4140–4156 Nucleic Acids Research, 2005, Vol. 33, No. 13
`doi:10.1093/nar/gki732
`
`Functional polarity is introduced by Dicer
`processing of short substrate RNAs
`Scott D. Rose, Dong-Ho Kim1, Mohammed Amarzguioui1, Jeremy D. Heidel2,
`Michael A. Collingwood, Mark E. Davis2, John J. Rossi1,3 and
`Mark A. Behlke*
`
`Integrated DNA Technologies, Inc. 1710 Commercial Park, Coralville, IA 52241, USA, 1Division of Molecular Biology,
`Beckman Research Institute of the City of Hope, 1450 East Duarte Road, Duarte, CA 91010-3011, USA,
`2Chemical Engineering, California Institute of Technology, 1200 E. California Boulevard, Pasadena, CA 91125, USA
`and 3Graduate School of Biological Sciences, City of Hope and Beckman Research Institute of the City of Hope,
`1450 East Duarte Road, Duarte, CA 91010-3011, USA
`
`Received May 13, 2005; Revised and Accepted July 6, 2005
`
`ABSTRACT
`
`Synthetic RNA duplexes that are substrates for Dicer
`are potent triggers of RNA interference (RNAi). Blunt
`27mer duplexes can be up to 100-fold more potent
`than traditional 21mer duplexes (1). Not all 27mer
`duplexes show increased potency. Evaluation of
`the products of in vitro dicing reactions using elec-
`trospray ionization mass spectrometry reveals that
`a variety of products can be produced by Dicer cleav-
`age. Use of asymmetric duplexes having a single
`2-base 30-overhang restricts the heterogeneity that
`results from dicing. Inclusion of DNA residues at
`the ends of blunt duplexes also limits heterogeneity.
`Combination of asymmetric 2-base 30-overhang with
`30-DNA residues on the blunt end result in a duplex
`form which directs dicing to predictably yield a single
`primary cleavage product. It is therefore possible to
`design a 27mer duplex which is processed by Dicer to
`yield a specific, desired 21mer species. Using this
`strategy, two different 27mers can be designed that
`result in the same 21mer after dicing, one where the
`30-overhang resides on the antisense (AS) strand and
`dicing proceeds to the ‘right’ (‘R’) and one where the
`30-overhang resides on the sense (S) strand and dicing
`proceeds to the ‘left’ (‘L’). Interestingly, the ‘R’ version
`of the asymmetric 27mer is generally more potent in
`reducing target gene levels than the ‘L’ version 27mer.
`Strand targeting experiments show asymmetric
`strand utilization between the two different 27mer
`forms, with the ‘R’ form favoring S strand and the
`‘L’ form favoring AS strand silencing. Thus, Dicer
`
`processing confers functional polarity within the
`RNAi pathway.
`
`INTRODUCTION
`
`RNA interference (RNAi) is a conserved pathway present
`in most eukaryotes where double-stranded RNA (dsRNA)
`triggers a series of biochemical events that culminates in
`sequence-specific suppression of gene expression (2–4).
`Long dsRNAs have been employed for many years as a means
`to modulate gene expression in plants (5), yeast (6) and
`Caenorhabditis elegans (7). Similar attempts in mammalian
`cells failed due to interferon activation. In Drosophila, dsRNA
`of >150 bp length efficiently induces an RNAi response. This
`response becomes weaker as RNA length decreases and
`25–38 bp duplexes are inactive; very short duplexes of 19–
`23 bp length regain activity (8). The biochemistry is slightly
`different in mammalian cells and 25–30 bp duplexes will
`strongly induce an RNAi response, but potency does decrease
`with length and 45 bp duplexes are mostly inactive (1). In vivo,
`long dsRNAs are cleaved by the RNase III class endoribo-
`nuclease Dicer into 21–23 base duplexes having 2-base
`30-overhangs (9,10). These species, called ‘small interfering
`RNAs’ (siRNAs), enter the RNA induced silencing complex
`(RISC) and serve as a sequence-specific guide to target
`degradation of complementary mRNA species (8).
`The discovery of siRNAs permitted RNAi to be used as an
`experimental tool in higher eukaryotes. Typically, siRNAs are
`chemically synthesized as 21mers with a central 19 bp duplex
`region and symmetric 2-base 30-overhangs on the termini.
`These duplexes are transfected into cells lines, directly mim-
`icking the products made by Dicer in vivo. Most siRNA
`sequences can be administered to cultured cells or to animals
`without eliciting an interferon response (11–13). There are
`
`*To whom correspondence should be addressed. Tel: +1 319 626 8432; Fax: +1 319 626 8444; Email: mbehlke@idtdna.com
`
`Ó The Author 2005. Published by Oxford University Press. All rights reserved.
`
`The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access
`version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press
`are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but
`only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oupjournals.org
`
`Alnylam Exh. 1086
`
`

`

`some reports that particular motifs can induce such a response
`when delivered via lipids (13–15), although a cyclodextrin-
`containing polycation system has been shown to deliver
`siRNA containing one such putative immunostimulatory
`motif that achieves target gene down-regulation in mice
`without triggering an interferon response (16), even in a
`disseminated tumor model.
`RNAi has rapidly become the favored method to knock
`down single genes for detailed study or hundreds to thousands
`of genes in high-throughput functional genomics surveys. The
`potential of 21mer siRNAs for use as therapeutic agents to
`reduce activity of specific gene products is also receiving
`considerable attention and successful knockdown of gene
`expression in mice has already been demonstrated by several
`groups (17–20).
`We recently described that chemically synthesized RNA
`duplexes of 25–30 base length can have as much as a
`100-fold increase in potency compared with 21mers at the
`same location. At the site most extensively examined in this
`study, EGFPS1, only minor differences in potency were seen
`between duplexes with blunt, 30-overhang or 50-overhang ends,
`and a blunt 27mer duplex was most potent (1). Increased
`potency has similarly been described for 29mer stem short
`hairpin RNAs (shRNAs) when compared with 19mer stem
`hairpins (21). While the primary function of Dicer is generally
`thought to be cleavage of long substrate dsRNAs into short
`siRNA products, Dicer also introduces the cleaved siRNA
`duplexes into nascent RISC in Drosophila (22–24). Dicer is
`involved in RISC assembly and is itself part of the pre-RISC
`complex (25). The observed increased potency obtained using
`longer RNAs in triggering RNAi is theorized to result from
`providing Dicer with a substrate (27mer) instead of a product
`(21mer) and that this improves the rate or efficiency of entry
`of the siRNA duplex into RISC.
`Unfortunately, not all 27mers show this kind of increased
`potency. It is well known that shifting a 21mer siRNA by a
`few bases along the mRNA sequence can change its potency
`by 10-fold or more (26–28). Different products that result from
`dicing can have different functional potency, and control of
`the dicing reaction may be necessary to best utilize Dicer–
`substrate RNAs in RNAi. The EGFPS1 blunt 27mer studied in
`Kim et al. (1) is diced into two distinct 21mers. Vermeulen
`and colleagues reported studies where synthetic 61mer duplex
`RNAs were digested using recombinant human Dicer in vitro
`and examined for cut sites using a 32P-end-labeled gel assay
`system. Heterogeneous cleavage patterns were observed and
`the presence of blunt versus 30-overhang ends altered precise
`cleavage sites (29). Dicing patterns for short 25–30mer RNA
`substrates have not been published and processing rules that
`enable accurate prediction of these patterns do not exist. We
`therefore studied dicing patterns at a variety of sites using
`different duplex designs to see if cleavage products could
`be predicted. We find that a wide variety of dicing patterns
`can result from blunt 27mer duplexes. An asymmetric duplex
`having a single 2-base 30-overhang generally has a more pre-
`dictable and limited dicing pattern where a major cleavage site
`is located 21–22 bases from the overhang. Including DNA
`residues at the 30 end of the blunt side of an asymmetric duplex
`further limits heterogeneity in dicing patterns and makes it
`possible to design 27mer duplexes that result in predictable
`products after dicing.
`
`Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`4141
`
`the 30-overhang influences
`We find that position of
`potency and asymmetric duplexes having a 30-overhang on
`the antisense strand are generally more potent than those
`with the 30-overhang on the sense strand. This can be attri-
`buted to asymmetrical strand loading into RISC, as the
`opposite efficacy patterns are observed when targeting the
`antisense transcript. Novel designs described here that
`incorporate a combination of asymmetric 30-overhang with
`DNA residues in the blunt end offer a reliable approach to
`design Dicer–substrate RNA duplexes for use in RNAi
`applications.
`
`MATERIALS AND METHODS
`
`Chemically synthesized siRNAs
`
`All RNA oligonucleotides described in this study were
`synthesized and purified using high-performance liquid
`chromatography (HPLC) at Integrated DNA Technologies
`(Coralville, IA). All oligonucleotides were examined by elec-
`trospray ionization mass spectrometry (ESI-MS) and were
`within ±0.02% predicted mass and were further examined
`by capillary electrophoresis and were >90% molar purity.
`Final duplexes were prepared in sodium salt form. Duplexes
`are named by site (EGFPS2 is Site 2 in enhanced green fluor-
`escent protein) and by strand length. EGFPS2 27/27 is a blunt
`27mer RNA duplex. EGFPS2 25/27 has a 25 base top (sense,
`‘S’) strand and 27 base bottom (antisense, ‘AS’) strand with a
`2-base 30-overhang. EGFPS2 27/25 has a 27 base sense strand
`with a 2-base 30-overhang and a 25 base antisense strand.
`DNA bases have been substituted at various locations and
`inclusion is indicated as ‘D’ in compound names and DNA
`residues are identified using bold lower case letters in
`sequence; ‘p’ represents 50-phosphate. Starting with a 21mer
`sequence, an ‘R’ 27mer has added bases extending to the right
`side of this sequence (30 with respect to the target) and an ‘L’
`27mer has added bases extending to the left side of this
`sequence (50 with respect to the target). All RNA duplexes
`used in this study are listed in Table S1 (online Supplementary
`Material).
`
`In vitro dicing assays
`RNA duplexes (100 pmol) were incubated in 20 ml of 20 mM
`Tris, pH 8.0, 200 mM NaCl, 2.5 mM MgCl2 with or without
`1 U of recombinant human Dicer (Stratagene, La Jolla, CA)
`at 37C for 24 h. Samples were desalted using a Performa SR
`96 well plate (Edge Biosystems, Gaithersburg, MD). Electro-
`spray ionization liquid chromatography mass spectroscopy
`(ESI-LC-MS) of duplex RNAs pre- and post-treatment
`with Dicer were done using an Oligo HTCS system (Novatia,
`Princeton, NJ) (30), which consisted of ThermoFinnigan
`TSQ7000, Xcalibur data system, ProMass data processing soft-
`ware and Paradigm MS4Ô HPLC (Michrom BioResources,
`Auburn, CA). The liquid chromatography step employed
`before injection into the mass spectrometer (LC-MS) removes
`most of the cations complexed with the nucleic acids; some
`sodium ion can remain bound to the RNA and are visualized as
`minor +22 or +44 species, which is the net mass gain seen with
`substitution of sodium for hydrogen. All dicing experiments
`were performed at least twice.
`
`

`

`4142 Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`EGFP RNAi assays
`
`HEK 293 cells were split in 24-well plates to 60% confluency
`in DMEM media one day prior to transfection. The indicated
`amounts of reporter and internal control DNAs and siRNAs
`were diluted in 50 ml of Opti-MEM I (Invitrogen, Carlsbad,
`CA) and mixed with 50 ml of Opti-MEM I diluted Lipofectam-
`ine 2000 (1 ml per well) (Invitrogen). After incubation at room
`temperature for 20–30 min, the complexes were added to cells
`in 0.4 ml of DMEM media. To normalize for transfection
`efficiency, either firefly luciferase or red fluorescent protein
`(RFP) reporter plasmids were included as internal controls.
`For the luciferase assay, the Steady Glo Luciferase Assay Kit
`(Promega, Madison, WI) was used according to manufacturer’s
`instruction. For RFP co-transfection, the indicated amount of
`EGFP reporter plasmid (pLEGFP-C1 vector, Clontech, Palo
`Alto, CA) was co-transfected with 20 ng of RFP reporter
`plasmid (pDsRed2-C1, BD Sciences, Franklin Lakes, NJ).
`After 24 h, RFP expression levels were monitored by fluor-
`escence microscopy. Only experiments where transfection
`efficiencies were >90% (as assessed by RFP expression)
`were evaluated. Levels of EGFP expression were measured
`24 h later. EGFP expression was determined either from the
`median number of EGFP-fluorescent cells determined by
`FACS (live cells) or by fluorometer readings (cell extracts).
`All transfections were minimally performed in duplicate and
`data averaged. Assays done in triplicate (or more) include
`error bars in data reporting.
`
`EGFP reporter assays of hnRNPH RNAi
`
`efficacy of hnRNPH-specific
`throughput,
`To facilitate
`duplexes were assayed using a synthetic reporter system where
`the complete coding region of the hnRNPH gene was cloned
`into the XhoI and BamHI sites of the EGFP gene in the expres-
`sion vector pLEGFP-C1 (Clontech, Palo Alto, CA) to make an
`EGFP-hnRNPH fusion protein. A PCR product was made
`from a human hnRNPH cDNA clone using the XhoI contain-
`ing forward primer 50-ACGCAGAACTCGAGTGTCTA-30
`and the BamHI containing reverse primer 50-TCACTGCTCC-
`TAGGTTACCT-30. The resulting PCR product was digested
`using XhoI and BamHI, and cloned into pLEGFP-C1 that
`had been similarly digested. Products were verified by DNA
`sequencing. The fusion protein reporter construct was used to
`directly measure activity of anti-hnRNPH reagents by change
`in EGFP fluorescence (31). Transfection and EGFP assays
`were performed as described above.
`
`Firefly luciferase RNAi assays
`
`HeLa cells were seeded at 5e4 cells per well in 24-well plates
`24 h prior to transfection in DMEM (Mediatech, Herndon,
`VA) containing 10% fetal bovine serum (Invitrogen). In
`each well, 100 ml of OptiMEM I containing 2.5 ml Lipofectam-
`ine was added to an equal volume of OptiMEM I containing
`1 mg pGL3-Control Vector (Promega), and the indicated
`duplex at either 20, 2 or 0.4 nM and incubated at room tem-
`perature for 30 min prior to transfection. Plates were washed
`with PBS and incubated for 4 h at 37C with lipoplex solution.
`Lipoplex solutions were replaced with 1 ml complete growth
`medium. At 48 h post-transfection, growth medium was
`replaced with 100 ml 1· Cell Culture Lysis Reagent (Promega).
`Plates were incubated at room temperature with gentle shaking
`
`for 1 h to allow complete cell lysis. The luminescence of cell
`lysates was determined using a Monolight 3010 luminometer
`(BD Pharmingen, San Diego, CA) and the Luciferase assay
`system (Promega). Substrate solution (100 ml) was added to
`10 ml cell lysate, and the resulting luminescent signal was
`integrated over 10 s. Results are presented in relative light
`units (RLU) as a percentage of the average signal for plasmid-
`only samples on the same plate. Mean and SD values for
`triplicate wells are presented for all treatments.
`
`La RNAi assays
`
`antigen
`human La
`the
`for
`specific
`RNA duplexes
`(NM_003142) were transfected into HEK293 cells in 6-well
`plates at 30% confluency as described above. Cellular extracts
`were prepared 72 h post-transfection. Western blots were per-
`formed as described previously (1). The anti-Enolase antibody
`was obtained from Biogenesis (Kingston, NH) and the anti-La
`antibody was kindly provided by Dr Ger Pruijn (Katholieke
`Universiteit Nijmergen, Netherlands).
`HeLa cells were split in 24-well plates at 35% confluency
`and were transfected the next day with Oligofectamine (Invit-
`rogen) using 1 ml per 65 ml OptiMEM I with RNA duplexes at
`the indicated concentrations. All transfections were performed
`in triplicate. RNA was harvested at 24 h post-transfection
`using SV96 Total RNA Isolation Kit (Promega). RNA was
`checked for quality using a Bioanalyzer 2100 (Agilent, Palo
`Alto, CA) and cDNA was prepared using 500 ng total RNA
`with SuperScript-II reverse transcriptase (Invitrogen) per
`manufacturer’s instructions using both oligo-dT and random
`hexamer priming. Real-time PCRs were done using an estim-
`ated 33 ng cDNA per 25 ml reaction using Immolase DNA
`Polymerase (Bioline, Randolph, MA) and 200 nM primers
`and probe. La-specific primers were La-For 50-GACCAACA-
`AGAATCCCTAAACA, La-Rev 50-CTTGCCCTGAAACT-
`GTACTT and probe La-P 50-FAM-AAGGGTAATAAAGC-
`TGCCCAGCCTGGGT-IowaBlackFQ. Cycling conditions
`employed were as follows: 50C for 2 min and 95C for
`10 min followed by 40 cycles of 2-step PCR with 95C for
`15 s and 60C for 1 min. PCR and fluorescence measurements
`were done using an ABI PrismÔ 7000 Sequence Detector
`(Applied Biosystems Inc., Foster City, CA). All data points
`were performed in triplicate. Expression data was normalized
`to internal control human acidic ribosomal phosphoprotein
`P0 (RPLP0) (NM_001002) levels which were measured
`in separate wells in parallel using primers RPLP0-For
`50-GGCGACCTGGAAGTCCAACT, RPLP0-Rev 50-CCAT-
`CAGCACCACAGCCTTC, and probe RPLP0-P 50-FAM-
`ATCTGCTGCATCTGCTTGGAGCCCA-IowaBlackFQ (32).
`
`Luciferase reporter vectors with EGFP and
`hnRNPH, S versus AS targeting
`
`A PCR generated fragment of the EGFP coding region span-
`ning sites EGFPS1 and EGFPS2 was cloned into the XhoI site
`located in the 30-untranslated region (30-UTR) of the human-
`ized Renilla luciferase gene of plasmid psiCHECKÔ-2
`(Promega). Primers containing an XhoI restriction site and
`EGFP nucleotides 67–85 (For, 50-TTTCTCGAGGTAAACGG-
`CCACAAGTTCA-30) and 291–311 (Rev, 50-TTTCTCGAG-
`TCGTCCTTGAAGAAGATGGTG-30) were used to generate
`a 245 bp EGFP PCR product. The PCR product was digested
`
`

`

`with XhoI and the fragment was cloned into a unique XhoI
`site located in the psiCHECKÔ-2 vector. Clones with both
`orientations (S and AS) were obtained and verified by DNA
`sequencing.
`A PCR generated fragment of the hnRNPH coding region
`spanning sites H1 and H3 was similarly cloned into the XhoI
`site located in the 30-UTR of the humanized Renilla luciferase
`gene of plasmid psiCHECKÔ-2. The fragment was 343 bp
`long including the region 90–432 of reference sequence
`NM_005520. Clones with both orientations (S and AS)
`were obtained. Maps of the ‘S’ and ‘AS’ psiCHECK-2 reporter
`vectors are shown in Figure S2 in the Supplementary Material.
`HEK293 cells were transfected with 150 ng of reporter
`vectors Luc-EGFP-‘S’ or Luc-EGFP-‘AS’ with the indicated
`amounts of EGFPS2 or control duplex RNAs as described
`above. HCT116 cells were transfected with 100 ng of reporter
`vectors Luc-hnRNPH-‘S’ or Luc-hnRNPH-‘AS’ with the
`indicated amounts of H3 or control duplex RNAs. Luciferase
`assays were performed 24 h post-transfection. Changes
`in expression of Renilla luciferase (target) were calculated
`relative to firefly luciferase (internal control).
`
`RESULTS
`
`Dicing patterns for EGFP RNA duplexes
`
`The products that result from in vitro digestion of various
`substrate RNA duplexes with recombinant human Dicer
`were
`visualized
`using
`ESI-MS
`(hereafter
`referred
`to as the ‘ESI-dicing assay’). A blunt 27mer derived from
`EGFP Site-1 (EGFPS1) was previously shown to produce
`two primary 21mer cleavage products by ESI-dicing (1).
`RNA duplexes derived from EGFP Site 2 (EGFPS2) sequence
`are studied here in greater detail and are shown in
`Figure 1 and Figure S1 (online Supplementary Material). In
`general, in vitro dicing of 27mer duplexes using recombinant
`human Dicer results in a heterogeneous set of products, most
`of which are 21mer or 22mer species with 50-phosphate. More
`rarely, 20mer and 23mer species are also seen; these species
`are usually inconsistent between repetitions. If a cleavage
`product includes the 50 end from the original chemically syn-
`thesized substrate duplex (i.e. no enzymatic cleavage event
`was needed to produce that end), the 50 end of the diced
`product remained identical to the substrate (50-phosphate or
`50-OH). 50 ends resulting from internal cleavage events show
`mass values consistent with the presence of 50-phosphate. This
`spectrum of diced products is consistent with the pattern seen
`by Elbashir in a collection of sequenced clones derived from
`diced longer dsRNAs (8). Peaks seen in a mass spectrometry
`trace can sometimes be identified to represent a single, unam-
`biguous, unique sequence. In other cases, it is not possible to
`unambiguously identify a specific sequence based solely upon
`mass if more than one 21mer with same base composition (and
`therefore the same mass) could be produced from cleavage of
`the substrate 27mer. In deconvolution of the mass spectra data,
`we have assigned duplex identity to unique species wherever
`possible and otherwise have identified products which are
`consistent with 21mer or 22mer duplexes with 2-base 30-over-
`hangs (instead of other combinations that would yield blunt
`duplexes or 1-base overhangs, etc.).
`
`Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`4143
`
`ESI-dicing of the blunt 27mer substrate duplex EGFPS2 R
`27/27 shows a variety of mass peaks which are consistent with
`five duplex products, shown in Figure 1A. Using other blunt
`27mer duplex sequences (from other sites), we have seen 2, 3,
`4 or more duplex products result from dicing (data not shown).
`Elbashir reported in vitro dicing of 39, 52 and 111 bp dsRNA
`substrates using Drosophila extracts. Products were cloned
`and a heterogeneous collection of 21–22 base fragments
`were identified by DNA sequencing. While dicing appeared
`to generally initiate from one or both ends, no single product
`or subset of products was dominant (8). We cannot identify
`any obvious pattern that allows prediction of what specific
`21mer(s) will result from dicing a given blunt 27mer, making
`rational design of 27mer RNAi reagents difficult. We therefore
`tested a series of design variants to see if this heterogeneity
`could be reduced with the goal of finding a design for which
`cleavage patterns could be reliably predicted.
`Modification of one end of a blunt 27mer duplex with
`fluorescein inhibits dicing. Modification of both 30 ends
`with fluorescein completely blocks dicing (1). We tested if
`substitution of three DNA residues at the 50 or 30 end of either
`strand of a blunt 27mer duplex affected dicing patterns.
`Results are shown in Figure S1 (online Supplementary
`Material). DNA residues cannot be cleaved by Dicer and
`therefore directly alter cleavage patterns. Further, the DNA
`residues seem to reduce the likelihood of Dicer binding
`that end (assuming that Dicer binds the RNA duplex via
`the PAZ domain and then cuts 21–22 bases away) (33), espe-
`cially when positioned on the 30 end. Curiously, EGFPS2 R
`27/27(30D), the duplex with DNA residues positioned at the
`30 end of the antisense strand, was not diced to produce any
`21–22mer cleavage product (identical results were obtained in
`three attempts). While the use of terminal DNA residues can
`change and somewhat simplify dicing patterns, this approach
`alone is insufficient
`to truly direct dicing to predictable
`patterns.
`Naturally occurring substrates for Dicer include micro-
`RNAs (miRNAs). These species typically originate in the
`nucleus as long primary transcripts where they are processed
`to 70mer pre-miRNAs by the RNase III class endonuclease
`Drosha (34,35). Drosha products are shRNAs, which structur-
`ally look like an asymmetric duplex, having a single 2-base
`30-overhang on one side and a hairpin loop on the other side.
`These species are exported to the cytoplasm (36) where final
`processing by Dicer takes place (37). The 30-overhang may
`be important for Dicer processing. The presence of a single
`2-base 30-overhang has been shown to help direct ‘correct’
`dicing of synthetic shRNAs (21) and alters dicing patterns of
`synthetic linear 61mer duplexes (29).
`Asymmetric duplexes having one 2-base 30-overhang
`and one blunt end were studied using the ESI-dicing assay.
`Duplex EGFPS2 R 25/27 has two bases removed from the
`50 end of the sense strand compared with blunt duplex
`EGFPS2 R 27/27, providing a single 2-base 30-overhang.
`This duplex showed a much simplified dicing pattern,
`Figure 1B. Only one 21mer and one 22mer duplex were pro-
`duced. The observed pattern is consistent with a model where
`Dicer binds the 30-overhang of the substrate RNA and cleavage
`takes place 21–22 bases distant from this site (33). In fact,
`this pattern could be considered a single event as dicing gen-
`erates a single pair of related duplexes as ‘the product’ made
`
`

`

`4144 Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`from digestion of the substrate duplex from a single unique
`binding or start site.
`The presence of a 30-overhang does not always restrict
`dicing to a simple pattern. Duplex EGFPS2 R 27/25 has
`two bases removed from the 50 end of the antisense strand
`compared with blunt duplex EGFPS2 R 27/27. This duplex
`showed a complex dicing pattern (Figure 1C) which is
`different from both the parent blunt duplex (Figure 1A)
`and the asymmetric duplex with the 30-overhang on the oppos-
`ite end (Figure 1B). Changing the bases present on the
`
`30-overhang from ‘CC’ to ‘GG’ did not reduce the complexity
`of dicing pattern seen for this duplex (data not shown).
`The two approaches of modifying dicing patterns were com-
`bined and DNA residues were placed at the 30 end (blunt end)
`of the antisense strand of asymmetric duplex EGFPS2 R 27/25,
`resulting in duplex EGFPS2 R 27/25D. This duplex gave a
`simplified dicing pattern with only a single 21mer and a single
`22mer species (Figure 1D). This new design strategy places an
`element on one end that is generally favorable for Dicer
`binding (2-base 30-overhang) and an element on the other
`
`A
`
`B
`
`

`

`Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`4145
`
`end that is generally unfavorable for Dicer binding (blunt end
`with 30-DNA residues). This type of duplex appears to offer
`Dicer a single favorable binding site and would be predicted to
`result in a single 21mer product.
`Dicing patterns from the new design having a 2-base
`30-overhang on one end and blunt with 30-DNA residues on
`the other end were examined using additional duplexes.
`Duplex EGFPS2 R 25D/27 has a 30-overhang on the antisense
`strand and two DNA bases at the 30 end (blunt end) of the sense
`strand. Duplex EGFPS2 L 27/25D has the opposite design
`with a 30-overhang on the sense strand and two DNA bases
`at the 30 end (blunt end) of the antisense strand. If the design
`strategy works as predicted, these two different duplexes
`
`should both be diced into the same 21mer product. EGFPS2
`R 25D/27 diced into the expected 21mer/22mer pair with
`cleavage of 21–22 bases from the 30-overhang (Figure 1E).
`Similarly, EGFPS2 L 27/25D diced into the expected 21mer/
`22mer pair with cleavage 21–22 bases from the 30-overhang
`(Figure 1F). Thus, the same 21mer duplex was produced from
`two different substrate RNAs. Note, however, that the sister
`product (22mer) is necessarily different for each substrate
`since the single base addition occurs on opposite sides of
`the 21mer.
`To confirm that this cleavage pattern holds at other sites
`and is truly predictable, a similar related pair of asymmetric
`27mers were tested at EGFP Site-1 (Supplementary Table S1).
`
`C
`
`D
`
`

`

`4146 Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`EGFPS1 R 25D/27 diced into the expected 21mer duplex with
`cleavage 21 bases from the 30-overhang (Figure 1G). In this
`instance, no 22mer was seen. Similarly, EGFPS1 L 27/25D
`diced into the expected 21mer/22mer pair with cleavage of
`21–22 bases from the 30-overhang (Figure 1H). Again, the
`same 21mer duplex was produced from two different 27mer
`substrate RNAs.
`
`Functional potency is different for ‘R’
`versus ‘L’ form RNA duplexes
`
`EGFPS2 duplexes were co-transfected into HEK293 cells
`with an EGFP expression plasmid. The blunt 27mer duplex
`(EGFPS2 R 27/27) was more potent in reducing EGFP expres-
`sion levels than the 21mer duplex (EGFPS2 21/21) (1). The
`L 27/25D asymmetric duplex was slightly less potent than
`the blunt 27mer while the R 25D/27 asymmetric duplex
`
`was significantly more potent than any other duplex tested at
`the EGFPS2 site (Figure 2A). A similarly matched pair of
`asymmetric 27mer duplexes was studied at EGFP Site-1
`(Figure 2B). As was observed previously for the EGFPS2 site,
`the ‘R’ asymmetric duplex was significantly more potent than
`the ‘L’ asymmetric duplex.
`We originally expected that the ‘R’-form and ‘L’-form
`duplexes would have similar functional potency since both
`result
`in the same antisense 21mer product after dicing
`(EGFPS2, Figure 1E and F; EGFPS1, Figure 1G and H).
`The ‘L’ and ‘R’ duplexes do produce different 22mer products
`after dicing. It is possible that the differential potency relates
`to these 22mer products. To test this possibility directly,
`RNA duplexes of 22mer length that correspond to the products
`identified from ESI-dicing of the ‘L’ and ‘R’-form 27mers
`for both EGFPS1 and EGFPS2 were co-transfected into
`HEK293 cells with an EGFP expression plasmid. For site
`
`E
`
`F
`
`

`

`Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`4147
`
`G
`
`H
`
`Figure 1. ESI-MS analysis of in vitro dicing reactions. Mass spectra of duplexes are shown before (top) and after (middle) digestion with recombinant human
`Dicer. Sequence of the substrate RNA duplex is provided with detected cleavage products along with calculated molecular weight and strand length (bottom). RNA
`bases are upper case, DNA bases are lower case bold and ‘p’ represents 50-phosphate. (A) Blunt duplex EGFPS2 R 27/27; (B) asymmetric duplex EGFPS2 R 25/27;
`(C) asymmetric duplex EGFPS2 R 27/25; (D) asymmetric duplex EGFPS2 R 27/25(30D); (E) asymmetric duplex EGFPS2 R 25D/27; (F) asymmetric duplex
`EGFPS2 L 27/25D; (G) asymmetric duplex EGFPS1 R 25D/27; (H) asymmetric duplex EGFPS1 L 27/25D.
`
`EGFPS2, the ‘R-derived’ 22mer was significantly more potent
`than the ‘L-derived’ 22mer. This species could contribute to
`the higher potency observed for the ‘R’ form EGFPS2 27mer.
`However, for site EGFPS1 the ‘L-derived’ 22mer was 3-fold
`more potent
`than the ‘R-derived’ 22mer (Supplementary
`Figure S3). In this case, 22mer products cannot contribute
`to the observed higher potency of the ‘R’ form EGFPS1
`27mer. Therefore, although the different 22mer species that
`result from 27mer dicing can contribute to the ultimate
`potency of a given duplex, we believe that the generalized
`difference in potency seen between ‘R’ and ‘L’-form 27mers
`results from other aspects of Dicer/RISC biochemistry.
`Vermeulen reported that asymmetric 21mer duplexes with a
`
`single 30-overhang on the antisense strand were generally
`than duplexes with a single 30-overhang on
`more potent
`the sense strand (for 3 out of 4 duplexes studied in one
`gene target) (29). In this case, the duplexes tested were not
`Dicer substrates.
`target
`Dicer–substrate duplexes were studied in other
`systems to see if the seeming functional asymmetry between
`duplexes that are processed to the same 21mer product was
`reproducible at additional sites. RNA duplexes targeting two
`adjacent sites in the hnRNPH gene were tested for potency
`in triggering RNAi. Site H3 and neighboring site H1 (two
`base 50-shift) were studied for RNAi-mediated suppression
`of EGFP fluorescence using a modified EGFP expression
`
`

`

`4148 Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`A
`
`B
`
`C
`
`

`

`D
`
`E
`
`Nucleic Acids Research, 2005, Vol. 33, No. 13
`
`4149
`
`Figure 2. ‘R’ form duplexes are more potent than ‘L’ form duplexes. Top: relative expression data are shown. Bottom: target sequences (S strand) are shown and
`transfected RNAi duplexes are aligned beneath in duplex form with sense strand top (50!30) and antisense strand bottom (30!50). RNA bases are upper case, DNA
`bases are lower case bold, and ‘p’ represents 50-phosphate. (A) Relative EGFP fluorescence was measured following co-transfection of an EGFP expression plasmid
`and EGFPS2 RNA duplexes; (B) relative EGFP fluorescence was measured following co-transfection of an EGFP expression plasmid and EGFPS1 RNA duplexes;
`(C) relative EGFP fluorescence was measured following co-transfection of an EGFP-hnRNPH fusion vector and hnRNPH RNA duplexes; (D) relative light emission
`from firefly luciferase was measured following co-transfection of a luciferase expression vector and luciferase RNA duplexes; (E) RNA duplexes specific for La
`antigen mRNA were transfected into cells at 2.5 nM concentration and protein extracts were prepared 72 h post-transfection. Western blots were done using anti-La
`antibodies and anti-enolase (control) antibodies