5
`0
`at the 5
`of the guide strand have been reported to reduce
`Ago2 binding and cleavage activity (17). Therefore, as a
`negative control, a separate set of oligos were synthesized
`
`with three abasic residues (i.e. ribose sugar without nu-
`cleotide base) at positions 1–3. The chemical structures
`0
`modifications are shown with adenosine as the
`of the 2
`representative nucleotide. All together, 30 different siRNA
`modification patterns were evaluated for each of the ApoB
`or SSB target sites (a total of 180 unique guide strands).
`Each of the different guide strands was evaluated as single
`strands or as duplexes with a passenger strand to create a
`conventional siRNA duplex. Passenger strands are largely
`0
`OMe uridine overhangs and
`ribonucleotides with 2
`inverted abasic residues on each end. Inverted abasics
`block the loading of these strands into Ago2, thus prevent-
`ing passenger strand competition with the guide strand
`(25,26). Sequence information for each guide strand and
`passenger strand are provided in Supplementary Tables S3
`and S4. For in vitro studies, all knockdown data were
`normalized to a chemically modified non-specific siRNA
`duplex which was designed to lack homology to nearly all
`mammalian transcripts (see ‘Materials and Methods’
`section).
`
`0
`F modifications are critical for the activity of single
`2
`strand RNA in vitro
`
`The mouse hepatocyte derived cell line Hepa1-6 was trans-
`fected with single strand and duplex siRNAs at 100 and
`10 nM concentrations. For dsRNAs, 10 nM is likely a
`saturating concentration as 100 nM did not result in sig-
`nificant improvements in observed target knockdown.
`Therefore,
`two siRNAs were also screened at 1 nM.
`Target knockdown was measured 24 h later
`and
`compared to unmodified controls. Figure 1 shows the
`results for ApoB (8786), though a similar trend was
`observed
`for
`the
`other
`five
`siRNAs
`tested
`(Supplementary Figure S1). Generally, dsRNAs were
`broadly tolerant of modifications
`though there was
`0
`-deoxy which is consistent with
`notable sensitivity to 2
`Ago2 preference for RNA (rather than DNA) substrates
`(13,14). High percentage methoxy modification of the
`guide strand was tolerated within dsRNAs though a loss
`in knockdown was seen at the submaximal 1 nM concen-
`tration. Such tolerance of methoxy modification is likely
`
`Table 2. Summary of modification strategy used to evaluate structure–activity relationship for single strand and duplex siRNAs
`0
`0
`p
`5
`aba
`5
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`+/ +/
`
`Key
`
`Modifications
`
`Y=r, R=r
`Y=d, R=d
`Y=f, R=f
`Y=m, R=m
`Y=f, R=r
`Y=f, R=d
`Y=f, R=m
`Y=r, R=f
`Y=d, R=f
`Y=m, R=f
`0
`
`Pyr (Y)
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`0
`2
`
`OH
`H
`F
`OMe
`F
`F
`F
`OH
`H
`OMe
`
`2
`2
`2
`2
`2
`2
`2
`2
`2
`2
`
`Pur (R)
`0
`0
`0
`0
`0
`0
`0
`0
`0
`0
`
`OH
`H
`F
`OMe
`OH
`H
`OMe
`F
`F
`F
`
`0
`0
`PO4 were tested and with or without three abasic
`p). Additionally, oligos containing a 5
`phosphate (5
`Oligos were prepared with or without 5
`0
`0
`0
`0
`0
`0
`0
`0
`0
`end of the guide strand (5
`aba). Unmodified ribose (2
`OH) is compared to 2
`-fluoro (2
`F), 2
`-deoxy (2
`H), and 2
`-methoxy (2
`OMe)
`residues at the 5
`modifications assigned to pyrimidines (Pyr) or purines (Pur). Key column lists nomenclature used in subsequent figures. Structures are shown with
`adenosine as the representative nucleotide. A total of 30 different siRNAs were designed for each target site. Sequences and modification patterns are
`listed in Supplementary Table S3.
`
`

`

`Nucleic Acids Research, 2012, Vol. 40, No. 9 4129
`
`dsRNAi (100nM)
`dsRNAi (10nM)
`dsRNAi (1nM)
`ssRNAi (100nM)
`ssRNAi (10nM)
`
`120
`
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`-20
`
`-40
`
`ApoB (8786) target site
`
`ApoB mRNA knockdown (%)
`
`Y=f, R=r
`
`Y=f, R=d
`
`Y=f, R= m
`
`Y=r, R=f
`
`Y=d, R=f
`Y= m, R=f
`
`-60
`
`Y=r, R=r
`Y=d, R=d
`
`Y=f, R=f
`Y= m, R= m
`
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`
`2'OH
`
`2'F
`
`2'H
`
`2'OMe
`
`Figure 1. Comparison of ssRNA and dsRNA oligos containing pyrimidine (Y) and purine (R) 2’ modifications. ApoB (8786) sequence shown.
`In vitro cell-based evaluation of ApoB mRNA knockdown (error bars represent standard deviation of four replicates). Unmodified ribose (‘r’) is
`compared to 2’-fluoro (‘f’), 2’-methoxy (‘m’), and 2’-deoxy (‘d’) in all combinations on purines (R) and pyrimidines (Y). Oligos do not contain a 5’
`phosphate. Guide strand oligo sequence and color-coded modification patterns are depicted. Five other siRNAs were tested and a similar require-
`ment of 2’F for ssRNA knockdown was observed (Supplementary Figure S1).
`
`0
`
`OH at position 14 of the
`due to our inclusion of 2
`guide strand which has previously been reported as
`highly sensitive to methoxy modifications (17).
`0
`F content exhibit
`Strikingly, single strands containing 2
`the most pronounced mRNA knockdown of the various
`single-strand modification patterns
`tested with 67%
`knockdown compared to the negligible activity seen
`for unmodified (Figure 1, Supplementary Table S3). Five
`other siRNA sequences were also evaluated and a similar
`0
`F dependence for ssRNA activity was observed, though
`2
`the magnitude of response was variable (Supplementary
`Figure S1). This variance is likely attributable to the fact
`that these RNA oligos were not phosphorylated, and there
`0
`phosphorylation of the
`is precedent in the literature that 5
`guide strand is key for ssRNA knockdown (11,12).
`0
`
`5
`
`phosphorylation improves the potency of ssRNA
`
`The same series of RNA modifications evaluated in
`0
`phosphor-
`Figure 1 were then tested in the presence of 5
`ylation of the guide strand. Figure 2 shows that addition
`0
`phosphate significantly improves the potency of
`of a 5
`ssRNA, resulting in significant knockdown even at a
`lower concentration of 10 nM. The overall effectiveness
`0
`of 2
`F-modified ssRNA activity for
`the other five
`siRNAs is also more evident when the ssRNAs are
`phosphorylated (Supplementary Figure S2).
`In the
`0
`presence of 5
`phosphorylation, the beneficial impact of
`0
`F content is significantly more obvious (compare to
`2
`Supplementary Figure S1). While the ssRNA knockdown
`at
`this
`lower
`concentration still
`lags
`that of
`the
`
`0
`
`F
`the combination of 2
`comparable duplex siRNAs,
`0
`modifications and 5
`phosphorylation are key to
`improving
`the
`potency
`of
`ssRNA knockdown:
`improving knockdown from negligible levels to 65–95%
`(Supplementary Table S3).
`0
`Pyrimidines modified with 2
`-fluoro appear to confer
`improved
`knockdown
`to
`ssRNAs
`than
`similarly
`0
`F
`modified purines. It does not seem that overall 2
`content of the RNA oligo can explain this difference as
`the ApoB siRNAs are either approximately evenly split
`between purine and pyrimidine content or are purine
`rich (58% purine). The Ssb sequences are predominantly
`pyrimidine rich (58–74%). The underlying mechanism for
`0
`F pyrimidine preference for ssRNA silencing activity
`2
`remains unclear. Overall, phosphorylated ssRNAs con-
`0
`F modifications of both purines and pyrimidines
`taining 2
`exhibited the best knockdown activity in vitro.
`
`ssRNA potency significantly underperforms
`duplex siRNAs
`
`At the two concentrations tested, ssRNAs appear less
`potent than their corresponding duplex siRNAs. Two
`ApoB siRNAs were selected for dose response measure-
`ments and calculation of EC50 values
`(Figure 3).
`0
`F-modified RNAs were compared as
`Unmodified and 2
`single strands and duplexes over an eight-point dose–
`response ranging from 40 nM down to 0.002 nM (all
`0
`F modifica-
`oligos were phosphorylated). The value of 2
`tions for ssRNA activity was immediately evident for both
`ApoB (8786) and (6981) sequences: unmodified ssRNAs
`
`

`

`dsRNAi (100nM)
`dsRNAi (10nM)
`dsRNAi (1nM)
`ssRNAi (100nM)
`ssRNAi (10nM)
`
`4130 Nucleic Acids Research, 2012, Vol. 40, No. 9
`
`120
`
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`-20
`
`-40
`
`ApoB (8786) target site
`
`ApoB mRNA knockdown (%)
`
`-60
`
`Y=f, R=f, 5'p
`Y=r, R=r, 5'p
`Y=d, R=f, 5'p
`Y=f, R=d, 5'p
`Y=r, R=f, 5'p
`Y=f, R=r, 5'p
`Y= m, R=f, 5'p
`Y=d, R=d, 5'p
`Y=f, R= m, 5'p
`Y= m, R= m, 5'p
`
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`U U A A G A G A A G C C U U A C U G G U U
`
`p
`
`p
`
`p
`
`p
`
`p
`
`p
`
`p
`
`p
`
`p
`
`p
`
`2'OH
`
`2'F
`
`2'H
`0
`0
`phosphorylation of the guide strand.
`modifications for ApoB (8786) in the presence of 5
`Figure 2. Comparison of purine versus pyrimidine 2
`In vitro cell-based evaluation of ApoB mRNA knockdown (error bars represent standard deviation of four replicates). Unmodified ribose (‘r’) is
`0
`0
`0
`0
`-fluoro (‘f’), 2
`-methoxy (‘m’), and 2
`-deoxy (‘d’) in all combinations on purines (R) and pyrimidines (Y). All RNA oligos contain 5
`compared to 2
`0
`0
`phosphate (5
`p). Guide strand oligo sequence and color-coded modification patterns are depicted. Five other siRNAs with were also tested with 5
`0
`phosphorylation for optimal in vitro ssRNA knockdown was observed
`phosphorylation (Supplementary Figure S2) and a similar requirement for 5
`(compare to Supplementary Figure S1).
`
`2'OMe
`
`0
`0
`Figure 3. In vitro potency of single strands compared to duplex siRNAs. (A) ApoB (8786) and (B) ApoB (6981) compared as 2
`OH and 2
`F guide
`phosphorylation. Single strands are 20- to 80-fold less potent than the corresponding duplex siRNAs containing identical guide
`0
`strands with 5
`strands. Potency and mRNA knockdown values listed in Supplementary Table S1.
`
`essentially had no knockdown over the dose range while
`0
`2
`F ssRNA exhibited a dose response. However, the
`duplex siRNAs were found to be 20- to 80-fold more
`potent
`than their corresponding single strands with
`picomolar potencies
`for dsRNAs compared to the
`nanomolar
`potencies
`measured
`for
`ssRNAs
`(Supplementary Table S1). ApoB (8786) duplexes
`0
`possess 15 pM potencies (for both unmodified and 2
`F)
`
`F ssRNA EC50 is 1.2 nM. The ApoB (6981)
`0
`while the 2
`F duplex is 2-fold more potent than the unmodified
`0
`2
`0
`duplex (2
`OH); however, both are significantly more
`F ssRNA (0.9 nM). Despite the
`0
`potent
`than the 2
`lower overall potencies relative to duplexes, it is important
`to note that ssRNAs are still capable of maximal mRNA
`knockdown >90% and that this level of knockdown was
`enabled by chemical modifications.
`
`

`

`Nucleic Acids Research, 2012, Vol. 40, No. 9 4131
`
`ApoB (6981) target site
`
`B
`
`120
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`(%) (mean +/- SD)
`
`ApoB mRNA Knockdown
`
`ApoB (8786) target site
`
`120
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`A
`
`(%) (mean +/- SD)
`
`ApoB mRNA Knockdown
`
`-20
`
`P B S
`
`single (2'F, 5'p)
`duplex (2'F, 5'p)duplex (2'O H, 5'p)single (2'O H, 5'p)
`
`
`
`
`
`-20
`
`P B S
`
`single (2'F, 5'p)
`duplex (2'F, 5'p)
`duplex (2'O H, 5'p)single (2'O H, 5'p)
`
`
`
`ApoB (6981) target site
`Day7
`
`Day14
`
`Day2
`
`D
`
`120
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`-20
`
`(%) (mean +/- SD)
`
`ApoB mRNA Knockdown
`
`ApoB (8786) target site
`Day2
`Day7
`Day14
`
`120
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`-20
`
`C
`
`(%) (mean +/- SD)
`
`ApoB mRNA Knockdown
`
`P B S
`
`P B S
`
`P B S
`
`duplex (2'F, 5'p)
`single (2'F, 5'p)
`duplex (2'F, 5'p)single (2'F, 5'p)
`
`duplex (2'F, 5'p)
`single (2'F, 5'p)
`
`P B S
`
`
`
`duplex (2'F, 5'p)single (2'F, 5'p)duplex (2'F, 5'p)
`
`
`
`P B S
`
`P B S
`
`
`
`single (2'F, 5'p)single (2'F, 5'p)duplex (2'F, 5'p)
`
`Figure 4. In vivo knockdown of single strand versus duplex siRNA demonstrates ssRNAi activity in vivo. RNAs formulated in a lipid nanoparticle
`(LNP) delivery vehicle. Modified guide strands (2’F) are compared to unmodified (2’OH). All oligos contain 5’ phosphorylation. Comparison of Day
`3 ApoB mRNA knockdown after an LNP delivered 3 mg/kg dose of dsRNA or ssRNA for target sites (A) ApoB (8786) and (B) ApoB (6981).
`Knockdown of 60–75% was observed for phosphorylated single strands containing 2’F modifications. Duration of mRNA knockdown after an LNP
`delivered 6 mg/kg dose was measured over 2 weeks (Day 2, Day 7, Day 14 time points) for (C) ApoB (8786) and (D) ApoB (6981). Knockdown
`values shown in Supplementary Table S2.
`
`ssRNA knockdown in vivo
`
`We extended our comparison of ssRNA and dsRNA into
`mice to identify whether ssRNAs were competent for
`knockdown in vivo. Single strand and duplex RNAs
`were formulated in lipid nanoparticles (LNP) which
`permitted side by side comparison of these RNAs while
`using a well-established delivery vehicle with proven
`capabilities (30,32). Since the nanoparticle encapsulates
`the siRNAs, it is likely shielded from serum nucleases
`which eliminated an additional
`round of
`siRNA
`chemical optimization for serum stability. The same
`ApoB siRNAs assayed for cell-based potency (Figure 3)
`were formulated in LNPs and intravenously dosed at
`3 mg/kg into mice. Three days later the mice were
`sacrificed and ApoB mRNA knockdown was measured
`from harvested livers. Compared to buffer control, the
`duplex siRNAs reduced ApoB expression by 85–98%
`(Figure 4A and B and Supplementary Table S2).
`Unmodified ssRNAs had no significant knockdown
`consistent with in
`vitro observations.
`which was
`0
`However 2
`F-modified single strands were quite effective,
`reducing ApoB expression levels by 61–74% and
`demonstrating that ssRNA functions in vivo.
`
`To evaluate duration of knockdown, a separate in vivo
`0
`F-modified ssRNA
`study was then conducted with the 2
`and dsRNA by using a higher dose (6 mg/kg) and con-
`ducting a 2-week time-course. Groups of mice were
`sacrificed at Days 2, 7, and 14, and ApoB mRNA
`knockdown was measured (Figure 4C and D). At the
`higher dose, ssRNA knockdown increased to nearly
`90% at Day 2, although the effect was transient—
`knockdown reduced to 71–77% at 7 days and returned
`to baseline after 14 days. In marked contrast, duplex
`siRNAs still had 95% knockdown two weeks after
`dosing. In summary, chemical modifications permit effec