Abstract
In animals, the double-stranded RNA-specific endonuclease Dicer produces two classes of functionally distinct, tiny RNAs: microRNAs (miRNAs) and small interfering RNAs (siRNAs). miRNAs regulate mRNA translation, whereas siRNAs direct RNA destruction via the RNA interference (RNAi) pathway. Here we show that, in human cell extracts, the miRNA let-7 naturally enters the RNAi pathway, which suggests that only the degree of complementarity between a miRNA and its RNA target determines its function. Humanlet-7 is a component of a previously identified, miRNA-containing ribonucleoprotein particle, which we show is an RNAi enzyme complex. Each let-7–containing complex directs multiple rounds of RNA cleavage, which explains the remarkable efficiency of the RNAi pathway in human cells.
Two types of 21- to 23-nucleotide (nt) RNAs are produced by the multidomain ribonuclease (RNase) III enzyme Dicer: small interfering RNAs (siRNAs) from long double-stranded RNA (1, 2) and microRNAs (miRNAs) from ∼70-nt hairpin precursor RNAs whose expression is often developmentally regulated (3–8). siRNAs direct the cleavage of complementary mRNA targets, a process known as RNA interference (RNAi) (9). Target RNA cleavage is catalyzed by the RNA-induced silencing complex (RISC), which acts as an siRNA-directed endonuclease, cleaving the target RNA across from the center of the complementary siRNA strand (10,11). Assembly of the RISC is adenosine triphosphate (ATP) dependent and precedes target recognition (10,12). Unlike siRNAs, miRNAs are single stranded and pair with target mRNAs that contain sequences only partially complementary to the miRNA and repress mRNA translation without altering mRNA stability (13–19). Although at least 135 miRNAs have been identified collectively from Caenorhabditis elegans,Drosophila melanogaster, and humans, none is fully complementary to any mRNA sequence in these organisms, which suggests that miRNAs do not function in the RNAi pathway because RNAi requires extensive complementarity between the siRNA and its mRNA target (20).
siRNAs and miRNAs have been proposed to act in distinct biochemical pathways, in part because distinct PPD (PAZ and Piwi domain) proteins are required in C. elegans for RNAi (21) and miRNA function (22). In this model, the specific PPD protein associated with an siRNA, miRNA, or miRNA precursor determines the pathway in which a small RNA functions. Unique features of miRNAs or their precursors might lead them to associate with miRNA-specific PPD proteins. Thus, the sequence or structure of a miRNA or its precursor would ensure that it functions as a translational repressor and not as a trigger of RNAi.
To determine whether an siRNA duplex in which one strand corresponded exactly to the miRNA let-7(let-7 siRNA) could function in the RNAi pathway (Fig. 1A), we incubated thelet-7 siRNA duplex with Drosophilaembryo lysate in an in vitro RNAi reaction containing a 5′32P-radiolabeled target mRNA. This target RNA contained sequence fully complementary to let-7 as well as the sequence from the 3′ untranslated region of the C.elegans lin-41 mRNA that mediateslet-7–directed translational repression (Fig. 1B) (23). The let-7 siRNA directed cleavage of the target RNA only within the sequence that was fully complementary to let-7 (Fig. 1C). Thus, the intrinsic sequence of let-7 does not preclude its functioning in RNAi. No cleavage was observed at any other site, including within the lin-41 sequences. Thelin-41 sequence differs from thelet-7 complementary sequence at only 4 of the 19 positions that determine siRNA specificity (Fig. 1B). Thelin-41 sequence was not refractory to RNAi, because an appropriately complementary siRNA directs its cleavage (Fig. 1D) (lin-41 siRNA).
Neither the sequence oflet-7 nor that of its precursor precludeslet-7 entering the RNAi pathway in vitro. (A) Sequences of RNAi triggers. let-7is in red. (B) Structure of the target RNA and its pairing with let-7. Red, let-7 orlet-7 complementary sequence; green,C. elegans lin-41 sequence. The 5′32P radiolabel is indicated by an asterisk. UTR, untranslated region. (C) In vitro RNAi reactions withDrosophila embryo lysate using the triggers in (A) and the target in (B). (D) In vitro RNAi reactions as in (C) usinglet-7– andlin-41–specific siRNAs.
Nor does the structure or sequence of the let-7precursor preclude entry of let-7 into the RNAi pathway. We added synthetic D. melanogasterpre–let-7 RNA (Fig. 1A) to an in vitro RNAi reaction mixture containing the let-7complementary target RNA. Again, the target RNA was cleaved within thelet-7 complementary sequences but not at any other site (Fig. 1C). Although pre–let-7 RNA promoted a lower level of RNAi than a let-7 siRNA duplex, it was surprising that RNAi occurred at all, because Dicer cleavage of pre–let-7 generates single-strandedlet-7 in vivo and in vitro (3,7, 22). It is not known how Dicer produces single-stranded miRNAs but double-stranded siRNAs. One model is that Dicer initially generates an siRNA-like, double-stranded intermediate whose non-miRNA strand is then selectively destroyed. In this model, Dicer would produce the “pre–let-7 siRNA” duplex shown in Fig. 1A. In an in vitro RNAi reaction, this siRNA duplex produced about the same amount of target cleavage as pre–let-7 itself (Fig. 1C). Only a small fraction (∼5%) of the input pre–let-7 RNA (100 nM) is converted to mature let-7 in vitro (3). Thus, when produced from pre–let-7, ∼5 nM maturelet-7 entered the RNAi pathway as efficiently as 100 nM pre–let-7 siRNA duplex, which suggests that production by Dicer cleavage enhances entry oflet-7 into the pathway. Dicer action may therefore be coupled to RISC formation, consistent with the interaction of Dicer with the RISC components Ago-2 in Drosophila(24) and Rde-1 in C. elegans(25).
These experiments suggest that the degree of complementarity between a miRNA and its target RNA is the sole determinant of its function, because single-strandedlet-7 can clearly act as an siRNA in vitro. Nonetheless, let-7 might be precluded from entering the RNAi pathway in vivo. Therefore, we tested whether the endogenous let-7 produced by cultured human HeLa cells (3) enters the RNAi pathway.
We tested both HeLa cytoplasmic (S100) and nuclear extracts for their ability to direct cleavage of the let-7complementary RNA target (Fig. 1B) (26). InDrosophila embryo lysate, which contains nolet-7, this target RNA is cleaved at thelet-7 complementary site only upon addition of exogenous let-7 siRNA duplex. In contrast, HeLa cytoplasmic S100—but not nuclear—extract (Fig. 2A) directed target RNA cleavage within the let-7 complementary sequences in the absence of exogenous siRNA. Target cleavage in HeLa S100 occurred between nt 541 and 542, exactly the same cleavage site directed by the exogenouslet-7 siRNA in Drosophila embryo lysate (Fig. 2B). No target cleavage occurred within thelin-41 sequences contained in the target RNA. These sequences do not pair with let-7 at positions 9 and 10 (Fig. 1B); mispairing between an siRNA and its target at these positions blocks RNAi (20) (27). These results suggest that the endogenouslet-7 in the HeLa cytoplasmic extract is associated with RISC, the enzyme complex that mediates endonucleolytic cleavage in the RNAi pathway.
Endogenous humanlet-7 enters the RNAi pathway. (A) In vitro RNA reactions with Drosophila embryo lysate, human HeLa nuclear or S100 extract, using the target RNA described in Fig. 1B, with or without the let-7 siRNA duplex described in Fig. 1A. (B) Denaturing gel analysis of the cleavage products from the RNAi reactions in (A). (C) Immunoprecipitation oflet-7–programmed RISC activity by antibodies (α) to components of the miRNP (eIF2C2, Gem3, Gem4) or to the SMN protein. (D) Relative amount of let-7RNA recovered, as determined by Northern hybridization, in the immunoprecipitates in (C).
To test directly whether let-7 was associated with RISC, we asked if the cleavage activity copurified with the protein eIF2C2, a member of the PPD family of proteins. PPD proteins are required for RNAi and posttranscriptional gene silencing in animals (21, 28, 29), plants (30, 31), and fungi (32). In flies, the PPD protein Ago-2 is a component of the RISC complex (24); in Neurospora (Qde-2) (33) and C. elegans (Rde-1) (25), PPD proteins are associated with siRNAs in multiprotein complexes likely to correspond to RISC. A human RISC has not yet been characterized, but human eIF2C2 is associated with an RNA-protein complex, the miRNA ribonucleoprotein particle (miRNP), that contains miRNAs and is nearly the same size as DrosophilaRISC (34). The let-7 paralog miR-98, which differs from let-7 at two positions, was previously shown to be in the miRNP (34). We immunoprecipitated eIF2C2 from HeLa S100 with monoclonal (8C7) or polyclonal (411-1) antibodies and tested the immunoprecipitates for their ability to cleave the target RNA at thelet-7 complementary sequence (Fig. 2C) (35). Both let-7 (Fig. 2D) and the nuclease activity specific for let-7complementary target sequences (Fig. 2C) copurified with eIF2C2. Therefore, we refer to this activity as alet-7–programmed RISC.
Next, we tested whether other components of the miRNP—Gemin4 and the putative DEAD-box RNA helicase Gemin3—were also components of thelet-7–programmed RISC complex (34). Monoclonal antibodies to Gemin3 and Gemin4, but not the survival of motor neurons (SMN) protein, immunoprecipitatedlet-7–programmed RISC activity (Fig. 2C). SMN is a component of a Gemin3/Gemin4-containing complex that restructures nuclear RNPs; this complex does not contain eIF2C2 (34). Consistent with the idea thatlet-7 was preassembled into the miRNP before the HeLa cells were lysed, target cleavage in S100 was not enhanced by exogenous let-7 siRNA (36).
Human Dicer protein has previously been shown to be localized to the cytoplasm (37). The experiments in Fig. 2 suggest that the rest of the human RNAi pathway is likewise cytoplasmic, because no RISC-associated let-7 was detected in the nuclear extract, nor could the nuclear extract be programmed with alet-7–containing siRNA to direct target cleavage. In contrast, an exogenous siRNA duplex complementary to firefly luciferase sequences successfully programmed the HeLa S100 to cleave the target (38). Although both the endogenous humanlet-7 and the exogenous luciferase siRNA triggered target cleavage, the two triggers differ in at least one respect: endogenous human let-7 is single stranded (7, 11, 19), whereas the siRNA was double stranded. Double-stranded siRNAs must be unwound in order to direct RNAi; this unwinding requires ATP (12). Once unwound, RISC-associated siRNAs can cleave their targets in the absence of high-energy cofactors (12). Becauselet-7 is single stranded, target cleavage by HeLalet-7 should not require ATP. To test this hypothesis, we depleted ATP (12) from HeLa S100 and then added the let-7 complementary target RNA. Thelet-7–programmed RISC cleaved the target in the absence of ATP (Fig. 3A).
(A) In vitro RNAi reactions using the target RNA and HeLa S100. ATP-depleted reaction mixtures (–) contained about 1/100th the ATP of nondepleted reaction mixtures (+). (B) Northern hybridization analysis to measure the concentration of let-7 in HeLa S100. (C) Quantification of (B). Hybridization signals are plotted for the synthetic let-7 standards (filled circles), or for 3.6 μl (open circle) or 7.2 μl (open square) of HeLa S100.
Therefore, as in Drosophila embryo lysates (12), RNAi in human HeLa cytoplasmic extracts does not require ATP for target cleavage. Accordingly, models for the RNAi pathway that invoke synthesis of new RNA as a prerequisite for target RNA destruction (39) do not accurately describe the mechanism of RNAi in human cells. New RNA synthesis is thought to be an important step for RNAi in C. elegans andDictyostelium discoideum, posttranscriptional gene silencing in plants, and quelling in Neurospora crassa (9). In each of these organisms, a member of a family of RNA-dependent RNA polymerases (RdRPs) is required for RNA silencing. In contrast, no such RdRP is encoded by the current release of the Drosophila or the human genome.
Why then is RNAi so efficient in flies and cultured mammalian cells? The concentration of let-7 in HeLa S100 is ≤900 pM (Fig. 3, B and C) (40), and therefore the concentration of let-7–programmed RISC in our reactions is ≤450 pM. Because the target RNA concentration in these experiments was ∼6 nM and 70% of the target was destroyed in 2 hours (Fig. 2A), each let-7–programmed RISC must catalyze the cleavage of ∼10 target molecules. (let-7 produced de novo during the reaction is negligible, because the concentration of pre–let-7 in HeLa S100 is 1/10th to 1/20th that of let-7.) Thus, thelet-7–programmed RISC is a true enzyme, catalyzing multiple rounds of RNA cleavage. It seems highly likely that all RISCs are multiple-turnover enzyme complexes.
Our results suggest that humanlet-7 is in the enzyme complex that mediates RNAi, yet human cells do not contain mRNAs that could function aslet-7 RNAi targets. Perhapslet-7 enters two separate complexes, one for RNAi and one for translational control. Such a model implies that a portion of let-7 enters a complex that serves no function in human cells. More likely is that the recently identified miRNP (34) is the human RISC (10,24) and that this one complex carries out both target cleavage in the RNAi pathway and translational control in the miRNA pathway (Fig. 4). Such a view does not preclude miRNAs or siRNAs from also being associated with smaller complexes that contain only a subset of RISC components but that nonetheless are capable of target RNA cleavage (12).
Model for a common pathway in which miRNAs direct translational repression and siRNAs direct target RNA destruction (RNAi). The RNA duplex in brackets is proposed to be a short-lived intermediate. We propose that Dicer cleavage of both miRNA precursors and double-stranded RNA is coupled to the formation of a miRNP/RISC complex. RISC is envisioned to act stoichiometrically in repressing translation but catalytically in RNA destruction via RNAi.
At least one plant miRNA (miR171) is perfectly complementary to a potential regulatory target mRNA, raising the possibility that miRNAs may naturally be used in plants for RNAi-based regulation of gene expression (8, 41). We anticipate that, in plants, artificial miRNAs will direct translational control when they do not pair with their mRNA targets at the site of cleavage in the RNAi pathway. Conversely, we predict that, in animals, synthetic siRNA duplexes with only partial complementarity to their corresponding mRNA targets will repress translation of the mRNA without triggering RNA degradation.
↵* To whom correspondence should be addressed. E-mail: phillip.zamore{at}umassmed.edu.
