Review

Several years ago it was discovered that plant transformation with a transcribed sense transgene could shut down the expression of a homologous endogenous gene. Moreover, it was shown that the introduction into the cell of dsRNA (double-stranded RNA) containing nucleotide sequence complementary to an mRNA sequence causes selective degradation of the latter and thus silencing of a specific gene. This phenomenon, called RNA interference (RNAi) was demonstrated to be present in almost all eukaryotic organisms. RNAi is also capable of silencing transposons in germ line cells and fighting RNA virus infection. Enzymes involved in this process exhibit high homology across species. Some of these enzymes are involved in other cellular processes, for instance developmental timing, suggesting strong interconnections between RNAi and other metabolic pathways. RNAi is probably an ancient mechanism that evolved to protect eukaryotic cells against invasive forms of nucleic acids.

protein maturation/degradation.The posttranscriptional level has recently attracted much attention because of the discovery of the phenomenon called RNA interference (RNAi) (Fire et al., 1998).Soon after this discovery it became clear that earlier reports about posttranscriptional gene silencing (PTGS) in plants (Napoli et al., 1990), gene quelling in fungi (Cogni et al., 1996) and gene silencing with antisense RNA (Fire et al., 1991) describe specific variants of RNAi observed in Caenorhabditis elegans (Fire et al., 1998).Gene silencing was first described in plants.Already twelve years ago Napoli et al. (1990) discovered that the introduction of transcribed sense transgenes could shut down the expression of homologous endogenous genes.At that time PTGS was named co-suppression since it silenced both the transgene and the endogenous gene.In 1996 Cogni et al. (1996) described trasngene-induced gene silencing (quelling) in Neurospora crassa using the carotenoid biosynthesis gene albino-1 as a visual marker.
This review describes three similar phenomena: RNAi, quelling and PTGS.It also presents recent opinions on RNAi mechanism.

RNAi
Over the last decade exogenous antisense RNA has been widely used to silence eukaryotic endogenous genes.The most widely accepted interpretation of these phe-nomena was that antisense RNA is able to hybridise with the mRNA and to inhibit translation (Fire et al., 1991).However, it was also discovered that injection of sense RNA into the cell leads to a similar effect (Guo & Kemphues, 1995).It was a big surprise when it was demonstrated that injection of double-stranded RNA had the most profound impact on gene silencing.The above described phenomenon, discovered in Caenorhabditis elegans by Craig C. Mello and his co-workers (Fire et al., 1998), was called RNA interference (RNAi).
The most important observations in this work were as follows: 1) RNAi can spread within the individual and be transmitted to offspring 2) only a few molecules of dsRNA are sufficient to trigger RNAi suggesting the presence of catalytic and amplification components in the interference process 3) RNAi occurs at the posttranscriptional level since dsRNA fragments corresponding to promoter and intron sequences do not activate the RNAi pathway 4) RNAi is a highly specific process: the injection of dsRNA segments homologous to particular gene exons eliminates or decreases only the level of the corresponding mRNA Further experiments carried out on C. elegans showed that the triggering of RNAi can be achieved via different ways: microinjection of dsRNA into the cytoplasm of the intestine (Fire et al., 1998), feeding the worms on engineered Escherichia coli producing dsRNA against the target gene (Timmons & Fire, 1998) and even simply soaking the worms in a solution containing dsRNA (Tabara et al., 1999).
Additional experiments conducted to clarify the genetic requirements for inheritance of RNAi in C. elegans revealed that the target locus is not needed for transmission of the interference effect to the offspring, suggesting a role in the transmission of an extragenic sequence-specific silencing factor (Grishok et al., 2000).
Further experiments using classical genetic tools and the establishment of in vitro systems in Drosophila and human cells have greatly broadened our knowledge about RNAi mechanism and enzymes involved in this process.

HOW DOES IT WORK?
RNAi can be divided into four stages: (i) double-stranded RNA cleavage, (ii) silencing complex formation, (iii) silencing complex activation, and (iv) mRNA degradation.The first one includes ATP-dependent, processive dsRNA cleavage into double-stranded fragments 21 to 25 nucleotides long.They contain 5¢ phosphate and 3¢ hydroxyl termini, and two additional overhanging nucleotides on their 3¢ ends (Hamilton & Baulcombe, 1999;Elbashir et al., 2001) (see Fig. 1).The gener-ated fragments were called small interfering RNAs (siRNAs).It was shown that the already mentioned structural features of siRNAs are crucial for the next stages of the RNAi pathway, e.g.modification of the 5¢ end of the antisense strand inhibits siRNA activity, whereas blunt-ended siRNAs are very inefficient intermediates in further steps (Martinez et al., 2002;Sijen et al., 2001).Genetical experiments revealed that RNase III-like nuclease named Dicer is involved in the production of siRNAs (Ketting et al., 2001).In the second step siRNAs are incorporated into a protein complex called RISC (RNA-induced silencing complex) which is inactive in this form to conduct RNAi (Nykänen et al. 2001).The third step involves unwinding of the siRNA duplex and remodelling of the complex to create an active form of RISC (Nykänen et al. 2001;Hammond et al., 2000).The final step in- dsRNA, double-stranded RNA; siRNA, small interfering RNA; RISC, RNAi-induced silencing complex; RdRp, RNA-dependent RNA polymerase.
cludes the recognition and cleavage of mRNA complementary to the siRNA strand present in RISC (guide strand of siRNA) (Nykänen et al. 2001;Hutvagner & Zamore, 2002).In some organisms (C.elegans, Arabidopsis thaliana) an additional step in the RNAi pathway has been described involving a population of secondary siRNAs derived from the action of a cellular RNA-dependent RNA polymerase (RdRp).They are most likely generated during cyclic amplification in which RdRp is primed on the target mRNA template by existing siRNAs (Sijen et al., 2001).Figure 1 shows the four stages of RNAi.
Screening for the genes encoding enzymes involved in PTGS, quelling and RNAi revealed strong homology between the proteins crucial in all three phenomena.

ENZYMES INVOLVED IN RNAi/PTGS/QUELLING
A large number of genes whose products are somehow implicated in RNA silencing have been identified in C. elegans, D. melanogaster, Homo sapiens, Dictyostelium discoideum, N. crassa, Chlamydomonas reinhardtii and A. thaliana.The identified genes encode dsRNases, RNA-dependent RNA polymerases, RNA-dependent helicases and proteins of unknown function.
The enzyme first discovered in Drosophila and called Dicer has dsRNase activity and is involved in the first step of RNA silencingthe production of siRNAs.It belongs to the RNase III-family.Dicer is ATP-dependent and contains several characteristic domains: an N-terminal helicase domain, a PAZ domain (a domain conservative throughout evolution found in Piwi/Argonaute/Zwille proteins in Drosophila and Arabidopsis and involved in developmental control), dual RNase III domains, and a double stranded RNA-binding domain (Bernstein et al., 2001a).A homologue of this gene was found and characterised in C. elegans-dcr-1 (Ketting et al., 2001).
Evolutionarily conserved homologues of Dicer were also identified and tested in plants, fungi and mammals (Bernstein et al., 2001 b).dcr-1 mutants show a defect in RNAi and, additional, phenotypic abnormalities connected with a role for Dcr-1 in a regulatory pathway in which small temporal RNA (stRNA let-7) and its target (lin-41) are involved (Ketting et al., 2001).
During studies on the biochemistry of RNAi several proteins engaged in RISC formation were characterised.Chromatographic purification of crude extracts from Drosophila embryo lysate and human HeLa cells exhibiting RNAi activity revealed the presence of a specific RNAi nuclease, distinct from Dicer, that degrades target mRNAs.Active fractions also contain 25 nucleotide long RNA species homologous to the target mRNA (Hammond et al., 2000).During further studies on Drosophila embryo lysate the formation of a protein complex on siRNAs was shown.A 360 kDa siRNP complex (RISC) was purified.It did not require ATP for the assemblage, contained ds-siRNAs but was not competent to cleave a target RNA.Another complex of 232 kDa was isolated from Drosophila embryo lysate that contained unwound siRNAs and all of the factors required for efficient, sequence-specific mRNA cleavage.This active complex has been termed RISC* (Nykänen et al., 2001).These results suggest that an RNA helicase can participate in the transformation of RISC into RISC*.Hannon and co-workers have reported the isolation of a 500 kDa complex from Drosophila S2 cells that displays the target mRNA cleavage activity (Hammond et al., 2000).Moreover, one protein was identified as a component of RISC -Argonaute 2 belonging to the Argonaute protein family.The authors give two possible explanations for the discrepancies between the estimated molecular sizes of RISC*: (1) the different chromatographic procedures used enable the isolation of larger precursor complexes and smaller active complexes, and (2) a dissociable cofactor is present in the larger RISC*, such as an ATP-dependent RNA helicase (Nykänen et al., 2001).As a result of the analysis of different protein fractions from HeLa cells extract, active human RISC was isolated.Its molecular size was estimated to be between 90 and 160 kDa.When the same methods of purification and analysis of RISC were used for D. melanogaster S2 cells cytoplasmic extract, the isolated complex was of similar molecular size as the human RISC.Two proteins of approximately 100 kDa each, both members of the Argonaute family, namely eIF2C1 and eIF2C2/GERp95, were identified in human RISC (Martinez et al., 2002).Argonaute proteins form a distinct class of proteins containing PAZ and PIWI domains and, as earlier reports suggested, they take part in posttranscriptional gene silencing.Recently it has been shown that PPW-1 protein, a PAZ/PIWI protein is required for efficient RNAi in C. elegans germ line (Tijsterman et al., 2002).The rat ortholog GERp95 of human eIF2C was identified as a component of the Golgi apparatus or endoplasmic reticulum (Cikaluk et al., 1999).eIF2C was found as a component of SMN complex (Survival of Motor Neurons complex), which is a regulator of ribonucleoprotein assembly (Mourelatos et al., 2002).The exact role of these proteins in RISC is still unclear.In human RISC it is still not possible to point to the candidate for the specific ribonuclease, probably present in limited amounts and thus escaping detection.The identification of other RISC components is still to come.An intensive search for mutants defective in RNAi revealed an additional protein family involved in this process: RNA helicases.The A. thaliana mutant SDE3, defective in the production of an RNA helicase, is unable to carry out RNAi response mediated by a transgene.The sde3 locus encodes a protein that was classified as RNA helicase from the Upf1p-like family.However, two subfamilies seem to emerge from this protein family: one including SDE3 and proteins from mouse, human and Drosophila and the other subfamily that contains Upf1p and its homologue in C. elegans-SMG-2.The latter contains a conserved cysteine-rich region and multiple SQ doublets near the C-terminus that are absent in the SDE3 RNA helicase.From these data it seems unlikely that SDE3 is a functional homologue of Upf1p and SMG-2 (Dalmay et al., 2001).The SMG-2 protein from C. elegans exhibits strong homology to the Upf1p RNA helicase from yeast and is involved in NMD (nonsense mediated decay) and in the persistence of silencing by RNAi (Domeier et al., 2000).Despite the similarities in the sequences of SMG-2 and Upf1p, expression of Upf1p in C. elegans does not rescue smg-2 mutants which can be explained by the fact that there is no RNAi pathway in Saccharomyces cerevisiae and some functions of Upf1 could be lost (Page et al., 1999;Aravind et al., 2000).The exact biochemical role of the SMG-2 helicase in RNAi is not clear.Another protein identified in C. reinhardtii called MUT 6 is a DEAH-box helicase family member and mutations in its gene activate transposons (Wu-Scharf et al., 2000).It is speculated that MUT 6 unwinds dsRNA in some steps of RNAi.Since Dicer contains a helicase domain that is probably involved in early step(s) of dsRNAs formation, it can be assumed that the above described RNA helicases operate at a downstream step(s) of proper RISC formation containing antisense siRNA.
As a result of screening for genes involved in RNAi a family of proteins that exhibit the activity of RNA-dependent RNA polymerase was also identified.The C. elegans nuclear genome contains four members of this gene family: ego-1, rrf-1, rrf-2 and rrf-3.Among these rrf-1 was found as the gene coding for RdRp involved in RNAi: this enzyme is necessary to enrich the initial pool of siRNA directed against the target mRNA with the fraction of secondary siRNAs that cannot derive directly from the input dsRNA.The secondary siRNAs exhibited distinct polarity (the antisense strands were complementary to an upstream part of the initially targeted mRNA fragment) suggesting a cyclic amplification process in which the 3¢ hydroxyl group of the antisense RNA serves as a primer (Sijen et al., 2001) (see Fig. 1).Clear experiments confirm the role of RdRp proteins in RNAi in A. thaliana (SDE1/SGS2) (Dalmay et al., 2000) and N. crassa (qde-1) (Cogni & Macino, 1999).However, no RdRp has been identified by homology in the genomes of either flies or humans.Two contradictory reports on RdRp activity have been published, one showing RdRp activity in Drosophila embryo lysates (Lipardi et al., 2001) and the other showing that, for each dsRNA, the positions at which the target mRNA sequence was cleaved were precisely confined to the region spanned by the sequence of the dsRNA, with no cleavages observed upstream from the sequence present in the dsRNA (Zamore et al., 2000).Also, a recent report concerning the mechanism of RNAi in human cells, using an in vitro system, convincingly contradicts the involvement of RdRp activity in this process: it was shown that modification of the 3¢ ends of antisense siRNA did not interfere with the reconstitution of RNA silencing (thus siRNA priming is not necessary to trigger gene silencing) while in an identical experiment in C. elegans such siRNAs were not functional (Martinez et al., 2002).Obviously more experiments have to be carried out to solve this problem but it should be kept in mind that both experiments contradicting the role of RdRp in RNAi in Drosophila and human were done using in vitro systems.Table 1 summarizes the functions of different proteins involved in RNAi/PTGS and quelling.

QUELLING
Much earlier than RNAi discovery, transgene-induced gene silencing was observed in the filamentous fungus N. crassa (Pandit & Russo, 1992;Romano & Macino, 1992).The phenomenon was called quelling.It affected both transgenes and endogenous genes.As in plants, methylation of DNA was observed during posttranscriptional gene silencing.However, it was not obligatory for this process, since mutants of N. crassa defective in DNA cytosine methylation still promoted quelling (Cogni et al., 1996).Several genes coding for proteins involved in quelling were identified, namely QDE-1, QDE-2 and QDE-3.QDE-1 encodes a protein with homology to tomato RNA-dependent RNA polymerase (Cogni & Macino, 1999a), QDE-3 encodes a DNA helicase whose role in quelling is unclear (Cogni & Macino, 2002).Mutants of N. crassa in this gene are not active in gene silencing.Catalanotto et al. (2002) reported that, like in RNAi, small sense and antisense RNAs are also involved in quelling.While the accumulation of these RNAs depends on the presence of functional qde-1 and qde-3 genes, the qde-2 product is involved in a downstream step of quelling.Accordingly it was found that QDE-2 protein can participate in RISC-like complex formation.The QDE-2 protein (as AGO1 in plants and RDE-1 in C. elegans) contains PAZ and PIWI domains thus belongs to the Argonaute protein family.The finding that RdRps are involved in posttranscriptional gene silencing in N. crassa, C. elegans and plants suggests that these phenomena are mechanistically related.

POSTTRANSCRIPTIONAL GENE SILENCING (PTGS)
PTGS is regarded as a mechanistic variant of RNAi.This term defines both the co-suppression of an endogenous gene by the introduction of a transgene into the nuclear plant genome and the cellular defence mechanism induced by RNA virus infection.
As in the case of quelling in fungi and RNAi in animals, PTGS operates via siRNAs and their accumulation requires either transgene sense or antisense transcription or RNA virus replication (Hamilton & Baulcombe, 1999).PTGS induced locally spreads systemically to other tissues of the plant.Such a process requires the presence of a signal molecule transporting information about PTGS triggering.siRNAs were found in both the site of silencing initiation and in distal tissues.Although the nature of the signalling molecule remains unknown, siRNAs are small enough to move through plasmodesmata and/or via the vascular system.Consequently, at present siRNAs are the best candidates for signal molecules ensuring systemic spreading and specificity of PTGS.Systemic gene silencing in plants was investigated using grafting procedure.It was shown that silencing was transmitted from silenced stocks to non-silenced scions.Transmission was unidirectional, always from stock to scion, transgene specific, locus independent and required a transcriptionally active transgene in the target scion (Palauqui et al., 1997).
Interestingly, it was observed that PTGS in transgenic tobacco plants correlates with site-specific DNA methylation (Ingelbrecht et al., 1994), which points to the link between PTGS and TGS (transcriptional gene silencing).In a brilliant work Jones et al. (1999) Table 1.Proteins implicated in RNAi/PTGS/quelling phenomena demonstrated sequence-specific methylation of a transgene when PTGS was induced using RNA virus-based vectors carrying different parts of the transgene coding sequence or transgene promoter.If vectors containing the transcribed region were used, PTGS affected both the transgene and viral RNA levels.The methylation profile of the transgene was similar irrespective of the part of the transgene coding sequence used in the viral vector.However, when the promoter region was inserted into the viral vector, the transgene RNA level was reduced but the level of viral RNA was unchanged.This supports the idea that, in plants, dsRNA homologous to the promoter does not induce PTGS but rather TGS, which reduces transcription efficiency due to promoter methylation.Taking together the described experiments it can be concluded that gene specific methylation is tightly connected to PTGS and occurs directly via RNA-DNA interaction (RNA-directed DNA methylation -RdDm).In the case of promoter methylation the only interaction leading to it would be between RNA and the promoter DNA as described above.Although PTGS and TGS have been long considered as separate pathways, now it is clear that both could form alternative, nonexclusive ways to regulate gene expression (Vaucheret & Fagard, 2001).Many papers have been devoted to virus induced gene silencing (VIGS) which can be classified as a specific variant of PTGS.It was proposed that the mechanism of virus related PTGS involves two separate stages: initiation and maintenance.The initiation of PTGS is dependent on the virus presence but maintenance is virus independent.The VIGS process, once initiated persists in the absence of the inducing virus (Ruiz et al., 1998).In addition, it was found that some plant RNA viruses (potyviruses and cucumoviruses) encode proteins that suppress PTGS.This discovery confirmed earlier suggestions that plants use PTGS as an important part of virus protection system.One of them, Hc-protease (HcPro), encoded in the potato virus Y (PVY) genome, is responsible for the lack of siRNAs.Moreover, grafting experiments showed that HcPro prevents the plant from responding to the mobile silencing signal but does not inhibits production or sending of the signal.These data contradict the assumption of siRNAs as PTGS signal molecules: the mobile signal is produced at a step preceding accumulation of siRNAs (Mallory et al., 2001).Another suppressor of gene silencing called 2b is encoded by cucumber mosaic virus (CMV) (Brigneti et al., 1998).Protein 2b inhibits PTGS at an earlier step than HcPro: it affects silencing initiation.The third identified silencing suppressor was found in potato virus X (PVX).It is the p25 protein responsible for the movement of PVX within a plant.This suppressor inhibits the generation of the systemic silencing signal and prevents spreading of PTGS in the plant (Vionnet et al., 2000).It is clear that plant RNA viruses developed different strategies to evade plant defence system.If a plant uses PTGS as the defence mechanism against RNA viruses, mutants defective in some proteins involved in gene silencing should be hypersensitive to viral infection.Indeed, this is the case: A. thaliana mutants sgs3 (impaired in PTGS) display enhanced susceptibility to viruses, showing clearly that PTGS represents an antiviral defence mechanism (Mourrain et al., 2000).Another A. thaliana mutant defective in PTGS (sde1/sgs2 mutant impaired in cellular RdRp activity) was unable to evoke transgene mediated gene silencing but VIGS was efficiently induced.This result can be explained by the fact that viruses encode their own RNA-dependent RNA polymerase responsible for the generation of double-stranded RNA and do not require the cellular enzyme (Dalmay et al., 2000).

RNAi/PTGS/QUELLING: SIMILARITIES AND DIFFERENCES
There are several lines of evidence that RNAi/PTGS/quelling are closely related phe-nomena which share several common features.All of them are induced by dsRNA, occur posttranscriptionally in the cytoplasm and spread in the whole organism.In each case siRNAs are generated and then employed in specific mRNA degradation.As a result the expression of a target gene is inhibited.
As mentioned in the previous chapter, plant RNA virus replication can be strongly inhibited by PTGS.On the other hand, RNA virus protects itself by producing PTGC suppressors.Recent data show that an insect RNA virus also induces RNAi in Drosophila cells.Additionally it encodes an RNA silencing inhibitor (Li et al., 2002).Since RNAi has also been observed in the case of mammalian cells it is possible that mammalian viruses can be also affected by RNAi (Alquist, 2002).
Enzymes crucial for posttranscriptional gene silencing with high sequence homology have been identified in the case of RNAi, quelling and PTGS (see the chapter in this review devoted to enzymes), suggesting strong similarities in the biochemical mechanisms of these processes.These enzymes have been proposed to function at the steps in the silencing pathway that are common to all three phenomena.However, the role of RdRp in all these processes seems to be unclear: its role in the signal amplification has been proved in the case of PTGS, VIGS, quelling and in some cases of RNAi -namely in C. elegans.However, RdRp is most likely redundant in the case of RNAi in Drosophila and mammals.In plants PTGS seems to be tightly connected with TGS that occurs via de novo DNA methylation.On the other hand, RNAi does not lead to methylation of silenced genes.For years it was thought that Drosophila and C. elegans nuclear DNA is not methylated.Although this is probably true in the case of C. elegans (Wassenegger, 2002), recent reports present data that nuclear DNA of D. melanogaster is methylated, however to a lesser extent than human and plant DNA (Lyko, 2002).It remains to be established whether DNA methylation in the fruit fly is connected to RNAi or not.PTGS/quelling/RNAi can be induced in plants, fungi and C. elegans by the delivery of long dsRNAs that function as substrates for Dicer.Interestingly such a procedure can not be applied for mammalian cells.The delivery of long dsRNAs into mammalian cells results in a general blockage of protein synthesis because of the interferon response and apoptosis.The only example of successful silencing by delivery of long dsRNAs was observed for early embryos suggesting that the cells from this stage of development are incapable of interferon response (Wianny & Zernicka-Goetz, 2000).To overcome these difficulties, short dsRNAs are used to trigger RNAi in mammalian cells (Martinez et al., 2002).
The discovery of RNAi/PTGS and quelling opens up completely new possibilities of research of the functioning of particular animal, plant and fungi genes without the necessity of altering the genome structure (Fjose et al., 2001;Wojtkowiak et al., 2002).

BIOLOGICAL ROLE OF RNAi/PTGS/QUELLING
The data collected untill now suggest that RNAi/PTGS/quelling derive from an ancestral mechanism that controlled nucleic acids invasions.In different eukaryotic kingdoms RNAi/PTGS/quelling are responsible for silencing of transposons, viruses and/or transgenes.These assumptions were supported by experiments showing that silent transposons are reactivated in PTGS or RNAi impaired Arabidopsis (Hirochika et al., 2000), C. elegans (Ketting et al., 1999) and Chlamydomonas mutants (Wu-Scharf et al., 2000).
Additionally, the same enzymes as in RNAi/PTGS are used in several other cellular processes.In Arabidopsis mutations in the Dicer-like gene have dramatic developmental consequences suggesting an important, yet undiscovered role of small RNAs in developmental processes (Hutvagner et al., 2001).In the case of C. elegans Grishok et al. (2001) have found that Dicer is required for both lin-4 and lin-7 maturation and thus functions in developmental timing.Also in C. elegans seven genes smg 1-7 are responsible for nonsense-mediated mRNA decay (NMD).Three of these genes are required for persistence of RNAi (Mango, 2001).Although the interconnections between gene silencing and other cellular processes seem to be well proved, it still remains to be clarified exactly how all of these metabolic pathways interplay and influence each other.RNAi/PTGS/quelling represent processes whose biochemistry was completely unknown until late 90-ties.In all of these processes small RNAs (siRNAs) are involved.Generally, small untranslated RNAs exhibit different functions in the eukaryotic cell (tRNAs, snRNAs, snoRNAs, stRNAs -small temporal RNAs).It seems now that a large number of additional micro-RNAs (miRNAs) exist in eukaryotes performing a multitude of functions (Grosshans & Slack, 2002).Challenges concerning the still undiscovered functions of small RNAs, beside their functions in gene silencing and developmental timing, are ahead of us and assure fascinating scientific adventure in the future.

Figure 1 .
Figure 1.Schematic representation of four-step gene silencing pathway.