Transport across the nuclear envelope is mediated by transport receptors from the Importin β family. We identified Exportin 1 from Arabidopsis (AtXPO1/AtCRM1) as the nuclear export receptor for proteins carrying leucine-rich nuclear export signals (NESs). AtXPO1 shares 42–50% identity with its functional homologues from humans and yeasts. We functionally characterised AtXPO1 by its interaction with NESs of animal and plant proteins, which is inhibited by the cytotoxin leptomycin B (LMB), and also by its interaction with the small GTPase Ran1 in the yeast two-hybrid system. Furthermore, we demonstrated the existence of a nuclear export pathway for proteins in plants. For the characterisation of nuclear export activities, we established an in vivo assay based on the localisation equilibrium of a GFP reporter protein fused to both a nuclear localisation signal (NLS) and an NES motif. Using this in vivo assay we demonstrated that the NES of the heterologous protein Rev is also functional in plants and that its export is inhibited by LMB. In addition, we identified a leucine-rich NES in the Arabidopsis protein AtRanBP1a. The NES, which is located at the carboxy terminus of the protein, is disrupted by mutating three long chain hydrophobic amino acid residues to alanine (L176A, L179A, V181A). In BY-2 protoplasts the NES of AtRanBP1a is functionally indistinguishable from the Rev NES. Our results demonstrate that the machinery for the nuclear export of proteins is functionally conserved in plants.
In eukaryotes, the nuclear envelope provides a physical barrier that allows spatial separation of fundamental processes such as gene transcription and translation of mRNAs. Molecules smaller than the size exclusion limit of about 60 kDa are able to diffuse through the channel of the nuclear pore complex (NPC) while larger macromolecules must be transported actively. The translocation across the nuclear envelope is mediated by nuclear transport receptors of the Importin β family ( Mattaj & Englmeier 1998). The small GTPase Ran regulates the directionality of the transport. The regulatory proteins of Ran are distributed asymmetrically across the nuclear envelope ( Görlich 1998). The GTPase-activating protein RanGAP1 and the Ran-binding protein RanBP1 are located in the cytoplasm while the nucleotide exchange factor RCC1 is bound to chromatin. Therefore, a gradient of a high nuclear and a low cytoplasmic concentration of RanGTP has been predicted to define the identity of the two compartments concerning the direction of the transport.
The nuclear import of proteins is a well characterised transport process in vertebrate and yeast cells and also in plants ( Görlich 1997; Merkle & Nagy 1997; Smith & Raikhel 1999). Until recently, much less was known about the functional mechanism of the export of macromolecules from the nucleus. The prediction of a nuclear export pathway for proteins arose from the identification of small transferable nuclear export signals (NESs) in the inhibitor of the cAMP-dependent protein kinase, PKI, and the HIV-1 protein Rev ( Fischer et al. 1995 ; Wen et al. 1995 ). This signal is characterised by a short sequence of amino acids with a certain spacing of leucine or other long chain hydrophobic residues.
The nuclear export receptor for proteins containing a leucine-rich NES, Exportin 1/XPO1 (CRM1) has recently been functionally characterised in humans, Saccharomyces cerevisiae, and Schizosaccharomyces pombe ( Fornerod et al. 1997a ; Fukuda et al. 1997 ; Ossareh-Nazari et al. 1997 ; Stade et al. 1997 ). Exportin 1 had been identified previously in S. pombe as Crm1p (chromosome region maintenance protein 1) in mutants showing deformed chromosome domains ( Adachi & Yanagida 1989). Exportin 1 directly binds to proteins carrying a leucine-rich NES and to RanGTP in a co-operative manner ( Fornerod et al. 1997a ). This triple complex is exported from the nucleus due to the ability of Exportin 1 to interact with nuclear pore proteins ( Fornerod et al. 1997b ; Neville et al. 1997 ). The complex is dissociated in the cytoplasm when Ran-bound GTP is hydrolysed to GDP by RanGAP/RanBP1-activated Ran ( Bischoff & Görlich 1997). Nuclear export mediated by Exportin 1 is inhibited by the antifungal antibiotic leptomycin B (LMB) by covalent modification of Exportin 1 at a cysteine residue in the conserved central region (K udo et al. 1997 ; Kudo et al. 1999 ).
Nucleo-cytoplasmic partitioning of many transcription factors or enzymes modulating their activity is determined by the relative accessibility of an NES and/or a nuclear localisation signal (NLS). This is used as a control mechanism in several signal transduction pathways. Leucine-rich NESs have been identified in many regulatory proteins in animals and yeast, e.g. Cyclin B1, IκBα, NF-AT, and the S. pombe and S. cerevisiae AP-1-like transcription factors Pap1p and Yap1p ( Hood & Silver 1999; Mattaj & Englmeier 1998; Ossareh-Nazari et al. 1997 ). The presence of an intact NES in these proteins has been shown to be very important for their function. Leucine-rich NESs have also been identified in two regulators of the Ran GTPase cycle, RanBP1 and RanGAP1, which are crucial for the maintenance of the RanGTP gradient between the cytoplasm and the nucleus ( Feng et al. 1999 ; Richards et al. 1996 ; Zolotukhin & Felber 1997).
In plants, regulated nucleo-cytoplasmic partitioning has also been described for a number of proteins, e.g. COP1, PHYB and GBFs ( Nagatani 1998; Yamamoto & Deng 1999). However, currently nothing is known about the export of proteins. Here we demonstrate for the first time the existence of a nuclear export pathway in plants. We report the functional characterisation of a homologue of the export receptor Exportin 1 (XPO1/CRM1) from Arabidopsis thaliana, designated AtXPO1. AtXPO1 interacts with NES-containing proteins and with Ran in the yeast two-hybrid system. The interaction of AtXPO1 with the NESs is inhibited by LMB. In addition, we developed an assay for the functional characterisation of nuclear export activity by which we identified a leucine-rich NES motif in the Arabidopsis protein AtRanBP1a.
Isolation of a full-length cDNA clone coding for the Arabidopsis export receptor AtXPO1
In order to identify an Arabidopsis homologue of the export receptor XPO1/CRM1, we searched the Arabidopsis database for EST clones with significant homology to human CRM1. By screening a genomic library of Arabidopsis thaliana with a probe comprising part of the EST clone 219H5T7 (GenBank accession number N38124) we isolated two overlapping genomic clones of AtXPO1/AtCRM1 (Haasen et al. 1999). The corresponding full-length cDNA clone was obtained by PCR amplification from a cDNA library and verified by sequencing. The open reading frame of AtXPO1 comprises 3228 bp and encodes an acidic protein (calculated pI 5.3) of 1075 amino acids with a predicted molecular mass of 123 kDa. An in-frame stop codon was found 54 bp upstream of the translational start in the cDNA sequence.
The deduced amino acid sequence of AtXPO1 shares 42–50% identity and 54–62% similarity over its entire length with its functional human, S. cerevisiae, and S. pombe homologues ( Fig. 1). The similarity is most significant in the central part of the proteins (CCR; amino acid residues 419–603 of AtXPO1). The cysteine residue that determines LMB sensitivity (C529 in S. pombe Crm1p; Kudo et al. 1999 ) is aligning with a cysteine residue in the AtXPO1 sequence. Besides Exportin 1 from other organisms, the most closely related member among the Importin β family of transport receptors is human Exportin (tRNA), the export receptor for tRNAs (data not shown). The amino terminal 150 residues comprise the CRIME-domain that is conserved between all members of the Importin β family and contains a characteristic tryptophan residue at position 137 of AtXPO1 ( Fornerod et al. 1997b ; Görlich et al. 1997 ). This domain probably constitutes the conserved core of the Ran binding domain which has been mapped to the amino terminal 364 residues in Importin β ( Kutay et al. 1997 ).
AtXPO1 is a single copy gene and is expressed in all tissues
Arabidopsis genomic DNA digested with two different restriction endonucleases was subjected to Southern blot analysis ( Fig. 2a). The hybridising fragments detected under high stringency conditions using the full-length AtXPO1 cDNA as a probe were due to the presence of the respective internal restriction sites in the AtXPO1 gene. This argues for AtXPO1 being a single copy gene.
The steady state level of mRNA expression in different tissues of flowering Arabidopsis plants was assayed by Northern blot experiments ( Fig. 2b). A transcript with a size of approximately 4.2 kb, which corresponds to the expected size of the AtXPO1 mRNA, was detected in all tissues analysed. However, the level of expression differed significantly, with the highest level of expression found in stems, followed by medium levels in inflorescence and root tissues. In leaves of 6-week-old plants only a very low amount of AtXPO1 mRNA was detected.
Protein export of GFP fusion proteins in tobacco BY-2 protoplasts
For the analysis of the nuclear export of proteins containing a leucine-rich NES, we developed an in vivo assay system that uses transient transformation of tobacco BY-2 protoplasts with green fluorescent protein (GFP) reporter constructs that are able to shuttle between the cytoplasm and the nucleus. The reporter constructs are schematically represented in Fig. 3. The cytosolic enzyme chalcone synthase (CHS) was fused to GFP to increase the size of the reporter protein. Both the SV40 large T antigen NLS and the leucine-rich NES from the HIV-1 protein Rev (NESRev) were introduced into GFP-CHS, resulting in a reporter fusion protein with a calculated molecular weight of 73 kDa. Confocal laser scanning microscopy revealed that the green fluorescence of the GFP-NLS-CHS-NESRev fusion protein was almost equally distributed between the cytoplasm and the nucleus ( Fig. 4a). However, when two leucine residues of the NES of Rev were replaced by alanine (L81A, L83A; NES(–)Rev, Meyer & Malim 1994), the green fluorescence accumulated exclusively in the nucleus ( Fig. 4b). In contrast, when one lysine of the NLS was mutated to an asparagine (K128N; NLS(–); Lanford & Butel 1984), the green fluorescence was clearly excluded from the nucleus and was observed only in the cytoplasm ( Fig. 4c). The introduction of both a mutated NLS and a mutated NES resulted in a more or less equal distribution of the green fluorescence between the cytoplasm and the nucleus ( Fig. 4d). This clearly demonstrates that the NES of the heterologous protein Rev is a functional export signal in BY-2 protoplasts. The localisation of the green fluorescence of GFP-NLS(–)-CHS- NES(–)Rev is comparable to the basic GFP-CHS reporter that also showed green fluorescence in the cytoplasm and in the nucleus, presumably due to passive diffusion through the NPC ( Fig. 4f). Diffusion of the GFP-NLS(–)-CHS-NES(–)Rev reporter was also observed using different transfection techniques such as electroporation and particle bombardment (data not shown). The intracellular localisation of the GFP fusion proteins was confirmed by 4′,6′-diamidino-2-phenylindole (DAPI) staining using epifluorescence. Figure 4(n–p) shows co-localisation of nuclear green fluorescence with the DAPI-stained nucleus in a protoplast transformed with GFP-NLS-CHS-NES(–)Rev. In contrast, the cytosolic green fluorescence of the GFP-NLS(–)-CHS-NESRev fusion protein and DAPI staining of the nucleus were mutually exclusive ( Fig. 4q–s).
The Arabidopsis protein AtRanBP1a contains a leucine-rich NES
After having demonstrated the functionality of the NES of the heterologous protein Rev in BY-2 protoplasts, we were then interested in the localisation of the plant protein AtRanBP1a from Arabidopsis ( Haizel et al. 1997 ). For human and mouse RanBP1 homologues it has been demonstrated that they are excluded from the nucleus due to a leucine-rich NES located at the carboxy terminus of the protein. The amino acid sequence 176KVAEKLEA- LSVR189 of mouse RanBP1 functions as an NES and the mutation of two or three leucine residues within this sequence (L183A, L186A and V188A) results in nuclear accumulation of the protein ( Richards et al. 1996 ; Zolotukhin & Felber 1997). A sequence comparison of the region spanning the NESs of mouse and human RanBP1, three RanBP1 sequences from Arabidopsis, and the NESs of Rev and mouse PKIα is shown in Fig. 5. The amino acid residues L176, L179 and V181 of AtRanBP1a were prominent components of a putative leucine-rich NES, with respect to the spacing of the long chain hydrophobic residues in the NESs of mammalian RanBP1, Rev and PKIα.
When AtRanBP1a was fused to GFP, the green fluorescence was clearly excluded from the nucleus ( Fig. 4g). However, when the putative NES was disrupted by substituting the three prominent amino acid residues by alanine (L176A, L179A, V181A; AtRanBP1NES(–)), the green fluorescence was detected in the cytoplasm and nucleus ( Fig. 4h). When the carboxy terminus of RanBP1a containing the putative NES but not the Ran binding domain (NESAtRanBP1a) was fused to GFP-CHS, the reporter again was clearly excluded from the nucleus ( Fig. 4i). In contrast, when the mutated NES of AtRanBP1a (NES(–)AtRanBP1a) was fused to GFP-CHS, the green fluorescence was observed in the cytoplasm and nucleus ( Fig. 4k). We further analysed if the NES of RanBP1a was comparable to the NES of Rev in its export activity in BY-2 protoplasts. When the NES of AtRanBP1a was fused to the GFP-NLS-CHS reporter, the green fluorescence was almost equally distributed between the cytoplasm and nucleus ( Fig. 4l), as was the construct containing the NESRev. In contrast, when the mutated NES of AtRanBP1a was fused to GFP-NLS-CHS, the green fluorescence accumulated exclusively in the nucleus ( Fig. 4m), as it was observed for the construct containing the mutated NES of Rev.
The nuclear export of the NES of Rev is inhibited by LMB
The NES-dependent nuclear export of proteins that is mediated by Exportin 1 is inhibited by LMB in S. pombe and mammalian cells. LMB has been demonstrated to directly interact with human and S. pombe Exportin 1 in vitro but not with Exportin 1 from S. cerevisiae, which may explain why this yeast species is insensitive to LMB ( Fornerod et al. 1997a ). In AtXPO1, a cysteine residue was aligning with C529 of S. pombe Crm1p that is alkylated by LMB ( Fig. 1; Kudo et al. 1999 ). Using the in vivo assay we therefore investigated whether LMB also inhibits the nuclear export of NES-containing proteins in plant cells. As shown above, the green fluorescence in BY-2 protoplasts that had been transiently transformed with the GFP-NLS(–)-CHS-NESRev reporter was excluded from the nucleus, demonstrating the functionality of the NES of Rev. However, treatment with 2 μm LMB for 4 h resulted in a shift of the green fluorescence into the nucleus. Green fluorescence was now detected in the nucleus and cytoplasm ( Fig. 6a). We also investigated the inhibitory effect of LMB on the localisation of the GFP-NLS-CHS-NESRev reporter that showed an almost equal distribution between the two compartments. After treatment with 2 μm LMB for 4 h, the green fluorescence completely accumulated in the nucleus due to the functionality of the NLS and the block in nuclear export ( Fig. 6c). Control treatments with ethanol did not effect the localisation of the GFP reporter proteins ( Fig. 6b,d).
AtXPO1 interacts with the leucine-rich NES and with Ran in the two-hybrid system
In order to test whether AtXPO1 is able to bind to leucine-rich NESs, we assayed the interaction of AtXPO1 with several proteins containing an NES in the LexA two-hybrid system in yeast. The NES-containing substrates were fused to the B42-activation domain and were tested for their interaction with AtXPO1 fused to the LexA-binding domain in β-galactosidase assays on induction plates containing X-Gal ( Fig. 7a). Transactivation was observed when the AtXPO1 hybrid protein was co-expressed with the hybrid protein containing the NESRev but not with the hybrid protein containing the mutated NES of Rev ( Fig. 7a, panels 1 and 2) nor when the hybrid proteins of NESRev, NES(–)Rev, or AtXPO1 were co-expressed with the vector controls ( Fig. 7a, panels 3, 4 and 13). We also tested the interaction of AtXPO1 with the NES of mouse PKIα. When the hybrid protein of the carboxy terminus containing the NES (NESmPKIα) was co-expressed with the AtXPO1 hybrid protein, transactivation could be observed ( Fig. 7a, panel 5). No β-galactosidase activity was detected when the hybrid proteins of NESmPKIα or AtXPO1 were co-expressed with the vector controls ( Fig. 7a, panels 6 and 13). Besides the interaction of AtXPO1 with the NESs of heterologous proteins, we also investigated its interaction with the NES of the Arabidopsis protein AtRanBP1a. β-galactosidase activity was only detected when the hybrid protein containing the NESAtRanBP1a was co-expressed with the AtXPO1 hybrid protein but not with the hybrid protein containing the mutated NES of AtRanBP1a ( Fig. 7a, panels 7 and 8) nor when the hybrid proteins of NES AtRanBP1a, NES(–)AtRanBP1a, or AtXPO1 were co-expressed with the vector controls ( Fig. 7a, panels 9, 10 and 13).
Substrate binding or dissociation of Importin β-like proteins is controlled by their binding to the regulatory GTPase Ran ( Weis 1998). Therefore, we also tested the interaction of AtXPO1 and AtRan1 in the yeast two-hybrid system. No β-galactosidase activity was detected when either the LexA-binding domain-AtXPO1 hybrid protein or the B42-activation domain-AtRan1 hybrid protein were co-expressed with the vector controls ( Fig. 7a, panels 12 and 13). However, when the hybrid proteins of AtXPO1 and AtRan1 were co-expressed transactivation was observed ( Fig. 7a, panel 11).
For quantitative evaluation of the protein interactions β-galactosidase activities were measured in liquid culture assays ( Fig. 7b). The transactivation of the AtXPO1 hybrid protein with the NESRev hybrid protein was approximately 50 times higher than the one measured with the vector control ( Fig. 7b, bars 1 and 13), whereas the mutant NES(–)Rev hybrid protein showed about 20 times lower transactivation compared to the NESRev hybrid protein ( Fig. 7b, bar 2). The transactivation with the NESmPKIα hybrid protein was about 20-fold higher than the one measured with the vector control ( Fig. 7b, bars 5 and 13.) The β-galactosidase activity obtained with the NESAtRanBP1a hybrid protein was two- to threefold higher than the one measured with the NES(–)AtRanBP1a hybrid protein or with the vector control ( Fig. 7b, bars 7, 8 and 13). The co-expression of the AtXPO1 hybrid protein with the AtRan1 hybrid protein resulted in an approximately 15 times higher β-galactosidase activity compared to the vector control ( Fig. 7b, bars 7, and 13).
The two-hybrid assay revealed that AtXPO1 is functional in yeast and interacts with mammalian and plant proteins containing a leucine-rich NES. As mentioned above, S. cerevisiae Exportin 1 does not bind to LMB, which allowed us to investigate the effect of LMB on the two-hybrid interaction of AtXPO1 and leucine-rich NESs in β-galactosidase assays on induction plates containing X-Gal and a different concentration of LMB ( Fig. 7c). The transactivation observed when the hybrid proteins of AtXPO1 and NESRev, NESmPKIα or NESAtRanBP1a were co-expressed was progressively weakened at increasing concentrations of LMB. Full inhibition of β-galactosidase activity was detected at concentrations of 400 n m LMB for the NESRev, 200 n m LMB for the NESmPKIα and 100 n m LMB for the NESAtRanBP1a. A control interaction that did not involve AtXPO1 was not affected by LMB.
In this study we identified an Arabidopsis cDNA coding for a protein with 42–50% overall identity to the recently described human CRM1, S. cerevisiae Xpo1p/Crm1p, and S. pombe Crm1p. We designated it Exportin 1 (AtXPO1) according to Stade et al. (1997) , who proposed Crm1p to be renamed Exportin 1/XPO1p because of its function in nuclear export. AtXPO1 is very probably a single copy gene that is expressed in all Arabidopsis tissues but at very different levels. In fully developed leaves of 6-week-old plants only a very low amount of AtXPO1 mRNA was detected. Results which were obtained from human CRM1 suggest that CRM1 is a component of the nuclear architecture that is duplicated in the late cell cycle and that its expression is strictly regulated by the cell cycle ( Kudo et al. 1997 ). We suggest that the very low expression of AtXPO1 mRNA in leaves of 6-week-old Arabidopsis plants may be explained by a presumably low mitotic activity in this tissue.
We developed an in vivo system to investigate the export of proteins in tobacco BY-2 protoplasts. It also allows us to quickly identify putative NLSs and NESs in vivo by function. The assay is based on the localisation equilibrium of a GFP-CHS reporter protein fused to both an NLS and NES motif. At steady state the GFP-NLS-CHS-NES reporter presumably shuttled between the nucleus and cytoplasm as a result of permanent cycles of import and export by its functional localisation signals. When we mutated either one of the localisation signals of the reporter, competition between these two processes was disturbed, which allowed us to investigate the functionality of import and export signals. The mutation of the NLS shifted the localisation of the reporter protein towards the cytoplasm and clearly excluded it from the nucleus due to its functional NES. In contrast, the mutation of the NES of the reporter resulted in a completely nuclear localisation by its functional NLS. A reporter containing both a mutated NLS and a mutated NES was equally distributed between the cytoplasm and nucleus. The distribution by diffusion of a reporter that should be bigger than the exclusion limit of the NPC has also been reported for mammalian cells ( Kudo et al. 1998 ) and may result from the massive overexpression of the reporter constructs and the sensitivity of the reporter system revealing even very low diffusion activity. Furthermore, we demonstrated that the nuclear export of an NES-containing reporter protein in vivo is inhibited by LMB, providing strong evidence for a role of Exportin 1 in this transport process. In case of the reporter containing a functional NES but a mutated NLS, the LMB-induced redistribution of the green fluorescence to the nucleus was passive by diffusion. In contrast, the GFP-NLS-CHS-NES reporter actively accumulated in the nucleus due to its functional NLS and the inhibition of the export process, which directly confirmed the dynamic cycling of a protein containing both NLS and NES.
The NES of the heterologous HIV-1 protein Rev is also functional in plants, which clearly demonstrates the strong evolutionary conservation of the export machinery between animals and plants. Furthermore, we investigated the localisation of the Arabidopsis protein AtRanBP1a that is strictly cytosolic and observed that the leucine-rich NES motif located at the carboxy terminus is necessary and sufficient to account for nuclear exclusion of AtRanBP1a. The functionality of the NES is disrupted by the mutation of three amino acid residues to alanine (L176A, L179A, V181A). The leucine-rich motif of RanBP1a is a transferable NES because it is sufficient in re-directing a chimeric nuclear protein (GFP-NLS-CHS) to the cytoplasm. We showed that in BY-2 protoplasts the NES of AtRanBP1a is functionally indistinguishable from the NES of Rev.
We functionally characterised AtXPO1 as the export receptor by its interaction in yeast with proteins containing a leucine-rich NES and the GTPase Ran. AtXPO1 was interacting with AtRan1 in the two-hybrid system like other members of the Importin β family of transport receptors. We demonstrated that the leucine-rich NES of the plant protein AtRanBP1 as well as the NESs of the heterologous proteins Rev and mPKIα interacted with AtXPO1, which provided further confirmation of the conservation of the export machinery between animals and plants. From quantification of the two-hybrid interaction tests we suppose that the NES substrates may be recognised by AtXPO1 with different efficiencies in yeast. In the test for β-galactosidase activity on plates, the transformed yeast were grown under induction conditions for a much longer time period than in the liquid culture assay. This may explain differences in the strength of the interactions in the two assays, e.g. in the two-hybrid interaction of AtXPO1 and AtRan1. In humans and yeasts, XPO1 has been demonstrated to form a triple complex with the NES substrate and Ran ( Fornerod et al. 1997a ). This suggests that in the two-hybrid interaction the endogenous yeast Ran forms a complex with AtXPO1 and the NES substrate. Similarly, a yeast protein with a leucine-rich NES may be required for efficient interaction of AtXPO1 with AtRan1. The interactions of AtXPO1 with the animal and plant NESs were specifically inhibited by LMB, revealing that AtXPO1 is the direct target of LMB in contrast to S. cerevisiae Exportin 1.
In summary, we characterised for the existence of a nuclear export pathway for proteins carrying a leucine-rich NES in plants. Our results demonstrate that the machinery of nuclear export of proteins is functionally conserved among animals, yeasts and plants, and that active export from the nucleus may play an important role in the partitioning of regulatory proteins in plants.
Plant growth and cell culture
Arabidopsis plants ecotype Col-0 were grown to flowering stage in a growth chamber at 16/8 h light/dark cycle at a temperature of 22°C. Tobacco BY-2 cell suspension culture was grown as described by Merkle et al. (1996) .
Screening and isolation of genomic and cDNA clones
Genomic clones of AtXPO1 were isolated from lambda EMBL3 SP6/T7 library (CLONTECH, Heidelberg, Germany) constructed from genomic DNA of Arabidopsis thaliana ecotype Col-0 by plaque hybridisation using a DIG-labelled 285 bp fragment from EST clone 219H5T7 (GenBank accession number N38124) as a probe. Labelling and detection were performed according to the manufacturer’s instructions (Boehringer, Mannheim, Germany). DNA fragments from lambda clones were subcloned into the pBluescript-KSII vector (Stratagene, Amsterdam, the Netherlands) and sequenced. The cDNA sequence was amplified by PCR from an Arabidopsis thaliana ecotype Col-0 matchmaker cDNA library (CLONTECH).
Recombinant DNA techniques: Southern and Northern blot analysis
Recombinant DNA techniques were essentially performed as described by Sambrook et al. (1989) . All constructs obtained by PCR were sequenced to confirm their integrity.
Genomic DNA was prepared from 3-week-old Arabidopsis seedlings by the CTAB DNA isolation method according to Murray & Thompson (1980). For Southern blot analysis 20 μg of genomic DNA were digested with the restriction enzymes EcoRI or HindIII, separated on a 0.8% agarose gel and transferred to a Hybond N+ membrane (Amersham Pharmacia Biotech, Freiburg, Germany). Labelling of the probe and detection were performed as described for the ECL direct system provided by Amersham.
Total RNA was isolated as described by Logemann et al. (1987) . For Northern analysis 20 μg of total RNA were separated on a 1% agarose gel containing 2% formaldehyde and transferred to a positively charged nylon membrane (Boehringer). Labelling of the probes and detection were performed as described for the DIG DNA detection system by Boehringer.
Amino terminal GFP fusions were generated by cloning the respective cDNAs into plasmid mAV4 5′-GFP that was derived from the plasmid mAV4 kindly provided by S. Kircher (University of Freiburg, Freiburg, Germany). mAV4 5′-GFP contains mGFP4 followed by a multiple cloning site and the NOS terminator to generate in-frame GFP fusion proteins under the control of the constitutive 35S promoter. Parsley chalcone synthase (CHS; accession v01538) was cloned into mAV4 5′-GFP by introducing the required restriction sites (EcoRI/BamHI) by PCR. The SV40 large T-antigen NLS or NLS(–) were introduced 5′ of CHS by PCR using specific oligonucleotides coding for 126PKKKRKVEDP135 (forward primer EcoRI-NLS) or 126PKNKRKVEDP135 (forward primer EcoRI-NLS(–)/K128N). The Rev NES or Rev NES(–) were created 3′ of CHS by specific oligonucleotides coding for 73LQLP- PLERLTLD84 (reverse primer BamHI-NESRev) or 73LQLPPLERA- TAD84 (reverse primer BamHI-NES(–)Rev/L81A, L83A). AtRanBP1a and AtRanBP1aNES(–) were cloned into mAV4 5′-GFP by introducing the required restriction sites (EcoRI/BamHI) by PCR. AtRanBP1aNES(–) was generated by a PCR using overlapping oligonucleotides forward primer NES(–) 5′-GGC CTT GCT GAG AAA GCG ACT GTG GAA GAG ACA-3′ and reverse primer NES(–)/L176A, L179A, V181A 5′-CTC CTT CGC AGT CGC TTT CTC AGC AAG GCC AGC-3′. Deletion constructs of AtRanBP1a and AtRanBP1aNES(–) (amino acid residues 158–225), which were termed NESAtRanBP1a and NES(–)AtRanBP1a for convenience, were cloned downstream of amino acid residue 292 of CHS into mAV4 5′-GFP-CHS and mAV4 5′-GFP-NLS-CHS by introducing the required restriction sites (HindIII/BamHI) by PCR.
For two-hybrid interaction, the AtXPO1 cDNA was cloned into the vector pGILDA LexA by introducing the respective restriction sites (XmaI/NotI) by PCR. The cDNAs coding described above for NLS-CHS-NESRev and NLS-CHS-NES(–)Rev, NESAtRanBP1a and NES(–)AtRanBP1a were cloned into plasmid pB42 AD. The deleted cDNA coding for NESmPKIα (coding for amino acid residues 25–76 of mouse PKIα) was cloned into pB42 AD by restriction sites (EcoRI/BamHI) introduced by PCR amplification from a mouse cDNA library (CLONTECH).
PEG transformation of BY-2 protoplasts and microscopy
Dark-grown tobacco BY-2 cells were harvested from suspension culture medium 3 days after subculture by centrifugation at 400 g. Protoplasts were prepared as described by Merkle et al. (1996) . Transient transformation of the protoplasts by PEG was performed according to the protocol of Negrutiu et al. (1987) . Protoplasts were incubated at 26°C in the dark overnight. Efficiency of the transformation was about 3% of the living cells and was determined using a Axioplan microscope (Zeiss, Oberkochem, Germany) equipped with a 20× objective and epifluorescence filters. Excitation of GFP was performed with standard fluorescein isothiocyanate (FITC) filters. For staining of nuclei, transformed protoplasts were fixed as described by Smith et al. (1997) . 4′,6′-Diamidino-2-phenylindole (DAPI) was added at 100 ng ml−1 to visualise DNA (excitation at 395 nm). Epi- fluorescence images of GFP and DAPI using a 40× objective were taken with a CCD-videocamera (AVT Horn, Aalen/Zeiss). For confocal laser scanning microscopy, samples were directly examined under oil with a 63× objective and a DM RBE TCS4D microscope (Leica, Bensheim, Germany) equipped with an argon-krypton laser (excitation at 488 nm excitation, beam splitter at 510 nm, filter at 515 nm) using Leica Scanware. Analysis of the localisation of green fluorescence was qualitative only, representing approximately 80–100 transformed protoplasts analysed in at least five independent experiments. Treatment with LMB was performed by adding LMB from a stock (1 m m in ethanol) 16–18 h after transformation and incubated for up to 4 h. Mock treatment was with ethanol alone.
Vectors (pGILDA BD, pB42 AD) and the yeast strain (EGY48) were purchased from CLONTECH. The plasmid p8op-lacZ (CLONTECH) was introduced into EGY48 to give EGY48[p8op-lacZ]. Yeast transformation and analysis of interactions were performed according to the manufacturer’s protocol. For interaction tests on induction plates containing X-Gal (5-Bromo-4-chloro-3-indolyl β- d-galactopyranoside), yeasts were grown for 40–48 h at 30°C in liquid culture assays using ONPG (o-Nitrophenyl β- d-galactopyranoside) as substrate, induction was for 4 h at 30°C. The control (LexA-binding domain-murine p53 and B42-activation domain-SV40 large T-antigen hybrid proteins) for inhibition experiments with LMB was provided by the manufacturer (CLONTECH). LMB was directly included in the plates.
We wish to thank Elke Faller for excellent technical help, Gabor Igloi and Elfi Schiefermayr for excellent sequencing service, Stefan Kircher for kindly providing plasmid mAV4 and Barbara Wolff (Novartis) for providing leptomycin B. We are grateful to Chris Lundberg for critical reading of the manuscript. This work was supported in part by a grant from SFB 388 to T.M. and G.N.