The 21-nucleotide small temporal RNA (stRNA) let-7 regulates developmental timing in Caenorhabditis elegans and probably in other bilateral animals. In vivo and in vitro evidence is presented that in Drosophila a developmentally regulated precursor RNA (see microRNA encoding gene let-7), related in sequence to C. elegans let-7, is cleaved by an RNA interference-like mechanism to produce mature let-7 stRNA. Targeted destruction in cultured human cells of the messenger RNA encoding the enzyme Dicer, which acts in the RNA interference pathway, leads to accumulation of the let-7 precursor. Thus, the RNA interference and stRNA pathways intersect. Both pathways require the RNA-processing enzyme Dicer to produce the active small-RNA component that represses gene expression (Hutvagner, 2001).
Two small temporal RNAs (stRNAs), lin-4 and let-7, regulate the timing of development in Caenorhabditis elegans. stRNAs encode no protein, but instead appear to block the productive translation of mRNA by binding sequences in the 3'-untranslated region of their target mRNAs. let-7 is present in most if not all bilaterally symmetric animals, including Drosophila melanogaster and humans. In Drosophila, let-7 first appears at the end of the third larval instar, accumulates to high levels in pupae, and persists in adult flies (Hutvagner, 2001).
The mechanism by which stRNAs are synthesized is unknown. The ~21-nucleotide (nt) let-7 RNA has been proposed to be cleaved from a larger precursor transcript. The generation of small RNAs from a longer, structured precursor -- double-stranded RNA (dsRNA) -- is an essential feature of the RNA interference (RNAi) pathway, raising the possibility that stRNAs are generated by mechanisms similar to the initial steps in RNAi and suggesting that enzymes such as the Drosophila protein Dicer might play a role in generating stRNAs (Hutvagner, 2001),
Examination of the developmental expression of let-7 in Drosophila revealed a candidate for a let-7 precursor RNA, let-7L (Pasquinelli, 2000). let-7L was detected at the end of the third larval instar and at the beginning of pupation, the same developmental stages where let-7 itself is first expressed. Consistent with the transcript being a let-7 precursor, the amount of let-7L RNA declines as let-7 accumulates. let-7L RNA is slightly shorter than a 76-nt RNA standard. Previous analysis of the genomic sequence flanking Drosophila let-7 led to the proposal that a 72-nt RNA hairpin might be a let-7 precursor (Pasquinelli, 2000; Hutvagner, 2001).
A let-7 homolog is also expressed in human tissues (Pasquinelli, 2000) and in cultured human HeLa cells, but not in Drosophila embryos or cultured Drosophila S2 cells. Primer extension analyses confirmed that the mature Drosophila let-7 RNA detected by Northern hybridization was bona fide let-7. Primer extension products corresponding to the 5' ends of mature let-7 RNAs were detected in total RNA from early and unstaged Drosophila pupae and from human HeLa cells. Primer extension analysis of total RNA from unstaged worms, as well as Northern hybridization experiments, indicated that worm let-7 is 1 nt longer than that in flies and humans. In early pupae, primer extension analysis also detected three longer extension products. The major (middle) product and the less abundant (lower) product comigrate with primer extension products templated by a synthetic 72-nt RNA corresponding to putative pre-let-7. This longer transcript from early pupae has the same 5' end as the 72-nt let-7 precursor and is therefore a good candidate for a let-7 precursor RNA (Pasquinelli, 2000; Hutvagner, 2001).
To determine if the let-7L RNA detected in vivo is, in fact, the direct precursor of mature let-7, processing of the proposed pre-let-7 stem-loop RNA into let-7 was tested in Drosophila embryo lysates, which contain no detectable let-7 RNA (Pasquinelli, 2000). These lysates recapitulate RNAi in vitro, prompting the question of whether the proposed precursor RNA is cleaved into mature let-7 by an RNAi-like mechanism. The 72-nt RNA was incubated with Drosophila embryo lysate for various times, then assayed for the production of let-7 by primer extension. As seen in vivo, mature let-7 RNA accumulates in the cell-free reaction. Thus, an RNA corresponding to the proposed let-7 precursor is converted to an RNA with precisely the same 5' ends as authentic let-7 by one or more factors in the Drosophila embryo lysate (Hutvagner, 2001).
Only let-7 RNA, not its complement, has been detected in vivo in worms, flies, and human tissues (Pasquinelli, 2000). Thus, it is expected that bona fide let-7 maturation in vitro would be asymmetric, yielding only let-7 and not small RNAs complementary to let-7, such as antisense let-7. In contrast, processing of long, dsRNA by the RNAi pathway is symmetric, yielding double-stranded 21- to 22-nt RNAs. Therefore, it was asked if processing of the proposed pre-let-7 RNA in vitro is symmetric or asymmetric, yielding let-7 but not its complement. Four pre-let-7 RNAs were prepared by in vitro transcription, each uniformly labeled with a different alpha-32P-nucleotide [adenosine 5'-triphosphate (ATP), cytidine 5'-triphosphate, guanosine 5'-triphosphate, or uridine 5'-triphosphate] and incubated separately in an in vitro reaction. Since let-7 contains no cytosine, accurate in vitro processing of pre-let-7 should produce a 21- to 22-nt product for RNAs labeled at A, G, or U but not at C. A product of the appropriate size for let-7 was produced for pre-let-7 transcripts labeled at A, G, and U. No 32P-labeled product accumulated from the 32P-C-labeled pre-let-7 RNA. Although pre-let-7 RNA continued to disappear with incubation in the lysate, mature-let-7 production rapidly reached a plateau. Because single-stranded 21-nt RNAs are generally unstable in the embryo lysate, this likely reflects degradation of let-7 in the lysate, which may lack factors required for let-7 stabilization and function. Nonetheless, it is remarkable that let-7 RNA accumulates at all, because exogenous, single-stranded, 21-nt RNAs are degraded by the lysate within minutes (Hutvagner, 2001).
Next, the products of an in vitro reaction were analyzed by Northern hybridization using three different deoxyoligonucleotide probes. Probe 2 was entirely complementary to mature let-7. Probe 3 was complementary to the first 21 nt of the precursor and therefore only partially complementary to mature let-7. Control experiments showed that probe 3 detected mature let-7 substantially less well than probe 2, whereas probe 3 detected as well or better than probe 2 products derived from the precursor sequence that is 5' to the region encoding let-7. Finally, probe 4 was complementary to the side of the stem of the precursor opposite the portion encoding let-7. Thus, probe 4 should detect the products of symmetric processing of the precursor RNA. Control experiments demonstrated that probe 4 readily detected synthetic antisense let-7 RNA, but not let-7 itself. Northern hybridization experiments were quantified by determining the amount of each probe that hybridized to the region of the blot corresponding to the ~21-nt reaction product and, as a control for hybridization efficiency, the amount of hybridization of each probe to the unreacted precursor remaining at 3 hours, because the full-length precursor is perfectly complementary to all three probes. Probe 2, which is complementary to let-7, readily detected an RNA that accumulated with time. In contrast, probe 3 detected only weakly an RNA that accumulated over the course of the reaction, consistent with it detecting by partial hybridization mature let-7 but not reaction products derived from the region of the precursor 5' to the let-7 sequence. Most important, probe 4, which was designed to detect reaction products like antisense let-7, did not detect products that accumulated upon incubation of pre-let-7 in the lysate. These data strongly imply that symmetric processing products such as antisense let-7 are either not generated at all or are far less stable than let-7 in the in vitro reaction. Thus, the in vitro reaction displays the same specificity and asymmetry that characterize let-7 biogenesis in vivo (Hutvagner, 2001).
It remained possible that the mechanisms of cleavage in vitro and in vivo differ. To assess the type of ribonuclease (RNase) that might be responsible for pre-let-7 processing, both in vitro and in vivo, the 5' and 3' ends of both the let-7 generated by the in vitro processing reaction and the let-7 from pupae were analyzed. Treatment with periodate, followed by ß-elimination (of either RNA from the in vitro processing reaction or total pupal RNA) increased the apparent mobility of let-7 by nearly 2 nt, a change diagnostic of RNAs bearing 2',3'-terminal hydroxyl groups. Treatment with calf intestinal phosphatase (CIP) of in vitro-generated let-7 or pupal RNA decreased the apparent mobility of let-7 by 1 nt, consistent with the removal of a charged phosphate group. Furthermore, treatment of the CIP-treated RNA with polynucleotide kinase and ATP restored its original mobility, demonstrating that let-7 contains a monophosphate. Because let-7 contains 2'- and 3'-terminal hydroxyls, this single phosphate must be at its 5' end. Thus, let-7 produced by in vitro processing and let-7 isolated from pupae have the same terminal structure: a 5' monophosphate and 2'- and 3'-terminal hydroxyls. Notably, such termini are characteristic of the products of cleavage of dsRNA by RNase III (Hutvagner, 2001).
The small interfering RNAs (siRNAs) that mediate RNAi also bear a 5' monophosphate and 2'- and 3'-terminal hydroxyls. In Drosophila, siRNA duplexes are produced by the cleavage of long dsRNA by the enzyme Dicer (Bernstein, 2001). Cleavage by Dicer is thought to be catalyzed by its tandem RNase III domains. Only two types of RNase III enzymes are predicted to occur in Drosophila : Drosha (Filippov, 2000) and Dicer. Dicer is the only RNase III domain protein in the publicly available sequence of the Drosophila genome that contains an ATP-binding motif, the DEAD-box RNA helicase domain (Bernstein, 2001). Cleavage of dsRNA by Dicer is strictly ATP-dependent (Bernstein, 2001). Cleavage of pre-let-7 into mature let-7 in Drosophila embryo lysates also requires ATP. Taken together, the chemical structure of mature let-7 RNA in vitro and in vivo and the ATP dependence of pre-let-7 processing in vitro strongly implicate Dicer in let-7 maturation. However, it is noted that expression of Dicer protein in Drosophila larvae or pupae has not yet been demonstrated, although the RNAi pathway, which requires Dicer, functions in larvae and pupae (Hutvagner, 2001).
A more stringent test of a role for Dicer in pre-let-7 processing would be to assay let-7 production in flies lacking Dicer protein. However, mutant alleles of Dicer have yet to be identified in Drosophila . As an alternative approach, a recently reported sequence-specific method was used in which cultured mammalian cells were transfected with synthetic 21-nt siRNA duplexes to suppress gene expression. Because they are <30 base pairs long, the siRNA duplexes do not trigger the sequence-nonspecific responses that complicate standard dsRNA-induced interference in mammalian cells (Hutvagner, 2001).
This method was used to evaluate the role of the human ortholog of Dicer (Helicase-MOI) in let-7 biogenesis. Human Dicer was identified by its unique domain structure, comprising an NH2-terminal DEXH-box ATP-dependent RNA helicase domain, PAZ domain, tandem RNase III motifs, and COOH-terminal dsRNA-binding domain, and by its sequence homology to Drosophila Dicer. HeLa cells were transfected with a single, synthetic siRNA duplex containing 19 nt of the coding sequence of human Dicer mRNA, beginning at position 183 relative to the start of translation. Three days after transfection, total RNA was prepared from the cells and analyzed by reverse transcriptase-polymerase chain reaction (RT-PCR) for Dicer and actin mRNA levels and by primer extension for the presence of let-7. The level of Dicer mRNA in the Dicer siRNA-treated cells was four- to six-fold lower than in the control samples, whereas actin mRNA levels were unchanged. Separate controls showed that ~70% to 80% of the cells were transfected. Thus, the observed decrease in Dicer mRNA levels demonstrates that the Dicer siRNA induced substantial degradation of Dicer mRNA in the fraction of the cells that were successfully transfected (Hutvagner, 2001).
Transfection of HeLa cells with the siRNA duplex corresponding to human Dicer, but not the control siRNA duplex, led to the accumulation of a longer let-7-containing RNA, let-7L. Primer extension analysis of RNA from cells transfected with the Dicer siRNA detected an RNA with a 5' end ~7 nt and ~11 to 12 nt upstream of the mature let-7 product. These products are consistent with the accumulation of the predicted human let-7 precursor RNA (Pasquinelli, 2000) and with a longer form of this precursor containing an extended stem. The mature human let-7 RNA was readily detected in control cells, but not in the cells transfected with the Dicer siRNA duplex, providing additional evidence for a role for Dicer in let-7 maturation. These findings, together with in vitro data, provide strong evidence that Dicer protein function is required for the maturation of let-7. Thus, the RNAi and stRNA pathways intersect; both require the RNA-processing enzyme Dicer to produce the active small-RNA component that represses gene expression. The two pathways must also diverge after the action of Dicer, because siRNA duplexes are generated from long, dsRNA direct mRNA cleavage, whereas the single-stranded stRNA let-7 represses mRNA translation (Hutvagner, 2001).
Recently, Grishok (2001) has shown that the Dicer homolog Dcr-1 is required for both lin-4 and let-7 function in C. elegans. Thus, Dicer is likely to have a broad role in the biogenesis of stRNAs and perhaps other small regulatory RNAs. Furthermore, mutations in the Arabidopsis homolog of Dicer, SIN-1/CARPEL FACTORY (SIN1/CAF), have dramatic developmental consequences (A. Ray, 1996; S. Ray, 1996; Jacobsen, 1999). Perhaps SIN1/CAF protein in plants, like Dicer in bilateral animals, processes structured RNA precursors into small RNAs that regulate development (Hutvagner, 2001).
Pre-let-7 is processed asymmetrically to yield only let-7. It is not yet known what structural or sequence features of pre-let-7 determine its asymmetric cleavage. RNase III enzymes cleave perfectly paired dsRNA on both strands, producing a pair of cuts, one on each strand, displaced by two nucleotides. For the R1.1 RNA hairpin of T7 bacteriophage, internal loops and bulges constrain the Escherichia coli RNase III dimer to cut only one strand of the stem. The proposed let-7 precursor contains such an internal loop at the site of 5' cleavage. It is possible that if the stem were uninterrupted by such distortions, a pair of 21- to 22-nt RNAs might be generated, rather than the single stRNA let-7. If so, it might be possible to design stem-loop RNA precursors that produce an siRNA duplex. The hope is that such an siRNA duplex, generated in vivo in a specific cell type or at a specific developmental stage, would be able to target an mRNA for destruction by the RNAi machinery, thereby extending the utility of RNAi to the study of mammalian development (Hutvagner, 2001).
In a diverse group of organisms that includes Caenorhabditis elegans, Drosophila, planaria, hydra, trypanosomes, fungi and plants, the introduction of double-stranded RNAs inhibits gene expression in a sequence-specific manner. These responses, called RNA interference or post-transcriptional gene silencing, may provide anti-viral defence, modulate transposition or regulate gene expression. A biochemical approach has been taken towards elucidating the mechanisms underlying this genetic phenomenon. 'Loss-of-function' phenotypes can be created in cultured Drosophila cells by transfection with specific double-stranded RNAs. This coincides with a marked reduction in the level of cognate cellular messenger RNAs. Extracts of transfected cells contain a nuclease activity (termed RISC for RNA-induced silencing complex) that specifically degrades exogenous transcripts homologous to transfected double-stranded RNA. This enzyme contains an essential RNA component. After partial purification, the sequence-specific nuclease co-fractionates with a discrete, approximately 25-nucleotide RNA species, which may confer specificity to the enzyme through homology to the substrate mRNAs (Hammond, 2000).
Although double-stranded RNAs (dsRNAs) can provoke gene silencing in numerous biological contexts including Drosophila, the mechanisms underlying this phenomenon have remained mostly unknown. It was therefore important to establish a biochemically tractable model in which such mechanisms could be investigated. Transient transfection of cultured, Drosophila S2 cells with a lacZ expression vector results in ß-galactosidase activity that is easily detectable by an in situ assay. This activity is greatly reduced by co-transfection with a dsRNA corresponding to the first 300 nucleotides of the lacZ sequence, whereas co-transfection with a control dsRNA (CD8) or with single-stranded RNAs of either sense or antisense orientation has little or no effect. This indicates that dsRNAs could interfere, in a sequence-specific fashion, with gene expression in cultured cells (Hammond, 2000).
To determine whether RNA interference (RNAi) could be used to target endogenous genes, S2 cells were transfected with a dsRNA corresponding to the first 540 nucleotides of Drosophila cyclin E, a gene that is essential for progression into S phase of the cell cycle. During log-phase growth, untreated S2 cells reside primarily in G2/ M. Transfection with lacZ dsRNA has no effect on cell-cycle distribution, but transfection with the cyclin E dsRNA causes a G1-phase cell-cycle arrest. The ability of cyclin E dsRNA to provoke this response is length-dependent. Double-stranded RNAs of 540 and 400 nucleotides are quite effective, whereas dsRNAs of 200 and 300 nucleotides are less potent. Double-stranded cyclin E RNAs of 50 or 100 nucleotides are inert in this assay, and transfection with a single-stranded, antisense cyclin E RNA has virtually no effect (Hammond, 2000).
One hallmark of RNAi is a reduction in the level of mRNAs that are homologous to the dsRNA. Cells transfected with the cyclin E dsRNA (bulk population) show diminished endogenous cyclin E mRNA as compared with control cells. Similarly, transfection of cells with dsRNAs homologous to fizzy, a component of the anaphase-promoting complex (APC) or cyclin A, a cyclin that acts in S, G2 and M, also causes reduction of their cognate mRNAs. The modest reduction in fizzy mRNA levels in cells transfected with cyclin A dsRNA probably results from arrest at a point in the division cycle at which fizzy transcription is low (Hammond, 2000).
These results indicate that RNAi may be a generally applicable method for probing gene function in cultured Drosophila cells. The decrease in mRNA levels observed upon transfection of specific dsRNAs into Drosophila cells could be explained by effects at transcriptional or post-transcriptional levels. Data from other systems have indicated that some elements of the dsRNA response may affect mRNAdirectly. Attempts were therefore made to develop a cell-free assay that reflected, at least in part, RNAi (Hammond, 2000).
S2 cells were transfected with dsRNAs corresponding to either cyclin E or lacZ. Cellular extracts were incubated with synthetic mRNAs of lacZ or cyclin E. Extracts prepared from cells transfected with the 540-nucleotide cyclin E dsRNA efficiently degraded the cyclin E transcript; however, the lacZ transcript was stable in these lysates. Conversely, lysates from cells transfected with the lacZ dsRNA degraded the lacZ transcript but left the cyclin E mRNA intact. These results indicate that RNAi ablates target mRNAs through the generation of a sequence-specific nuclease activity (Hammond, 2000).
This enzyme has been termed RISC (RNA-induced silencing complex). Although possible intermediates in the degradation process were occasionally observed, the absence of stable cleavage end-products indicates an exonuclease (perhaps coupled to an endo-nuclease). However, it is possible that the RNAi nuclease makes an initial endonucleolytic cut and that non-specific exonucleases in the extract complete the degradation process. In addition, the ability to create an extract that targets lacZ in vitro indicates that the presence of an endogenous gene is not required for the RNAi response (Hammond, 2000).
To examine the substrate requirements for the dsRNA-induced, sequence-specific nuclease activity, a variety of cyclin-E- derived transcripts was incubated with an extract derived from cells that had been transfected with the 540-nucleotide cyclin E dsRNA. Just as a length requirement was observed for the transfected dsRNA, the RNAi nuclease activity showed a dependence on the size of the RNA substrate. Both a 600-nucleotide transcript that extends slightly beyond the targeted region and an ~1-kilobase (kb) transcript that contains the entire coding sequence are completely destroyed by the extract. Surprisingly, shorter substrates are not degraded as efficiently. Reduced activity is observed against either a 300- or a 220-nucleotide transcript, and a 100-nucleotide transcript is resistant to nuclease in this assay. This was not due solely to position effects because ~100- nucleotide transcripts derived from other portions of the transfected dsRNA behave similarly. As expected, the nuclease activity (or activities) present in the extract can also recognize the antisense strand of the cyclin E mRNA. Again, substrates that contained a substantial portion of the targeted region are degraded efficiently whereas those that contained a shorter stretch of homologous sequence (~130 nucleotides) were recognized inefficiently. For both the sense and antisense strands, transcripts that had no homology with the transfected dsRNA were not degraded (Hammond, 2000).
Although the possibility that nuclease specificity could have migrated beyond the targeted region cannot be excluded, the resistance of transcripts that do not contain homology to the dsRNA is consistent with data from C. elegans. Double-stranded RNAs homologous to an upstream cistron have little or no effect on a linked downstream cistron, despite the fact that unprocessed, polycistronic mRNAs can be readily detected. Furthermore, the nuclease is inactive against a dsRNA identical to that used to provoke the RNAi response in vivo. In the in vitro system, neither a 5' cap nor a poly(A) tail is required, since such transcripts are degraded as efficiently as uncapped and non-polyadenylated RNAs (Hammond, 2000).
Gene silencing provoked by dsRNA is sequence specific. A plausible mechanism for determining specificity would be incorporation of nucleic-acid guide sequences into the complexes that accomplish silencing. In accord with this idea, pre-treatment of extracts with nuclease (micrococcal nuclease) abolishes the ability of these extracts to degrade cognate mRNAs. Activity can not be rescued by addition of non-specific RNAs such as yeast transfer RNA. Although micrococcal nuclease can degrade both DNA and RNA, treatment of the extract with DNAse I has no effect. Sequence-specific nuclease activity, however, does require protein. Together, these results support the possibility that the RNAi nuclease is a ribonucleoprotein, requiring both RNA and protein components. Biochemical fractionation is consistent with these components being associated in extract rather than being assembled on the target mRNA after its addition (Hammond, 2000).
In plants, the phenomenon of co-suppression has been associated with the existence of small (~25-nucleotide) RNAs that correspond to the gene that is being silenced. To address the possibility that a similar RNA might exist in Drosophila and guide the sequence-specific nuclease in the choice of substrate, the Drosophila activity was followed through several fractionation steps. Crude extracts contain both sequence-specific nuclease activity and abundant, heterogeneous RNAs homologous to the transfected dsRNA. The RNAi nuclease fractionates with ribosomes in a high-speed centrifugation step. Activity can be extracted by treatment with high salt, and ribosomes can be removed by an additional centrifugation step. Chromatography of soluble nuclease over an anion-exchange column results in a discrete peak of activity. This peak retains specificity since it is inactive against a heterologous mRNA. Active fractions also contain an RNA species of 25 nucleotides that is homologous to the cyclin E target. The band observed on Northern blots may represent a family of discrete RNAs because it could be detected with probes specific for both the sense and antisense cyclin E sequences and with probes derived from distinct segments of the dsRNA. At present, it cannot be determine whether the 25-nucleotide RNA is present in the nuclease complex in a double-stranded or single-stranded form (Hammond, 2000).
RNA interference allows an adaptive defence against both exogenous and endogenous dsRNAs, providing something akin to a dsRNA immune response. These and other data are consistent with a model in which dsRNAs present in a cell are converted, either through processing or replication, into small specificity determinants of discrete size in a manner analogous to antigen processing. The results suggest that the post-transcriptional component of dsRNA-dependent gene silencing is accomplished by a sequence-specific nuclease that incorporates these small RNAs as guides that target specific messages based upon sequence recognition. The identical size of putative specificity determinants in plants and animals predicts a conservation of both the mechanisms and the components of dsRNA-induced, post-transcriptional gene silencing in diverse organisms. In plants, dsRNAs provoke not only post-transcriptional gene silencing but also chromatin remodelling and transcriptional repression. It is now critical to determine whether conservation of gene-silencing mechanisms also exists at the transcriptional level and whether chromatin remodelling can be directed in a sequence-specific fashion by these same dsRNA-derived guide sequences (Hammond, 2000).
Double-stranded (ds) RNA causes the specific degradation of homologous RNAs in a process called 'RNA interference (RNAi)'; this process is called 'posttranscriptional gene silencing (PTGS)' in plants. The duplex RNA becomes processed by Dicer or another RNase III-like enzyme to short dsRNA fragments of about 21-23 nucleotides (nt), which are incorporated in the RNA-induced silencing complex (RISC) that directs target-specific RNA degradation. Different synthetic dsRNA cassettes, consisting of two 5'-phosphorylated RNA strands of 22 nt each, can initiate RNAi in Drosophila embryos. The cassettes were active at similar quantities required to initiate RNAi by conventional dsRNA. Their sequence specificity was confirmed using synthetic dsRNA cassettes for two different genes, Notch and hedgehog; each time, only the relevant embryonic phenotype was observed. Introduction of point mutations had only a moderate effect on the silencing potential, indicating that the silencing machinery does not require perfect sequence identity. 5'-phosphorylated synthetic RNA was more active than its hydroxylated form. Substitution of either RNA strand by DNA strongly reduces activity. Synthetic cassettes of siRNA will provide a new tool to induce mutant phenotypes of genes with unknown function (Boutla, 2001).
The Notch gene was chosen for the first target, because it is ubiquitously expressed in the early embryo and loss of function produces a characteristic 'neurogenic' phenotype. The expressivity of the neurogenic phenotype can be used as a rough quantitative estimate of the severity of Notch function disruption. An in vitro-synthesized 985-bp dsRNA fragment of the Notch mRNA was injected into Drosophila precellular embryos at 5 µM, which is a typical concentration for RNAi, of which about 100 pl, equivalent to about 0.5 fmole, were actually transferred. The dsRNA induced a strong Notch phenotype with high penetrance, indicative of an almost complete inactivation of the Notch mRNA, both zygotic and maternal (Boutla, 2001).
Next, two Notch-specific RNAs of 22 nucleotides were synthesized. Selection of the sequence within the 985-bp cDNA fragment was based solely on structural considerations, to avoid self-dimerization or undesired intramolecular basepairing of each RNA molecule. In contrast with previous methods, a simple all-RNA cassette was used, without deoxynucleotides in the 3' protruding ends and without special considerations of which nucleotide would form the 3' end. However, a 5' phosphate was additionally introduced as an authentic RNase III product and it was compared with the nonphosphorylated RNA for its silencing potential. Both cassettes were adjusted to 100 µM and were used for injection. It should be noted that this is a 20-fold higher molar concentration than that of the dsRNA; however, in terms of absolute amount of RNA, it is less than half. The phosphorylated cassette was able to induce a strong Notch phenotype, exactly as observed after the injection of the long dsRNA. The nonphosphorylated cassette gave phenotypes with decreased penetrance, but the expressivity remained strong. At present, it cannot be distinguish whether the reduced efficiency is a general property or whether the phosphorylated 5' terminus simply provided protection against exonucleases. It is noteworthy that the phosphorylated cassette had a higher penetrance at a 10-fold dilution compared to the undiluted nonphosphorylated dsRNA cassette, but under these conditions, its expressivity was slightly lower (Boutla, 2001).
To test for the specificity and the general applicability of inducing RNAi, a second synthetic phosphorylated RNA cassette of the same general design was used, this time directed against the hedgehog (hh) gene. At a concentration of 100 µM, the hh dsRNA cassette induced a strong mutant phenotype in 88% of the 268 injected embryos. As for Notch, the strength of the observed phenotype suggested complete silencing of the hh gene (Boutla, 2001).
According to current models, the antisense RNA confers sequence specificity upon the RNAi-mediated RNA degradation process. In view of this, a test was made as to the extent one of the RNA strands of the Notch siRNA cassette could be substituted by DNA. The combination DNA sense/RNA antisense was the most promising, since it left the antisense RNA intact. However, substitution of either sense or antisense strand by DNA resulted in a dramatic drop in both penetrance and expressivity of the Notch phenotypes to levels comparable to those obtained with ordinary antisense RNA. A phosphorylated dsDNA cassette had an even weaker effect, inducing only a very mild Notch phenotype that had not been observed with any of the other samples tested (Boutla, 2001).
Subsequently, tests were performed to see whether the 3' protruding ends, as generated by RNase III, are required. It was reported that blunt-ended RNAs were less active in insect tissue culture, and it has been reported that a synthetic blunt-ended 26mer dsRNA is about 250-times less effective than an 81mer dsRNA, although a 26mer might be too large to act as a siRNA. In this study, after injection with a blunt-ended RNA cassette an increased number of viable embryos were obtained and, in accordance with this, a reduced expressivity. At 10-fold dilution, it became evident that this construct was less active than the proper siRNA. Thus, the protruding 3' ends are not mandatory to elicit RNAi; although, in this case, the difference is not as pronounced as reported earlier. A potential difference to the blunt-ended cassette used previously is the presence of the 5' phosphate in the construct used in this study (Boutla, 2001).
Tests were performed of several RNA cassettes that carried mutations. In the first example, a single nucleotide exchange was introduced that would be likely to interfere as much as possible with substrate binding. Therefore, the mutation was positioned centrally and was simultaneously introduced into the sense and the antisense strand, so that the RNA cassette remained double stranded. In previous reports, nonmatching nucleotides greatly impaired the silencing potential when introduced to the antisense strand of longer dsRNAs. Surprisingly, this synthetic cassette was also able to induce a strong Notch phenotype with high penetrance, indicating that a perfect match to the target RNA is not necessary to initiate the RNAi response. The 10-fold-diluted sample was still active, but penetrance and, in particular, expressivity were reduced. Next, each of these mutated sense and antisense RNAs was tested in combination with the wild-type sequence. Either of the combinations, characterized by a G:U pair or a mismatch, was highly active. As a third example, an RNA cassette with a double mutation in the antisense strand was paired with the nonmutated 22mer. Even this RNA construct, with its central bulge loop, had some silencing potential. However, both penetrance and expressivity dropped significantly compared with the single mutant. It will require a more detailed analysis to determine at what position and to what degree sequence deviations can be tolerated without loss of silencing function (Boutla, 2001).
Duplexes of 21-23 nucleotide (nt) RNAs are the sequence-specific mediators of RNA interference (RNAi) and post-transcriptional gene silencing (PTGS). Synthetic, short interfering RNAs (siRNAs) were examined in Drosophila embryo lysate for their requirements regarding length, structure, chemical composition and sequence in order to mediate efficient RNAi. Duplexes of 21 nt siRNAs with 2 nt 3' overhangs were the most efficient triggers of sequence-specific mRNA degradation. Substitution of one or both siRNA strands by 2'-deoxy or 2'-O-methyl oligonucleotides abolished RNAi, although multiple 2'-deoxynucleotide substitutions at the 3' end of siRNAs were tolerated. The target recognition process is highly sequence specific, but not all positions of a siRNA contribute equally to target recognition; mismatches in the center of the siRNA duplex prevent target RNA cleavage. The position of the cleavage site in the target RNA is defined by the 5' end of the guide siRNA rather than its 3' end. These results provide a rational basis for the design of siRNAs in future gene targeting experiments (Elbashir, 2001b).
The structural determinants of siRNA duplexes required to promote efficient target RNA degradation in D.melanogaster embryo lysate were systematically analyzed, thus providing rules for the design of most potent siRNA duplexes. A perfect siRNA duplex is able to silence gene expression with an efficiency comparable to a 500 bp dsRNA, given that comparable quantities of total RNA are used (Elbashir, 2001b).
Efficiently silencing siRNA duplexes are composed of 21 nt sense and 21 nt antisense siRNAs and must be selected to form a 19 bp double helix with 2 nt 3'-overhanging ends. 2'-deoxy substitutions of the 2 nt 3'-overhanging ribonucleotides do not affect RNAi, but help to reduce the costs of RNA synthesis and may enhance RNase resistance of siRNA duplexes. More extensive 2'-deoxy or 2'-O-methyl modifications reduce the ability of siRNAs to mediate RNAi, probably by interfering with protein association for siRNP assembly (Elbashir, 2001b).
Target recognition is a highly sequence-specific process, mediated by the siRNA complementary to the target. The 3'-most nucleotide of the guide siRNA does not contribute to the specificity of target recognition, while the penultimate nucleotide of the 3' overhang affects target RNA cleavage and a mismatch reduces RNAi 2- to 4-fold. The 5' end of the guide siRNA also appears more permissive for mismatched target RNA recognition when compared with the 3' end. Nucleotides in the center of the siRNA, located opposite the target RNA cleavage site, are important specificity determinants and even single nucleotide changes reduce RNAi to undetectable levels. This suggests that siRNA duplexes may be able to discriminate mutant or polymorphic alleles in gene targeting experiments, which may become an important feature for future therapeutic developments (Elbashir, 2001b).
Sense and antisense siRNAs, when associated with the protein components of the endonuclease complex or its commitment complex, were suggested to play distinct roles; the relative orientation of the siRNA duplex in this complex defines which strand can be used for target recognition (Elbashir, 2001a). Synthetic siRNA duplexes with an equal number of overhanging nucleotides have dyad symmetry with respect to the double-helical structure, but not with respect to sequence. The association of siRNA duplexes with the RNAi proteins in the D. melanogaster lysate leads to the formation of two asymmetric complexes. In such hypothetical complexes, the chiral environment is distinct for sense and antisense siRNA, hence their function. The prediction obviously does not apply to palindromic siRNA sequences or to RNAi proteins that could associate as homodimers. To minimize sequence effects that may affect the ratio of sense- and antisense-targeting siRNPs, using siRNA sequences with identical 3'-overhanging sequences is suggested. Adjusting the sequence of the overhang of the sense siRNA to that of the antisense 3' overhang is recommended because the sense siRNA does not have a target in typical knock-down experiments. Asymmetry in the reconstitution of sense- and antisense-cleaving siRNPs is partially responsible for the variation in RNAi efficiency observed for various 21 nt siRNA duplexes with 2 nt 3' overhangs used in this study. Alternatively, the nucleotide sequence at the target site and/or the accessibility of the target RNA structure may be responsible for the variation in efficiency observed for these siRNA duplexes. It should be noted that all siRNAs used in this study are derived from a short region of one gene. Thus, it is more likely that differences in siRNA efficiency are a consequence of the primary sequences of the siRNAs and the respective target sites, rather than the secondary or tertiary structure of the targeted RNA (Elbashir, 2001b).
In Drosophila, siRNA duplexes are produced in vitro and in vivo from long dsRNAs. About 45% of these short RNAs are precisely 21 nt long, 28% are 22 nt long and a few percent are shorter or longer RNAs (Elbashir, 2001a). This length distribution correlates with the finding that 21 nt siRNA duplexes are the most efficient mediators of mRNA degradation. In addition to the length, the paired structure and overhang are also important. This structural feature may explain why siRNA duplexes isolated from the dsRNA processing reaction under denaturing conditions are less potent for RNAi than longer dsRNAs that are processed to siRNAs during the targeting reaction. Presumably, denaturation followed by renaturation favours the formation of the thermodynamically more stable, blunt-ended, but less active, siRNA duplexes. Isolation of siRNAs under native conditions does not reduce siRNA activity (Elbashir, 2001b).
Production of siRNAs from long dsRNA requires the RNase III enzyme Dicer (Bernstein, 2001). Dicer is a bidentate RNase III, which also contains an ATP-dependent RNA helicase domain and a PAZ domain, presumably important for dsRNA unwinding and mediation of protein-protein interactions, respectively. Dicer is evolutionarily conserved in worms, flies, plants, fungi and mammals, and has a second cellular function important for the development of these organisms. At present, it is uncertain whether Dicer activity in species other than Drosophila produces siRNAs of predominantly 21 nt in length. The estimates of siRNA size vary in the literature between 21 and 25 nt (Elbashir, 2001b).
In posttranscriptional gene silencing (PTGS) and RNA interference (RNAi), 21-25 nucleotide RNA fragments are produced from the initiating dsRNA. Short interfering RNAs (siRNAs) mediate RNAi by an unknown mechanism. GFP and Pp-Luc siRNAs, isolated from a protein complex in Drosophila embryo extract, target mRNA degradation in vitro. Most importantly, these siRNAs, as well as a synthetic 21-nucleotide duplex GFP siRNA, serve as primers to transform the target mRNA into dsRNA. The nascent dsRNA is degraded to eliminate the incorporated target mRNA while generating new siRNAs in a cycle of dsRNA synthesis and degradation. Evidence is presented that mRNA-dependent siRNA incorporation to form dsRNA is carried out by an RNA-dependent RNA polymerase activity (RdRP) (Lipardi, 2001).
This study demonstrates the template-specific incorporation of the 21-25 nucleotide RNAs, or siRNAs, to generate dsRNA that is subsequently cleaved by RNase III activity into new siRNAs. In this way, mRNA is degraded through a cycle of 'degradative-PCR'. Evidence for RdRP activity in Drosophila extracts and it is suggested siRNA incorporation into dsRNA involves RdRP, the crucial step in the amplification of the target RNA for rapid degradation by RNase III-type activity. Although it cannot be excluded that siRNAs may be incorporated into dsRNA by a direct 'guide' mechanism not involving RdRP, such a process would not give the sufficient amplification of the double-stranded RNA target. This would be needed to trigger efficient RNAi with substoichiometric levels of the initiating double-stranded trigger RNA. Consistent with the genetic screens in other lower eukaryotes, the results suggest a role for RdRP in Drosophila RNAi as well (Lipardi, 2001).
The requirement for a dsRNA trigger as the effector for silencing can be partially explained by the nature of the dsRNA cleavage step required for siRNA production. Any factor that significantly alters the double-stranded nature of the dsRNA trigger, such as sequence divergence or chemical modification, affects silencing substantially. In the model proposed here, any changes in strand complementarity could presumably reduce the susceptibility of the triggering dsRNA to RNase III-type cleavage. The nature of the sense and antisense strands in the triggering dsRNA would also play a role in the efficacy of silencing since the RdRP amplification step would depend upon the production and functionality of the siRNAs. Both strands of a synthetic 21 nucleotide GFP duplex siRNA function as primers to give the expected RNA products when the appropriate GFP template strand is used (Lipardi, 2001).
The length of the siRNAs may be an important aspect of their function. Previous reports indicated that 29-36 nucleotide dsRNAs transcribed in vitro do not direct RNAi efficiently in Drosophila extracts, and that a 26 nucleotide dsRNA, also transcribed in vitro, when injected into worms, triggers lower than expected levels of RNA interference at 25°C and none at 16°C. However, chemically synthesized 21 and 22 nucleotide siRNAs can mediate targeted RNA cleavage in Drosophila embryo extracts and in Schneider cells. The siRNAs produced in Drosophila embryo extract by micrococcal nuclease and CIP treatment are essentially as efficient on a weight basis in RNAi as the full-length dsRNA from which they were derived, suggesting there is some optimal length for siRNAs in RNAi. The conservation in the size range for the small RNAs associated with silencing in all the species examined proposes that it may be closely correlated with primer function. This could be due to some unique property of primer activity in a protein complex that has yet to be identified (Lipardi, 2001).
siRNAs require a 3' hydroxyl group for function in RNAi. The authentic siRNAs, produced in Drosophila extracts by RNase III-related enzymes such as Dicer (Bernstein, 2001), have been chemically characterized and shown to have a 5' phosphate and a 3' hydroxyl group. The micrococcal nuclease generated siRNAs described in this study are functional in RNAi and can be incorporated into dsRNA after phosphatase treatment to remove the 3' phosphate group produced by the nuclease digestion. The 21 nucleotide synthetic siRNA primer also has a 3' hydroxyl group that would be required for incorporation into dsRNA by RdRP activity. Whether or not the 3' hydroxyl group is also used in a primer ligation step remains to be determined (Lipardi, 2001).
Cleavage of the GFP target RNA occurs after the synthetic GFP siRNA is incorporated into dsRNA. If cleavage occurred in the template RNA immediately upon binding to the synthetic GFP siRNA, no full-length GFP dsRNA would have been observed. Therefore, cleavage occurs in the nascent dsRNA in regions inside and outside the zone represented by the initial siRNA since the primers are extended to make dsRNA. The fact that the synthetic GFP siRNA is extended to the 5' end of the sense strand template would also restrict cleavage, in this instance, to the region upstream of the 3' terminus of the siRNA. Any region of the target RNA converted into duplex by a given siRNA would be subject to digestion by RNase III activity (Lipardi, 2001).
It has been proposed previously that there is no amplification of the trigger dsRNA in RNAi in C. elegans, based upon the effects of asymmetric strand substitutions in the input dsRNA. This study provides evidence that both single- and double-stranded RNAs can serve as templates for siRNA incorporation into dsRNA in Drosophila extract. However, the rapid degradation of dsRNA suggests that amplification of the trigger dsRNA is of limited value. The antisense siRNA strand would be the most important for the synthesis of new dsRNA from the mRNA template (Lipardi, 2001).
Genetic studies have identified several mutants in C. elegans, Neurospora crassa, and Arabidopsis thaliana that resist RNA interference. These include mutants that affect the initiation of silencing activity (rde-1 and rde-4 in C. elegans, qde-2 in Neurospora, and AGO-1 in Arabidopsis, mutants in the effectors of silencing (rde-2 and mut-7 in C. elegans, the latter related to RNase D), mutants in helicase (qde-3 in Neurospora, and SDE-3 in Arabidopsis, and mutants in RNA-dependent RNA polymerase (ego-1 in C. elegans, qde-1 in Neurospora, and SGS-2/SDE-1 in Arabidopsis). Although by sequence comparison an RdRP homolog in Drosophila has not yet been identified, the results presented in this paper suggest the presence of an RdRP gene (Lipardi, 2001 and references therein).
RdRP-dependent as well as -independent mechanisms may be involved in the generation of dsRNA up to the full-length of the target RNA, according to one of the following schemes: (1) a single siRNA primer would be extended from various positions along individual template strands by RdRP to generate dsRNAs; and (2) different siRNAs would associate along a single template RNA and be extended by RdRP to the adjacent siRNA primer. The extension products would be ligated to generate dsRNAs. This model would require RdRP activity as well as an RNA ligase step; and (3) dsRNAs would be formed by a primer 'guide' mechanism where they would align along the template for subsequent ligation. All these mechanisms could generate dsRNA of sufficient length to be cleaved by RNase III-type activity since this requires a minimum of 39 base pairs. The results favor the first and second models since RdRP activity would be required to amplify the target dsRNA sufficiently when substiochiometric amounts of the trigger dsRNA are involved in initiating RNAi. A primer 'guide' mechanism would require the involvement of an RNA ligase in order to generate larger dsRNAs, and genetic screens have not identified related genes as candidates essential for RNAi. As previously noted, RdRP genes have been shown to be involved in posttranscriptional gene silencing in three different lower eukaryotes. In addition, the 'guide' primer ligation model is not supported by observations using synthetic siRNAs. siRNAs generated from dsRNAs greater than 111 nucleotides in length are not well defined and are derived from several overlapping regions of different lengths (18-24 nucleotides) to make the siRNA population heterogeneous in composition. It is unlikely that a 'guide' mechanism could sort out the precise siRNAs for ligation along the target RNA to rapidly generate the full-length dsRNA. The most convincing evidence for the involvement of RdRP activity in Drosophila RNAi comes from results using the synthetic 21 nucleotide GFP duplex siRNA, where full-length GFP dsRNA is produced from a single primer with the same time course of synthesis and degradation as dsRNA produced using the micrococcal-nuclease-generated GFP siRNAs. The extension of both strands of the synthetic siRNA in a template-dependent manner to yield the expected dsRNA products would specifically require an RdRP. The role for helicase activity in RNAi, as shown for qde-3 in Neurospora and SDE-3 in Arabidopsis in the genetic screens, may be to unwind the primers or the dsRNA trigger, but this remains to be demonstrated (Lipardi, 2001).
Double-stranded RNA is processed into siRNA primers that convert the target mRNA into dsRNA for subsequent degradation and the formation of new siRNAs. Since the siRNA primers are double stranded, they should direct the degradation of either sense or antisense cognate target RNAs. This is exactly what is observed when either sense or antisense GFP mRNA is incubated in extract with GFP dsRNA. Therefore, dsRNA representing transcripts derived from opposite strands of a complementary template would be targeted simultaneously to effect silencing of more than a single gene in some instances. The siRNA primer model also suggests a single siRNA should target transcript degradation as long as the primer extended product is of sufficient length to be cleaved by RNase III activity, roughly 39 nucleotides. When the siRNA-21 is used in a silencing assay, GFP mRNA is selectively degraded, as predicted, since the extended primer produces dsRNA 44 bp long. Therefore, a single primer is sufficient to target mRNA silencing. This result also predicts that a primer producing the longest dsRNA product would be the most efficient since the second generation of siRNAs would represent more of the target RNA (Lipardi, 2001).
In Drosophila, Fmr1 binds to and represses the translation of an mRNA encoding of the microtuble-associated protein Futsch. A Fmr1-associated complex has been isolated that includes two ribosomal proteins, L5 and L11, along with 5S RNA. The Fmr1 complex also contains Argonaute2 (AGO2) and a Drosophila homolog of p68 RNA helicase (Dmp68). AGO2 is an essential component for the RNA-induced silencing complex (RISC), a sequence-specific nuclease complex that mediates RNA interference (RNAi) in Drosophila. Dmp68 is also required for efficient RNAi. Fmr1 is associated with Dicer, another essential component of the RNAi pathway, and microRNAs (miRNAs) in vivo, suggesting that Fmr1 is part of the RNAi-related apparatus. These findings suggest a model in which the RNAi and Frm1-mediated translational control pathways intersect in Drosophila. The findings also raise the possibility that defects in an RNAi-related machinery may cause human disease (Ishizuka, 2002).
The connection that has been established between Fmr1, components of RNAi, miRNAs, and the general translation machinery is of considerable significance because they provide intriguing clues and possible connections to the function of Fmr1 and the pathways with which it may intersect. Recent work in numerous organisms has shown that RNAi shares features with a developmental gene regulatory mechanism that involves miRNAs. For example, both the foreign dsRNAs that trigger RNAi and the endogenous miRNA precursors that function in development are processed into small RNAs by Dicer. Members of the Argonaute gene family are also involved in both the siRNA and miRNA pathways. In C. elegans, Dicer, the dsRNA-binding protein RDE-4, and a conserved DExH-box RNA helicase (DRH-1) are in a complex with RDE-1, an AGO2 ortholog. Furthermore, the human AGO2 ortholog, eIF2C2, is in a complex, the miRNP, that contains the DEAD-box RNA helicase Gem3. Therefore, Argonaute proteins appear to be in a complex that contains an RNA helicase(s), Dicer and small guide RNAs, and function in a variety of homology-dependent mechanisms that involve base-pairing between small guide RNAs and target mRNAs. The findings that Fmr1 interacts with AGO2, Dmp68, Dicer, miRNAs, and the general translation machinery, provide a means to link RNAi enzymes to translational control pathways, and are also consistent with the fact that the RISC nuclease fractionates with ribosomes (Ishizuka, 2002).
microRNAs (miRNAs) are a large family of 21- to 22-nucleotide non-coding RNAs that interact with target mRNAs at specific sites to induce cleavage of the message or inhibit translation. miRNAs are excised in a stepwise process from primary miRNA (pri-miRNA) transcripts. The Drosha-Pasha/DGCR8 complex in the nucleus cleaves pri-miRNAs to release hairpin-shaped precursor miRNAs (pre-miRNAs). These pre-miRNAs are then exported to the cytoplasm and further processed by Dicer to mature miRNAs. Drosophila Dicer-1 interacts with Loquacious, a double-stranded RNA-binding domain protein. Depletion of Loquacious results in pre-miRNA accumulation in Drosophila S2 cells, as is the case for depletion of Dicer-1. Immuno-affinity purification experiments revealed that along with Dicer-1, Loquacious resides in a functional pre-miRNA processing complex, and stimulates and directs the specific pre-miRNA processing activity. Efficient miRNA-directed silencing of a reporter transgene, complete repression of white by a dsRNA trigger, and silencing of the endogenous Stellate locus by Suppressor of Stellate, all require Loqs. In loqsf00791 mutant ovaries, germ-line stem cells are not appropriately maintained. Loqs associates with Dcr-1, the Drosophila RNase III enzyme that processes pre-miRNA into mature miRNA. Thus, every known Drosophila RNase-III endonuclease is paired with a dsRBD protein that facilitates its function in small RNA biogenesis. These results support a model in which Loquacious mediates miRNA biogenesis and, thereby, the expression of genes regulated by miRNAs (Forstemann, 2005; Saito, 2005).
It was asked if Loqs forms a complex in vivo with Dicer-1. For these studies, Dicer-1 tagged with the Flag epitope and Loqs tagged with the myc epitope were simultaneously expressed in S2 cells. Then Dicer-1 was immunoprecipitated with anti-Flag antibodies, and Loqs with anti-myc antibody and then the precipitates were analyzed by immunoblotting. In reciprocal assays, Dicer-1 and Loqs were found to co-precipitate. Consistent with these findings that Dicer-1 and Loqs form a complex in vivo, both proteins are localized predominantly in the cytoplasm of S2 cells (Saito, 2005).
It was further investigated whether Loqs can bind to Dicer-1 in vitro. Dicer-1 was produced by an in vitro translation system and used in binding assays with recombinant Loqs fused to glutathione S-transferase (GST). GST-Loqs interacts with Dicer-1 even in the presence of RNase A, whereas GST itself shows no detectable binding. These results demonstrate that the association of Loqs with Dicer-1 occurs both in vivo and in vitro, and that RNA molecules do not appear to mediate the association (Saito, 2005).
To examine the functional connection between the Dicer-1-Loqs complex and pre-miRNA processing, whether depletion of Dicer-1 or Loqs has any effect on the production of mature miRNA from the precursor was investigated. First whether cytoplasmic lysates of S2 cells are capable of processing synthetic Drosophila let-7 precursor RNA into functional mature let-7 was investigated. In this experiment, the synthetic let-7 precursor RNA was converted to mature let-7 in S2 cytoplasmic lysates, as is the case in embryo lysates. In an in vitro RNAi assay, target RNA harboring a sequence perfectly complementary to mature let-7 was cleaved efficiently within the let-7 complementary sequence, thus showing production of functional let-7 in S2 cell lysates. Cytoplasmic lysates from Dicer-1- or Loqs-depleted cells were then subjected to the pre-let-7 processing assay. Both Dicer-1 and Loqs depletion led to reductions of mature let-7 compared with controls, showing that both Dicer-1 and Loqs function in pre-miRNA processing (Saito, 2005).
Next, pre-miR-ban was used as a substrate for pre-miRNA processing assays. It has been shown that S2 cell extracts contain primary-miRNA processing activity that cleaves pri-miRNA into an approximately 60- to 70-bp pre-miRNA precursor. This processing is known to occur in the nucleus; thus pre-miR-ban was prepared by in vitro processing of pri-miR-ban incubated with S2 nuclear extracts. Uniformly labeled pre-miR-ban was then gel-purified and used as a substrate for analysis of pre-miRNA processing. Incubation of the pre-miRNA with S2 cytoplasmic extracts results in the appearance of a mature 21-nucleotide miR-ban. Then the requirement of Dicer-1 and Loqs in pre-miR-ban processing was examined. Incubation of pre-miRNA with Dicer-1- and Loqs-depleted S2 cytoplasmic extracts results in a marked reduction in mature miRNA levels. In contrast, depletion of Dicer-2 or R2D2 shows no measurable reduction of mature miRNA levels. Then the pre-miRNA processing activity of the purified complexes (both Flag-Dicer-1 and Flag-Loqs complexes) was assayed. That the Flag-Loqs complex contains Dicer-1 was confirmed by immunoblotting. Both Dicer-1 and Loqs complexes are capable of generating maturemiR-ban from pre-miR-ban. Several steps in the RNAi and miRNA pathways are known to require a divalent metal ion. In addition, it is well known that RNase III-type enzymes require divalent metals for cleavage. Flag-Dicer-1 complex was employed and the processing was performed in the presence of magnesium ions or EDTA in a buffer. No pre-miRNA processing activity is detected at 10 mM EDTA. These results demonstrate that the Dicer-1-Loqs complex converts pre-miRNAs into mature miRNAs in a divalent metal ion-dependent manner (Saito, 2005).
To further examine the requirement for Loqs in pre-miRNA processing, Flag-Dicer-1 complex was purified under a harsher condition (high salt), where Dicer-1 is stripped of most Loqs protein, and this Dicer-1 complex was used in pre-miRNA processing assays with or without supplement of recombinant GST-Loqs. Without any supplement, the Flag-Dicer-1 complex purified under the harsh condition showed less activity than that under mild condition. Then GST-Loqs was added in the assay mixture. The addition of GST-Loqs to the Dicer-1 complex stimulates the processing of pre-miRNA. GST-Loqs alone does not show any significant pre-miRNA processing activity. These results show that Loqs is required for stimulating the processing of pre-miRNAs. Interestingly, it was found that the Dicer-1 complex purified under the harsh condition displays considerable siRNA-generating activity on the long dsRNA substrate in vitro, although previous genetic studies have shown that Dicer-1 is not required for siRNA production. The addition of GST-Loqs inhibits this effect. Western blot analysis shows that the Dicer-1 complex used in this experiment does not contain appreciable Dicer-2. GST-Loqs alone shows no activity for generating siRNAs from long dsRNAs. These results suggested that Dicer-1, stripped of much of its bound Loqs, processes both dsRNA and pre-miRNA substrates, but re-addition of recombinant Loqs suppresses dsRNA processing activity and enhances pre-miRNA processing activity. These findings thus imply that much of the apparent substrate specificity of Dicer-1 in vivo results from its association with Loqs. Although very unlikely, it is, however, formally possible that the Dicer-1 immunoprecipitates may contain very small amounts of Dicer-2 protein that can catalyze long dsRNA cleavage, and that addition of a large amount of dsRBD-containing Loqs may block the activity of Dicer-2 in this experiment (Saito, 2005).
The presence of endogenous miRNA was examined in RNA preparations from Flag-Dicer-1 and Flag-Loqs complexes obtained from S2 cells using anti-Flag antibodies. The Dicer-1 complex contains both the pre- and mature form of miR-ban, and the complex seems to preferentially bind the precursor form of miR-ban. In contrast, the precursor form of miR-ban is barely detectable in the Loqs complex, though it contains mature miR-ban. However, EDTA treatment, which inhibits pre-miRNA processing activity, results in an accumulation of pre-miR-ban in the Loqs complex. This may suggest that part of Flag-tagged Loqs protein interacts with Dicer-1 or pre-miRNAs or both. Alternatively, Flag-Loqs complexes may rapidly process pre-miRNAs into mature miRNAs and, therefore, may only transiently interact with them. Nonetheless, these results suggest that Dicer-1-Loqs complexes associate with both pre- and mature miRNAs in vivo (Saito, 2005).
Argonaute protein AGO1 is required for stable production of mature miRNAs and associates with Dicer-1. Thus, attempts were made to ascertain if Loqs is also present in an AGO1-associated complex, and if so, if the AGO1 complex is capable of processing pre-miRNA in vitro. Flag-Loqs and AGO1 tagged with TAP were simultaneously expressed in S2 cells, and the AGO1-TAP complex was purified through immunoglobulin G (IgG) bead-binding. The IgG bound was then subjected to Western blot analysis using anti-Dicer-1, anti-AGO1, or anti-Flag (for Loqs detection) antibodies. Not only Dicer-1 but also Loqs was detected in the AGO1 complex. These results indicate that all three proteins are present in the same complex, although they cannot exclude the possibility that there is one complex that contains AGO1 and Dicer-1 but not Loqs, and another complex that contains AGO1 and Loqs but not Dicer-1. The pre-miRNA processing activity of the AGO1 complex was then examined. Pre-miR-ban was utilized as a substrate. The AGO1 complex is able to efficiently process pre-miR-ban into the mature form. In contrast, another Argonaute protein AGO2-associated complex shows no such activity, which is consistent with the finding that the AGO2-associated complex does not contain Dicer-1. Considered together, these results showed that Dicer-1 and Loqs form a functional complex that mediates the genesis of mature miRNAs from pre-miRNAs, and suggested that the resultant mature miRNAs are loaded onto an AGO1-associated complex, which probably is miRNA-associated RISC, through specific interaction of AGO1 with Dicer-1 and Loqs (Saito, 2005).
Reduction of R2D2 protein by RNAi destabilizes Dcr-2; conversely, RNAi of Dcr-2 renders R2D2 unstable. In contrast, RNAi of loqs in S2 cells reduced Dcr-1 protein levels by no more than 15%, suggesting that Loqs functions together with Dcr-1 in pre-miRNA processing, rather than that Loqs is simply needed to stabilize Dcr-1 protein. However, loqsf00791 mutant ovaries, which lack detectable Loqs protein, contain 70% less Dcr-1 than wild-type. A role for Loqs in both Dcr-1 function and in Dcr-1 stability suggests that the two proteins physically interact, like R2D2 and Dcr-2. Therefore, tests were performed to see if Dcr-1 and Loqs are components of a common complex (Forstemann, 2005).
Myc-tagged versions for two protein isoforms of Loqs, Loqs PA and Loqs PB, were expressed in S2 cells, and the tagged proteins were immunoprecipitated with anti-myc monoclonal antibodies. The immunoprecipitated protein was analyzed by Western blotting using a polyclonal anti-Dcr-1 antibody. Dcr-1 protein co-immunoprecipitates with myc-tagged Loqs. When myc-tagged GFP was expressed in place of myc-tagged Loqs, no Dcr-1 protein was recovered in the anti-myc immunoprecipitate. Similarly, an affinity purified, polyclonal antibody directed against the N-terminus of endogenous Loqs protein also co-immunoprecipitated Dcr-1 protein. This interaction is resistant to treatment with RNase A. No co-immunoprecipitation of Dcr-2 with myc-tagged Loqs PB was detected under conditions where Dcr-1 was readily detected, but it cannot be excluded that Dcr-2 is a substoichiometric component of a complex that contains both Dcr-1 and Loqs (Forstemann, 2005).
When immunoprecipitated with anti-Dcr-1 antibody, both myc-tagged Loqs protein isoforms—PA and PB—associate with Dcr-1. Moreover, the antibody against endogenous Loqs protein detected two bands corresponding in size to Loqs PA and Loqs PB in the proteins immunoprecipitated with the anti-Dcr-1 antibody. Loqs PB comprises only approximately 22% of the total Loqs protein in S2 cells, but corresponds to approximately 95% of the Loqs associated with Dcr-1. Loqs PA, which is expressed at comparable levels in S2 cells, accounts for most of the remaining Loqs associated with Dcr-1. In contrast, the putative Loqs PC protein comprises the majority of S2 cell Loqs, but is not recovered in the Dcr-1 immunoprecipitate. Intriguingly, Loqs PA and PB contain a third dsRBD that Loqs PC lacks; perhaps this third dsRBD is required for the association of Loqs with Dcr-1 (Forstemann, 2005).
The immunoprecipitated Dcr-1-Loqs complexes accurately convert pre-miRNA to mature miRNA. Pre-miRNA processing by the immunoprecipitates is efficient and accurate when the anti-Dcr-1 antibody was used, and also when anti-myc antibody and myc-tagged Loqs was used. Thus, Dcr-1 and Loqs co-associate in a complex capable of converting pre-miRNA into mature miRNA. The data also demonstrate that an N-terminal tandem myc tag does not perturb Loqs function in pre-miRNA cleavage (Forstemann, 2005).
Next, the size of the pre-miRNA processing complex was estimated by gel filtration chromatography. Pre-miRNA processing activity chromatographs as a broad approximately 525-kDa peak that overlaps the peaks of both Dcr-1 and Loqs proteins. Dcr-1 protein chromatographs as an approximately 480-kDa complex that overlaps the peak of Loqs PB, which chromatographs as an approximately 630-kDa complex. The Loqs PB isoform accounts for most of the Dcr-1-associated Loqs in S2 cells. The apparent size of the Dcr-1 complex suggests that it is either associated with proteins in addition to Loqs or that the complex has an elongated shape that increases its apparent molecular weight. Pre-miRNA processing activity, Loqs, and Dcr-1 were all well resolved from the approximately 230-kDa peak of Dcr-2, which corresponds to the Dcr-2/R2D2 heterodimer. Although the peaks of Loqs and Dcr-1 do not co-migrate, Dcr-1 is stably associated with Loqs after gel filtration: Dcr-1 and Loqs reciprocally co-immunoprecipitates from the pooled peak Dcr-1 fractions. Loqs was not detected in the Dcr-2 peak by this method. Loqs PC, which does not associate with Dcr-1 in immunoprecipitation, chromatographs as a 58-kDa protein, suggesting that it is a free monomeric protein (Forstemann, 2005).
In Drosophila melanogaster, Dicer-2/R2D2 and Dicer-1 generate small interfering RNA (siRNA) and microRNA (miRNA), respectively. A novel dsRNA-binding protein, R3D1-L, forms a stable complex with Dicer-1 in vitro and in vivo. While depletion of R3D1-L by RNAi causes accumulation of precursor miRNA (pre-miRNA) in S2 cells, recombinant R3D1-L enhances miRNA production by Dicer-1 in vitro. Furthermore, R3D1 deficiency causes miRNA-generating defect and severe sterility in male and female flies. Therefore, R3D1-L functions in concert with Dicer-1 in miRNA biogenesis and is required for reproductive development in Drosophila (Jiang, 2005).
Based on studies of Dicer-2/R2D2 in the siRNA pathway, it was hypothesized that Dicer-1 also functions in concert with an unknown dsRNA-binding protein in the miRNA pathway. Through a bioinformatics approach, an open reading frame (ORF, CG6866) was identified in the fly genome that showed considerable homology with R2D2 and RDE-4, an R2D2 homolog in C. elegans. Furthermore, PSI-Blast ranked this ORF as the best hit among R2D2-like proteins in FlyBase, and vice versa. This protein was named R3D1 because it contained three putative dsRNA-binding domains (R3) and was later found to associate with Dicer-1 (D1). The R3D1 gene encodes two alternatively spliced proteins, R3D1-L (long; 465 amino acids) and R3D1-S (short; 419 amino acids) (Jiang, 2005).
To test physical association of endogenous Dicer-1 and R3D1, coimmunoprecipitation (co-IP) experiments were performed by using anti-Dicer-1 or anti-R3D1 antibodies in the cytosolic (S100) extracts of S2 cells. R3D1-L (~55 kDa) was present in the IPs of anti-Dicer-1 but not anti-Dicer-2 antibodies. The presence of R3D1-S (~50 kDa) was not detected probably because it was absent or masked by the Immunoglobin (IgG) heavy chain. Reciprocal IP using anti-R3D1 antibody brought down Dicer-1 but not Dicer-2, whereas anti-R2D2 antibody only brought down Dicer-2. In addition, both Dicer-1 and R3D1 interacted with AGO1, a critical component of miRISC. These studies indicate that endogenous R3D1-L specifically associates with Dicer-1 and AGO1, which are key components of the initiation and effector complexes of the miRNA pathway (Jiang, 2005).
To study miRNA biogenesis, the miRNA-generating enzyme was purified from S2 cell extracts by biochemical fractionation. A single peak of activity was observed on all columns tested and was followed throughout the purification. Western blots were performed to detect the presence of Dicer-1, R3D1, Dicer-2, and R2D2 among individual fractions following each step of purification. Both Dicer-1 and R3D1-L showed perfect correlation with the miRNA-generating activity after every chromatography step. This was not the case for R3D1-S, nor for Dicer-2/R2D2, which has been shown to generate siRNA in S2 cells. Therefore, these results indicate that Dicer-1/R3D1-L and Dicer-2/R2D2 represent distinct initiation complexes of the miRNA and siRNA pathways in Drosophila cells (Jiang, 2005).
To determine if R3D1-L is required for miRNA biogenesis in vivo, Dicer-1, R3D1, or both were depleted in S2 cells by RNAi followed by Northern blotting to measure the levels of pre-bantam and bantam miRNA. Surprisingly, only R3D1-L (~55 kDa), but not R3D1-S (~50 kDa), protein was efficiently knocked down by treatment of R3D1 dsRNA. The R3D1 (~1.1 kb) dsRNA should efficiently target both R3D1-L and R3D1-S mRNA, which differ by 138 nt. It was likely that R3D1-S comigrated with a cross-reacting protein on the Western blot. Depletion of AGO1, a key component of miRISC, results in a specific reduction of bantam miRNA in S2 cells. In contrast, knocking down Dicer-1 causes accumulation of pre-bantam but no reduction in bantam. Targeting R3D1 produces a similar phenotype as Dicer-1 depletion. When Dicer-1 and R3D1 were both targeted, there was a greater accumulation of pre-bantam and a modest reduction in bantam. Since RNAi is transient and rarely a complete knockout, the lack of significant bantam reduction is probably because the remaining Dicer-1 is sufficient to maintain the level of miRNA production. Consistent with these results, the miRNA-generating activity was reduced by approximately twofold in Dicer-1- or R3D1-L-depleted cells and by ~3.5-fold in cells of double RNAi treatment. Thus, like Dicer-1, R3D1 is required for miRNA maturation in S2 cells (Jiang, 2005).
Genetic and biochemical studies have suggested that Dicer-1 and Dicer-2 may possess different biochemical activities. It is also possible that associated proteins, such as R3D1-L and R2D2, can help define the functional specificity for Dicer-1 and Dicer-2. To distinguish the two possibilities, polyhistidine-tagged Dicer-1 or Dicer-1/R3D1-L and Dicer-2 or Dicer-2/R2D2 recombinant proteins were expressed by using an insect cell expression system. These recombinant proteins were highly purified by Ni-NTA columns followed by SP-Sepharose and Q-Sepharose chromatography (Jiang, 2005).
Despite sharing extensive sequence homology, Dicer-1 and Dicer-2 display different substrate specificities. Dicer-1 demonstrates striking pre-miRNA processing activity, whereas miRNA generation is not detected for Dicer-2 at these concentrations. In contrast, Dicer-2 is much more active than Dicer-1 in processing long dsRNA into siRNA. In addition, Dicer-1 and Dicer-2 had different ATP requirements. Like human Dicer, Dicer-1 or Dicer-1/R3D1-L generates miRNA or siRNA in an ATP-independent manner, whereas Dicer-2 or Dicer-2/R2D2 required ATP hydrolysis for efficient siRNA production. Taken together, these reconstitution studies establish that Drosophila Dicer-1 and Dicer-2 are functionally distinct enzymes with different substrate specificities and ATP requirements (Jiang, 2005).
Recombinant Dicer-1 and R3D1-L formed a stable complex and cofractionated on multiple columns. Purified recombinant Dicer-1/R3D1-L complex is at least fivefold more active than Dicer-1 alone when measured in the pre-miRNA-processing assay. Consistent with this, addition of purified R3D1-L to Dicer-1 greatly enhances its miRNA-generating activity in a dose-dependent manner. Purified R3D1-S has a similar role but to a lesser degree. To compare the substrate affinity of Dicer-1 and Dicer-1/R3D1-L, gel-shift experiments were performed in the absence of Mg2+; this treatment blocks cleavage of pre-miRNA by Dicer-1. Addition of R3D1-L to Dicer-1 greatly enhances its affinity for pre-miRNA in a dose-dependent manner. These studies suggest that R3D1-L can enhance Dicer-1's miRNA-generating activity by increasing its substrate affinity (Jiang, 2005).
To study the physiological function of R3D1 in flies, a piggyBac (PB) fly strain was obtained in which the piggyBac transposon was inserted in the vicinity of the R3D1 gene. By cloning and sequencing the flanking sequences, it was found that the PB-element was inserted within the first exon and 221 nt upstream of the translational start codon of the R3D1 gene. The levels of R3D1-L and R3D1-S mRNA were much reduced in the homozygous flies when compared with wild type or heterozygotes by semiquantitative RT-PCR. However, corresponding reductions in R3D1 proteins could not be verified by Western blots due to masking by cross-reacting proteins. Nevertheless, this suggests that the PB-insertion creates a hypomorphic mutant allele of the R3D1 gene by attenuating its transcription (Jiang, 2005).
To examine if miRNA biogenesis is defective in the r3d1PB mutant flies, the levels of pre-miR277 were compared in wild-type, heterozygous, and homozygous adult flies. As shown by Northern blots, there was significant accumulation of pre-miR277 in both male and female homozygotes. Consistent with this, there was an approximately sixfold reduction in the miRNA-generating activity in the whole fly extracts of r3d1PB/r3d1PB mutant females. The lack of reduction in mature miR277 can be explained by the fact that r3d1PB is a partial loss-of-function allele. In addition, these data suggest that miRNA production may not be the rate-limiting step in the Drosophila miRNA pathway. Importantly, the miRNA-generating defect in the mutant lysates can be rescued by addition of recombinant R3D1-L, but not R3D1-S. Together, these results indicate that r3d1PB mutant flies are defective for miRNA biogenesis (Jiang, 2005). It was suspected that r3d1 mutant flies might display developmental phenotypes because miRNAs play essential roles in animal development. Since r3d1PB mutants survived to adulthood, it was decided to examine their fertility by setting up crosses between r3d1PB homozygous males or females and their wild-type counterparts. Interestingly, while r3d1PB mutant females are completely sterile, the males are ~60%-70% sterile when compared with the control crosses between heterozygotes and wild-type flies. To further analyze this phenotype, the testes and ovaries were dissected and examined from r3d1PB mutant flies. Although mutant testes appeared normal, mutant ovaries contained a few maturing egg chambers and a shriveled germarium with few healthy germline stem cells. This is a classic 'germ cell loss' phenotype because a few egg chambers can develop from primordial germ cells when the adult ovary first forms. However, the mutant ovary did not sustain continuous egg chamber production since germline stem cells could not be properly maintained. These results indicate that R3D1 is required for normal reproductive development in male and female flies and suggest that specific miRNAs may play essential roles in maintaining germline stem cells in the fly ovary. It is suspected that R3D1 deficiency, and hence the miRNA biogenesis defect, is far more severe in mutant testes and ovaries than other parts of the fly body (Jiang, 2005).
miRNA and siRNA can be viewed as two parallel branches of the RNAi pathway. The biochemical studies demonstrate that Dicer-1/R3D1-L and Dicer-2/R2D2 are used as distinct initiation complexes for the miRNA and siRNA pathways, respectively, in Drosophila cells. The same concept can also be applied to species containing a single Dicer, such as C. elegans and humans. In C. elegans, DCR-1/RDE-4 functions as the initiation complex for the siRNA pathway. However, RDE-4 is not required for the miRNA pathway. It is likely that DCR-1 functions in concert with another dsRNA-binding protein in the miRNA pathway (Jiang, 2005).
By reconstitution, this study has established that Drosophila Dicer-1 and Dicer-2 enzymes are functional distinct enzymes with different ATP requirements and substrate specificities. Like Dicer-1, human Dicer generates miRNA or siRNA in an ATP-independent manner. Like Dicer-2, the C. elegans DCR-1 requires ATP hydrolysis for efficient siRNA production. While Dicer-1 is more suited for processing pre-miRNA, Dicer-2 favors long dsRNA as its ideal substrate. Thus, it will be important to identify the sequence and structural features that determine the evolutionary and functional differences between Dicer-1 and Dicer-2 (Jiang, 2005).
The Dicer-2/R2D2 complex not only generates siRNA, but also binds siRNA and facilitates siRNA loading onto the siRISC complex. It is likely that the Dicer-1/R3D1-L complex plays a similar role in facilitating miRNA loading onto the miRISC complex. Consistent with this hypothesis, recombinant Dicer-1/R3D1-L complex efficiently binds to the synthetic miRNA/miR* duplex. Since the majority of the miRNA/miR* duplexes have different stability at the two ends, this thermodynamic asymmetry is believed to cause preferential loading of miRNA onto miRISC and destruction of the miR* strand. It is reasonable to speculate that the Dicer-1/R3D1-L complex also functions as a sensor for the asymmetry of nascent miRNA/miR* duplex and helps to select the miRNA strand as the guide RNA for miRISC. Since R3D1-L interacts with both Dicer-1 and AGO1, it may play a similar role as R2D2 by bridging the initiation and effector steps of the miRNA pathway (Jiang, 2005).
While Dicer-2/R2D2 cleaves long dsRNA into siRNA, Drosha/Pasha (DGCR8 in human) and Dicer-1/R3D1-L catalyze sequential steps of miRNA biogenesis, processing of pri-miRNA into pre-miRNA and of pre-miRNA into miRNA, respectively. Although R2D2 does not regulate siRNA production, it facilitates the role of Dicer-2 in loading siRNA onto siRISC. While Pashafly/DGCR8human is essential for Drosha to process pri-miRNA, R3D1-L greatly enhances miRNA generation by Dicer-1. Taken together, these studies indicate that all known RNase III enzymes (Drosha, Dicer-1, and Dicer-2) are paired with specific dsRNA-binding proteins (Pasha, R3D1-L, and R2D2) in catalyzing small RNA biogenesis and/or function in Drosophila. It remains uncertain if the same pattern will repeat in other species (Jiang, 2005).
Short interfering RNAs (siRNAs) guide mRNA cleavage during RNA interference (RNAi). Only one siRNA strand assembles into the RNA-induced silencing complex (RISC), with preference given to the strand whose 5' terminus has lower base-pairing stability. In Drosophila, Dcr-2/R2D2 processes siRNAs from longer double-stranded RNAs (dsRNAs) and also nucleates RISC assembly, suggesting that nascent siRNAs could remain bound to Dcr-2/R2D2. In vitro, Dcr-2/R2D2 senses base-pairing asymmetry of synthetic siRNAs and dictates strand selection by asymmetric binding to the duplex ends. During dsRNA processing, Dicer (Dcr) liberates siRNAs from dsRNA ends in a manner dictated by asymmetric enzyme-substrate interactions. Because Dcr-2/R2D2 is unlikely to sense base-pairing asymmetry of an siRNA that is embedded within a precursor, it is not clear whether processed siRNAs strictly follow the thermodynamic asymmetry rules or whether processing polarity can affect strand selection. This study used a Drosophila in vitro system in which defined siRNAs with known asymmetry can be generated from longer dsRNA precursors. These dsRNAs permit processing specifically from either the 5' or the 3' end of the thermodynamically favored strand of the incipient siRNA. Combined dsRNA-processing/mRNA-cleavage assays indicate that siRNA strand selection is independent of dsRNA processing polarity during Drosophila RISC assembly in vitro (Preall, 2006).
In Drosophila, Dcr enzymes are required for RISC assembly as well as dsRNA processing, suggesting that the two phases of RNAi might be functionally coupled in a manner that affects siRNA strand selection. However, these experiments indicate that Drosophila RISC assembly and siRNA strand selection are not significantly influenced by the dsRNA processing step and that the thermodynamic asymmetry rules apply equally well with processed and unprocessed siRNAs in this system. This suggests that Drosophila Dcr enzymes do not channel newly generated siRNAs directly into RISC, but rather release the siRNAs into solution (or onto another factor) before they enter the RISC assembly pathway (Preall, 2006).
Several observations have suggested that thermodynamic asymmetry governs strand selection for processed RNAi triggers. MicroRNAs (miRNAs) are diced from stem-loop precursors, and in most instances only one strand of the processed miRNA duplex is stably incorporated into RISC. The mature strand can be present at either the 5' or the 3' end of the stem-loop, but either way, the selected strand is generally compatible with the thermodynamic asymmetry guidelines. In addition, artificial dsRNAs introduced into plant cells give rise to a stable set of siRNAs that adhere to the asymmetry rules. Similar results have been reported with natural dsRNAs in plants. However, interpretation of these results is difficult because the Dcr processing polarities were not defined, and it is also not clear whether small RNA stability is always a suitable surrogate measure of RISC assembly. Furthermore, plant cells (unlike insect and mammalian cells) export siRNAs into the vasculature to enable systemic RNAi, and therefore the plant dsRNA processing machinery may have specifically evolved the propensity to release newly processed siRNAs. Thus the applicability of the plant analyses to insects and mammals has not been clear (Preall, 2006).
While this work was in progress, Rose (2005) characterized modified ~27 nt duplexes that force a defined Dcr processing polarity and give rise to specific, predictable 21 nt siRNAs. Experiments with these Dcr substrate RNAs revealed that hDcr processing polarity can in fact influence siRNA strand selection in transfected human cells, although it does not completely supercede thermodynamic asymmetry. The reasons for the discrepancy between the results and those of Rose are not clear, although one possibility is that different Dcr enzymes may vary in their tendencies to remain associated with newly generated siRNAs. It is noteworthy that Drosophila Dcr-2 (which is primarily devoted to the siRNA pathway) appears to lack the canonical PAZ domain that normally provides Dcr with a binding pocket for 2 nt 3' overhangs. A PAZ domain is present in hDcr, and mutational analyses indicate that the hDcr PAZ domain assists with dsRNA binding and processing when a 2 nt 3' overhang is present. The apparent lack of a PAZ domain in Dcr-2 may compromise its ability to remain bound to newly cleaved siRNA. It is curious that Drosophila Dcrs are required for RISC assembly but do not appear to couple dsRNA processing to siRNA strand selection, whereas mammalian Dcrs are not required for RISC assembly but apparently do couple dsRNA processing to siRNA strand selection (Preall, 2006).
Finally, it remains to be determined whether Dcr enzymes associate with long dsRNA processing substrates and siRNA RISC-assembly substrates in the same way. This issue is undoubtedly important for understanding the functional relationship between Dcr's roles in the initiator and effector phases of RNAi. Crystal structures of E. coli RNase III, an ancestor of eukaryotic Dcrs, are likely to be informative. The structural data, and models derived from them, depict a protein that can engage dsRNAs in a dynamic fashion. A single dsRBD on each subunit of the RNase III homodimer is tethered to the endonuclease domain by a flexible linker that can rotate roughly 90° around the catalytic core. Thus, there are likely to be at least two binding modes for dsRNA in complex with an RNase III enzyme: one in which the dsRBD braces the RNA helix from either side as it is channeled into the catalytic cleft, and another where the dsRBD holds the dsRNA above and orthogonal to the active site. It is possible that Dcr enzymes also exhibit alternate dsRNA binding modes depending on whether they are actively processing dsRNA or channeling siRNA into RISC. Interconversion between these two conformations may require at least transient release of the siRNA product. Additional dynamic dsRNA/protein interactions during dsRNA processing and RISC assembly presumably involve the dsRNA binding proteins Loquacious/R3D1, R2D2, and TRBP, which associate with Dcr-1, Dcr-2, and hDcr, respectively. Further functional analysis of Dcr's PAZ, RNase III, and dsRNA binding domains, aided by recent advances in the structural biology of Dcr, will be necessary to understand Dcr's roles in the transition between the initiation and effector phases of RNAi (Preall, 2006).
The canonical microRNA (miRNA) pathway converts primary hairpin precursor transcripts into 22 nucleotide regulatory RNAs via consecutive cleavages by two RNase III enzymes, Drosha and Dicer. This study characterizes Drosophila small RNAs that derive from short intronic hairpins termed 'mirtrons.' Their nuclear biogenesis appears to bypass Drosha cleavage, which is essential for miRNA biogenesis. Instead, mirtron hairpins are defined by the action of the splicing machinery and lariat-debranching enzyme, which yield pre-miRNA-like hairpins. The mirtron pathway merges with the canonical miRNA pathway during hairpin export by Exportin-5, and both types of hairpins are subsequently processed by Dicer-1/loqs. This generates small RNAs that can repress perfectly matched and seed-matched targets, and evidence is provided that they function, at least in part, via the RNA-induced silencing complex effector Ago1. These findings reveal that mirtrons are an alternate source of miRNA-type regulatory RNAs (Okamura, 2007).
This study has characterized a class of intronic hairpins, termed mirtrons, that generate ~22 nt regulatory RNAs in Drosophila. The biogenesis of mirtrons is distinct from that of canonical miRNAs. Although alternate mechanisms are not excluded, the data points to a mechanism in which mirtron maturation bypasses cleavage by the pre-miRNA-generating enzyme Drosha but is instead initiated by splicing and intron lariat debranching. This differs explicitly from the processing of canonical intronic miRNA genes, whose cleavage by Drosha occurs prior to host intron splicing. However, the mirtron pathway merges with the canonical miRNA pathway to generate active regulatory RNAs, since debranched mirtrons are productive substrates of Exportin-5 and the Dicer-1/loqs system, yielding small RNAs that can repress target transcripts. This study showed specifically that mirtron-derived small RNAs can associate with Ago1 and require Ago1 to regulate seed-matched targets (Okamura, 2007).
The functional similarity between mirtrons and miRNA precursors is bolstered by the observation that miR-10-3p and the small-RNA product of a mirtron hairpin in Vha-SFD are extensively related across their 5' halves, are derived from the same (right-hand) hairpin arm, are the most abundant products of their respective hairpins, and have the same seed (positions 2-8, AAAUUCG). The small-RNA products of mirtrons are catagorized as a novel subclass of miRNAs (Okamura, 2007).
Fourteen mirtron loci were identified from a high-throughput sequencing effort that confidently identified 133 canonical miRNA genes; thus, mirtrons constitute a considerable fraction of total miRNA genes in Drosophila. In contrast, while a majority of canonical miRNA genes are well-conserved among the sequenced Drosophilids, most mirtrons arose recently during evolution. Since newly evolved miRNAs are thought to have fewer targets than highly conserved miRNAs, the regulatory networks involving mirtrons may be proportionally smaller than those mediated by canonical miRNAs. Still, the findings that both 'old' and 'young' mirtrons (1) produce miRNAs that associate with Ago1, (2) can actively repress minimally paired seed targets, and (3) display patterns of divergence on microevolutionary scales that indicate their incorporation into endogenous regulatory networks together suggest that mirtrons exert appreciable effects on biological networks. Indeed, the relative ease with which mirtrons have been born and/or lost raises the intriguing possibility that the changing mirtronic content of Drosophila genomes has contributed to fly speciation (Okamura, 2007).
The existence of mirtrons has implications for the interpretation of miRNA genetics. It is now recognized that the Dicer mutant condition does not solely reflect the loss of miRNAs, since Dicer has additional roles in chromatin dynamics and/or processing of exogenous or other endogenous dsRNA, depending on the organism. Drosha mutant cells do not accurately reflect the loss of miRNAs either; since Drosha processes other ncRNAs, including rRNAs. It has been suggested that DGCR8/Pasha mutant cells more purely reflect a 'miRNA null' state. This may not be the case either, because the mirtron pathway generates a subclass of miRNAs via a nuclear pathway that is largely, if not completely, distinct from the microprocessor. Therefore, caution should be exercised when using processing-enzyme mutants to assess the contribution of small RNAs to a given biological process (Okamura, 2007).
The data demonstrate that the Drosophila mirtron pathway merges the splicing/debranching pathway with the dicing pathway to generate functional miRNAs. Since the key parts of this hybrid small-RNA pathway are deeply conserved mechanisms for RNA processing, it seems plausible that mirtrons may exist outside of Drosophila. Since debranched introns are normally quite labile; however, it is hypothesize that critical to the operation of the mirtron pathway is a dedicated mechanism to hand-off debranched introns to the hairpin export machinery. Having such a mechanism in place may prove key to the existence of mirtrons in other species (Okamura, 2007).
Home page: The Interactive Fly © 1997 Thomas B. Brody, Ph.D.
The Interactive Fly resides on the
Society for Developmental Biology's Web server.