Nuclear Export of the DEAD Box An3 Protein by CRM1 Is Coupled to An3 Helicase Activity*

We have recently identified the Xenopus laevis An3 protein as a bona fide substrate for the nuclear export receptor CRM1 (Exportin 1). An3 binds directly to CRM1 with high affinity via a leucine-rich nuclear export signal located in the extreme N terminus. An3 is a member of the DEAD box family of RNA helicases, which unwind RNA duplexes. RNA unwinding is coupled to hydrolysis of nucleoside triphosphates by the helicase, and the ATPase activity of several helicases is greatly stimulated by various polynucleotides. Here we report that dATP hydrolysis by An3 is stimulated ; 6-fold by total RNA from X. laevis oocytes, whereas poly(U) RNA fails to enhance hydrolysis, suggesting the existence of a specific RNA activator for An3. Kinetic analysis reveals that a mutation within the conserved DEAD box motif reduces the rate of dATP hydrolysis by ; 6-fold. In accordance with this, the DEAD box mutant is unable to unwind double-stranded RNA. Microinjection of the An3 DEAD box mutant into X. laevis oocytes nuclei reveals a significantly lower export rate as compared with wild-type An3 protein. This is not because the mutant has lower affinity toward CRM1, nor is it due to altered RNA binding capacity. This suggests that nuclear export of An3 protein by CRM1 is coupled to An3 helicase activity. The multiple steps in gene expression require that RNA molecules interact with specific RNA and protein partners in a timely and synchronized manner. Some of these changes require alteration of RNA conformation. Chaperones have evolved that either stabilize or destabilize RNA structures and RNA-protein the Sma I site of pGEX-GTH (44). The PCR reaction was performed with Pfu I polymerase (Stratagene), primers 5 9 -CCGTTGCCCCACCAGTCAACC and 5 9 -GATGAGTCATGTGGCCGTGGAAAATG, and pET21a-An3 (40) as template. The fusion protein encoded by this plasmid also contains a recognition sequence for the catalytic subunit of cAMP-dependent heart muscle kinase at the C terminus and a thrombin cleavage site between the GST and An3 sequences. pGEX-GTH without insert was used for synthesis of GST protein. pGEX-GTH-An3 E389Q, which encodes an An3 fusion protein with an amino acid substitution at position 389 in An3 (a Glu to Gln substitution), was constructed by whole plasmid PCR using TaqPlus polymerase (Stratagene), primers 5 9 -ATCTAGTACCAA-ATACTTGC and 5 9 -CAAGCAGACAGAATGCTTGAG and pGEX-GTH- An3 as template. The ends of the resulting ; 7-kilobase PCR fragment were subsequently made blunt, phosphorylated, and finally ligated tog- ether. pET21a-An3 E389Q was made using the same strategy as for pGEX-GTH-An3 E389Q with pET21a-An3 as template. Similarly, pET- 21a-An3 L19A/L21A was constructed using primers 5 9 -ATGCGAACT-CAGCCGATGCTGAAAG and 5 9 -CTGCGCCAGCAAACTGCTGGTC and pET21a-An3 as template. This resulted in substitution of An3 amino acids Leu 19 and Leu 21 with alanines. The triple mutant pET21a- An3 L19A/L21A/E389Q was cloned using the same primers with pET21a-An3 E389Q as template. The plasmids pBS( 1 ) D EcoXma and pBS( 1 ) D XbaSph were made by digestion of pBS( 1 ) with Eco RII 1 Xma I and Xba I 1 Sph

The multiple steps in gene expression require that RNA molecules interact with specific RNA and protein partners in a timely and synchronized manner. Some of these changes require alteration of RNA conformation. Chaperones have evolved that either stabilize or destabilize RNA structures and RNA-protein interactions, with RNA helicases being a major class of modulators of RNA base pairings and likely also RNAprotein interactions (reviewed in Refs. 1 and 2). RNA helicases contain seven conserved motifs, including the so-called "DEAD box" (from the amino acid sequence Asp-Glu-Ala-Asp). The first two residues of this motif are invariant, while DEAD, DEAH, and DEXH box helicases are categorized in three separate subgroups. The best characterized eukaryotic RNA helicase is the translation initiation factor, eIF-4A, 1 which is regarded as a prototype for DEAD box RNA helicases. eIF-4A is thought to play an important role in unwinding RNA secondary structures in mRNA, thereby promoting binding of the small ribosomal subunit (3). Unwinding of RNA duplexes is dependent on the intrinsic ATPase activity of RNA helicases as shown for mammalian (4 -6), Xenopus laevis (7) Drosophila (8,9), Saccharomyces cerevisiae (10 -13), and viral enzymes (14 -16). Thus, although two Escherichia coli RNA helicases have been reported to disrupt RNA base pairing in the absence of nucleoside triphosphates (17,18), ATPase-dependent RNA unwinding is expected to be a general feature of DEXD/H box RNA helicases. Similarly, for most helicases, ATP hydrolysis is greatly stimulated by polynucleotides (e.g. Refs. 6, 11, 12, and 19 -27).
The messenger RNA encoding the X. laevis An3 protein was identified by Rebagliati and co-workers (28) in a search for mRNAs that localize to specific parts of unfertilized eggs. An3 is expressed throughout oogenesis and embryogenesis (28,29), as well as in most adult tissue (30). During the early stages of oogenesis An3 protein is found in both the nucleoli and the cytoplasm, whereas no signal is detected in the nucleus of stage VI oocytes (29,31). In early embryos An3 protein is still excluded from the nucleus but is later found in both the nucleus and the cytoplasm (31). The central part of the 697-amino acid sequence of An3 contains all the seven characteristically motifs of DEAD box RNA helicases and, furthermore, recombinant An3 protein has been shown to posses ATPase and RNA helicase activity (7).
Proteins highly related to An3 have been cloned from human (32), mouse (33,34), zebrafish (35), S. cerevisiae (36,37), and Schizosaccharomyces pombe (GenBank accession number AF084222), suggesting that the biological function of An3 is evolutionarily conserved. Since the mouse PL10 gene was the first to be identified we will refer collectively to the aforementioned proteins as members of the "PL10 family." Through biochemical and genetic analyses Chuang and co-workers (38) showed that the S. cerevisiae Ded1p is required for translation in vivo and in vitro (38). The related human DBX and mouse PL10 proteins each can rescue the growth of yeast cells with a ded1 deletion (38,39) and it therefore seems plausible that all proteins in the PL10 family are involved in protein synthesis. However, other aspects of gene expression might be regulated by these proteins. In agreement with the immunolocalization experiments showing that An3 is a partly nuclear protein (29,31), we have recently shown that the An3 protein shuttles between the nucleus and the cytoplasm (40). Nuclear export is mediated by direct binding to the transport receptor CRM1 (41) via an N-terminal leucine-rich nuclear export signal (NES) and, interestingly, an alignment of the PL10 family reveals that the NES sequence is conserved in all members. This suggests that PL10-like proteins may also have a nuclear function distinct from translation.
Transport of proteins and protein-RNA complexes across the nuclear envelope is an active and highly regulated process, which involves numerous factors (reviewed by Ref. 42). The translocations take place through large nuclear pore complexes that consist of 50 -100 distinct proteins, the nucleoporins (43). Trafficking of cargo through the nuclear pore complexes is mediated by soluble transport receptors, that interact with the nucleoporins. The transport receptors are capable of binding directly to their cargo or, alternatively, via adapters. Formation of export complexes containing CRM1-like transport receptors in the nucleus requires the GTPase Ran. Because of an asymmetric distribution of proteins that regulate the nucleotide bound state of Ran, Ran is predominantly GTP-bound in the nucleus and in the GDP form in the cytoplasm. Since only RanGTP, but not RanGDP, supports formation of a stable heteromeric export complex, the gradients of GDP-and GTPbound Ran across the nuclear envelope are essential for export of multiple cargoes from the nucleus. Conversely, nuclear import complexes consisting of importin ␤-like proteins are only stable in the absence of RanGTP, i.e. in the cytoplasm. Hence, Ran is a key player in defining directionality in nucleocytoplasmic transport (42).
In this report we show that the ATPase and helicase activities of An3 are coupled to nuclear export. An amino acid substitution within the characteristic DEAD box motif of An3 not only inhibits ATP hydrolysis and RNA unwinding activities in vitro but also diminishes nuclear export of An3 protein in X. laevis oocytes. We demonstrate that the mutant protein is capable of forming a heterotrimeric complex consisting of An3, CRM1, and RanGTP with the same efficiency as the wild-type protein. This excludes the possibility that the mutation leads to an aberrant protein conformation in which the NES of An3 is inaccessible.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The plasmid pGEX-GTH-An3 for expression of An3 fused to the C terminus of glutathione S-transferase (GST) was constructed by inserting a blunt-end PCR fragments into the SmaI site of pGEX-GTH (44). The PCR reaction was performed with PfuI polymerase (Stratagene), primers 5Ј-CCGTTGCCCCACCAGTCAACC and 5Ј-GATGAGTCATGTGGCCGTGGAAAATG, and pET21a-An3 (40) as template. The fusion protein encoded by this plasmid also contains a recognition sequence for the catalytic subunit of cAMP-dependent heart muscle kinase at the C terminus and a thrombin cleavage site between the GST and An3 sequences. pGEX-GTH without insert was used for synthesis of GST protein. pGEX-GTH-An3 E389Q, which encodes an An3 fusion protein with an amino acid substitution at position 389 in An3 (a Glu to Gln substitution), was constructed by whole plasmid PCR using TaqPlus polymerase (Stratagene), primers 5Ј-ATCTAGTACCAA-ATACTTGC and 5Ј-CAAGCAGACAGAATGCTTGAG and pGEX-GTH-An3 as template. The ends of the resulting ϳ7-kilobase PCR fragment were subsequently made blunt, phosphorylated, and finally ligated together. pET21a-An3 E389Q was made using the same strategy as for pGEX-GTH-An3 E389Q with pET21a-An3 as template. Similarly, pET-21a-An3 L19A/L21A was constructed using primers 5Ј-ATGCGAACT-CAGCCGATGCTGAAAG and 5Ј-CTGCGCCAGCAAACTGCTGGTC and pET21a-An3 as template. This resulted in substitution of An3 amino acids Leu 19 and Leu 21 with alanines. The triple mutant pET21a-An3 L19A/L21A/E389Q was cloned using the same primers with pET21a-An3 E389Q as template. The plasmids pBS(ϩ)⌬EcoXma and pBS(ϩ)⌬XbaSph were made by digestion of pBS(ϩ) with EcoRII ϩ XmaI and XbaI ϩ SphI, respectively, followed by Klenow fill-in in the presence of dNTP and religation of the plasmids. All constructs were verified by sequencing. The plasmids pQE-32-Ran, pQE-60-Rna1p, pQE-60-zzRBP1, which were a gift from Dr. Dirk Görlich, and pET-His-CRM1-H are described elsewhere (45)(46)(47). These plasmids encode Histagged Ran, Rna1p, RanBP1⌬NES (referred to as RanBP1 in the following), and CRM1, respectively.
Expression and Purification of Proteins-GST-An3 and GST-An3 E389Q proteins were prepared in parallel from E. coli strain BL21. 1.5 liters of bacterial cultures were grown at 37°C until A 600 reached ϳ1, transferred to room temperature, and induced with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. Bacteria were recovered by centrifugation and resuspended in 60 ml of ice-cold buffer A (20 mM HEPES/KOH, pH 7.9, 200 mM NaCl, 20% glycerol, 0.1 mM EDTA, 10 mM ␤-mercaptoethanol), containing proteinase inhibitors. The suspensions were gently sonicated followed by addition of Triton X-100 to 1% and agitation for 30 min at 4°C. The suspensions were cleared by centrifugation and the supernatants were mixed with 600 l of glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The proteins were allowed to bind to the matrix for 45 min at 4°C with gentle agitation. The matrix was washed extensively with buffer A and bound proteins were eluted with reduced glutathione (100 mM Tris/HCl, pH 8.0, 120 mM NaCl, 20 mM glutathione). Eluates were dialyzed against PBS containing 9% (v/v) glycerol and 1 mM dithiothreitol and purified further on a SP-Sepharose column using fast protein liquid chromatography equipment (Amersham Pharmacia Biotech) and PBS as running buffer. In a linear NaCl gradient proteins eluted at 100 mM NaCl, giving a total NaCl concentration of ϳ220 mM. Eluates were analyzed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) revealing that the GST-An3 and GST-An3 E389Q preparations were devoid of contaminants detectable at the level of Coomassie staining. Glycerol and dithiothreitol was added to the eluates at final concentrations of 9% and 1 mM, respectively, and eluates were stored at Ϫ80°C. GST protein was expressed from pGEX-GTH in E. coli BL21 according to standard procedures (Amersham Pharmacia Biotech). Expression and preparation of CRM1, GTP-bound Ran, Rna1p, and RanBP1 has been described previously (45).
RNA Preparation-Strand 1 and 2 RNAs were produced by standard run-off transcription using HindIII linearized pBS(ϩ)⌬EcoXma in combination with T7 RNA polymerase and EcoRI linearized pBS(ϩ)⌬XbaSph in combination with T3 RNA polymerase, respectively. For synthesis of randomly radiolabeled RNA [␣-32 P]UTP (Amersham Pharmacia Biotech, 800 Ci/mmol) was included in the reaction mixture. All RNAs were gel purified and quantified by the specific activity of 32 P incorporation or by UV absorption at 260 nm. Total RNA from X. laevis oocytes was prepared by incubating 100 oocytes in 5 ml of homogenization buffer (50 mM Tris/HCl, pH 7.4, 300 mM NaCl, 5 mM EDTA, 1.5% (w/v) SDS, 1.5 mg/ml Proteinase K) for 3 h at 37°C with vigorous shaking followed by phenol extraction twice and ethanol precipitation. 3Ј-end labeling of total X. laevis RNA was done using [␣-32 P]pCp prepared according to standard protocols and RNA ligase. Poly(U) RNA was purchased from Amersham Pharmacia Biotech.
ATP Hydrolysis Assays-Reactions containing 0.6 -2.9 pmol (60 -290 ng) of GST-An3 or GST-An3 E389Q in 5 l of An3 storage buffer (220 mM NaCl, 3 mM KH 2 PO 4 , 5 mM Na 2 HPO 4 , 9% (v/v) glycerol, 1 mM dithiothreitol, pH 7.3) was mixed on ice with 670 M MgCl 2 , 0 -1.7 g/l RNA, and 5-625 M unlabeled dATP (all final concentrations) in a total volume of 15 l. In competition experiments various unlabeled nucleoside triphosphates were added to final concentrations of 800 M on a background of 80 M dATP. In all experiments 0.1 l of [␣-32 P]dATP (Amersham Pharmacia Biotech, 750 Ci/mmol) was included. Following incubation at 37°C for 10 -90 min the reactions were stopped by placing the tubes on ice and adding 1 l of 0.5 M EDTA. 3.5 l of the reactions were spotted on PEI cellulose F thin layer chromatography plates (E. Merck). The plates were air dried shortly, pre-wetted with ethanol, and developed in 0.5 M KH 2 PO 4 . Finally, the plates were dried and subjected to PhosphorImager analysis (Molecular Dynamics).
Unwinding Experiments-Per reaction approximately 800 fmol of cold strand 1 RNA and 25 fmol of radiolabeled strand 2 RNA was renatured in 10 l of renaturation buffer (10 mM HEPES/KOH, pH 7.6, 100 mM KCl) by heating to 95°C followed by slow cooling to 37°C and then placed on ice. To the renatured RNA the following was added: 2 l of unwinding buffer (400 mM HEPES/KOH, pH 7.6, 500 mM CH 3 COOK, 30 mM (CH 3 COO) 2 Mg, 10 mM dithiothreitol, 0.5 mg/ml bovine serum albumin), 5 l of An3 storage buffer with or without ϳ2 pmol (200 ng) of GST-An3 or GST-An3 E389Q, and either 1 l of 30 mM ATP or 1 l of H 2 O. Reactions were placed at 30°C for 30 min after which 15 l of stop buffer (20 mM EDTA, 9% (v/v) glycerol, 0.4% (w/v) SDS, 0.1 mg/ml Proteinase K, bromphenol blue) were added and incubation continued at 30°C for 5 min. 20 l of the reactions were loaded on a pre-run nondenaturing 8% polyacrylamide gel containing 100 mM Tris borate, pH 8.3, and 1 mM EDTA. The gels were run at 10 mA at room temperature followed by autoradiography and PhosphorImager analysis.
Oocyte Injections- 35 S-Labeled wild-type and mutant An3 proteins were synthesized from the pET21a-An3 plasmid and derivatives hereof by coupled in vitro transcription and translation in rabbit reticulocyte lysate according to the manufacturer (Promega). GST protein produced from the pGEX-GTH plasmid was labeled with 35 S as described previously (40). Buffer exchange of An3 and GST proteins to PBS containing 9% (v/v) glycerol was performed using Nanosep concentrators (Palfiltron). 18 nl of a mixture of An3 and GST proteins were microinjected into the nuclei of X. laevis stage VI oocytes incubated in modified Barths' saline (15 mM HEPES/NaOH, pH 7.6, 88 mM NaCl, 1 mM KCl, 2.4 NaHCO 3 , 0.30 mM CaNO 3 , 0.41 mM CaCl 2 , 0.82 MgSO 4 , 10 g/ml sodium penicillin, 10 g/ml streptomycin sulfate). Upon incubation for 0 -100 min at 18°C injected oocytes were manually dissected in J buffer (10 mM HEPES/KOH, pH 7.6, 70 mM NH 4 Cl, 7 mM MgCl 2 , 0.1 mM EDTA, 10% (v/v) glycerol) and cytoplasms and nuclei were transferred immediately to TNE (10 mM Tris/HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA) and 96% ethanol, respectively. The fractions of seven oocytes per sample were pooled. Cytoplasmic fractions were homogenized in 250 l of TNE by repeated pipetting and cleared by centrifugation. The supernatants were mixed with 1 ml of acetone and placed at Ϫ70°C for 1 h along with the nuclear fractions. Cytoplasmic and nuclear proteins were recovered by centrifugation and resuspended in 49 l of SDS-PAGE sample buffer by vigorous shaking for 15 min and repeated pipetting. Samples were heated to 95°C for 3 min and 11 l were loaded on a 10% SDS-PAGE gel. The gels were fixed, equilibrated in Amplify solution (Amersham Pharmacia Biotech), dried, and subjected to autoradiography and PhosphorImager analysis.
Co-precipitation Assay-18 pmol (1.8 g) of GST-An3 or GST-An3 E389Q or 90 pmol (2.5 g) of GST was bound to glutathione-Sepharose 4B and mixed with 5 l of CRM1 buffer (45) (40), and 2.5 pmol (100 ng) of RanBP1 was added to each column followed by incubation at 4°C for 45 min with gentle agitation. Alternatively, complexes were dissolved by addition of 5 pmol (200 ng) of RanBP1 and 5 pmol (225 ng) of Rna1p and incubation at room temperature for 15 min. Released proteins were collected and columns were washed with washing buffer. Proteins remaining on the column were recovered by incubation in SDS-PAGE sample buffer at 37°C for 10 min. All samples were adjusted with SDS-PAGE sample buffer and resolved on a 10% SDS-PAGE gel. Proteins were visualized by silver staining.

RESULTS
ATPase Activities of An3 Protein-We have recently identified the An3 protein as a nuclear export substrate for the transport receptor CRM1 in X. laevis oocytes (40). This suggested that An3 has a nuclear function and in order to test this hypothesis we decided to investigate if a connection exists between An3 helicase function and nuclear export mediated by CRM1. Furthermore, we wished to extend the characterization of the enzymatic activities of recombinant An3. We expressed and purified full-length An3 protein fused to the C terminus of GST using glutathione-Sepharose and cation exchange chromatography. The ATPase activity of An3 has previously been reported to be independent of the presence of polynucleotides (7), which is in contrast to the behavior of most other RNA helicases (e.g. Ref. 6, 11, 12, and 19 -27). To re-investigate this we incubated GST-An3 protein from different batches with [␣-32 P]dATP and increasing concentrations of total RNA prepared from X. laevis oocytes. Following incubation at 37°C for 1 h the level of hydrolysis was detected by thin layer chromatography and quantified on a PhosphorImager. The combined results from several independent experiments are shown in Fig. 1A. Maximal ATPase activity was obtained with a total concentration of X. laevis RNA of 40 ng/l. At this concentration hydrolysis was stimulated 6-fold as compared with reactions without RNA, whereas at higher RNA concentrations the rate of hydrolysis decreased gradually to basal levels. For comparison we also tested whether poly(U) RNA stimulated the ATPase activity of GST-An3. However, at all concentrations tested poly(U) did not influence the ATPase activity, indicating the existence of a specific An3 activator among the total population of X. laevis RNA. To be sure that the stimulation seen upon addition of X. laevis RNA is RNA-mediated, we assayed whether prior treatment of the RNA with RNase A would abolish the observed enhancement. Indeed, titration with RNase A-digested RNA samples showed no stimulation of An3 ATPase activity (data not shown), leading us to conclude that this activity is RNA-dependent.
To examine the ability of various nucleoside triphosphates to function as substrates for the ATPase activity of An3 we added a 10-fold molar excess of unlabeled NTPs and dNTPs to standard hydrolysis reactions containing 80 M [␣-32 P]dATP and 100 nM GST-An3 (Fig. 1B). In the absence of competitor nucleotide 21% hydrolysis was obtained while excess ATP and dATP reduced hydrolysis of the radiolabeled dATP by approximately 5-fold (Fig. 1B, compare lane 2 with lanes 3 and 7). In contrast, the six other ribonucleotides and deoxyribonucleotides tested reduced dATP hydrolysis only slightly (Fig. 1B, compare lane 2  with lanes 4 -6 and 8 -10). Thus, although this assay measures the capacity of the individual nucleoside triphosphates to titrate out hydrolysis of [␣-32 P]dATP it presumably reflects their ability to function as nucleotide substrates for An3. We therefore conclude that An3 preferentially hydrolyses ATP and dATP. This notion is in agreement with analysis of nucleotide specificity in RNA unwinding experiments (7).
Having found an optimal concentration of oocyte RNA we next determined the kinetic parameters of GST-An3. In the presence of 40 ng/l X. laevis RNA and various concentrations of dATP the rate of hydrolysis was determined as before. The combined results of five independent experiments are shown in Fig. 1C, and in the double reciprocal Lineweaver-Burk plot in Fig. 1D. From these data the K m and k cat values for dATP as substrate for GST-An3 were calculated to 65 M and 9 min Ϫ1 , respectively.
A Mutation within the DEAD Box Motif of An3 Reduces ATPase Activity-The existing mutational and structural data on RNA helicases clearly show that the DEXD/H box motif (also called the Walker B motif or motif II) is essential for ATP binding and hydrolysis (e.g. Ref. (4, 19, 22, 48 -50)). We therefore decided to replace the glutamate residue at position 389 with glutamine (DEAD 3 DQAD), giving rise to a mutant protein that we name An3 E389Q. The mutant GST-An3 E389Q protein has the same properties as the wild-type proteins with regard to expression and purification from E. coli (see Fig. 5A). To compare the ATPase activities of the wild-type and the mutant protein we performed time course experiments with fixed concentrations of protein and dATP. Again, multiple experiments were performed using independent protein preparations and at different protein and dATP concentrations. In the presence of either wild-type GST-An3 or GST-An3 E389Q the amount of dADP produced increased as a linear function of incubation time, but the rate whereby GST-An3 E389Q hydrolyzed dATP was reduced about 6-fold relative to GST-An3 ( Fig.  2A). Thus, as expected, the DEAD box motif of An3 is important to ATP hydrolysis.
Unwinding of Double-stranded RNA by An3 Requires ATP Hydrolysis-To test if the ATPase activity of An3 is required for unwinding of RNA duplexes we constructed two plasmids from which in vitro transcription produces RNA molecules that can form extensive intermolecular base parings (Fig. 3A). When incubated with GST-An3 and ATP ϳ9% of the heterodimeric RNA was unwound to RNA monomers (Fig. 3B, lane  3), whereas the same amount of GST-An3 E389Q produced almost no detectable monomeric RNA (Fig. 3B, lane 4). ATP is an absolute prerequisite for RNA unwinding as no monomeric RNA was detected when ATP was omitted (Fig. 3B, lane 5). The absolute level of unwinding seen in the presence of GST-An3 is relatively low, which might be due to inefficient association of GST-An3 with the RNA substrate employed. The unwinding activity of RNA helicases requires that the substrate is partially single-stranded, presumably to provide an initial binding site from which helicase movement can proceed (e.g. Refs. 6, 11, and 51). Hence, if the putative RNA structure proposed in Fig.  3A is correct, only short unpaired regions are available for initial binding of An3, perhaps making this a limiting step in RNA unwinding.
Retention of An3 E389Q in the Nucleus of X. laevis Oocytes-Since An3 is a nucleocytoplasmic shuttling protein we speculated that this property may be coupled to its activities as an RNA helicase. To test this, we prepared An3 and An3 E389Q proteins by in vitro translation in rabbit reticulocyte lysate, followed by microinjection into the nuclei of X. laevis oocytes. Upon incubation for up to 100 min the oocytes were dissected manually and nuclear and cytoplasmic fractions were recovered. As a control for nuclear integrity and proper fractionation of the two cellular compartments GST protein was co-injected. Although the molecular mass of GST in principle is small enough to allow passive diffusion through the nuclear pore complexes, the majority of the protein remains in compartment of injection (Fig. 4), presumably because of multimerization. Using this method we have previously demonstrated that An3 is exported from the nucleus in a CRM1-dependent manner (40). To further study the rate of export we dissected the oocytes as early as 20 min after injection and found that nearly 50% of the injected wild-type An3 protein had left the nucleus (Fig. 4A, lanes 3 and 4, Fig. 4B), revealing a high export rate of An3. The percentage of An3 protein exported to the cytoplasm increased to approximately 90% after 100 min (Fig. 4, A and B). The amount of control GST protein in the cytoplasm was Յ12% at all time points. In contrast to the efficient export of wild-type protein, we repeatedly observed a significant lower export rate of the An3 E389Q mutant. 90 min after injection into the nucleus only 25% of the An3 E389Q protein had left the nucleus, as compared with ϳ80% export of wild-type protein (Fig.  4C, compare lanes 3 and 4 and lanes 5 and 6, Fig. 4D). An An3 mutant in which 2 leucine residues at positions 19 and 21   Fig. 1. Hydrolysis is expressed as nanomole of dATP hydrolyzed per nmol of protein allowing data from differential experimental conditions to be combined in one graph.
The Impaired Export of An3 E389Q Is Not Due to Inefficient Interaction with CRM1-Although we have mapped the NES of An3 to the N terminus, the reduced export rate of An3 E389Q might still be caused by an inefficient interaction with the nuclear export receptor CRM1, e.g. because of misfolding of the An3 E389Q protein. To exclude this possibility we examined whether wild-type GST-An3 and GST-An3 E389Q can form ternary complexes with CRM1 and RanGTP using purified recombinant proteins only. Equal amounts of GST-An3 and GST-An3 E389Q protein (see Fig. 5A) were bound to glutathione-Sepharose beads and incubated with CRM1 and RanGTP. GST protein was included as a control for specificity of CRM1 retention. Following extensive washes to remove unbound protein, retained CRM1 and RanGTP was released from the columns by challenging the complexes with RanBP1 and an NES peptide from the minute virus of mice NS2 protein (40). RanBP1 destabilizes RanGTP-containing export complexes while the excess NES peptide competes for binding to CRM1, thereby preventing re-association of released CRM1 with GST-An3. This approach has a major advantage, in that it specifically detects heterotrimeric An3⅐CRM1⅐RanGTP complexes mimicking export complexes. As is evident from Fig. 5B, GST-An3 and GST-An3 E389Q supported to a similar extent formation of CRM1-and Ran-containing complexes that could be disassembled by RanBP1 and NS2 NES peptide (Fig. 5B, lanes   4 and 5). In contrast, neither CRM1 nor Ran was released from the control GST column (Fig. 5B, lane 6). The CRM1 and Ran present in Fig. 5B, lane 9 reflects unspecific binding to the matrix or to GST itself. As seen in Fig. 5B, lanes 7 and 8, not all CRM1 and Ran was released from the GST-An3 columns upon incubation with RanBP1 and NES peptides, probably reflecting that free NES peptides are recognized less efficiently as compared with NES-containing polypeptides (40). Alternatively, residual binding might be due to protein-protein interactions not involving the NES of An3. To further substantiate the direct binding of CRM1 to GST-An3 and GST-An3 E389Q the pull-down experiment was repeated, but this time complexes were disassembled by triggering hydrolysis of Ran-bound GTP with RanBP1 and the Ran GTPase activating protein, Rna1p (40). As before, this approach specifically detects ternary An3⅐CRM1⅐RanGTP complexes. Fig. 5C shows that equal amounts of CRM1 and Ran were released from columns containing GST-An3 (lane 1) and GST-An3 E389Q (lane 2) by the concerted action of RanBP1 and Rna1p. We therefore conclude that wild-type An3 and An3 E389Q are recognized by CRM1 with the same efficiency.
Another possible explanation for the nuclear retention of An3 E389Q could be that the mutant protein has a dramatically higher affinity toward nucleic acids and therefore aggregates nonspecifically with RNA or DNA within the cell nucleus.
To examine this possibility we compared the binding of 3Ј-end labeled X. laevis RNA to GST-An3 versus GST-An3 E389Q by nitrocellulose filter binding as well as co-precipitation on glutathione-Sepharose columns. However, in either assay we could not detect a significant difference in RNA affinity of the two proteins (data not shown). The use of X. laevis oocyte extracts in co-precipitation experiments also failed to reveal any differences in retention of protein or RNA (data not shown). Thus, the impaired export of An3 E389Q from the nucleus is not due to a general increase in nucleic acid affinity, but rather is linked to an association with an yet unidentified specific nuclear RNA or protein target. DISCUSSION In this report, we have investigated the biochemical properties of the X. laevis DEAD box RNA helicase An3. An3 preferentially hydrolyzes ATP and dATP, and a 6-fold stimulation of dATP hydrolysis is seen with total RNA from X. laevis oocytes whereas poly(U) RNA has no effect. Intriguingly, a single amino acid substitution within the DEAD box motif of An3 that impedes the ATPase and RNA unwinding activities also leads to inhibition of nuclear export of An3 in X. laevis oocytes.
ATPase Activity-Most commonly, hydrolysis of nucleoside triphosphates by RNA helicases is dramatically stimulated by a variety of polynucleotides including homogenous polymers (e.g. Refs. 6, 11, 12, 19 -21, and 24 -27). However, An3, human CAP-Rf, and Drosophila Vasa proteins have been reported to display RNA-independent ATPase activity (7,9,52). The data presented here contradicts such observation in that dATP hydrolysis by An3 can be stimulated 6-fold by total RNA from X. laevis oocytes. At higher RNA concentrations the stimulatory effect decreases and finally disappears. Therefore, since Gururajan and Weeks (7) also tested total X. laevis RNA, but observed no effect, we suggest that the discrepancy between our and their data is caused by differences in RNA concentrations. The observation that poly(U) RNA fails to stimulate An3 ATPase activity indicates that a specific An3 activator is present in the total population of X. laevis RNA. Some RNA specificity in stimulation of ATP hydrolysis has also been demonstrated for a putative yeast homologue of An3, Ded1p (23) (see below), while the ATPase activity of E. coli DpbA and S. cerevisiae Prp5 is specifically stimulated 2400-fold by ribosomal 23 S RNA and 7-fold by U2 small nuclear RNA, respectively (53,54). The maximal dATP turnover number by An3 was estimated to 9 min Ϫ1 , which is comparable to human eIF-4A (3 min Ϫ1 (27)) and RNA helicase II (2 min Ϫ1 (55)), whereas the ATPase activity of other RNA helicases like S. cerevisiae Ded1p and Prp22, E. coli DbpA and hepatitis C virus NS3 is ϳ10 -100-fold higher (12,19,20,23,24,26,54). An3 seems to distinguish between the base of the nucleoside triphosphates as ATP and dATP are preferred substrates (Fig. 1B and Ref. 7). This feature is shared with eIF-4A and RNA helicase II (51,55), whereas for instance, RNA helicase A and hepatitis C virus NS3 can utilize all common nucleotides equally well (6,26).
Characterization of an An3 ATPase Mutant-Substitution of the conserved glutamate residue within the DEAD box of An3 by glutamine (An3 E389Q) reduces the ATPase activity ϳ6fold. Crystallographic data of RNA and DNA helicases reveal that the first, invariant aspartate residue of the helicase motif II (the DEAD box) is involved in co-ordination of a magnesium ion that contacts the terminal phosphate group of ATP (48, 56 -57). Furthermore, it is proposed that the succeeding glutamate residue activates the attacking water molecule during hydrolysis of ATP (56 -58). Due to the high degree of conservation of helicase motifs I-VI in An3 as compared with other helicases it is reasonable to assume that a similar mechanism is true for An3. In agreement with its reduced ATPase activity, An3 E389Q is severely impaired in unwinding of doublestranded RNA. We found that an excess of wild-type An3 protein is able to dissociate a stable RNA heterodimer into RNA monomers in an ATP-dependent process. The requirement for excess protein over RNA was also observed by Gururajan and Weeks (7) based on similar experiments with GST-An3, as well as in studies with most other RNA helicases (e.g. Ded1p (23), eIF-4A (4, 21, 51), Prp16p (11), Prp22p (12,24), p68 (5), Vasa (9), NS3 (59)). In fact, catalytic unwinding activity with subamounts of protein to RNA has so far only been demonstrated for RNA helicase A (6) and vaccinia virus NPH-II (14). The mechanism behind RNA unwinding is not well understood and perhaps it may differ among individual RNA helicases. In nearly all cases examined ATP hydrolysis is necessary for RNA unwinding, most likely reflecting that disruption of base pairings is an active, energy-dependent process (4 -16). A possible explanation of the requirement for excess protein in in vitro assays could then be that the helicase has to occupy the unwound single-stranded RNA to prevent reannealing of the RNA, and that several proteins are bound to each RNA molecule. Similarly, E. coli CsdA and DbpA, which both have been reported to unwind RNA in the absence of ATP (17,18), might function by binding to single-stranded RNA regions appearing due to "breathing" of the RNA duplex and thereby trapping the RNA in an unwound conformation (23).
A Nuclear Function of An3-We have recently shown that An3 shuttles continuously between the nucleus and the cytoplasm of X. laevis oocytes (40). Moreover, probing of early-stage oocytes as well as embryonic cells with anti-An3 antibodies gives rise to both nuclear and cytoplasmic staining (29,31). Export from the nucleus is mediated by the export receptor FIG. 4. An3 E389Q has a diminished nuclear export rate. Nuclear export of An3, An3 E389Q (An3 EQ), and An3 L19A/L21A (An3 LLAA) proteins was investigated by injection of 35 S-labeled proteins into the nuclei of X. laevis oocytes. Cytoplasmic and nuclear fractions were subsequently recovered and analyzed by SDS-PAGE. In all experiments 35 S-labeled GST protein was included as an internal control for proper injection and fractionation. A, a time course experiment of nuclear export of wild-type An3 protein was performed by dissecting the oocytes 0 -100 min after injection as indicated above the lanes. C and N denote cytoplasmic and nuclear fractions, respectively. The migration of An3 and GST is indicated on the right. B, the experiment shown in panel A was quantified by PhosphorImager analysis. Percentage of protein exported was calculated as the amount of protein located in the cytoplasmic fraction as compared with the combined signal from the cytoplasmic and nuclear fraction. Gray columns represent An3 protein, while GST is indicated by hatched columns. C, a comparison of nuclear export of wild-type An3 (lanes 3 and 4) to the export of An3 E389Q (lanes 5 and 6) and An3 L19A/L21A (lanes 7 and 8) mutants. Lanes 1 and 2 show the distribution immediately after injection, while the remaining lanes represent the cytoplasmic (C) and nuclear (N) fractions 90 min after injection. D, the average nuclear export observed in three independent experiments as the one shown in panel C. Calculations and representation of data are as described for panel B. Error bars indicate standard deviations. CRM1, requiring an N-terminal NES (40) (Fig. 4C). Alignment of putative An3 homologues from multiple species within the PL10 family (see Introduction) reveals that they all contain a similar motif with characteristic spacing of four hydrophobic residues at their N termini (40). Thus, we expect that other PL10 family members are nucleocytoplasmic shuttling proteins as well. Biochemical and genetic data have assigned an essential role of Ded1p in translation initiation (38). Human DBX and mouse PL10 can both substitute for Ded1p in a genetic complementation assay, which suggests that a function in translation initiation is conserved in all PL10-like proteins (38,39). However, due to the shuttling behavior of An3, we anticipate that PL10 family members may have a nuclear function as well. Based on localization studies, Weeks and colleagues have suggested that An3 is involved in rRNA processing (29,31), but there is no direct evidence available that might provide clues as to which function An3 fulfils in the nucleus. This led us to look for a correlation between nuclear export and the enzymatic activities of An3. Interestingly, we consistently observed a re-duction in the export rate of An3 E389Q as compared with wild-type protein, indicating that such a connection does exist. The similar mutation in eIF-4A gives rise to protein that acts as a dominant negative over wild-type protein in in vitro translation (50). We therefore speculated that nuclear injection of GST-An3 E389Q might inhibit the function of wild-type An3 in X. laevis oocytes. However, preliminary experiments have shown no effect on the nuclear export of several classes of RNA (data not shown). A likely explanation for this is the fact that we have been unable to obtain active protein preparations of higher concentrations than 0.5 M and consequently only 10 fmol of protein could be injected per nuclei. In comparison, a concentration of 50 -100 M of the human Dbp5 RNA helicase harboring a corresponding E389Q mutation was needed in order to detect an RNA export phenotype in similar oocyte injections (60); see below.
In addition to An3, RNA helicase A has been found to shuttle between the nucleus and the cytoplasm (61). However, in contrast to An3, export of RNA helicase A is not mediated by CRM1 and the nuclear export signal has no homology to known NESs (62). RNA helicase A associates with viral RNA transport elements in vivo and influence the expression of reporter genes regulated by these elements (63). Overexpression of RNA helicase A stimulates expression of genes encoded within introns harboring these RNA elements (i.e. where nuclear export of intron-containing pre-mRNA is required), while injection of anti-RNA helicase A antibodies reduces their expression (63). The mechanism underlying these effects remains unknown, although it has been suggested that RNA helicase A might stimulate nuclear export of RNA either directly (62), or by releasing the pre-mRNA from spliceosomes (63). In yeast, the Dbp5p RNA helicase was found by two groups to be essential for mRNA export (10,64). Conditional dbp5 mutant strains show a rapid accumulation of poly(A) ϩ RNA in the nucleus when shifted to the restrictive temperatures, concomitant with a decrease in the cytoplasmic poly(A) ϩ RNA level. Pre-mRNA splicing, RNA stability, and translation are unaffected upon the temperature shift, indicative of a direct role of Dbp5p in mRNA export (10,64). At normal growth conditions, Dbp5p is located in the cytoplasm and accumulates on the cytoplasmic side of the nuclear envelope (10,64). A similar distribution was recently found for the human homologue of Dbp5p, and is apparently mediated by an interaction with the nucleoporin CAN/Nup159p, which is part of the cytoplasmic fibrils of the nuclear pore complex (60,65). Interestingly, when mutant Dbp5 protein containing an amino acid substitution corresponding to the An3 E389Q mutation was injected into the cytoplasm of X. laevis oocytes it resulted in an inhibition of mRNA export (60). In addition, a more severe mutant inhibited export of U small nuclear RNA as well (60). Based on the accumulation of Dbp5 at the nuclear envelope, Dbp5 is proposed to act by releasing mRNA from export factors once the export complexes reaches the cytoplasmic face of the nuclear pore complexes. Likewise, as An3 ATPase and unwinding activities are required for An3 protein export, a nuclear RNA unwinding function of An3 may be coupled to transport through the nuclear pore complex. We speculate that An3 might be involved in stripping nuclear RNA processing factors off the RNA molecules and/or in melting of RNA secondary structures during the translocation process. However, future studies are required to elucidate to nuclear function of An3 and other PL10 family members.  1, 4, and 7), GST-An3 E389Q (lanes 2, 5, and 8), and GST-coated beads (lanes 3, 6, and 9). Approximately 1 M GST-An3 or GST-An3 E389Q or 4.5 M GST bound to glutathione-Sepharose beads was incubated for 45 min at 4°C with 1 M CRM1 and 5 M RanGTP in the presence of 1 mg/ml bovine serum albumin as nonspecific competitor. Non-bound proteins were collected (lanes 1-3) whereafter complexes were challenged with RanBP1 and NES peptide (lanes 4 -6). Proteins remaining on the beads after competition were eluted with SDS sample buffer (lanes 7-9). 12.5% of the flow-through were loaded on the gel, while 50% of the remaining samples were loaded. C, silver-stained SDS-PAGE gel of CRM1 pull down on GST-An3 (lane 1) and GST-An3 E389Q (lane 2). Complexes were assembled as in B and released by challenging with RanBP1 and Rna1p, thereby inducing hydrolysis of Ran-bound GTP. Migrations of markers (M) and applied proteins are shown on the left and right, respectively.