RNA InterferenceRegulates Gene Action

219

Extraordinary advances had been made in understanding the intricate

workings of the cell using recombinant DNA techniques, and so

biologists were amazed when a completely unsuspected pathway for the

control of gene expression was discovered in the 1990s. Researchers

studying plants, fungi, worms, flies, and even trypanosomes came to real-

ize that some unexpected results were the consequence of RNA molecules

silencing the expression of genes in a process now known by the general

term of RNA interference (RNAi). New examples of RNAi processes con-

tinue to be discovered but all share a common element—small RNAs that

pair with other RNAs and possibly DNA, and silence gene expression

through a variety of mechanisms, including mRNA degradation, DNA

methylation and histone modification, and translational inhibition.

If RNAi were simply an unexpected new mechanism of gene regulation,

it would still be a remarkable addition to our understanding of life, but the

importance of RNAi extends beyond its biological significance. Just as the

polymerase chain reaction (PCR) brought about a revolution in our abil-

ity to manipulate and detect nucleic acids, so RNAi is revolutionizing the

ways in which we can experimentally control the expression of any gene.

RNAi provides a general tool for reverse genetics—discovering the function

of a gene by interfering with its activity and observing the resulting phe-

notype, rather than by finding an atypical phenotype and then trying to

determine which gene is mutated. As we saw in Chapter 6, biologists have

C H A P T E R 9

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 219

220 � CHAPTER 9

invented clever methods for making specific changesto genes, but many of these are laborious and slow.RNAi is simple and rapid; all that is needed is thesequence of just a portion of the gene to be target-ed, the synthesis of the corresponding RNA mole-cule, and a method to introduce it into cells andorganisms.

The technical advance of RNAi could not havecome at a more opportune time. Large-scalesequencing is now producing vast amounts ofgenomic data that are being subjected to computa-tional analysis to predict genes and other genetic ele-ments. However, this is only the first step. As biolo-gists we want to understand the functions of the pre-dicted genes and their roles in the cell.Bioinformaticians are finding thousands—or tens ofthousands—of genes. How can the experimentalistsbegin to make sense of them all? We will reviewother methods of genome-scale functional analysisin Chapter 13, but here we first examine the fasci-nating biology of RNAi and then look at how RNAiknockdown of gene expression in plants, yeast,mammals, and fungi provides a powerful tool forstudying gene function.

Cosuppression of Transgenes Is Observedin Plants

The introduction of recombinant DNA techniquesled to the modern biotechnology industry, promot-ed by endeavors to create commercially desirablealterations in organisms. Richard Jorgensen and hiscolleagues undertook the project to produce morerichly colored petunia flowers. An additional copy ofa petunia pigment gene encoding chalcone synthasewas introduced into a petunia plant under the con-trol of a very powerful promoter with the expectationthat increased production of the encoded enzymewould produce flowers with a deeper purple color.Remarkably, some of the transgenic plants producedwhite flowers devoid of any pigment, despite the factthat both the endogenous and transgenic copies ofchalcone synthase gene were present (Fig. 9-1).Other transgenic lines had variegated patterns of pig-ment expression—petals displaying pigmented andunpigmented areas. This surprising phenomenon wasnamed cosuppression, because the transgene triggerednot only its own silencing but also that of theendogenous chalcone synthase gene, which had thesame sequence. This occurred even though the genes

were driven by different promoters, thus showingthat the effect required only homology between themessenger RNAs (mRNAs).

Interestingly, the cosuppression phenotypes werenot stable. Differently patterned flowers appeared onseparate branches of a single plant, and seeds grownfrom these newly patterned flowers sometimesreverted to the original phenotype of the parentplant. Occasionally, progeny of the variant flowershad entirely new patterns different from either theoriginal plant or the variant parent flower. Similarphenotypic instability was also seen in progeny fromtransgenic plants having fully purple flowers; thesecould give rise to flowers that lacked pigment.However, the DNA sequence of cosuppressing trans-genes was unchanged, even in parts of the plants withaltered flowers. This phenotypic instability withoutchange in sequence was similar to many of the unsta-ble epigenetic phenomena we discussed in Chapter 8.Biologists therefore began to regard cosuppression asan epigenetic process.

Further analysis showed that the level of trans-gene transcription is an essential determinant ofcosuppression. Transgenes under the control ofstrong promoters, transcribed at high levels, aremuch more likely to cause cosuppression than trans-genes transcribed at lower levels. Also, the numberof transgene copies affects the likelihood that cosup-pression will occur. By crossing transgenic plantswith themselves or to wild-type plants, investigatorsshowed that the number of copies of the chalconesynthase transgene determined the cosuppressionphenotype. For example, some petunias producedvariegated flowers when just a single copy of the chal-cone synthase transgene was present, but were purewhite when the plants carried two copies of the trans-gene.

But biologists were using transgenic techniquesto do more than produce prettier flowers—theyhoped to create disease-resistant plants. Experimentswith Escherichia coli expressing an altered form of aphage replicase protein had shown that the alteredprotein could compete with the infecting phage’sown replicase and so prevent phage replication.Plant virologists planned to mimic this process inplants by expressing dominant negative forms of keyproteins from plant viruses, in the expectation thatthese proteins would interfere with viral replication.Thus, transgenic plants were created that expresseda mutant form of viral replicase of potato virus X(PVX), and these transgenic plants were indeed

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 220

35Spromoter

Isolate pieces ofleaves from plant

Treat withhormonesto regenerateplants

Backcross to fullypurple parental plant

Observephenotypesof somaticgrowth

Observephenotypesof progeny

Transgenote218.43 andsomatic variants

Progeny

Transgenote218.56 andsomatic variants

Progeny

Select for transgeneNPTII using kanamycin

(a)

(b)

(c)

LB BB

Chalconesynthase

TransformAgrobacterium

NPTIIAgrobacteriumvir genes Infect with

Agrobacterium

FIGURE 9-1

An overexpressed transgene triggers silencing of the homologous cellular gene. (a) In an attempt to createmore brightly pigmented flowers, a cDNA encoding the pigment producing enzyme chalcone synthase wasplaced under the control of the very strong 35S promoter from the cauliflower mosaic virus. This constructwas placed in a binary vector and transferred into purple petunias using Agrobacterium-mediated gene trans-fer. The binary vector also contained a gene encoding kanamycin resistance to permit selection of transgenicseedlings. (For more details on this method, see Chapter 6.) (b) Overexpression of chalcone synthase in petu-nia resulted surprisingly not in the production of deeply colored flowers but instead in plants bearing whiteor variegated flowers. To test the pattern of inheritance of the transgene, variegated transgenic progeny werebackcrossed to the fully purple parental plant. (c) The flower phenotypes are unstable, both within a singleplant (somatically) and through multiple generations.

RNAi Regulates Gene Action � 221

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 221

222 � CHAPTER 9

resistant to the virus. The mechanism of viral repli-cation is highly conserved among different strains ofviruses, so virologists had hoped that this approachwould interfere with the replication of many viralstrains. However, these transgenic plants were resist-ant only to PVX and its very closely related strains.This was not the only surprising result of theseexperiments—remarkably, the most resistant linesexpressed the transgene mRNA at nearly unde-tectable levels! Nor was the replicase protein neededfor inducing resistance—transgenic lines made withan altered replicase gene carrying a frameshift muta-tion were as resistant as lines expressing full-lengthprotein. It seemed that the infecting virus was beingsilenced by a mechanism similar to cosuppression.

At what level was cosuppression occurring? Thecosuppressed endogenous genes were transcribed atnormal levels, as shown in nuclear runoff assays oftranscription (Fig. 9-2). Experiments isolating nucleifrom cosuppressing plants and radiolabeling thetranscripts clearly showed that the promoters of thetransgenes and endogenous genes were active.However, the mRNA accumulated to very low lev-els, which suggest that cosuppression was the resultof some kind of posttranscriptional gene silencing inwhich the mRNA was being destroyed. Other exam-ples of cosuppression, involving a variety of endoge-nous and viral genes, were found, but cosuppressionseemed to be an intriguing peculiarity of only theplant world. Then the unexpected results of anexperiment in Caenorhabditis elegans revealedanother case of unusual gene silencing, and itbecame clear that a new, perhaps universal, biologi-cal process had been discovered.

An Antisense Experiment Gone Awry Pointsthe Way to RNAi in Worms

Embryonic development has fascinated biologistsfor centuries. How does a single-celled zygote divideand differentiate to produce an adult organism madeup of many different cell types arranged in specificpatterns? Developmental biologists study simpleorganisms to tease out the pathways leading fromthe zygote to the adult animal, and the nematodeworm C. elegans has proved to be especially valuable.

Specific proteins and RNAs are distributed asym-metrically in the worm zygote. During cell division,different cytoplasmic components are partitionedinto specific cells, eventually directing the differenti-

ation of the many specialized cell types that make upthe worm. This process can be disrupted, and thereare many C. elegans mutants with altered or defec-tive embryogenesis. For example, one such mutant,par-1, fails to make an asymmetric cleavage in thefirst cell division; instead, the first division occursevenly down the center of the zygote, giving rise toan unpatterned embryo that arrests early in develop-ment.

The par-1 mutation was mapped to a specificlocation in the worm genome, but in 1995, theworm genome had not yet been sequenced and thepar-1 gene had first to be cloned before its analysiscould advance. At that time, a standard approachwas used to determine which of the several predict-ed genes in a genetically mapped region was defec-tive in mutants. In this method, genomic DNA or acDNA corresponding to each candidate gene wasfirst isolated and then expressed in par-1 mutants. Ifthe DNA was capable of restoring the normal phe-notype, the sequence of that candidate gene wasanalyzed in the mutant. Unfortunately, neither com-plete genomic clones encompassing the region norcomplete cDNA clones of the candidate par-1 genecould be recovered from cDNA libraries, and so theidentity of par-1 could not be confirmed using thethen-standard method.

Instead, Su Guo and Kenneth Kemphues used anantisense RNA to identify the par-1 gene fromamong the possible candidates. For each candidategene, they created a plasmid containing a portion ofits cDNA and transcribed that in vitro to produce anantisense copy of the gene. The resulting RNA wasthen injected into the gonads of wild-type adult her-maphrodites. An antisense RNA is complementaryto the mRNA produced from a gene and candecrease levels of the corresponding protein, pre-sumably because the antisense RNA binds themRNA and inhibits its translation or in some waydestabilizes it. In this experiment, the progeny ofworms injected with an antisense RNA to one of thecandidate genes died with a phenotype very similarto that of the par-1 mutants. Guo and Kemphueshad identified the par-1 gene!

However, the investigators were surprised at theresult of a key control for the experiment. In this con-trol, the worms were injected with the sense par-1mRNA, from a partial cDNA clone that could not betranslated into a complete protein. Injecting the senseRNA into worms produced a phenotype in theirprogeny very similar to that seen with the antisense

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 222

RNAi Regulates Gene Action � 223

35Spromoter

Par

enta

l(s

usce

ptib

le)

Tran

sgen

ic #

1(s

usce

ptib

le)

Tran

sgen

ic #

2(re

sist

ant)

(a)

(b) (c)

PVX RdRPtransgene

Ubiquitincontrol gene

PVX RdRPtranscripts

Northern blot

Transgenic #1 (susceptible)

Parental(susceptible)

Transgenic #2 (resistant)

PVX RdRP

Isolate RNA, run gel,and prepare Northernblot. Hybridize withPVX RdRP probe

Isolate nuclei and addradioactive nucleotidesAllow transcription to proceed

Transcriptionincorporateslabelednucleotides

Apply unlabeledDNA of transgeneand control tomembrane

Hybridize labeled RNAto prepared membranes“Seal-a-Meal” bag

Tobaccoplant

Agrobacterium

TransformAgrobacterium

Infect tobacco plant;select transgenic progeny

*

*

FIGURE 9-2

Nuclear run-on assays show that cosuppression occurs posttranscriptionally. (a) The RNA-dependent RNApolymerase (RdRP) from potato virus X (PVX) was placed under the control of a strong promoter, cloned intoa binary vector, and transferred into tobacco plants by Agrobacterium-mediated gene transfer. Transgenicplants were isolated and tested for their resistance to PVX. (b) RNA was isolated from the plants and trans-gene expression was analyzed by Northern blotting. Surprisingly, those transgenic lines resistant to the virusdid not accumulate the transgene transcript. (c) However, the transgene was in fact transcribed. Nuclei wereisolated from the parental and transgenic plants, and radioactive nucleotides were added so that transcrip-tion could proceed. The labeled RNA was then hybridized to RNA immobilized on filters. (The highlyexpressed gene, ubiquitin, was assayed as a control.) Both susceptible and resistant lines expressed the trans-gene at equally high levels, thereby demonstrating that the transgene transcript was being eliminated in theresistant lines. This suggested that when the resistant lines were infected with PVX, the viral RNA was quick-ly degraded, blocking viral replication.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 223

224 � CHAPTER 9

RNA. It was possible that the simple act of injectingthe worms might have disturbed their developmentand, if so, injection of any RNA would cause the samephenotype (Table 9-1). Injection of sense and anti-sense RNAs from other genes had no effect, howev-er, thus demonstrating that the act of injection didnot affect development. But how could these resultsbe interpreted? How could both the sense and anti-sense RNAs silence expression of par-1?

Double-stranded RNA Is the Trigger forRNAi

The search to understand how both sense and anti-sense RNA suppress gene expression in wormsrevealed a completely unsuspected phenomenon,one of the most remarkable discoveries of modernbiology. Andrew Fire and Craig Mello (who wereawarded the Nobel Prize in 2006) found, with theircolleagues, neither sense nor antisense RNA alonewas responsible, but both were needed. The in vitrotranscription method produces a small amount ofdouble-stranded RNA (dsRNA), and it is this prod-uct, not the much more abundant single-strandedsense or antisense RNA, that is the active agent insilencing. Used for injection, dsRNA was very effec-tive in reproducing knockout phenotypes of manygenes (Table 9-2). Only sequences derived from themature RNA had this effect—sequences from thepromoter or intronic regions of genes did not alterthe phenotype of injected worms.

At about the same time, biologists studying post-transcriptional gene silencing of plant viruses real-ized that plants carrying both forward- and reverse-

oriented transgenes corresponding to viralsequences were much more resistant to viral chal-lenge than plants carrying a transgene oriented injust one direction. This observation suggested thatdsRNA produced in the plants carrying transgenesoriented in both directions might silence RNAsmore effectively than sense or antisense RNAs alone.A plasmid was constructed that contained an invert-ed repeat corresponding to a potato virus Y (PVY)protease gene essential for viral transmission (Fig. 9-3). When such sequences are transcribed, the result-ing RNA can fold back on itself to produce adsRNA. Plants with these inverted repeats weremore effectively protected from infection by PVYthan even those plants that carried separate senseand antisense transgenes.

Another clue that dsRNA might trigger silencingwas the result of a chance observation made in exper-iments to make the transgenic tobacco plants resist-ant to the tobacco etch virus (TEV) by expressing theTEV coat protein as a transgene. When the trans-genic plants were infected with TEV, there were signsof infection on inoculated leaves, but new leaves didnot become infected and they were resistant to fur-ther virus infection. It was found that transcriptsfrom the coat protein transgene were degraded post-transcriptionally in these new leaves. Somehow, virusreplication in the old, infected leaves had triggeredthis silencing of the transgene in the new leaves. Thelatter were resistant to infection because the TEVRNA genome was being destroyed by the homology-dependent mechanism that was degrading the trans-gene. Soon, it was found that if a virus was engi-neered to include a portion of any plant gene, theendogenous copy was silenced along with the repli-cating virus.

One experiment made use of the phytoene desat-urase gene, whose expression is easily tracked becauseleaves deficient in the phytoene desaturase enzyme arewhite. White leaves were observed whether the phy-toene desaturase cDNA was present in the antisenseor sense orientation in the virus. When this work wasfirst done, a decade before the observations in worms,silencing was believed to be caused by antisense RNAsproduced either directly (in the case of the antisense-oriented clone) or by the process of viral replication.However, the viral replication intermediate is adsRNA, and once the role of dsRNA in RNAi inworms was apparent, biologists re-interpreted theseand other related results from plant viruses. It wasnow clear that silencing is triggered by the double-

TABLE 9-1. Various sense and antisense RNAs were injectedinto Caenorhabditis elegans and scored for induction ofembryonic lethality

Molecule No. of worms Embryonic injected injected lethality (%)

par-1 antisense 16 52par-1 sense 12 54Drosophila cofilin gene, 8 0

antisenseCaenorhabditis elegans 8 0

zygotic gene, antisenseWater 4 0

From Guo and Kemphues, Cell 81: 611–620 (1995).Adult hermaphrodite worms were injected, and the number of

hatched versus unhatched embryos was counted.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 224

stranded replication intermediates rather than just theantisense strand. (There are plant DNA viruses thatcan trigger similar silencing even though they do notreplicate through a dsRNA intermediate. The mecha-nism of this silencing remains a topic of debate.)

RNAi Is Used to Block Gene Expression inMany Organisms

No general methods have yet been developed for cre-ating targeted mutations or knockouts in C. elegans,so finding mutants in a specific gene requires labori-ous screening of mutagenized populations of worms.

RNAi Regulates Gene Action � 225

In contrast, knocking down gene expression usingRNAi is straightforward—all that is needed is thesequence of the target gene so that it can be used tosynthesize a dsRNA. Fortuitously, RNAi was discov-ered just as the worm genome was being sequenced,and soon the sequences of all C. elegans genes wereavailable. Very quickly, worm biologists seized onRNAi as a new and precise tool for controlling geneexpression.

Initially, RNAi experiments were carried out justas the initial antisense experiments had been—bytedious injection of dsRNA into the gonads of adulthermaphrodites. But investigators soon discoveredthat RNAi worked equally well if worm larvae were

TABLE 9-2. Double-stranded RNA derived from a given gene induces phenotypes resembling a null mutation

Gene Segment Size (kb) Injected RNA F1 phenotype

unc-22 unc-22 null mutants: strong twitchersunc22A Exons 21–22 742 Sense Wild type

Antisense Wild typeSense + antisense Strong twitchers (100%)

unc22B Exon 27 1033 Sense Wild typeAntisense Wild typeSense + antisense Strong twitchers (100%)

unc-22C Exon 21–22 785 Sense + antisense Strong twitchers (100%)

fem-1 fem-1 null mutants: female (no sperm)fem1A Exon 10 531 Sense Hermaphrodite (98%)

Antisense Hermaphrodite (98%)Sense + antisense Female (72%)

fem1B Intron 8 556 Sense + antisense Hermaphrodite (>98%)

unc-54 unc-54 null mutants: paralyzedunc54A Exon 6 576 Sense Wild type (100%)

Antisense Wild type (100%)Sense + antisense Paralyzed (100%)

unc54B Exon 6 651 Sense Wild type (100%)Antisense Wild type (100%)Sense + antisense Paralyzed (100%)

unc54C Exon 1–5 1015 Sense + antisense Arrested embryos and larvae (100%)unc54D Promoter 567 Sense + antisense Wild type (100%)unc54E Intron 1 369 Sense + antisense Wild type (100%)unc54F Intron 3 386 Sense + antisense Wild type (100%)

hih-1 hih-1 null mutants: lumpy-dumpy larvaehih1A Exons 1–6 1033 Sense Wild type (<2% lpy-dpy)

Antisense Wild type (<2% lpy-dpy)Sense + antisense Lpy-dpy larvae (>90%)

hih1B Exons 1–2 438 Sense + antisense Lpy-dpy larvae (>80%)hih1C Exons 4–6 299 Sense + antisense Lpy-dpy larvae (>80%)hih1D Intron 1 697 Sense + antisense Wild type (<2% lpy-dpy)

From Fire et al., Nature 391: 806–811 (1998).Sense, antisense, and double-stranded mixtures of sense and antisense RNA from several different genes were inject-

ed into adult hermaphrodites. The progeny were examined and compared to null phenotypes of the given gene. dsRNAstargeted to exonic sequences generated mutant phenotypes, whereas promoter and intronic sequences did not result inabnormal phenotypes. This suggested that the RNA interference was acting on processed mRNAs and not pre-mRNAs oruntranscribed sequences.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 225

226 � CHAPTER 9

mixed directly with dsRNA or, even more remark-ably, if worms were fed bacteria expressing dsRNA(Fig. 9-4)! Soon, worm biologists constructedstrains of E. coli in which each strain expressed adsRNA homologous to a different C. elegans gene.Now E. coli strains for feeding to worms are availablethat express dsRNAs corresponding to every gene inthe worm genome.

RNAi has also been enthusiastically embraced byplant biologists. As in worms, there are no generaltools for creating specific mutations in plants, buttransgenes transcribed as inverted repeats are veryeffective at suppressing gene expression in plants.RNAi is an especially important tool in plantsbecause many agriculturally important plants arepolyploid. When a plant is tetraploid or hexaploid, it

can be extremely difficult to isolate a plant withmutation in every copy of a gene. RNA interferencesilences all copies of a gene homologous to thetransgene.

Because RNAi requires close sequence similaritybetween the target gene and the dsRNA, a dsRNAtargeted to a particular gene can inhibit relatedgenes. Therefore, the sequences of the dsRNA usedfor RNAi must be carefully selected so that they areidentical only to the gene or genes of interest. Thisis typically not a problem. There are usually manysilent changes in codon usage that make even close-ly related genes sufficiently different at the mRNAlevel to permit precise silencing. Even when a greatdeal of sequence similarity exists in the coding regionof related genes, all is not lost. The 5´- and 3´-

(a)

(b)

35S promoter

PVY Pro ORFSense

Terminator

35S promoter

PVY Pro ORFAntisense

Terminator

35S promoter

PVY Pro ORF PVY Pro ORFSense andantisense

Create transgenic linewith each construct

Infect with PVY

Construct

Immune Resistant Susceptible

Sense 2 2 23 27

Antisense 1 0 24 25

Sense andantisense

10 2 15 27

Number ofimmunelines

Number ofresistantlines

Number ofsusceptiblelines

Totalnumber oflines

Terminator Terminator

FIGURE 9-3

Expressing an inverted repeat trans-genic construct in plants suggests thatdouble-stranded RNA (dsRNA) caninitiate cosuppression. (a) Constructsexpressing the protease gene of potatovirus Y (PVY) were prepared thatexpressed a sense copy of the gene, anantisense copy, or both sense and anti-sense copies from the same construct.Each of these PVY protease expressionconstructs was integrated into a tobac-co plant by Agrobacterium-mediatedgene transfer. (b) Transgenic plantswere challenged by infecting themwith sap from PVY-infected plants andobserved for symptoms of leaf mottlingand leaf stunting. The constructexpressing both sense and antisense ofthe protease gene was much moreeffective at producing transgenicplants immune or resistant to viralinfection than the sense- or antisense-only constructs. ORF, open readingframe.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 226

RNAi Regulates Gene Action � 227

untranslated regions of mRNAs are typically muchless conserved between these genes and so can beused to target RNAi to a specific gene.

A Nuclease Destroys RNAs Homologous tothe Silenced Gene

How does dsRNA silence genes? There were fourlines of evidence that the silencing dsRNAs wereinteracting with mRNA. First, transcription ofsilenced genes typically proceeds at normal rates.Second, mRNA levels were nevertheless decreased inorganisms undergoing silencing. Third, dsRNA (ortransgenes in the case of plants) could cause silenc-ing only when complementary to the exonic por-tions of genes, which suggests that the dsRNA wasprobably interacting in some way with maturemRNA. Finally, it was unlikely that any change hadbeen made in the DNA sequence, because the RNAieffect declined over time.

Biochemical analysis of the mechanism of RNAibegan with the use of extracts of Drosophila embryosand cells. dsRNA was incubated in these extracts andwhen RNA homologous to the dsRNA was subse-quently added to the extracts, it was quickly degrad-ed. RNAs not homologous to the dsRNA were com-pletely stable. The protein machinery responsible forthis sequence-specific nuclease activity was called theRNA-induced silencing complex (RISC). Any RNAcould be targeted by RISC if its homologous dsRNAwas present in the extract. Drosophila cells could evenbe pretreated with dsRNA. When extracts were pre-pared from these cells, they contained RISC activitycorresponding to the transfected dsRNA.

Small RNAs Are Found in Plants UndergoingSilencing

Because dsRNA triggered the creation of thesequence-specific RISC nuclease, it seemed probablethat RISCs would contain an antisense RNA thatdirected the nuclease to the homologous mRNA.Many investigators used Northern blots to search forthese antisense RNAs but without success, as theelusive antisense RNAs were lost on the agarose gelsused to separate RNA for standard Northern blotanalysis. Agarose gels are excellent for separatingRNAs of 500 nucleotides or more in size, but small-er RNAs diffuse rapidly and do not form tight bands,thereby evading detection.

Ligate intoplasmid

Transform each plasmid from thelibrary into E. coli strainexpressing T7 RNA polymeraseand deficient in RNase III

Array strains withdifferent plasmidsinto wells

Screen for phenotype of interest

Add worms toeach well in plate

PCR amplify portion of geneof interest from cDNA library:use Taq polymerase to leaveA residues for TA cloning

T7 RNApolymerase

lacpromoter

T7T7T7

AmpAmpAmpAmp

T7T7T7

A

AT T

A

A

FIGURE 9-4

RNA interference can be induced in worms simply by feedingthem Escherichia coli expressing dsRNAs. A DNA sequencefrom a gene is amplified by PCR using Taq polymerase toleave A overhangs for TA cloning (Chapter 6). The PCR prod-ucts were cloned into a vector containing two T7 RNA poly-merase promoters in inverted orientations flanking the cloningsite. PCR products are ligated into a plasmid and transformedinto a special strain of E. coli deficient in RNase III, a dsRNase.The strain also contains T7 RNA polymerase under the controlof the lac promoter, so that it can be induced by addition ofisopropyl thiogalactoside (IPTG). Individual strains of bacteriaare prepared, each transformed with a plasmid containing adifferent gene. Each strain is plated on agar in a well of a 96-multiwell microtiter plate. The agar contains IPTG to induceexpression of the dsRNA. Then, worms are added and allowedto consume the bacteria. The worms or their progeny areexamined for interesting phenotypes.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 227

228 � CHAPTER 9

Polyacrylamide gels provide much better resolu-tion of small molecules. Andrew Hamilton and DavidBaulcombe used these gels to examine RNAs in plantsundergoing RNAi and found a population of smallRNAs homologous to the gene undergoing silencing(Fig. 9-5). These small RNAs were of uniform length,21–25 nucleotides long, and were observed in plantswhere silencing had been triggered by single trans-genes, by inverted repeats, and by the replication ofRNA viruses. Both antisense and sense RNAs werepresent, indicating that dsRNA is the common ele-ment of all these silencing processes. Surprisingly,antisense RNAs were present even in those plantswhere cosuppression had been triggered by a single,sense-oriented transgene. We will see below thatgenetic evidence is beginning to explain how theseantisense RNAs are produced.

These small RNAs were not unique to plants;they were soon observed in worms and flies under-going RNAi. Their origins and functions were inves-tigated by adding radiolabeled dsRNA to extractsfrom Drosophila embryos. The dsRNA was quicklydigested into fragments 21–22 nucleotides long,and when RISC was purified from these extracts, theradiolabeled RNAs copurified with the RISC nucle-ase activity. Presumably, the small RNAs were guid-ing the RISC nuclease to homologous mRNAs.

Dicer Chops Double-stranded RNA toProduce Small RNAs

The discovery of small RNAs in organisms undergo-ing silencing immediately led to a search for thegene or genes encoding the dsRNA RNases. Thisresearch was carried out in Drosophila, because theDrosophila genome sequence was nearly complete.Computer searches of predicted proteins identifiedproteins containing RNase III domains, the onlytype of RNases known to be capable of cuttingdsRNA.

One approach to studying these candidate geneswould have been to clone the genes, express them inbacteria, purify the proteins, and test their ability tochop dsRNAs into short RNAs. There are limita-tions to this approach. Not the least is that it presup-poses that only a single protein is responsible for thenuclease activity, thus ignoring any cofactors thatmight be required for assembly of an active complex.A different approach is to add epitope tags to thecDNAs of the candidate proteins and transfect these

(a)

(b)

(c)

35Spromoter

Agrobacterium-mediated transfer;select multiple transgenic lines

Purify RNA and blotwith indicated probes

Antisenseprobe to ACO

1500 nt1000 nt

1500 nt1000 nt

30 nt

20 nt

30 nt

20 nt

Pare

ntTr

ansg

enic

1Tr

ansg

enic

2

Tran

sgen

ic 1

Tran

sgen

ic 2

Sense ACOprobe

Northern blotagarose gel

Northern blotpolyacrylamidegel

Pare

nt

LB

AMP RB

ACO

NPTII

FIGURE 9-5

Small RNAs homologous to the silenced gene are present inplants undergoing silencing. (a) Transgenic tomato plantswere constructed by Agrobacterium-mediated gene transferto overexpress the enzyme 1-aminocyclopropane-1-car-boxylate oxidase (ACO), whose gene is under the control ofthe very strong 35S promoter. (ACO catalyzes a key step inethylene synthesis, a gas that is important for fruit ripening.)(b) Some lines exhibited silencing of the endogenous ACOmRNA, as observed by Northern blotting on agarose gels(the ACO transcript is roughly 1.2 kb). (c) However, whenNorthern blots were prepared from samples electrophoresedon polyacrylamide gels (which are capable of resolving verysmall RNAs), a population of 22–25-nucleotide RNAs wasfound. Curiously, both sense and antisense RNAs werefound, even though the transgene initiating ACO silencingwas present in the sense orientation only.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 228

tagged cDNAs into cells. Extracts prepared from thetransfected cells are immunoprecipitated with anti-bodies recognizing the epitope tags. The expressionand purification is performed in Drosophila cells thatsupport RNAi. Thus, if some other accessory factoris needed in addition to the protein product of thecandidate gene, it is likely to be immunoprecipitatedalong with the tagged protein. The experiment wasperformed using the candidate RNase III proteins.

RNAi Regulates Gene Action � 229

Immunoprecipitated proteins were tested, and oneof the tagged candidates displayed a nuclease activi-ty that cut long dsRNAs into small RNAs. Thesewere equal in size to those observed in organismsand in purified RISC from in vitro experiments (Fig.9-6). The candidate gene, a predicted Drosophilagene of unknown function, was named Dicer for theability of its protein to chop double-stranded RNAinto small, equally sized pieces. The initial identifica-

Unt

reat

ed d

sRN

A

RIII b

Transfect constructs individuallyinto Drosophila cells

Lyse cellswith detergents

Add anti-T7beads

Incubate to bindprotein; wash awayunbound protein; addradiolabeled dsRNA Isolate RNA

and run poly-acrylamide gel.Detect dsRNAsubstrate byautoradiography

M Cel

l ext

ract

bef

ore

pre

cip

itatio

n

Dro

sha

CG

4792

(Dic

er)

Hom

eles

s

β-ga

l

24 nucleotides22 nucleotides20 nucleotides

T7

dsrm

Paz RIII aCG4792 Helicase

T7 Helicase

RIII bT7 RIII a

β-gal

Drosha

Homeless

β-gal

FIGURE 9-6

Dicer is discovered by a candidate gene approach. The genes encod-ing two RNase III proteins, Drosha and a candidate gene encoding aprotein named CG4792, were PCR-amplified using primers that addeda 5′ T7 peptide sequence and cloned into an expression construct.Because CG4792 contained a helicase domain, a related protein with ahelicase domain but no RNase III domain, Homeless, was tagged. Thetagged constructs were individually transfected into Drosophila S2cells. The cells were lysed, and agarose beads covalently linked to anti-T7 antibodies were added so that the tagged proteins became bound tothe beads. The beads were washed to remove all other components ofthe incubation mixture and radiolabeled dsRNA was added. RNA waspurified from these incubations and analyzed by polyacrylamide gelelectrophoresis. Labeled small RNA fragments were compared to simi-lar preparations from the control transfected S2 cells. Only extracts fromcells expressing tagged CG4792 could digest dsRNA into small RNAs,and so the candidate gene encoding CG4792 was dubbed Dicer.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 229

230 � CHAPTER 9

tion was carried out using immunoprecipitates thatmight have contained factors associated with Dicer.But it was later shown, by expressing and purifyingthe protein from a baculovirus system, that Dicerprotein alone digests dsRNA.

With Drosophila Dicer in hand, biologists usedcomputers to search for homologous sequences inother organisms. Some organisms have multipleDicer family members—Drosophila has two,Arabidopsis has four, and C. elegans has but one. Asearch of the not-yet-complete human genomesequence revealed a copy of the Dicer gene. Thisfinding suggested that RNAi might well occur inmammalian cells. Indeed, the protein specified byhuman Dicer cDNA was later found to have exactlythe same double-stranded RNase activity as that ofDrosophila Dicer.

Mimicking Dicer Products Induces RNAi inMammalian Cells

Dicer produces small RNAs from a dsRNA, andthese small RNAs are very similar to those observedin organisms undergoing RNAi, which suggests thatDicer is a key player in RNAi. One of the first ques-tions asked was, what exactly is the nature of thesesmall RNAs produced by Dicer? To answer this,dsRNAs of known sequence were digested inextracts of Drosophila embryos that contained abun-dant Dicer activity. The resulting small RNAs werepurified, cloned, and sequenced. Because they wereso small and their ends were unknown, the RNAswere cloned directly by using RNA ligase, anenzyme that ligates single-stranded RNA and DNA,to link primers to the ends of the small RNA. Theseprimers were used for PCR amplification andcloning. These clones were sequenced, revealingthat Dicer attacks dsRNAs from the ends, choppingoff 21- and 22-nucleotide RNA molecules and leav-ing 3´ overhangs, 2 bp long, with a phosphate at the5´ end and a hydroxyl group at the 3´ end.

Based on this information, synthetic RNAoligonucleotides were made to mimic Dicer prod-ucts—these oligonucleotides were double stranded,with 2-bp 3´ overhangs and 3´-hydroxyl groups(Fig. 9-7). When these small RNAs were added toDrosophila embryo extracts, they worked nearly aswell as long dsRNAs in triggering mRNA degrada-tion. Therefore, the products of Dicer nucleaseactivity are sufficient to induce RNAi.

Long dsRNAs were very successful for inducingRNAi in flies and worms, but in most mammaliancells, dsRNA completely stops all protein translation.This is because dsRNA is recognized by antiviraldefenses. These defenses likely evolved becausemany viruses have RNA genomes that replicatethrough a dsRNA phase. Yet, there were hints thatRNAi might work to knock down gene expression inmammalian cells. Mammalian genomes containedhomologs of many of the genes required for RNAiin flies, worms, and plants, including Dicer. Also,dsRNA could silence genes in embryonic cell typeslacking the antiviral responses and in cells in whichthe dsRNA antiviral response was inhibited. Butthese seemed exceptional cases, and researchersfeared that RNAi would not show the same poten-tial for knocking out gene expression in mammalsthat had been demonstrated in plants, worms, andflies.

The enzyme protein kinase R (PKR) is a key trig-ger of the dsRNA translational arrest in mammaliancells. When PKR binds dsRNA, its kinase activity isactivated. The activated PKR phosphorylates essen-tial translation initiation factors, thereby blockingtranslation. The smallest RNA that can be bound byPKR is roughly 30 bp in length, so it seemed possi-ble that synthetic 21-bp small RNAs resemblingDicer products might be small enough to avoidbinding by PKR. If this were the case, then it mightbe possible to induce RNAi in mammalian cells. Thereasoning was correct; when 21-nucleotide dsRNAswith 2-bp 3´ overhangs were transfected into mam-malian cells, they decreased the level of the comple-mentary mRNA without triggering a general trans-lation arrest. These small RNAs were named shortinterfering RNAs (siRNAs).

Biologists around the world seized on this newmethod for directly testing the function of mam-malian genes by transfecting the appropriate siRNAsinto cells. Other researchers leapt into action to tryto find ways to introduce siRNAs into cells to pre-vent or treat disease. It is difficult to coax normalmammalian cells into taking up DNA and RNA, butthere has been recent success in preventing herpes-virus infection by treating the skin of mice with anantiherpes siRNA (Fig. 9-8). These applications ofRNAi technology to answer research questions andsolve health problems are exciting, but the basic bio-logical questions remains. Why do organisms carryout RNAi? For what purpose did silencing bydsRNA evolve?

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 230

Add luciferinsubstrate

Optimalstructure

Add small RNAto Drosophilaembryo extract

Add luciferasemRNA, NTPs,and aminoacids

QuantitateluciferaseactivityIncubate

Test direction of overhangAmount of luciferase

overhang, 1 bp

Incubate

overhang, 1 bp

A U C A C G U A C G C G G A A U A C U U C

A G U G C A U G C G C C U U A U G A A G C

Blunt U C A C G U A C G C G G A A U A C U U C G

A G U G C A U G C G C C U U A U G A A G C

C A C G U A C G C G G A A U A C U U C G A

A G U G C A U G C G C C U U A U G A A G C

overhang, 3 bp

overhang, 2 bpA C G U A C G C G G A A U A C U U C G A AA C G U A C G C G G A A U A C U U C G A A

A G U G C A U G C G C C U U A U G A A G CA G U G C A U G C G C C U U A U G A A G C

C G U A C G C G G A A U A C U U C G A A A

A G U G C A U G C G C C U U A U G A A G C

Optimallength

Test length of duplex

C G U A C G C G G A A U A C U U C G

G U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A

G U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A

G U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A AC G U A C G C G G A A U A C U U C G A A A

G U G C A U G C G C C U U A U G A A G C UG U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A A UC G U A C G C G G A A U A C U U C G A A A U

G U G C A U G C G C C U U A U G A A G C UG U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A A U G

G U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A A U G U

G U G C A U G C G C C U U A U G A A G C U

C G U A C G C G G A A U A C U U C G A A A U G U C

G U G C A U G C G C C U U A U G A A G C U

0

–1

0

1

2

3

0.2 0.4 0.6 0.8 1

Amount of luciferase

0

–1

0

1

2

3

4

5

6

18

19

20

21

22

23

24

25

0.2 0.4 0.6 0.8 1

FIGURE 9-7

Synthesized siRNAs mimic Dicer products and trigger RNAi. Small RNA cloning experiments showed thatthe RNA fragments produced by Dicer were 22 nucleotides in length and, like the products of other RNaseIII enzymes, had a 2-nucleotide, 3´ overhang. Short siRNA duplexes corresponding to the gene encoding fire-fly luciferase were synthesized and added to a Drosophila embryo extract. After incubation, mRNA for fire-fly luciferase was added, and the incubation was continued. The extracts contain all of the factors necessaryfor translation of the luciferase mRNA into luciferase. The amount of luciferase made was assayed by addingluciferin, its substrate, and measuring the luminescence produced as luciferase hydrolyzed luciferin. Extractswith active RISC degraded the luciferase mRNA, and so there was less luciferase activity. Different smallRNAs were tested to confirm the importance of both the type of overhang on the duplexes and the length ofduplexes on the efficiency of RNAi.

219-246_Ch09_rDNA 10/18/06 3:36 PM Page 231