Fluorescence correlation spectroscopy and fluorescence cross-correlation spectroscopy reveal the cytoplasmic origination of loaded nuclear RISC in vivo in human cells

Cell culture

ER293 cells stably transfected with the pERV3 vector (Stratagene) were cultured at 37°C in DMEM (high glucose, Sigma) with 10% FCS (PAA Laboratories GmbH), 2 mM glutamine (Gibco) and 0.3 mg/ml G418 (50 mg/ml, Gibco).

For the generation of the stable EGFP-Ago2 cell line, the pEGSH-EGFP-Ago2 vector was linearized with AseI. ER293 cells were seeded at 4 × 10⁴ cells/ml, 24 h before transfection. Transfection was carried out using linearized DNA and Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocols. After 24 h cells were treated with 0.4 mg/ml selection antibiotic Hygromycin B (50 mg/ml, Gibco) for 3 weeks. Cells were diluted to 1 cell/0.2 ml and transferred onto 96-well plates for recloning. Wells with single colonies were cultured, resulting in EGFP-Ago2 positive clones that were confirmed via western blot, immunofluorescence and PCR. The generated cell line 10G stably expressing EGFP-Ago2 was cultured at 37°C in DMEM (high glucose, Sigma) with 10% fetal calf serum, 2 mM glutamine, 0.3 mg/ml G418 and 0.4 mg/ml Hygromycin B. All cells were regularly passaged at subconfluency and were plated with 1–5 × 10⁴ cells/ml density.

siRNA sequences and calculation of the 5′-end free energy

All siRNA strands were obtained from IBA GmbH. The RNA oligonucleotides were synthesized with a 5′-phosphate and a 3′-amino group on a C6-carbon linker and were labelled with Cy5 succinimidyl ester (Cy5-NHS, Amersham Biosciences) as described previously (17). The nucleotides 18–21 were 2′-O-methyl modified to protect the label from degradation except for silencing assays were the siRNAs were not modified. Target-RNA and siRNA sequences are listed in the Supplementary data.

For siRNA control experiments, we used Silencer™ Negative control #2 (NegsiRNA) from Ambion.

The free energies of the first 4 bp of each siRNA strand of the duplex at the 5′-end was calculated in kcal/mol by using the nearest-neighbour method and the mfold algorithm (32,33).

Microinjection

For microinjection, 7–9 × 10⁴ ER293/10G cells were transferred onto MatTek chambers coated with Fibronectin (25 µg/ml in PBS including CaCl₂ and MgCl₂, Roche) 24 h before microinjection. The micropipette (Femtotip 2, Eppendorf) is loaded with 1.5–4 µM labelled siRNAs (WGA 2.25 µg/µl, Biomeda) in 110 mM K-gluconate; 18 mM NaCl; 10 mM HEPES pH 7.4 and 0.6 mM MgSO₄. The micromanipulator consists of a FemtoJet and InjectMan NI2 which is mounted directly on a microscope. Working pressure for injection was between 25 and 80 hPa for 0.1 s and a holding pressure of 15 hPa.

Fluorescence correlation spectroscopy setup

Fluorescence correlation spectroscopy and laser scanning microscopy (LSM) were carried out on a commercial system consisting of a LSM510 and a ConfoCor3 (Zeiss, Jena, Germany). The 488 nm line of a Ar-Ion laser and the 633 nm line of a HeNe laser were attenuated by an acousto-optical tunable filter to 3.5 and 1.05 kW/cm² and directed via a 488/633 dichroic mirror onto the back aperture of a Zeiss C-Apochromat 40×, N.A. = 1.2, water immersion objective. Fluorescence emission light was collected by the same objective and split into two spectral channels by a second dichroic (LP635). To remove any residual laser light, a 505–610 nm bandpass or 655 nm longpass emission filter, respectively, was employed. The fluorescence was recorded by avalanche photodiode detectors (APDs) in each channel. For EGFP-Ago2 autocorrelation measurements, a mirror substituted the second dichroic and a 505-nm longpass emission filter was used in a one channel setup. Out-of-plane fluorescence was reduced with a 70-μm pinhole. The fluorescence signals were software-correlated and evaluated with MATLAB (Mathworks, Natick, MA, USA) by using weighted Marquardt non-linear least-square fitting routine.

Laser scanning microscopy was performed using the APD's of the ConfoCor3 on the same setup. Cell measurements were performed in air-buffer (150 mM NaCl, 20 mM HEPES pH 7.4, 15 mM glucose, 15 µg/ml BSA, 10 mM trehalose, 5.4 mM KCl, 0.85 mM MgCl₂, 0.7 mM CaCl₂) at RT.

Preparation and fractionation of cell extracts

Fractionated cell extracts were prepared from 10G, 293 and ER293 cells from four confluent T175 chambers essentially performed as previously described by Robb et al. (23). Sucrose gradients were exactly performed as has been described in (34).

Immunoprecipitation

ER293 and 10G cells were lysed with IP buffer (0.5% NP-40, 150 mM KCl, 25 mM Tris–HCl pH 7.5, 2 mM EDTA, 1 mM NaF, 0.5 mM DTT, protease inhibitors, Roche) and centrifuged at 10 000g for 10 min. Extracts were incubated with anti-GFP (Roche) or anti-p53 for 2 h at 4°C. Antibodies were pulled down with protein G Sepharose 4 fast flow beads (GE Healthcare) for 1 h at 4°C. Beads were subsequently washed with buffer (0.1% NP-40, 300 mM KCl, 25 mM Tris–HCl pH 7.5, 2 mM EDTA, 1 mM NaF, 0.5 mM DTT, protease inhibitors, Roche) and with PBS.

Western blotting

For western blot analysis, the cells were lysed with ×1.5 loading buffer for denaturating polyacrylamide gel electrophoresis. For immunoblotting, proteins were run on 7.5–10% PAGE and transferred onto Protran® Nitrocellulose Transfer Membrane (Schleicher & Schuell). Antibodies: α-EGFP (Abcam ab290, 1 : 2000), α-Ago2 clone 11A9 (1:50), α-GAPDH (Abcam 9484, 1:2000), α-Actin (Abcam AC15). Immunoreactive signals were detected using ECL™ Western Blotting Detection Reagents according to the manufacturer's protocol (Amersham Biosciences).

qPCR

Total RNA was isolated using the Prep Ease kit (USB, USA), according to the manufacturer's instructions, using DNaseI digestion on column. cDNA was synthesized with random hexamer primers from 2 μg of total RNA using the First Strand cDNA synthesis kit (Fermentas, Canada), according to the manufacturer's protocol. qPCR was performed on a MyiQ cycler (BioRad, USA) using the Mesa Green qPCR mastermix (Eurogenetec, Belgium) with primers at 100 nM final concentration and a two-step PCR protocol (5 min initial denaturation at 95°C, 40 cycles with 10 s at 95°C and 1 min elongation at 60°C). The primers were (forward, reverse): GAPDH, 5′ TGGTATCGTGGAAGGACTCATGAC, 5′ ATGCCAGTGAGCTTCCCGTTCAGC; 7SK, 5′ CCTGCTAGAACCTCCAAACAAG, 5′ GCCTCATTTGGATGTGTCTG. Data were evaluated using the ΔΔCt method.

Nuclear RISC differs from cytoplasmic RISC

Although the human RISC complex has been intensely studied, the protein composition is still not entirely clear, with reported molecular weights from ∼160 kDa to ∼2 MDa (37,38). The huge deviation in molecular mass might depend on extract generation or on the experimental procedure for their estimation. To avoid preparation artefacts, we measured hAgo2 containing RISC complexes in thermal equilibrium in vivo in 10G cells and determined the molecular mass of the individual RISC in different cellular compartments by FCS (for details see Supplementary materials). The autocorrelation curves of EGFP in 293 cells resulted in almost identical diffusion times in the cytoplasm and the nucleus of τ_EGFP(cytopl) = (354 ± 13) μs and τ_EGFP(nucl) = (368 ± 8) μs, corresponding to a diffusion coefficient of D_EGFP(cytopl) = (25.5 ± 0.9) μm²/s and D_EGFP(nucl) = (24.5 ± 0.5) μm²/s, indicating that the viscosity and therefore the motility is approximately the same in both compartments (Figure 1A) (39). In contrast, measurements in 10G cells displayed a significant difference in diffusion time for EGFP-Ago2 between the cytoplasm (τ_{EGFP-Ago2(cytopl)} = (1678 ± 58) μs, D_{EGFP-Ago2(cytopl)} = (5.4 ± 0.2) μm²/s) and the nucleus (τ_{EGFP-Ago2(nucl)} = (657 ± 22) μs, D_{EGFP-Ago2(nucl)} = (13.7 ± 0.5) μm²/s). With respect to EGFP, these diffusion times correspond to a molecular weight of (3.0 ± 0.6) MDa for cRISC and (158 ± 26) kDa for nRISC. While the obtained molecular weight for cRISC is in the range of a previously isolated RISC of about 2 MDa (37), the molecular weight of nRISC resembles almost the distinct EGFP-Ago2 alone. It should be noted that the molecular mass for cRISC could be overestimated due to a higher effective viscosity a larger molecule experiences in a crowded environment. Yet, due to the ∼20-fold difference in molecular mass, we conclude that cRISC represents a different complex compared to nRISC. Further analyses of the measurements indicate that within the accuracy of the FCS technique no distinct EGFP-Ago2 is present in the cytoplasm (Supplementary data).

Western blot analysis of nuclear and cytoplasmic 10G-extracts with an EGFP-specific antibody yielded a single specific band for EGFP-Ago2 at 130 kDa, demonstrating the absence of contaminating EGFP-Ago2 fragments in the nucleus (Figure 1B). Therefore, nRISC is not formed by a truncated version of EGFP-Ago2. To corroborate the FCS data, sucrose gradients with nuclear and cytoplasmic extracts from 293 cells were performed. The cytoplasmic extracts displayed the aforementioned hAgo2 containing complexes I–III with a large portion of complexes being bigger than 350 kDa. The large complexes II–III were missing in nuclear extracts (Figure 1C). The same result was obtained for hAgo1 (data not shown). The discrepancy between the obtained molecular weights and the proportions of different RISC in sucrose density gradients and FCS measurements lie within the different experimental techniques. Whereas FCS is a real-time acquisition technique in vivo, the gradient extract preparation includes a centrifugation step which excludes the analysis of high molecular weight complexes. Furthermore, the sucrose gradients are performed over a period of 20 h which might lead to the disassembly of the majority of RISC complexes. However, the results obtained by FCS are supported by sucrose gradients illustrating the absence of higher molecular weight complexes in the nucleus.

To characterize the sub-cellular distribution of hAgo2 containing complexes, we quantified the amount of EGFP-Ago2 in the nucleus and the cytoplasm by FCS and imaging. EGFP-Ago2 preferentially localized to the cytoplasm with a cytoplasmic to nuclear ratio of (4.2 ± 0.5) : 1 determined by FCS and (4.7 ± 0.5) : 1 determined by imaging (for details see Supplementary materials). The higher value for imaging might be caused by the accumulation of EGFP-Ago2 into P-bodies, which are not accessible with FCS.

We could directly characterize RISC sizes in the nucleus in thermal equilibrium by FCS in vivo without preparation artefacts and contaminations by cRISC. Our data demonstrate that cytoplasmic EGFP-Ago2 is part of a large complex and differs from nRISC which most probably represents the discrete EGFP-Ago2 protein. We obtained a cytoplasmic to nuclear EGFP-Ago2 concentration of ∼4–5, indicating that hAgo2 is not as underrepresented in the nucleus as estimated by previous imaging studies. This might result from the high expression levels in transient systems in previous reports, compared to the almost endogenous expression levels of EGFP-Ago2 in 10G cells (35,36).

Asymmetric guide strand incorporation into nuclear and cytoplasmic RISC

The incorporation of siRNAs into nRISC and the effects of target interactions on the distribution of activated RISC in the cytoplasm and the nucleus in mammalian cells are largely unknown. To assess siRNA incorporation levels and asymmetric loading of the guide strand into RISC by FCCS in vivo (for details see Supplementary materials), we covalently labelled siRNAs on the 3′-end of the guide or passenger strand with Cy5 (illustrated in Figure 2C and D top) and delivered them directly into 10G cells via microinjection to avoid siRNA segregation within endocytic compartments. This resulted in silencing active and homogeneous localized cytoplasmic siRNAs required for FCCS (17). The siRNA siTK3 targets the mRNA of Renilla luciferase encoded on the plasmid pRL-TK. The 5′-end hybridization energy of siTK3 was calculated as described (12) to define the guide and passenger strand of the siTK3 duplex (Figure 2A). The siTK3 siRNA was used as it displayed high levels of Renilla luciferase silencing, demonstrating the incorporation of the guide strand into endogenous RISC (Figure 2B).

Microinjection of the passenger strand labelled siTK3 resulted in cross-correlation amplitudes between 0% and 5% in the cytoplasm and nucleus up to 12 h, indicating the exclusion of the passenger strand from RISC (Figure 2C). On the contrary, guide strand labelled siTK3 lead to significant cross-correlation amplitudes in the nucleus and the cytoplasm gradually increasing for up to 6 h with a slight decline thereafter (Figure 2D). As the relative amplitude of the cross-correlation function is directly proportional to the concentration of double labelled species, this result demonstrates the specific incorporation of the guide strand into nuclear and cytoplasmic RISC whereas the passenger strand is excluded during the loading process. The specific incorporation of the guide strand could be observed down to a 5′-hybridization energy difference of 0.5 kcal/mol (siGL2) (data not shown). Interestingly, the incorporation levels in the nucleus and the cytoplasm synchronistically increased and declined, in the nucleus however to a lower level, reaching saturation after 6 h (Figure 2D, bottom panel). Similar results were obtained with other siRNAs: siTK2 (Supplementary Figure S3), siRNA targeting Firefly luciferase (siGL2) and siGAPDH (data not shown). Guide strand labelled siRNAs could be examined for longer time periods compared to passenger strand labelled siRNAs, due to the loss of the Cy5 signal in passenger strand containing cells. This most probably results from a stabilization of the guide strand caused by the interaction with RISC whereas the passenger strand gets degraded. Control experiments with EGFP supplemented with guide strand labelled siTK3 exhibited no cross-correlation amplitude, whereas a double labelled control resulted in 80% cross-correlation amplitude in vitro (Supplementary Figure S4A and B).

For the first time, we could monitor the asymmetry dependent incorporation into nRISC and show that nuclear levels of guide strand incorporation correlate with cytoplasmic incorporation levels in time in vivo. The specific guide strand incorporation into nRISC provides additional evidence for the specific loading and function of nRISC.

Cleavage impaired RISC-target interactions mediate RISC accumulation in the nucleus

To directly investigate the effect of RISC-target-RNA interactions in the nucleus and cytoplasm by FCCS, we microinjected a 50 nt long RNA into 10G cells that is labelled with Cy5 on the 5′-end, contains the siTK3 target site and is modified with 2′-O-methyl on each end to protect the target-RNA molecule against cellular RNases. The interaction of the EGFP-labelled RISC containing an unlabelled siRNA with the target-RNA can then be visualized by increased cross-correlation amplitudes (Figure 4A). The target-RNA localized to the cytoplasm and displayed a slight accumulation of fluorescence signal in the nucleus and even stronger in nucleolar structures (Figure 4B). FCCS measurements in non-transfected and NegsiRNA transfected 10G cells after 1 and 3 h of target-RNA delivery showed no interaction of RISC with the target-RNA illustrating assay specificity and the absence of miRNA target sites within the target-RNA (Figure 4C). Longer incubation times were not tested as already 30% of target-RNA was degraded after 3 h in S20 extracts as determined by PAGE (data not shown). Despite that, transfection of perfectly matching siTK3 resulted in very low levels of EGFP-Ago2-target-RNA interactions (Figure 5C). In contrast, transfection of siTK3-B2 which forms a central bulge with the siTK3-target site yielded high levels of RISC-bound target-RNA in the cytoplasm. In the nucleus however, levels were 3-fold lower after 1 h with an increase after 2 and 3 h almost reaching cytoplasmic interaction levels (Figure 4C). The complete inactivation of the siTK3-B2 cleavage activity was verified by a dual luciferase assay (Figure 4D).

These results demonstrate (i) the localization of functional guide strand loaded RISC to the nucleus, (ii) the short interaction time of RISC loaded with a perfectly matching guide strand with the target-RNA and that (iii) bulges in the centre of the guide-target-RNA hybridizations increase the lifetime of RISC-target-RNA interaction, resulting in the accumulation of the complex in the nucleus over time. This is in strong agreement with our previous results showing the accumulation of specifically loaded RISC in a target- and duration of the interaction dependent fashion. Therefore, the target-dependent accumulation of nRISC is caused by the stable interaction of RISC with its target and not by the nuclear localization of 7SK snRNA, again underlining the dynamic levels of specifically loaded nRISC.

Nuclear activated RISC originates from the cytoplasm

The fact that we could detect a target dependent accumulation of nRISC raises the question of the origination of nRISC. Two mechanisms are possible: First, RISC is loaded both inside the nucleus and the cytoplasm or second, RISC is loaded in the cytoplasm and can than be imported into the nucleus to regulate gene expression/function. The first hypothesis is less likely for several reasons. The nRISC that was characterized by FCS in vivo displayed a molecular mass of ∼158 kDa which is too small for the RLC (containing EGFP-Ago2, TRBP and Dicer which should result in a complex of approximately 380 kDa). In addition, siRNAs are too short to contain localization signals, therefore the RLC is incapable to locate the appropriate cellular compartment for individual silencing triggers until RISC interacts with its target RNA. And last, miRNAs and siRNAs depend on their cytoplasmic localization, mediated by Exp5, for effective silencing to occur (17,18,42). To elucidate the cytoplasmic origin of nRISC, we microinjected si7SK1 into the cytoplasm of 10G cells and allowed for exchange between nucleus and cytoplasm. This resulted in a preferential cytoplasmic localization of the siRNAs with 24% of EGFP-Ago2 loading levels in the cytoplasm with the corresponding nuclear levels of 21% after 3 h (Figure 5). In the second experiment, we microinjected siRNAs into the cytoplasm together with wheat germ agglutinin (WGA), an inhibitor of nuclear pore-mediated active transport. The sub-cellular localization of the siRNAs was still preferentially in the cytoplasm, but due to their small size could still be detected in the nucleus (Figure 5). The cytoplasmic incorporation levels were unaffected compared to WGA-free microinjected cells showing that WGA did not interfere with the RISC loading process. The nuclear incorporation level decreased 3-fold down to 7%, indicating that most of nRISC is originating from the cytoplasm. The same decrease in EGFP-Ago2 nuclear incorporation levels were detected by nuclear microinjection of si7SK1 together with WGA. The drop of loaded EGFP-Ago2 with si7SK1 in the cytoplasm results from reduced siRNA levels in the cytoplasm mediated by WGA. The preferential delivery of siRNAs via microinjection into the nucleus could be detected by the higher fluorescence intensity in the nucleus in comparison with the cytoplasm (Figure 5). The same results were obtained for si7SK3-bulge (Supplementary Figure S6). The inhibitory function of injected WGA on import and export processes was verified with the stress kinase Mk2 tagged with EGFP. Upon stress this predominantly nuclear protein shuttles into the cytoplasm (43). We could show that EGFP-MK2 shuttles into the cytoplasm after treatments with Anisomycin for 3 h that could be stopped with the addition of Leptomycin B which is an inhibitor of the exporter of MK2 (Supplementary Figure S7A, upper panel and S7B). The microinjection with WGA inhibited the export of EGFP-MK2 very efficiently (Supplementary Figure S7A, bottom panel and S7B). Treatment with cycloheximide did not change incorporation levels of EGFP-Ago2 in the nucleus, thereby excluding that the accumulation of specifically loaded RISC in the nucleus results from de novo synthesis of EGFP-Ago2 which is transported in the nucleus and subsequently loaded (Supplementary Figure S8).

Our results show that RISC is loaded in the cytoplasm and is imported into the nucleus. Therefore, the RLC is strictly localized in the cytoplasm. The accumulation of the si7SK1 and si7SK3-bulge siRNA indicates a dynamic exchange of RISC between the nucleus and the cytoplasm. Furthermore, the interaction with the target RNA seems to interfere with the export of RISC leading to the accumulation of this particularly loaded RISC in the nucleus. RISC shuttling between the cytoplasm and the nucleus represents a mechanism to access targets in both cellular compartments.