Exportin-5 Mediates Nuclear Export of Minihelix-containing RNAs

The adenovirus VA1 RNA (VA1), a 160-nucleotide (nt)-long RNA transcribed by RNA polymerase III, is efficiently exported from the nucleus to the cytoplasm of infected cells, where it antagonizes the interferon-induced antiviral defense system. We recently reported that nuclear export of VA1 is mediated by a cis-acting RNA export motif, called minihelix, that comprises a double-stranded stem (>14 nt) with a base-paired 5′ end and a 3–8-nt protruding 3′ end. RNA export mediated by the minihelix motif is Ran-dependent, which indicates the involvement of a karyopherin-related factor (exportin) that remained to be determined. Here we show using microinjection in Xenopus laevis oocytes that VA1 is transported to the cytoplasm by exportin-5, a nuclear transport factor for double-stranded RNA binding proteins. Gel retardation assays revealed that exportin-5 directly interacts with VA1 RNA in a RanGTP-dependent manner. More generally, in vivo and in vitro competition experiments using various VA1-derived, but also artificial and cellular, RNAs lead to the conclusion that exportin-5 preferentially recognizes and transports minihelix motif-containing RNAs. The adenovirus VA1 RNA (VA1), a 160-nucleotide (nt)-long RNA transcribed by RNA polymerase III, is efficiently exported from the nucleus to the cytoplasm of infected cells, where it antagonizes the interferon-induced antiviral defense system. We recently reported that nuclear export of VA1 is mediated by a cis-acting RNA export motif, called minihelix, that comprises a double-stranded stem (>14 nt) with a base-paired 5′ end and a 3–8-nt protruding 3′ end. RNA export mediated by the minihelix motif is Ran-dependent, which indicates the involvement of a karyopherin-related factor (exportin) that remained to be determined. Here we show using microinjection in Xenopus laevis oocytes that VA1 is transported to the cytoplasm by exportin-5, a nuclear transport factor for double-stranded RNA binding proteins. Gel retardation assays revealed that exportin-5 directly interacts with VA1 RNA in a RanGTP-dependent manner. More generally, in vivo and in vitro competition experiments using various VA1-derived, but also artificial and cellular, RNAs lead to the conclusion that exportin-5 preferentially recognizes and transports minihelix motif-containing RNAs. Nucleo-cytoplasmic transport of most RNAs and proteins is dependent on soluble receptors called karyopherins that can dock at and translocate through the nuclear pore complex. Interaction between cargo and karyopherin β is governed by the GTPase Ran. The asymmetric distribution of the Ran regulatory proteins provides a steep gradient of RanGDP (cytoplasmic)/RanGTP (nuclear) across the nuclear envelope that ensures the directionality of nuclear transport (1Gorlich D. EMBO J. 1998; 17: 2721-2727Google Scholar, 2Smith A.E. Slepchenko B.M. Schaff J.C. Loew L.M. Macara I.G. Science. 2002; 295: 488-491Google Scholar). Nuclear import receptors unload their cargo upon binding to RanGTP in the nucleus, whereas RanGTP is used to assemble export complexes which are in turn destabilized by dissociation of RanGTP in the cytoplasm (3Mattaj I.W. Englmeier L. Annu. Rev. Biochem. 1998; 67: 265-306Google Scholar,4Ossareh-Nazari B. Gwizdek C. Dargemont C. Traffic. 2001; 2: 684-689Google Scholar). Our understanding of the nuclear export of RNAs has been greatly facilitated by the study of viral RNAs. For this reason, we focused our attention on the adenovirus VA1 RNA (VA1), a 160-nt 1The abbreviations used are: nt, nucleotide; PKR, protein kinase R; EMSA, electrophoretic mobility shift assay; dsRBD, double-stranded RNA binding domains -long RNA transcribed by RNA polymerase III that massively accumulates in the cytoplasm of infected cells. It serves to antagonize the interferon-induced cellular antiviral defense system. Indeed, VA1 binds and inhibits the double-stranded RNA-dependent protein kinase R (PKR), which otherwise phosphorylates eIF2α and leads to the inhibition of protein synthesis (5Reichel P.A. Merrick W.C. Siekierka J. Mathews M.B. Nature. 1985; 313: 196-200Google Scholar, 6O'Malley R.P. Mariano T.M. Siekierka J. Mathews M.B. Cell. 1986; 44: 391-400Google Scholar). Adenovirus VA1 RNA contains a new cis-acting RNA export motif that comprises a double-stranded stem (>14 nt) with a base-paired 5′ end and a 3–8-nt protruding 3′ end and that can tolerate some mismatches and bends (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). This export signal, called minihelix, is present not only in VA1 but in a large family of small viral and cellular RNAs transcribed by polymerase III. RNA export mediated by the minihelix motif is Ran-dependent, which indicates the involvement of a karyopherin-related factor (exportin). This exportin is distinct from Crm1 and exportin-t (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar, 8Rutjes S.A. Lund E. van der Heijden A. Grimm C. van Venrooij W.J. Pruijn G.J. RNA (N. Y.). 2001; 7: 741-752Google Scholar). Therefore, we sought to identify cellular factors that bind to and mediate the export of minihelix-containing RNAs. VAΔIV, Mut10 mutants, and artificial stems have been described previously (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). VARdm and Mut9 mutants are derived from VAΔIV. VARdm contains a 3′-terminal oligouridine stretch replaced by a UAG sequence. Mut9 presents a 6-base mispairing in the 5′ end. These mutants have been generated by PCR and confirmed by sequencing. In vitro transcription was performed as described previously (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). t-RNA iMet with additional GGG at the 5′ end and a mature 3′ end presents a pseudo-minihelix structure but is exported to the cytoplasm only after the processing of the 5′ end (9Lund E. Dahlberg J.E. Science. 1998; 282: 2082-2085Google Scholar, 10Ossareh-Nazari B. Maison C. Black B.E. Levesque L. Paschal B.M. Dargemont C. Mol. Cell. Biol. 2000; 20: 4562-4571Google Scholar). EMSA was performed in binding buffer (20 mm Hepes, pH 7.9, 50 mm KCl, 5 mm NaCl, 0.2 mm EDTA, 0.5 mmdithiothreitol, and 10% glycerol) containing 250 nmMut10 to reduce unspecific binding, in a final sample volume of 20 μl. Recombinant proteins were preincubated with unlabeled competitors for 10 min at room temperature and incubated with the radiolabeled probe for 25 additional minutes. Then, 1 μl of loading DTT-dithiothreitol buffer (0.6 mg ml−1 heparin, 1 mg ml−1 bromphenol blue) was added, and incubation was pursued for another 5 min. Half of each sample was loaded on a 5% non-denaturing polyacrylamide gel, and electrophoresis was carried out at a constant voltage of 13.3 V cm−1 at 4 °C in 0.5× TBE (45 mm Tris borate, 1 mmEDTA). Oocytes injections and analysis of microinjected RNA by denaturing gel electrophoresis and autoradiography were performed as described previously (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). Stability of competitor VA RNA has been verified previously (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). VA1 consists of three functional domains, an apical stem-loop required for PKR binding, a central domain responsible for PKR inhibition, and a terminal stem, which brings together the 5′ and 3′ ends of the RNA. The 3′ end consists in an unpaired oligouridine stretch characteristic of polymerase III transcripts that not only acts as a transcription termination signal but also as a primary binding site for the La protein. To analyze the role of the oligouridine stretch in the nuclear export of VA RNA, this sequence was replaced by a UAG sequence (Fig. 1 A). Corresponding wild type (VAΔIV) and mutant (VARdm) radiolabeled RNAs were transcribed in vitro and injected intoXenopus oocyte nuclei. Radiolabeled U1 small nuclear RNA with a mutated Sm-binding site (U1ΔSm) and tRNAPhewere co-injected as internal control of exported RNA and U6Δss RNA that is not transported from the nucleus was used as a control of injection. Nucleocytoplasmic distribution of these RNAs was analyzed following a 3-h incubation (Fig. 1 B). As described previously, about 40–50% of VAΔIV was detected in the cytoplasm 3 h after injection (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). In contrast, a complete nuclear export of VARdm was observed in the same experimental conditions. This result indicates that the 3′-oligouridine stretch is not necessary for export of VA1 and rather retains this RNA in the nucleus as it has been previously shown for human Y1 RNA and U6 RNA (11Simons F.H. Rutjes S.A. van Venrooij W.J. Pruijn G.J. RNA (N. Y.). 1996; 2: 264-273Google Scholar, 12Boelens W.C. Palacios I. Mattaj I.W. RNA (N. Y.). 1995; 1: 273-283Google Scholar). Study of the determinant responsible for the cytoplasmic localization of VA1 led to the identification of a new cis-acting RNA export motif within the terminal stem. This structural motif consists in a double-stranded stem (>14 nt) with a base-paired 5′ end and a 3–8-nt protruding 3′ end (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar). We thus used either VA1-derived or artificial RNA containing a functional or altered minihelix structure and tested their ability to compete out the export of VARdm in Xenopusoocyte nuclei. Both VAΔIV and Stem20, which form a terminal stem mimicking that of VA1, efficiently prevented this process (Fig. 1,A and B) but did not affect the nuclear export of U1ΔSm and tRNAPhe, which are exported by Crm1 and exportin-t, respectively. It should be noted that cellular RNA Y1, which also presents a minihelix motif required for its nuclear export, also competes out the transport of VA1 but not tRNA or U1ΔSm (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar, 8Rutjes S.A. Lund E. van der Heijden A. Grimm C. van Venrooij W.J. Pruijn G.J. RNA (N. Y.). 2001; 7: 741-752Google Scholar). In contrast, mutant RNAs that display mispairing of the 5′ end (Mut9 and MM3), a shorter stem (Stem12), or no terminal stem (Mut10) poorly interfere with nuclear export (Fig. 1, A and B). The correspondence between the capacity of a given RNA to be exported from the nucleus (7Gwizdek C. Bertrand E. Dargemont C. Lefebvre J.C. Blanchard J.M. Singer R.H. Doglio A. J. Biol. Chem. 2001; 276: 25910-25918Google Scholar) and to inhibit VARdm nuclear export strongly suggested that RNAs containing a minihelix motif utilize a unique, saturable, and structure-dependent nuclear export pathway. We previously reported that RNA export mediated by the minihelix motif is controlled by the GTPase Ran, which indicates the involvement of a karyopherin-related factor (exportin). In addition, we also found that exportin-5, a karyopherin related to human Crm1 and theSaccharomyces cerevisiae Msn5/Kap142p, acts as a nuclear export receptor for proteins containing double-stranded RNA binding domains (dsRBD) (13Brownawell A.M. Macara I.G. J. Cell Biol. 2002; 156: 53-64Google Scholar). Since the structural basis of the minihelix motif corresponds to a double-stranded RNA stem, we tested the ability of exportin-5 to interact with VARdm in a EMSA. No detectable binding of recombinant exportin-5 or RanGTP could be observed on a VARdm probe (Fig. 2 A, lanes 2and 3). In contrast, exportin-5 was able to directly bind VARdm in a RanGTP-dependent manner (Fig. 2 A, compare lanes 4 and 5). The specificity of the exportin-5/minihelix interaction was further confirmed by the ability of VARdm and Stem 20 to compete out the formation of this complex (Fig.2 A, lanes 6 and 9), whereas minihelix-derived mutants that are not exported did not interfere with this complex (Fig. 2 A, lanes 7, 8,10, and 11). In agreement with these results, EMSA experiments using either VARdm or Mut9 as probe indicate that exportin-5 interacts better with VARdm than with Mut9 (data not shown). These data clearly indicate that exportin-5 specifically recognizes the minihelix motif, and this interaction is prevented by a 5′ protruding end or limited stem length. We then compared the competition efficiency of VARdm and various cellular RNAs on the exportin-5-VARdm complex. Addition of 1 pmol of unlabelled VARdm already affected the formation of the radiolabeled probe-exportin-5-RanGTP complex that was completely prevented with 50 pmol of VARdm (Fig. 2, B andC). No significant competition could be observed using U1Δsm that do not contain the minihelix motif, and a poor competition was observed using Mut10. hY1 RNA presents a 18-nt-long double-stranded stem with mismaches, a base-paired 5′ end, and a protruding 3′ end, thus corresponding to a minihelix structure less optimal than the VA1. Consistently, 5-fold more hY1 RNA than VARdm were required to compete out the formation of the radiolabeled probe-exportin-5-RanGTP complex. Finally tRNAPhe that contains a highly degenerated minihelix motif displayed a very weak competition effect (Fig. 2,B and C). Similar results were obtained with purified Escherichia coli tRNA than in vitrotranscribed human tRNAPhe. Together, these data indicate that exportin-5 preferentially recognizes minihelix-containing RNAs with affinity varying as a function of the minihelix structure. To determine whether the ability of exportin-5 to interact with minihelix structure in a RanGTP-dependent manner correlates with its ability to mediate the nuclear export of minihelix-containing RNAs, we examined the effect of microinjection of recombinant exportin-5 on the transport of various RNA in Xenopusoocytes (Fig. 3). For this purpose, a mixture of labeled VARdm, U3, U1ΔSm, Mut10, and tRNAMetRNAs were injected into oocyte nuclei following preinjection of control buffer or recombinant exportin-5, and their nucleocytoplasmic distribution was analyzed after short incubation periods (12–30 min). It should be noted that t-RNAMet used in this experiment contained an additional GGG at the 5′ end allowing the formation of a pseudo-minihelix structure 2 nt longer than mature tRNA. This tRNA is normally exported to the cytoplasm only after the processing of the GGG 5′ end (9Lund E. Dahlberg J.E. Science. 1998; 282: 2082-2085Google Scholar, 10Ossareh-Nazari B. Maison C. Black B.E. Levesque L. Paschal B.M. Dargemont C. Mol. Cell. Biol. 2000; 20: 4562-4571Google Scholar). As shown on Fig. 3, A and B, preinjection of exportin-5 dramatically accelerated VARdm export and weakly stimulates the export of unprocessed tRNA. Importantly, the exportin-5 did not significantly affect the export of other RNAs. Altogether, our results demonstrate that exportin-5 promotes the nuclear export of the viral VA1 RNA. More generally, this exportin appears to specifically interact with the minihelix RNA motif in a RanGTP-dependent manner and is likely responsible for the nuclear transport of viral or cellular RNAs containing such a motif. Based on our results, an optimal exportin-5-interacting minihelix presents a 20-nt-long double-stranded stem with a base-paired 5′ end and a protruding 3′ end. It has been reported recently that exportin-5 is able to stimulate nuclear export of tRNA in microinjectedXenopus oocytes nuclei when transport of endogenous tRNA was artificially saturated (14Calado A. Treichel N. Muller E.C. Otto A. Kutay U. EMBO J. 2002; 21: 6216-6224Google Scholar). However, in normal experimental conditions, we found that exportin-5 does not affect tRNA transport. In addition, although exportin-5 can bind tRNA, our results show that its affinity for the optimal minihelix structure is much higher. This strongly suggest that tRNAs, which contain a degenerate minihelix, do not represent a preferential cargo for exportin-5 but can eventually use this transport receptor when their own transport pathway using exportin-t is deficient. Interestingly, the predicted secondary structures of intermediates of the recently discovered class of micro-RNAs and siRNAs resemble the minihelix motif and might mediate their nuclear export by the exportin-5 pathway. In addition, exportin-5 acts as a nuclear export receptor for proteins containing dsRBD. This could occur through an indirect interaction mediated by dsRNA similarly to exportin-5/eEF1A interaction via tRNA (15Bohnsack M.T. Regener K. Schwappach B. Saffrich R. Paraskeva E. Hartmann E. Gorlich D. EMBO J. 2002; 21: 6205-6215Google Scholar), but one can certainly not exclude that certain dsRBD-containing proteins might participate in the nuclear export of minihelix RNAs. We thank members of the laboratory for helpful and stimulatory discussions and M. Rodriguez, U. Nehrbass, and J. Salamero for critical reading of the manuscript.