Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`RNA interference is mediated by 21-
`and 22-nucleotide RNAs
`
`Sayda M. Elbashir, Winfried Lendeckel, and Thomas Tuschl1
`Department of Cellular Biochemistry, Max-Planck-Institute for Biophysical Chemistry, Am Fassberg 11,
`D-37077 Go¨ ttingen, Germany
`
`Double-stranded RNA (dsRNA) induces sequence-specific posttranscriptional gene silencing in many
`organisms by a process known as RNA interference (RNAi). Using a Drosophila in vitro system, we
`demonstrate that 21- and 22-nt RNA fragments are the sequence-specific mediators of RNAi. The short
`interfering RNAs (siRNAs) are generated by an RNase III–like processing reaction from long dsRNA.
`Chemically synthesized siRNA duplexes with overhanging 3ⴕ ends mediate efficient target RNA cleavage in
`the lysate, and the cleavage site is located near the center of the region spanned by the guiding siRNA.
`Furthermore, we provide evidence that the direction of dsRNA processing determines whether sense or
`antisense target RNA can be cleaved by the siRNA–protein complex.
`[Key Words: RNAi; posttranscriptional gene silencing; dsRNA; siRNA]
`
`Received October 25, 2000; revised version accepted November 28, 2000.
`
`The term RNA interference (RNAi) was coined after the
`discovery that injection of dsRNA into the nematode
`Caenorhabditis elegans leads to specific silencing of
`genes highly homologous in sequence to the delivered
`dsRNA (Fire et al. 1998). RNAi was also observed sub-
`sequently in insects (Kennerdell and Carthew 1998), frog
`(Oelgeschlager et al. 2000), and other animals including
`mice (Svoboda et al. 2000; Wianny and Zernicka-Goetz
`2000) and is likely to also exist in human. RNAi is
`closely linked to the posttranscriptional gene-silencing
`(PTGS) mechanism of cosuppression in plants and quell-
`ing in fungi (Cogoni and Macino 1999; Catalanotto et al.
`2000; Dalmay et al. 2000; Ketting and Plasterk 2000;
`Mourrain et al. 2000; Smardon et al. 2000), and some
`components of the RNAi machinery are also necessary
`for posttranscriptional silencing by cosuppression (Cat-
`alanotto et al. 2000; Dernburg et al. 2000; Ketting and
`Plasterk 2000). The topic has been reviewed recently
`(Fire 1999; Sharp 1999; Bass 2000; Bosher and Labouesse
`2000; Plasterk and Ketting 2000; Sijen and Kooter 2000;
`see also the entire issue of Plant Molecular Biology, Vol.
`43, issue 2/3, 2000).
`The natural function of RNAi and cosuppression ap-
`pears to be protection of the genome against invasion by
`mobile genetic elements such as transposons and vi-
`ruses, which produce aberrant RNA or dsRNA in the
`host cell when they become active (Jensen et al. 1999;
`Ketting et al. 1999; Ratcliff et al. 1999; Tabara et al. 1999;
`Malinsky et al. 2000). Specific mRNA degradation pre-
`
`1Corresponding author.
`E-MAIL ttuschl@mpibpc.gwdg.de; FAX 49-551-201-1197.
`Article and publication are at www.genesdev.org/cgi/doi/10.1101/
`gad.862301.
`
`vents transposon and virus replication, although some
`viruses are able to overcome or prevent this process by
`expressing proteins that suppress PTGS (Anandalakshmi
`et al. 2000; Lucy et al. 2000; Voinnet et al. 2000).
`DsRNA triggers the specific degradation of homolo-
`gous RNAs only within the region of identity with the
`dsRNA (Zamore et al. 2000). The dsRNA is processed to
`21–23-nt RNA fragments (Zamore et al. 2000). These
`short fragments were also detected in extracts prepared
`from Drosophila melanogaster Schneider 2 cells that
`were transfected with dsRNA before cell lysis (Ham-
`mond et al. 2000) or after injection of radiolabeled
`dsRNA into D. melanogaster embryos (Yang et al. 2000)
`or C. elegans adults (Parrish et al. 2000). RNA molecules
`of similar size also accumulate in plant tissue that ex-
`hibits PTGS (Hamilton and Baulcombe 1999). It has been
`suggested that the 21–23-nt fragments are the guide
`RNAs for target recognition (Hamilton and Baulcombe
`1999; Hammond et al. 2000), which is supported by the
`finding that the target mRNA is cleaved in 21–23-nt in-
`tervals (Zamore et al. 2000).
`Here, we use the established Drosophila in vitro sys-
`tem (Tuschl et al. 1999; Zamore et al. 2000) to explore
`further the mechanism of RNAi. It is demonstrated that
`synthetic 21- and 22-nt RNAs, when base paired with 3⬘
`overhanging ends, act as the guide RNAs for sequence-
`specific mRNA degradation. Short 30-bp dsRNAs are in-
`efficiently processed to 21- and 22-nt RNAs, which
`may explain why they are ineffective in mediating RNAi.
`Furthermore, we define the target RNA cleavage sites
`relative to the 21- and 22-nt short interfering RNAs
`(siRNAs) and provide evidence that the direction of dsRNA
`processing determines whether a sense or an antisense
`
`188
`
`GENES & DEVELOPMENT 15:188–200 © 2001 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/01 $5.00; www.genesdev.org
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`RNA interference is mediated by 21-
`and 22-nucleotide RNAs
`
`Sayda M. Elbashir, Winfried Lendeckel, and Thomas Tuschl1
`Department of Cellular Biochemistry, Max-Planck-Institute for Biophysical Chemistry, Am Fassberg 11,
`D-37077 Go¨ ttingen, Germany
`
`Double-stranded RNA (dsRNA) induces sequence-specific posttranscriptional gene silencing in many
`organisms by a process known as RNA interference (RNAi). Using a Drosophila in vitro system, we
`demonstrate that 21- and 22-nt RNA fragments are the sequence-specific mediators of RNAi. The short
`interfering RNAs (siRNAs) are generated by an RNase III–like processing reaction from long dsRNA.
`Chemically synthesized siRNA duplexes with overhanging 3ⴕ ends mediate efficient target RNA cleavage in
`the lysate, and the cleavage site is located near the center of the region spanned by the guiding siRNA.
`Furthermore, we provide evidence that the direction of dsRNA processing determines whether sense or
`antisense target RNA can be cleaved by the siRNA–protein complex.
`[Key Words: RNAi; posttranscriptional gene silencing; dsRNA; siRNA]
`
`Received October 25, 2000; revised version accepted November 28, 2000.
`
`The term RNA interference (RNAi) was coined after the
`discovery that injection of dsRNA into the nematode
`Caenorhabditis elegans leads to specific silencing of
`genes highly homologous in sequence to the delivered
`dsRNA (Fire et al. 1998). RNAi was also observed sub-
`sequently in insects (Kennerdell and Carthew 1998), frog
`(Oelgeschlager et al. 2000), and other animals including
`mice (Svoboda et al. 2000; Wianny and Zernicka-Goetz
`2000) and is likely to also exist in human. RNAi is
`closely linked to the posttranscriptional gene-silencing
`(PTGS) mechanism of cosuppression in plants and quell-
`ing in fungi (Cogoni and Macino 1999; Catalanotto et al.
`2000; Dalmay et al. 2000; Ketting and Plasterk 2000;
`Mourrain et al. 2000; Smardon et al. 2000), and some
`components of the RNAi machinery are also necessary
`for posttranscriptional silencing by cosuppression (Cat-
`alanotto et al. 2000; Dernburg et al. 2000; Ketting and
`Plasterk 2000). The topic has been reviewed recently
`(Fire 1999; Sharp 1999; Bass 2000; Bosher and Labouesse
`2000; Plasterk and Ketting 2000; Sijen and Kooter 2000;
`see also the entire issue of Plant Molecular Biology, Vol.
`43, issue 2/3, 2000).
`The natural function of RNAi and cosuppression ap-
`pears to be protection of the genome against invasion by
`mobile genetic elements such as transposons and vi-
`ruses, which produce aberrant RNA or dsRNA in the
`host cell when they become active (Jensen et al. 1999;
`Ketting et al. 1999; Ratcliff et al. 1999; Tabara et al. 1999;
`Malinsky et al. 2000). Specific mRNA degradation pre-
`
`1Corresponding author.
`E-MAIL ttuschl@mpibpc.gwdg.de; FAX 49-551-201-1197.
`Article and publication are at www.genesdev.org/cgi/doi/10.1101/
`gad.862301.
`
`vents transposon and virus replication, although some
`viruses are able to overcome or prevent this process by
`expressing proteins that suppress PTGS (Anandalakshmi
`et al. 2000; Lucy et al. 2000; Voinnet et al. 2000).
`DsRNA triggers the specific degradation of homolo-
`gous RNAs only within the region of identity with the
`dsRNA (Zamore et al. 2000). The dsRNA is processed to
`21–23-nt RNA fragments (Zamore et al. 2000). These
`short fragments were also detected in extracts prepared
`from Drosophila melanogaster Schneider 2 cells that
`were transfected with dsRNA before cell lysis (Ham-
`mond et al. 2000) or after injection of radiolabeled
`dsRNA into D. melanogaster embryos (Yang et al. 2000)
`or C. elegans adults (Parrish et al. 2000). RNA molecules
`of similar size also accumulate in plant tissue that ex-
`hibits PTGS (Hamilton and Baulcombe 1999). It has been
`suggested that the 21–23-nt fragments are the guide
`RNAs for target recognition (Hamilton and Baulcombe
`1999; Hammond et al. 2000), which is supported by the
`finding that the target mRNA is cleaved in 21–23-nt in-
`tervals (Zamore et al. 2000).
`Here, we use the established Drosophila in vitro sys-
`tem (Tuschl et al. 1999; Zamore et al. 2000) to explore
`further the mechanism of RNAi. It is demonstrated that
`synthetic 21- and 22-nt RNAs, when base paired with 3⬘
`overhanging ends, act as the guide RNAs for sequence-
`specific mRNA degradation. Short 30-bp dsRNAs are in-
`efficiently processed to 21- and 22-nt RNAs, which
`may explain why they are ineffective in mediating RNAi.
`Furthermore, we define the target RNA cleavage sites
`relative to the 21- and 22-nt short interfering RNAs
`(siRNAs) and provide evidence that the direction of dsRNA
`processing determines whether a sense or an antisense
`
`188
`
`GENES & DEVELOPMENT 15:188–200 © 2001 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/01 $5.00; www.genesdev.org
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`target RNA can be cleaved by the siRNP endonuclease
`complex.
`
`Length requirements for processing of dsRNA to 21-
`and 22-nt RNA fragments
`
`Lysate prepared from D. melanogaster syncytial em-
`bryos recapitulates RNAi in vitro, providing a tool for
`biochemical analysis of the mechanism of RNAi (Tuschl
`et al. 1999; Zamore et al. 2000). In vitro and in vivo
`analysis of the length requirements of dsRNA for RNAi
`has revealed that short dsRNA (<150 bp) are less effective
`than longer dsRNAs in degrading target mRNA (Ngo et
`al. 1998; Tuschl et al. 1999; Caplen et al. 2000; Ham-
`mond et al. 2000). The reasons for reduction in mRNA
`degrading efficiency are not understood. We therefore ex-
`amined the precise length requirement of dsRNA for tar-
`get RNA degradation under optimized conditions in the
`Drosophila lysate. Three series of dsRNAs were synthe-
`sized and directed against firefly luciferase (Pp-luc) re-
`porter RNA. The dual luciferase assay was used to moni-
`tor specific suppression of target RNA expression (Tus-
`chl et al. 1999; (Fig. 1A,B). Specific inhibition of target
`RNA expression was detected for dsRNAs as short as 38
`bp, but dsRNAs of 29–36 bp were not effective in this
`process. The effect was independent of the target posi-
`tion and the degree of inhibition of Pp-luc mRNA ex-
`pression correlated with the length of the dsRNA; that
`is, long dsRNAs were more effective than short dsRNAs.
`It has been suggested that the 21–23-nt RNA frag-
`ments generated by processing of dsRNAs are the me-
`diators of RNA interference and cosuppression (Hamil-
`ton and Baulcombe 1999; Hammond et al. 2000; Zamore
`et al. 2000). We therefore analyzed the rate of 21–23-nt
`
`RNA interference is mediated by 21- and 22-nt RNAs
`
`fragment formation for a subset of dsRNAs ranging in
`size from 501 to 29 bp. Formation of 21–23-nt fragments
`in Drosophila lysate (Fig. 2) was readily detectable for
`39–501 bp dsRNAs but was significantly delayed for the
`29-bp dsRNA. This observation is consistent with a role
`of 21–23-nt fragments in guiding mRNA cleavage and
`provides an explanation for the lack of RNAi by 30-bp
`dsRNAs. The length dependence of 21–23 mer formation
`is likely to reflect a mechanism to prevent the undesired
`activation of RNAi by short intramolecular base-paired
`structures of cellular RNAs.
`
`Mapping of the cleavage sites on sense and antisense
`target RNAs
`
`Addition of dsRNA and 5⬘-capped target RNA to the
`Drosophila lysate results in sequence-specific degrada-
`tion of the target RNA (Tuschl et al. 1999). The target
`mRNA is only cleaved within the region of identity with
`the dsRNA, and many of the target cleavage sites are
`separated by 21–23 nt (Zamore et al. 2000). Thus, the
`number of cleavage sites for a given dsRNA was expected
`to roughly correspond to the length of the dsRNA di-
`vided by 21. We mapped the target cleavage sites on a
`sense and an antisense target RNA that was 5⬘ radiola-
`beled at the cap (Zamore et al. 2000; Fig. 3A,B). Stable 5⬘
`cleavage products were separated on a sequencing gel,
`and the position of cleavage was determined by compari-
`son with a partial RNase T1 and an alkaline hydrolysis
`ladder from the target RNA.
`Consistent with the previous observation (Zamore et
`al. 2000), all target RNA cleavage sites were located
`within the region of identity to the dsRNA. The 39-bp
`dsRNA produced a strong and a weak (often hardly de-
`
`Figure 1. Double-stranded RNA as short
`as 38 bp can mediate RNAi. (A) Graphic
`representation of dsRNAs used for target-
`ing Pp-luc mRNA. Three series of blunt-
`ended dsRNAs covering a range of 29–504
`bp were prepared. The position of the first
`nucleotide of
`the sense strand of
`the
`dsRNA is indicated relative to the start
`codon of Pp-luc mRNA (p1). (B) RNA inter-
`ference assay (Tuschl et al. 1999). Ratios of
`target Pp-luc to control Rr-luc activity were
`normalized to a buffer control (black bar).
`DsRNAs (5 nM) were preincubated in Dro-
`sophila lysate at 25°C for 15 min before the
`addition of 7-methyl-guanosine-capped Pp-
`luc and Rr-luc mRNAs (∼50 pM). The incu-
`bation was continued for another hour and
`then analyzed by the dual luciferase assay
`(Promega). The data are the average from at
`least four independent experiments ±S.D.
`
`GENES & DEVELOPMENT
`
`189
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`Elbashir et al.
`
`Figure 2. A 29-bp dsRNA is only slowly processed to 21–23-nt
`fragments. Time course of 21–23-mer formation from process-
`ing of internally 32P-labeled dsRNAs (5 nM) in the Drosophila
`lysate. The length and source of the dsRNA are indicated. An
`RNA size marker (M) has been loaded in the left lane, and the
`fragment sizes are indicated. Double bands at time zero are
`caused by incompletely denatured dsRNA.
`
`tectable) cleavage site in the sense target RNA separated
`by 19 nt. The antisense target was only cleaved once, by
`the 39-bp dsRNA. The predominant cleavage site of the
`sense strand and the cleavage site of the antisense strand
`are located 10 nt from the 5⬘ end of the region covered by
`the dsRNA (Fig. 3B). The 52-bp dsRNA, which shares the
`same 5⬘ end as the 39-bp dsRNA, produces the same
`strong cleavage site on the sense target, located 10 nt
`from the 5⬘ end of the region of identity with the dsRNA
`in addition to two weaker cleavage sites 23 and 24 nt
`downstream of the first site. The antisense target was
`only cleaved once, again 10 nt from the 5⬘ end of the
`region covered by its respective dsRNA. Mapping of the
`cleavage sites for the 38–49-bp dsRNAs shown in Figure
`1 revealed that the first and predominant cleavage site
`
`190
`
`GENES & DEVELOPMENT
`
`was always located 7–10 nt downstream from the 5⬘
`boundary of the region covered by the dsRNA (data not
`shown). This suggests that the point-of-target RNA
`cleavage can be determined by the end of the dsRNA and
`could imply that processing to 21–23mers starts from the
`ends of the duplex.
`Cleavage sites on sense and antisense targets for the
`longer 111-bp dsRNA were much more frequent than
`anticipated, and most of them appear in clusters sepa-
`rated by 20–23 nt (Fig. 3A,B). As for the shorter dsRNAs,
`the first cleavage site on the sense target is 10 nt from
`the 5⬘ end of the region spanned by the dsRNA, and the
`first cleavage site on the antisense target is located 9 nt
`from the 5⬘ end of region covered by the dsRNA. It is
`unclear what causes this disordered cleavage, but one
`possibility could be that longer dsRNAs may not only
`get processed from the ends but also internally, or there
`are some specificity determinants for dsRNA processing
`that we do not yet understand. Some irregularities to the
`21–23 nt spacing were also noted previously (Zamore et
`al. 2000).
`
`dsRNA is processed to 21- and 22-nt RNAs
`by an RNase III–like mechanism
`
`To understand better the molecular basis of dsRNA pro-
`cessing and target RNA recognition, we decided to ana-
`lyze the sequences of the 21–23-nt fragments generated
`by processing of 39-, 52-, and 111-bp dsRNAs in the Dro-
`sophila lysate. We first examined the 5⬘ and 3⬘ termini of
`the RNA fragments. Periodate oxidation of gel-purified
`21–23-nt RNAs followed by ␤-elimination indicated the
`presence of a terminal 2⬘ and 3⬘ hydroxyl (data not
`shown). The 21–23mers were also responsive to alkaline
`phosphatase treatment, implying the presence of a 5⬘ ter-
`minal phosphate (data not shown). The presence of 5⬘
`phosphate and 3⬘ hydroxyl termini suggests that the
`dsRNA could be processed by an enzymatic activity
`similar to Escherichia coli RNase III (for reviews, see
`Dunn 1982; Nicholson 1999; Robertson 1982, 1990).
`To directionally clone the 21–23-nt RNA fragments, 3⬘
`and 5⬘ adapter oligonucleotides were ligated to the puri-
`fied 21–23 mers using T4 RNA ligase. The ligation
`products were reverse transcribed, PCR-amplified, con-
`catamerized, cloned, and sequenced. Over 220 short RNAs
`were sequenced from dsRNA processing reactions of the
`39-, 52-, and 111-bp dsRNAs (Fig. 4A). We found the
`following length distribution: 1% 18 nt, 5% 19 nt, 12%
`20 nt, 45% 21 nt, 28% 22 nt, 6% 23 nt, and 2% 24 nt.
`Sequence analysis of the 5⬘ terminal nucleotide of the
`processed fragments indicated that oligonucleotides
`with a 5⬘ guanosine were underrepresented. This bias
`was most likely introduced by T4 RNA ligase, which
`discriminates against 5⬘ phosphorylated guanosine as do-
`nor oligonucleotide (Romaniuk et al. 1982); no signifi-
`cant sequence bias was seen at the 3⬘ end. Many of the
`∼21-nt fragments originating from the 3⬘ ends of the
`sense or antisense strand of the duplexes include 3⬘
`nucleotides that are derived from untemplated addition
`of nucleotides during RNA synthesis using T7 RNA
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`RNA interference is mediated by 21- and 22-nt RNAs
`
`isk.
`circleforweakcleavage).The32P-radiolabeledphosphategroupismarkedbyanaster-
`quences.Cleavagesitesareindicatedbycircles(largecircleforstrongcleavage,small
`indicatebydifferentlycoloredbarspositionedbetweensenseandantisensetargetse-
`tarysequenceareopposingeachother.TheregiontargetedbythedifferentdsRNAsare
`antisensetargetRNAsarerepresentedinantiparallelorientationsuchthatcomplemen-
`senseandantisensetargetRNAs.Thesequencesofthecapped177-ntsenseand180-nt
`dicatesunspecificcleavagenotcausedbyRNAi.(B)Positionofthecleavagesiteson
`predominantcleavagesitesforthe111-bpdsRNAisshown.Thehorizontalarrowin-
`thedsRNAsareindicatedasblackbarsonbothsides.The20–23-ntspacingbetweenthe
`partialalkalinehydrolysis(OH)ofthecap-labeledtargetRNA.Theregionstargetedby
`Drosophilalysate.LengthmarkersweregeneratedbypartialnucleaseT1digestionand
`senseorantisenseRNA32P-labeledatthecapwith10nMdsRNAsofthep133seriesin
`electrophoresisofthestable5⬘cleavageproductsproducedby1hincubationof10nM
`Figure3.MappingofsenseandantisensetargetRNAcleavagesites.(A)Denaturinggel
`
`GENES & DEVELOPMENT
`
`191
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`Elbashir et al.
`
`Figure 4. 21- and 22-nt RNA fragments
`are generated by an RNase III–like mecha-
`nism. (A) Sequences of ∼21-nt RNAs after
`dsRNA processing. The ∼21-nt RNA frag-
`ments generated by dsRNA processing
`were directionally cloned and sequenced.
`Oligoribonucleotides originating from the
`sense strand of the dsRNA are indicated as
`blue lines; those originating from the an-
`tisense strand are red lines. Thick bars are
`used if the same sequence was present in
`multiple clones, the number at the right
`indicating the frequency. The target RNA
`cleavage sites mediated by the dsRNA are
`indicated as orange circles: large circle for
`strong cleavage, small circle for weak
`cleavage (see Fig. 3B). Circles on top of the
`sense strand indicated cleavage sites
`within the sense target, and circles at the
`bottom of the dsRNA indicate cleavage
`site in the antisense target. Up to five ad-
`ditional nucleotides were identified in
`∼21-nt fragments derived from the 3⬘ ends
`of the dsRNA. These nucleotides are ran-
`dom combinations of predominantly C, G,
`or A residues and were most likely added
`in an untemplated fashion during T7 tran-
`scription of
`the
`dsRNA-constituting
`strands. (B) Two-dimensional TLC analy-
`sis of the nucleotide composition of ∼21-nt
`RNAs. The ∼21-nt RNAs were generated
`by incubation of internally radiolabeled
`504-bp Pp-luc dsRNA in Drosophila ly-
`sate, gel purified, and then digested to
`mononucleotides with nuclease P1 (top
`row) or ribonuclease T2 (bottom row). The
`dsRNA was internally radiolabeled by
`transcription in the presence of one of the
`indicated ␣-32P nucleoside triphosphates.
`Radioactivity was detected by phospho-
`rimaging. Nucleoside 5⬘-monophosphates,
`nucleoside 3⬘-monophosphates, nucleo-
`side 5⬘,3⬘-diphosphates, and inorganic
`phosphate are indicated as pN, Np, pNp,
`and pi, respectively. Black circles indicate
`UV-absorbing spots from nonradioactive
`carrier nucleotides. The 3⬘,5⬘-diphos-
`phates (red circles) were identified by
`comigration with radiolabeled standards
`prepared by 5⬘-phosphorylation of nucleo-
`side 3⬘-monophosphates with T4 poly-
`nucleotide kinase and ␥-32P-ATP (data not
`shown).
`
`polymerase. Interestingly, a significant number of endog-
`enous Drosophila ∼21-nt RNAs were also cloned, some
`of them from LTR and non-LTR retrotransposons (data
`not shown). This is consistent with a possible role for
`RNAi
`in transposon silencing (Ketting et al. 1999;
`Tabara et al. 1999).
`The ∼21-nt RNAs appear in clustered groups (Fig. 4A)
`that cover the entire dsRNA sequences. For the 39-bp
`dsRNA, two clusters of ∼21-nt RNAs were found from
`each dsRNA-constituting strand (including overhanging
`
`3⬘ ends). Only one of the clusters from each strand can be
`correlated with a strong cleavage hot spot on the target
`sense or antisense RNA (Fig. 3A,B), indicating that
`dsRNA processing produced primarily two functional
`small RNAs originating from the 3⬘ ends of the duplex.
`Perhaps the ∼21-nt RNAs are present in double-stranded
`form in the endonuclease complex, but only one of the
`strands can be used for target RNA recognition and
`cleavage.
`The ∼21-mer clusters for the 52- and 111-bp dsRNA
`
`192
`
`GENES & DEVELOPMENT
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`are less well defined when compared to the 39-bp
`dsRNA. The clusters are spread over regions of 25–30 nt
`most likely representing several distinct subpopulations
`of ∼21-nt duplexes and, therefore, guiding target cleavage
`at several nearby sites. These cleavage regions are still
`predominantly separated by 20–23-nt intervals. The
`rules determining how dsRNA can be processed to ∼21-
`nt fragments are not yet understood, but it was observed
`previously that the ∼21–23-nt spacing of cleavage sites
`could be altered by a run of uridines (Zamore et al. 2000).
`The specificity of dsRNA cleavage by E. coli RNase III
`appears to be mainly controlled by antideterminants,
`that is, excluding some specific base pairs at given posi-
`tions relative to the cleavage site (Zhang and Nicholson
`1997). The sequence dependence of dsRNA processing
`and target RNA cleavage in RNAi needs to be examined
`further.
`To test whether sugar-, base-, or cap-modification
`were present in processed ∼21-nt RNA fragments, we
`incubated radiolabeled 505-bp Pp-luc dsRNA in lysate
`for 1 h, isolated the ∼21-nt products, and digested it with
`P1 or T2 nuclease to mononucleotides. The nucleotide
`mixture was then analyzed by two-dimensional thin-
`layer chromatography (Fig. 4B). None of the four ribo-
`nucleotides were modified, as indicated by P1 or T2 di-
`gestion. We have previously analyzed adenosine to ino-
`sine conversion in the ∼21-nt fragments (after a 2 h
`incubation) and detected a small extent (<0.7%) deami-
`nation (Zamore et al. 2000); shorter incubation in lysate
`(1 h) reduced this inosine fraction to barely detectable
`levels. RNase T2, which cleaves 3⬘ of the phosphodiester
`linkage, produced nucleoside 3⬘-phosphate and nucleo-
`side 3⬘,5⬘-diphosphate, thereby indicating the presence of
`a 5⬘-terminal monophosphate. All four nucleoside 3⬘,5⬘-
`diphosphates were detected and indicate that the inter-
`nucleotidic linkage was cleaved with little or no se-
`quence specificity for the residue 3⬘ to the cleavage site,
`and according to the sequence analysis of the cloned ∼21-
`nt fragments, no significant sequence bias was observed
`for the residue 5⬘ of the cleavage site. In summary, the
`∼21-nt fragments are unmodified and were generated
`from dsRNA such that 5⬘-monophosphates and 3⬘-hy-
`droxyls were present at the 5⬘-ends. Analysis of the prod-
`ucts of dsRNA processing indicated that the ∼21-nt frag-
`ments are generated by a reaction with all the character-
`istics of an RNase III cleavage reaction (Dunn 1982;
`Robertson 1982, 1990; Nicholson 1999).
`
`Synthetic 21- and 22-nt RNAs mediate target
`RNA cleavage
`
`We chemically synthesized 21- and 22-nt RNAs, identi-
`cal in sequence to some of the cloned ∼21-nt fragments,
`and tested them for their ability to mediate target RNA
`degradation (Fig. 5A–C). The 21- and 22-nt RNA du-
`plexes were incubated at 100 nM concentrations in the
`lysate, a 10- to 20-fold higher concentration than the
`52-bp control dsRNA. Under these conditions, target
`RNA cleavage was readily detectable. Tenfold reduced
`concentrations of 21- and 22-nt duplexes (10 nM) still
`
`RNA interference is mediated by 21- and 22-nt RNAs
`
`caused target RNA cleavage but to a smaller extent (data
`not shown). Increasing the duplex concentration from
`100 to 1000 nM, however, did not further increase target
`degradation (data not shown), perhaps because of a lim-
`iting protein factor within the lysate. Single-stranded
`sense or antisense 21- and 22-nt RNAs at 100 nM con-
`centration did not affect target RNA expression, most
`likely because single-stranded RNAs are not stable in the
`lysate and degraded to mononucleotides within minutes
`(data not shown). We also found that preannealing of the
`short antisense RNAs to the target mRNA before the
`addition of lysate had no effect on target RNA expression
`(data not shown).
`RNase III makes two staggered cuts in both strands of
`the dsRNA, leaving a 3⬘ overhang of 2 nt. The 21- and
`22-nt RNA duplexes with 2- or 3-nt overhanging 3⬘ ends
`(duplexes 1, 4, 6) were more efficient in reducing the
`target RNA expression than the corresponding blunt-
`ended dsRNAs (duplexes 2, 5, 7) or the dsRNA with 4 nt
`overhang (duplex 3). Duplexes 6 and 7 are generally more
`effective for RNAi than duplexes 1–5, probably as a con-
`sequence of target RNA accessibility (because of RNA
`self-structure or RNA-coating proteins) or because of se-
`quence-specific effects in the reconstitution of the RNA-
`degrading complexes. The interference effects deter-
`mined in the translation-based assay (Fig. 5B) correlate
`well with the intensity of the cleavage bands observed by
`targeting 5⬘ radiolabeled model substrates with the 21-
`and 22-nt RNA duplexes (Fig. 5C). Together, these data
`suggest that 2–3 nt of overhanging 3⬘ ends are beneficial
`for reconstitution of the RNAi nuclease complex and
`may be required for high-affinity binding of the short
`RNA duplex to the protein components. A 5⬘ terminal
`phosphate, although present after dsRNA processing,
`was not required to mediate target RNA cleavage and
`was absent from the short synthetic RNAs.
`The synthetic 21- and 22-nt duplexes guided cleavage
`of sense as well as antisense targets within the region
`covered by the short duplex. This is interesting, consid-
`ering that the presumably base-paired clusters of ∼21-nt
`fragments derived from the 39-bp dsRNA (Fig. 2) can
`only be correlated to a predominant cleavage site on ei-
`ther the sense or the antisense target but not both. We
`interpret this result by suggesting that only one of two
`strands present in the ∼21-nt duplex is able to guide tar-
`get RNA cleavage and that the orientation of the ∼21-nt
`duplex in the nuclease complex is determined by the
`initial direction of dsRNA processing. It also implies
`that the processed short RNAs are present in a tight ri-
`bonucleoprotein complex and do not dissociate and re-
`bind during the time scale of the experiment. The pre-
`sentation of an already perfectly processed ∼21-nt duplex
`to the in vitro system, however, does allow formation of
`the active sequence-specific nuclease complex with two
`possible orientations of the symmetric RNA duplex.
`This results in cleavage of sense as well as antisense
`targets within the region of identity with the ∼21-nt
`RNA duplex.
`The target cleavage site is located near center of the
`region covered by the 21- or 22-nt RNAs, 11 or 12 nt
`
`GENES & DEVELOPMENT
`
`193
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`withintheregionofidentityofthedsRNAs.Forprecisedeterminationofthecleavagesitesoftheantisensestrand,alowerpercentagegelwasused(datanotshown).
`inFigure5A.Theregiontargetedbythe52-bpdsRNAorthesense(s)orantisense(as)strandsareindicatedbytheblackbarstothesideofthegel.Thecleavagesitesarealllocated
`duplexes1–7(100nM)wereincubatedwithtargetRNAat25°Cfor2.5hinDrosophilalysate.Thestable5⬘cleavageproductswereresolvedonthegel.Thecleavagesitesareindicated
`(C)PositionofthecleavagesitesonsenseandantisensetargetRNAs.ThetargetRNAsequencesareasdescribedinFigure3B.Control52-bpdsRNA(10nM)or21-and22-ntRNA
`luminescenceoftargettocontrolluciferasenormalizedtoabuffercontrol(buf)isblotted;errorbarsindicatestandarddeviationscalculatedfromatleasttwoindependentexperiments.
`(5nM)or21-and22-ntRNAduplexes1–7(100nM)targetingfull-lengthPp-lucmRNAweretestedinthetranslation-basedRNAiassayasdescribedinFigure1B.Therelative
`circles(seelegendtoFig.4A)andweredeterminedasshowninFigure5B.(B)RNAinterferenceassay.ToevaluatetheefficiencyoftargetRNAdegradation,control52-bpdsRNA
`transcription,andafractionoftranscriptsmaycontainuntemplated3⬘nucleotideaddition.ThetargetRNAcleavagesitesdirectedbythesiRNAduplexesareindicatedasorange
`nucleotidesindicatedingreenarepresentinthesequenceofthesyntheticantisensestrandofduplexes1and3.Bothstrandsofthecontrol52-bpdsRNAwerepreparedbyinvitro
`dsRNAs(Fig.4A),exceptforthe22-ntantisensestrandofduplex5.ThesiRNAsinduplexex6and7wereuniquetothe111-bpdsRNA-processingreaction.Thetwo3⬘overhanging
`of21-and22-ntshortinterferingRNAs(siRNAs)isshowninblue,theantisensestrandinred.ThesequencesofthesiRNAswerederivedfromtheclonedfragmentsof52-and111-bp
`Figure5.Synthetic21-and22-ntRNAsmediatetargetRNAcleavage.(A)Graphicrepresentationofcontrol52-bpdsRNAandsynthetic21-and22-ntdsRNAs.Thesensestrand
`
`Elbashir et al.
`
`194
`
`GENES & DEVELOPMENT
`
`

`
`
`
`Downloaded from genesdev.cshlp.org on November 11, 2015 - Published by Cold Spring Harbor Laboratory Press
`
`
`
`
`
`
`
`downstream of the first nucleotide that is complemen-
`tary to the 21- or 22-nt guide sequence (Fig. 4A,B). Dis-
`placing the sense strand of a 22-nt duplex by two nucleo-
`tides (cf. duplexes 1 and 3 in Fig. 5A) displaced the cleav-
`age site of only the antisense target by two nucleotides.
`Displacing both sense and antisense strands by two
`nucleotides shifted both cleavage sites by two nucleo-
`tides (cf. duplexes 1 and 4). We predict that it will be
`possible to design a pair of 21- or 22-nt RNAs to cleave a
`target RNA at almost any given position.
`The specificity of target RNA cleavage guided by 21-
`and 22-nt RNAs appears exquisite, as no cleavage sites
`are detected outside of the region of complementarity to
`the 21- and 22-nt RNAs (Fig. 5C). It should, however, be
`noted that the nucleotides present in the 3⬘ overhang of
`the 21- and 22-nt RNA duplex may contribute less to
`substrate recognition than the nucleotides near the
`cleavage site. This is based on the observation that the
`3⬘-most nucleotide of the antisense strand of active du-
`plexes 1 or 3 (Fig. 5A) is not complementary to the target.
`A detailed a