siRNA in human cells selectively localizes to target RNA sites

To investigate these mechanistic possibilities, we used confocal imaging to directly visualize the localization of fluorescently labeled siRNA after transfection in human Huh-7 hepatoma cells harboring the subgenomic hepatitis C virus (HCV) replicon. HCV replication occurs at the virus-induced specialized membrane structures derived from the endoplasmic reticulum, which are dispersed in the cytoplasm (14). In the imaging experiments, we separately examined the intracellular locations of two siRNA duplexes, 7SK siRNA targeting the small nuclear 7SK RNA (6, 15–17) and NS3 siRNA that targets the cytoplasmic HCV replicon RNA (18). Each siRNA duplex was labeled with Alexa Fluor 488 and Cy5 fluorophores at the 3′ ends of sense and antisense strands, respectively, to visualize the localizations of the two strands in the cells. All siRNAs were determined to be active, and labeled and unlabeled siRNAs had identical activities within error (Figs. 6 and 7, which are published as supporting information on the PNAS web site).

Two hours after transfection, both sense and antisense strands of 7SK siRNA clearly accumulate in the nuclei of Huh-7 HCV cells (Fig. 1). Antisense and sense siRNA strands were visualized by using their fluorescent labels, depicted in red (Fig. 1a) and green (Fig. 1b), respectively, and the nuclei of the cells were stained with DAPI (blue, Fig. 1c). In sharp contrast to these observations, Huh-7 cells identically transfected with NS3 siRNA exhibited a completely different localization behavior (Fig. 2). In the latter case, no visible accumulation of NS3 siRNAs was seen in the nucleus. Furthermore, significant strand-separation of NS3 siRNA was observed. The antisense NS3 siRNA strand was observed to create perinuclear ring-like patterns of clustered dots (seen in red in Fig. 2d around nuclei in blue), whereas the sense strand nearly disappeared and was only sparsely visible as distinct dots (Fig. 2b).

Colocalization of the three signal profiles (DAPI, Alexa Fluor 488, and Cy5) along multiple lines through the cells (examples are given in Figs. 1f and 2f) provided additional information about the distributions of these siRNAs. In the case of 7SK siRNA, a minor amount of the RNA could be found as a diffuse background in the cytoplasm but clusters were typically not observed to concentrate in the perinuclear area. Within the nuclear boundaries, the two strands showed different levels of accumulation in the nucleoli (Fig. 1f, arrows), which could signify partial strand separation. In the case of NS3 siRNA, sharp peaks of the red signal at the edges of nuclear boundaries verify that the two strands are separated, and the antisense strand tends to accumulate in the perinuclear area in distinct loci. Also, a very low level of NS3 siRNA sense strand was seen in the nucleus as a diffusely distributed pattern (Fig. 2f, the green line inside the nuclear boundaries marked by a wide blue plateau). In these experiments, no localization of the antisense NS3 siRNA strand to the nucleus was observed within the sensitivity of the detectors, using the highest available laser powers for excitation of the dyes. These contrasting 7SK siRNA and NS3 siRNA behaviors did not vary significantly at various times after transfection. In particular, the nuclear-targeted 7SK siRNA was clearly observed to accumulate in the nucleus as quickly as 40 min after transfection, whereas cytoplasmically targeted NS3 siRNA remains distributed largely in the cytoplasm 4 h after transfection, showing no nuclear accumulation. In addition, we also carried out 7SK siRNA transfection for HeLa cells, and observed similar behavior.

RISC assembly involves incorporation of the antisense siRNA strand into the complex, and therefore requires duplex siRNA strand separation (12, 13). Recent experiments have shown that Ago2 protein likely cleaves the sense strand after siRNA binding, resulting in the sense strand dissociation and only antisense strand incorporation into active RISC (10, 11). Therefore, RISC formation will be correlated with siRNA duplex strand separation in our imaging experiments. Furthermore, because the incorporated siRNA single strand is known to be tightly bound to Ago proteins in RISC (19), the localization of RISC is expected to parallel the observed siRNA antisense strand localization. Based on this premise, the observed patterns of Alexa Fluor 488 and Cy5 fluorescence in the NS3 siRNA images (Fig. 2) clearly demonstrate that strand separation and RISC formation occur in the cytoplasm. In addition, the nearly complete exclusion of sense strand from the cytoplasm and low levels of only this strand in the nucleus are also consistent with the idea of siRNA duplex dissociation and sense strand cleavage during NS3 siRNA RISC formation all occurring in the cytoplasm. Therefore, we would expect that, if active RISC formation for 7SK siRNA were also taking place in the cytoplasm followed by transport of this active RISC into the nucleus, the images in Fig. 1 would show similar strand separation in the cytoplasm and sense strand distribution in the cell. Instead, we observe predominantly nuclear localization of both 7SK siRNA strands, leading us to conclude that this siRNA locates as a duplex to the nucleus in an efficient manner, without prior strand separation. Strand separation and formation of active RISC must then occur within the nucleus, resulting in the observed 7SK down-regulation (see Fig. 6).

The 7SK and NS3 siRNA localization trends also provide another intriguing clue about the RNAi mechanisms. In the images with these siRNAs (Figs. 1 and 2), we observed that the NS3 and 7SK siRNAs localize selectively in the cytoplasm and nucleus, respectively, precisely the sites of the respective silencing target RNA molecules. Hence, these observations provide evidence for the existence of a mechanism by which the RNAi machinery influences the passive or active transport and localization of the siRNA and corresponding RISC formation to subcellular compartments containing target RNA. To understand how siRNA duplexes are distributed in cells when no complementary target mRNA exists in the cell, we carried out similar experiments using CHO cells, which lack the cytoplasmic silencing target for NS3 siRNA, but carry the small nuclear 7SK and U6 target RNAs. After transfection in these cells, the “target-less” NS3 siRNA is distributed throughout both the cytoplasm and the nucleus of these cells, exhibiting only a low tendency for strand separation (Fig. 3d), a behavior very different from that of the same siRNA in Huh-7 cells. This result demonstrates that, when cells do not contain silencing target RNA, selective cytoplasmic or nuclear localization of siRNA and RISC complexes is no longer sustained.

Consistent with its behavior in Huh-7 and HeLa cells, 7SK siRNA once again localizes to the nucleus of CHO cells (Fig. 4). Additionally, to understand whether the observed nuclear localization of nuclear-targeted siRNA is specific to 7SK, analogous experiments were also carried out with an siRNA targeting the small nuclear U6 RNA (6, 17) in HeLa cells. A similar selective nuclear localization was also observed in this case (Fig. 8, which is published as supporting information on the PNAS web site), as well as in Huh-7 and CHO cells (data not shown), showing that this effect is a more general feature of nuclear-targeted siRNAs. Taken together, the imaging results with the three siRNA duplexes in multiple cell lines support a mechanism where the target RNA location determines the localization and function of siRNA and RISC.

Imaging experiments carried out 24 h after transfection provided additional information on later stages of the RNAi process. Huh-7 and CHO cells transfected with NS3 siRNA do not display significant changes in siRNA distributions at 24 h after transfection, still showing either clear strand separation and only clustered antisense strand throughout the cytoplasm (Huh-7 cells), or the accumulation of largely intact duplexes throughout the cytoplasm and the nucleus (CHO cells). On the other hand, Huh-7 and CHO cells transfected with 7SK siRNA, and examined 24 h after transfection, revealed an altered pattern of siRNA distribution. The antisense strand remains somewhat localized to the nucleus, being trapped mostly in the nucleoli (Fig. 5), but also shows significant spreading into the cytoplasm. Most remarkably, the sense strand was only observed at very low levels in the cytoplasm of a few cells, showing its absence not only in the nucleus but also nearly completely from the entire cell (Fig. 5d). Although not a direct observation of RISC movement, these observations may signify that nuclear-formed active RISCs are translocated into the cytoplasm at later stages of the nuclear RNAi pathway, as a consequence of or during the completion of the nuclear RNA decay pathway. Additional studies will be needed to further clarify this aspect of the nuclear RNAi mechanism.

Although the specific molecular basis for a target RNA biased siRNA localization is yet to be understood, our observations are consistent with a mechanism where the siRNA/RISC are trapped at cellular sites or in compartments containing the target RNA, via a specific targeting machinery or repeated cycles of target-binding, cleavage, and release. Along these lines, after incorporation of the antisense strand of NS3 siRNA into active RISC in the cytoplasm, it could be targeted as a RISC constituent to cytoplasmic centers of RNA decay, thus “trapping” the antisense strand of cytoplasmic targeted siRNA in the cytoplasm. Processing (P) or cytoplasmic bodies have recently been described as such centers for final degradation of mRNA by decapping enzymes and exonucleases (20, 21), and the studies have demonstrated the colocalization of RISC associated Ago1 and Ago2 proteins with these cytoplasmic centers, proposing a link between P-bodies and RNAi pathway in the cytoplasm (7–9, 22). On the other hand, our results show that the absence of a cytoplasmic target RNA results in a lack of 7SK and U6 siRNA target-associated trapping in the cytoplasm. Instead, these siRNAs effectively translocate to the nucleus, where a core cleavage protein (Ago2) is also located, as shown (6–9), and confirmed in our imaging experiments of Ago2 intracellular distribution by expressing Myc-tagged Ago2 in HeLa cells and visualizing it by using anti-Myc Alexa Fluor 555-conjugated antibody (data not shown). Nuclear active siRNAs likely form active RISCs directly in the nucleus, and therefore may be “trapped” at sites containing nuclear 7SK and U6 target RNAs. Additionally, the extent of this nuclear trapping appears reduced 24 h after transfection, and the antisense siRNA strand redistributes into the cytoplasm, whereas the sense strand, presumably cleaved by the Ago2 protein at the onset of RISC assembly (10, 11), fully decays. In line with the above scenario, the lack of a target RNA at any location in CHO cells eliminates all such trap sites for NS3 siRNA, and this siRNA is distributed throughout these cells.

In conclusion, our findings provide insights into the functioning of the RNAi machinery in mammalian cells. We find that cytoplasmically active RISC formation is likely not a common step in the cytoplasmic and nuclear RNAi pathways. Rather, for nuclear RNAi, active RISC appears to be formed in the nucleus. Our results also provide evidence for a mechanism by which the RNAi machinery influences the transport and localization of the corresponding siRNA and RISC to target RNA sites. Such a mechanism could serve a functional role in RNAi pathways by colocalizing the active RISC and its target RNA. Our studies reinforce the idea of RNAi as a phenomenon with multiple cellular pathways. They also provide evidence that subcellular compartmentalization may have a significant impact on RISC assembly and function, while raising the question of what specific molecular mechanisms are responsible for the observed selective localization.

We thank G. B. Robb and T. M. Rana (University of Massachusetts, Amherst) for providing information on the 7SK and U6 siRNA sequences, R. Parker (University of Arizona, Tucson) for Myc-tagged Ago2 plasmids, R. Bartenschlager (University of Heidelberg, Heidelberg) for HCV replicon plasmid, members of the A.A.D. laboratory for discussions, and Emily Remba for help with the figures. This work was supported by The Scripps Research Institute, National Institutes of Health Grant GM 073104 (to A.A.D.), and Department of Energy Grant ER46051 (to P.G.S.).