Nucleocytoplasmic shuttling of heterodimeric splicing factor U2AF.

The U2 small nuclear ribonucleoprotein auxiliary factor (U2AF) is a heterodimeric splicing factor composed of 65-kDa (U2AF(65)) and 35-kDa (U2AF(35)) subunits. The large subunit of U2AF recognizes the intronic polypyrimidine tract, a sequence located adjacent to the 3' splice site that serves as an important signal for both constitutive and regulated pre-mRNA splicing. The small subunit U2AF(35) interacts with the 3' splice site dinucleotide AG and is essential for regulated splicing. Like several other proteins involved in constitutive and regulated splicing, both U2AF(65) and U2AF(35) contain an arginine/serine-rich (RS) domain. In the present study we determined the role of RS domains in the subcellular localization of U2AF. Both U2AF(65) and U2AF(35) are shown to shuttle continuously between the nucleus and the cytoplasm by a mechanism that involves carrier receptors and is independent from binding to mRNA. The RS domain on either U2AF(65) or U2AF(35) acts as a nuclear localization signal and is sufficient to target a heterologous protein to the nuclear speckles. Furthermore, the results suggest that the presence of an RS domain in either U2AF subunit is sufficient to trigger the nucleocytoplasmic import of the heterodimeric complex. Shuttling of U2AF between nucleus and cytoplasm possibly represents a means to control the availability of this factor to initiate spliceosome assembly and therefore contribute to regulate splicing.

The U2 small nuclear ribonucleoprotein auxiliary factor (U2AF) is a heterodimeric splicing factor composed of 65-kDa (U2AF 65 ) and 35-kDa (U2AF 35 ) subunits. The large subunit of U2AF recognizes the intronic polypyrimidine tract, a sequence located adjacent to the 3 splice site that serves as an important signal for both constitutive and regulated pre-mRNA splicing. The small subunit U2AF 35 interacts with the 3 splice site dinucleotide AG and is essential for regulated splicing. Like several other proteins involved in constitutive and regulated splicing, both U2AF 65 and U2AF 35 contain an arginine/ serine-rich (RS) domain. In the present study we determined the role of RS domains in the subcellular localization of U2AF. Both U2AF 65 and U2AF 35 are shown to shuttle continuously between the nucleus and the cytoplasm by a mechanism that involves carrier receptors and is independent from binding to mRNA. The RS domain on either U2AF 65 or U2AF 35 acts as a nuclear localization signal and is sufficient to target a heterologous protein to the nuclear speckles. Furthermore, the results suggest that the presence of an RS domain in either U2AF subunit is sufficient to trigger the nucleocytoplasmic import of the heterodimeric complex. Shuttling of U2AF between nucleus and cytoplasm possibly represents a means to control the availability of this factor to initiate spliceosome assembly and therefore contribute to regulate splicing.
Pre-mRNA splicing is an essential step in eukaryotic gene expression that is often regulated in a cell type-specific or developmental manner. The splicing reaction takes place in the spliceosome, a multicomponent RNA-protein complex containing five uracil-rich small nuclear ribonucleoproteins (snRNP) 1 and 50 -100 non-snRNP protein splicing factors (1,2). To ensure the precise excision of introns, components of the spliceosome recognize weakly conserved sequences in the pre-mRNA (3). The decision to splice a given pre-mRNA and the selection of splice sites takes place during the first stages of spliceosome assembly. These early steps are therefore crucial for splicing regulation (3)(4)(5). The stable association of U2 snRNP with the 3Ј region of the intron (3,6,7) requires an auxiliary factor, U2AF (8). U2AF consists of two subunits that form a stable heterodimer (9). U2AF 65 binds directly to the intronic polypyrimidine tract and is essential for splicing (10). U2AF 35 recognizes the 3Ј splice site dinucleotide AG (11)(12)(13) and is required for the regulated splicing of a subset of pre-mRNAs (14).
Regulatory mechanisms underlying the selection of alternative exons remain poorly defined (5,15). A significant advance in understanding splice site recognition was provided by the observation that exon sequences also play a critical role in splicing. Constitutive as well as regulated exon splicing enhancers have been identified (16,17). These sequences contain binding sites for SR proteins (18,19), a family of essential splicing factors with a modular structure consisting of one or two RRMs and a C-terminal arginine/serine (RS)-rich domain, involved in protein-protein interactions (20,21). According to a current model, SR proteins may activate splice site choice by binding to an enhancer sequence and recruiting the splicing machinery to the adjacent intron (16,17,22), acting as bridging factors between the 5Ј and 3Ј splice sites (23)(24)(25).
The subcellular distribution of splicing factors has been shown to be complex and dynamic, and it is increasingly believed that it may play a role in the regulation of splicing. Splicing factors accumulate in the nucleus, concentrating in structures commonly designated as nuclear speckles (26). Several lines of evidence suggest that nuclear speckles represent sites where splicing factors accumulate when they are not actively engaged in the splicing reaction. By an as yet unknown mechanism, splicing factors are recruited from the speckles to the sites of transcription, where splicing of most introns takes place (27).
A major determinant of the subcellular localization of SR proteins is the RS domain. The RS domain can function as a NLS, and it is also thought to be responsible for targeting the proteins to the nuclear speckles (28 -30). Changes in the phosphorylation of the RS domain are sufficient to alter the proteinprotein interactions established by SR proteins (31,32) and to modulate their subcellular distribution (33,34). Two novel transport receptors, termed transportin-SR and transportin-SR2, bind specifically to the phosphorylated RS domain of SR proteins (35,36). Although splicing factors are predominantly detected in the nucleus, at least some members of the SR protein family have the ability to shuttle between the nucleus and the cytoplasm (37). The RS domain has been shown to be necessary, but not sufficient, for export of the shuttling SR proteins to the cytoplasm (30). Based on the observation that SR proteins can influence splice site selection in a concentrationdependent manner (19,38,39), it has been proposed that transport to the cytoplasm may provide a way to control the nuclear availability of these proteins and, hence, contribute to regulate splicing (37,40).
Although they are not members of the SR protein family, both U2AF 65 and U2AF 35 contain an RS-like domain (10,41). The RS domain of U2AF 65 is located in the N-terminal region of the protein and has never been shown to mediate the establishment of protein-protein interactions. Rather, it facilitates base pairing between the U2 small nuclear RNA and the branch point (42) and plays a role in the targeting of this protein to active splicing sites (43). In human U2AF 35 , the RS domain is C-terminal and interacts with the RS domain of SR proteins (20).
In the present work we analyzed the role of the RS domain in the subcellular distribution of the heterodimeric splicing factor U2AF. Despite the fact that RS residues in U2AF are more dispersed and have less periodicity than in SR proteins, the data presented here indicate that they play a similar role in subcellular localization. We show that the RS domain on either U2AF 65 or U2AF 35 acts as a nuclear and subnuclear localization signal. We also show that both U2AF 65 and U2AF 35 shuttle continuously between the nucleus and the cytoplasm by a mechanism that involves carrier receptors and is independent from binding to mRNA. Finally, we provide evidence suggesting that the presence of an RS domain in either U2AF subunit is sufficient to trigger the nucleocytoplasmic import of the heterodimeric complex. Taken together these results suggest that nucleocytoplasmic transport of RS-containing splicing factors may represent a common pathway to control the availability of these proteins to initiate spliceosome assembly and, therefore, contribute to regulate splicing.

EXPERIMENTAL PROCEDURES
Cell Culture, Heterokaryon Assays, and Fluorescence Microscopy-Human HeLa and murine NIH 3T3 cells were cultured as monolayers in modified Eagle's medium and Dulbecco's modified Eagle's medium, respectively (Life Technologies, Inc.). Drosophila SL2 cells were grown in suspension in Schneider's medium at room temperature. All media were supplemented with 10% fetal calf serum (Life Technologies, Inc.). Heterokaryons were obtained as previously described (44). Briefly, 1.5 ϫ 10 6 3T3 cells or 2 ϫ 10 7 SL2 cells were plated over subconfluent HeLa cells grown on coverslips on P35 Petri dishes. 3T3 cells were allowed to adhere for 3 h at 37°C, 5% CO 2 in the presence of 20 g/ml emetine. Adhesion of SL2 cells to the coverslips was induced by incubation in serum-free medium for 15 min at 29°C, 5% CO 2 . The cells were then placed in complete modified Eagle's medium with 20 g/ml emetine and incubated for an additional 3 h. To induce cell fusion, coverslips were rinsed in phosphate-buffered saline (PBS) and placed on a drop of polyethylene glycol (PEG 1500; Roche Molecular Biochemicals) for 2 min. The coverslips were then washed in PBS and further incubated in culture medium at 37°C for 4 h (for HeLa-3T3 heterokaryons) or at 29°C for 7 h (for HeLa-SL2 heterokaryons) Cells on coverslips were briefly rinsed in PBS, fixed in 3.7% formaldehyde (freshly prepared from paraformaldehyde) diluted in PBS for 10 min at room temperature, and washed in PBS. The cells were then permeabilized with 0.5% Triton X-100 in PBS for 10 min at room temperature and washed in PBS. Immunofluorescence and confocal microscopy were performed as described (44). Endogenous U2AF 65 was detected using monoclonal antibody MC3 (43), and HA-tagged constructs were visualized using monoclonal antibody 12CA5-I (Berkeley Antibody Co., Richmond, VA). Additionally, the following antibodies were used: monoclonal antibody anti-hnRNP C 4F4, (45) and monoclonal antibody anti-SC35 splicing factor (46).
Construction and Expression of Fusion Proteins-The HA-U2AF 65 ⌬1-150 construct, containing U2AF 65 amino acids 151 to 475, was obtained by polymerase chain reaction amplification of pBluescript-␤globin-U2AF 65 (10) with the following primers: primer 1, 5Ј-CC GGT ACC GAC ACC ATG TAC GAC TAC GAC GTC CCT GAT TAT GCA CGC CTC TAC GTG GGC-3Ј; primer 2, 5Ј-GG ACG CGT CTA CCA GAA GTC CCG GCG GTG-3Ј (Life Technologies, Inc.). Primer 1 con-tains a KpnI restriction site and encodes an HA tag in-frame with the U2AF 65 sequence from nucleotide 504, corresponding to the beginning of the RNA binding domain. Primer 2 contains an MluI restriction site and a Stop codon. After treatment to obtain phosphorylated blunt ends, the amplification product was digested with KpnI and cloned in the pCMV5 expression vector (47) digested with KpnI and SmaI. The ligated product was amplified and sequenced.
Green fluorescent protein (GFP) fusion constructs were obtained by restriction digestion and subcloning into the appropriate pEGFP-C vector (CLONTECH, Palo Alto, CA). GFP-U2AF 65 was obtained by digestion of pBluescript-␤globin-U2AF 65 with NotI and EcoRI and cloning in pEGFP-C2 digested with BglII and EcoRI. NotI and BglII ends were filled in to generate blunt ends. GFP-RS-U2AF 65 , containing U2AF 65 amino acids 1 to 55, was obtained by digestion of GFP-U2AF 65 with SmaI and religation. The U2AF 65 RRM1 mutant was obtained by polymerase chain reaction-based site-directed mutagenesis of pGEX-2T GST-U2AF 65 ⌬1-94 (42) as described (48). Mutations introduced at the RNP-1 octamer sequence of the first RRM consisted of substitution of lysine 195 by asparagine, phenylalanine 197 by alanine, and phenylalanine 199 by alanine and were obtained by introducing the following

FIG. 1. The RS domain of U2AF is a nuclear localization signal.
HeLa cells were either immuno-labeled with monoclonal antibody MC3 directed against human U2AF 65 (A) or transfected with GFP-U2AF 65 (B) and GFP-U2AF 35 (C). Note that both U2AF subunits are exclusively detected in the nucleus, excluding the nucleoli; the proteins are diffuse in the nucleoplasm with additional concentration in speckles (arrowheads). Expression of GFP alone results in diffuse staining throughout the nucleus and the cytoplasm (D) and does not alter the accumulation of splicing factors in nuclear speckles, as shown by the anti-SC35 monoclonal antibody (G). Chimeric proteins resulting from fusion of GFP to the RS domain of either U2AF 65 or U2AF 35 are exclusively detected in the nucleus (E and F). Labeling with anti-SC35 monoclonal antibody shows that these proteins accumulate in nuclear speckles (H and I, arrowheads). In contrast with full-length proteins, these polypeptides are localized in the nucleolus (E and F, arrows). Bar, 10 m. J, immunoblotting of transfected HeLa cell extracts with an anti-GFP antibody shows that a single band is observed for each GFP construct. Expected molecular masses are GFP ϳ27 kDa, GFP-U2AF 35 ϳ62 kDa, GFP-U2AF 35 ⌬RS ϳ57 kDa, GFP-RS35 ϳ33 kDa, GFP-U2AF 65 and GFP-U2AF 65 RRM1 mut ϳ92 kDa, GFP-RS65 ϳ33 kDa. wt, wild type. nucleotide changes: wild type sequence AAG AAT TTT GCC TTT TTG GAG TTC; RRM1 mutant AAT AAT GCT GCC GCT TTG GAG TTC. GFP-U2AF 65 RRM1mut was obtained by digesting both GFP-U2AF 65 and the pGEX-2T GST-U2AF 65 ⌬1-94 RRM1mut with SgrAI (amino acid 136) and EcoRI and replacing the C-terminal region of the wild type GFP construct from amino acid 136 with the corresponding region of the mutant.
DNA for transfection assays was purified using the Qiagen plasmid DNA midi-prep kit (Qiagen GmbH, Hilden, Germany). HeLa cells were transiently transfected with FuGene6 Reagent (Roche Molecular Biochemicals) using 1 g of DNA. For double transfections, 0.5 g of each DNA was used. Cells were analyzed at 16 -24 h after transfection. Western blot analysis of transfected cells was performed for all constructs as described (43)  RNA Binding Assays-Wild type and mutant RRM1mut proteins were expressed and purified from Escherichia coli as fusions with glutathione S-transferase from amino acids 95 to 475 of U2AF 65 . These proteins lack the arginine-serine (RS)-rich domain, which was deleted to facilitate the analysis of RNA binding exclusively through the RRM motifs of the protein and minimize nonspecific interactions (50). The purified proteins were incubated with 5 fmol (10,000 cpm) of a radioactively labeled RNA corresponding to a region of Drosophila sl-2 pre-mRNA containing a stretch of 11 consecutive uridines, which constitute a high affinity binding site for U2AF 65 (10,(51)(52)(53). After incubation for 15 min on ice, the mixtures were fractionated by electrophoresis on nondenaturing polyacrylamide gels as described (54).

RESULTS
The RS Domain on Either U2AF Subunit Functions as a NLS-The RS domain of SR splicing factors has been shown previously to function as a NLS (28,29) and to be necessary for export of the shuttling SR proteins to the cytoplasm (30). To test whether the RS domain of the U2AF polypeptides plays a similar role, we tagged the two U2AF subunits with GFP and examined their subcellular localization (Fig. 1). GFP expression levels vary widely in the transfected population, with very high expressors showing aberrant protein distribution patterns (data not shown). Therefore, we selected for analysis cells that expressed the GFP fusion proteins close to endogenous levels. As depicted in Fig. 1, both GFP-U2AF 65 (B) and GFP-U2AF 35 (C) show a distribution similar to that of endogenous U2AF 65 detected using a specific antibody (A). In all cases the staining is diffuse throughout the nucleoplasm, excluding the nucleoli, with a higher concentration in speckles, as confirmed by double-labeling with SC35 monoclonal antibody (data not shown). Contrasting with the results previously obtained using purified polyclonal antibodies raised against synthetic peptides of U2AF 65 and U2AF 35 (41), we observe no evidence for the localization of either U2AF subunit in the coiled body. Considering that similar results were obtained with an anti-U2AF 65 monoclonal antibody (43) and epitope-tagged U2AF 65 (43) and U2AF 35 (data not shown) as well as GFP-U2AF 65 and GFP-U2AF 35 fusion proteins (data not shown), it is most likely that U2AF does not associate with coiled bodies.
As expected, expression of GFP alone produces a uniform staining in both the nucleus and the cytoplasm (Fig. 1D) F). The cells were treated with a protein synthesis inhibitor (20 g/ml emetine) for 3 h and then incubated in medium containing emetine for another 3 h at 37 or 4°C, as indicated. At 37°C all proteins are exclusively detected in the nucleus due to active import. At 4°C, GFP-NLS was detected in the cytoplasm because it passively diffused out of the nucleus while active nuclear import was blocked; in contrast, U2AF 65 and GFP-U2AF 35 remained restricted to the nucleus, indicating that these proteins do not leak to the cytoplasm by passive diffusion. Bar, 10 m. sufficient to target the chimeric protein to the nucleus and to nuclear speckles (Fig. 1, E-H and F-I), demonstrating that this domain functions as both a NLS and a speckle targeting signal. Surprisingly, the GFP-RS chimeras also accumulate in the nucleoli. Since these are small proteins, they probably enter the nucleolus by diffusion and are retained by nonspecific binding of the RS domain to rRNA (50).
Immunoblotting analysis of transfected cells using an anti-GFP antibody (Fig. 1J) shows that a single protein band is observed for each GFP construct used in this work. This observation excludes the possibility that truncated GFP chimeras are contributing to the observed localization patterns. The GFP-RS35 construct migrates with an apparent weight that is slightly heavier than expected, possibly due to phosphorylation of this domain.
Both U2AF Subunits Shuttle Continuously between Nucleus and Cytoplasm-Having established that the RS domain of U2AF 65 and U2AF 35 plays a role in nuclear import similar to the RS domain of shuttling SR proteins, we next asked whether U2AF is also exported from the nucleus to the cytoplasm. To address this question, we have made use of an interspecies heterokaryon assay (55,56). The migration of human U2AF 65 was monitored in human/Drosophila heterokaryons produced by polyethylene glycol-induced fusion of HeLa and SL2 cells (Fig. 2). Heterokaryons were kept in culture for 4 -7 h in the presence or absence of the protein synthesis inhibitor emetine. Specific identification of human U2AF 65 was possible because the monoclonal antibody MC3 does not cross-react with Drosophila proteins ( Fig. 2A; Ref. 43). As a control, heterokaryons were double-labeled with a monoclonal antibody specific for human hnRNP C, a protein that is always restricted to the nucleus (56). In the absence of emetine, both human U2AF 65 and hnRNP C proteins synthesized in the cytoplasm of heterokaryons progressively accumulated in the Drosophila nuclei (data not shown). In contrast, when protein synthesis was inhibited, no human hnRNP C protein could be detected in Drosophila nuclei, whereas human U2AF 65 was readily visible (Fig. 2, B and C). From this we conclude that during the course of the experiment human U2AF 65 molecules migrated from the HeLa nucleus to the cytoplasm of the heterokaryon and from there they were imported into the Drosophila nucleus. Thus, U2AF 65 shuttles between nucleus and cytoplasm.
To determine whether the small subunit of U2AF also shuttles, we performed a HeLa/3T3 cell heterokaryon assay. HeLa cells expressing GFP-U2AF 35 were fused with mouse 3T3 cells in the presence of the protein synthesis inhibitor emetine. As depicted in Fig. 2, D-F, GFP-U2AF 35 migrates to the 3T3 nuclei, whereas the nonshuttling protein hnRNP C remains exclusively detected in the human nuclei. Thus, GFP-U2AF 35 is exported from the nucleus to the cytoplasm.
Given that nuclear pore complexes allow the diffusion of molecules up to 60 -70 kDa (57), the results obtained with interspecies heterokaryons could be due to passive leakage through the pores. To investigate whether U2AF is exported by diffusion or requires a receptor-mediated energy-driven mechanism, we performed a temperature-shift assay, as described by Michael et al. (58). The rational for the assay is that at 4°C both receptor-mediated nuclear import and export are blocked, whereas diffusion is unaffected. As a positive control we fused GFP to a classical NLS (the SV40 large T antigen NLS) and expressed it in HeLa cells (Fig. 3). Because GFP is a small protein (ϳ27 kDa), it may diffuse across the pores, but due to the presence of an NLS, it is actively imported. At physiological temperature in the presence of a protein synthesis inhibitor, GFP-NLS diffuses out of the nucleus. However, because it is rapidly re-imported, the protein is exclusively detected in the nucleus (Fig. 3A). In contrast, at low temperature in the absence of protein synthesis, GFP-NLS continues to leak out of the nucleus, and because NLS-mediated import is blocked, the protein progressively accumulates in the cytoplasm (Fig. 3B). Contrasting with the results obtained with GFP-NLS, both U2AF 65 and GFP-U2AF 35 are exclusively localized in the nucleus at either 37 or 4°C (Fig. 3, C-F). This implies that transport of U2AF to the cytoplasm involves an energy-dependent carrier-mediated pathway.
RNA Binding Is Dispensable for U2AF Shuttling-Two different mechanisms may account for the shuttling of an RNA-binding protein. Either it is a passive cargo of the mature mRNAs in transit to the cytoplasm, or it is actively exported through the interaction with a transporter molecule. To investigate if the migration of U2AF to the cytoplasm depends on RNA transport, HeLa cells were incubated with the transcription inhibitor actinomycin D for 3 h before fusion to Drosophila SL2 cells. As shown in Fig. 4, A-C, export of U2AF 65 continues under these conditions. A similar result was observed when HeLa cells expressing GFP-U2AF 35 were incubated with actinomycin D for 3 h before fusion to mouse 3T3 cells (Fig. 4, D-F). Thus, nuclear export of both U2AF subunits continues in the absence of RNA synthesis, suggesting that mRNA traffic to the cytoplasm is dispensable for U2AF shuttling. A direct prediction from this model is that binding of U2AF to RNA is not required for shuttling. To address this idea, we constructed a GFP fusion U2AF 65 RNA binding mutant (GFP-U2AF 65 RRM1mut) (Fig. 5). To obtain a minimal mutant U2AF 65 with compromised RNA binding properties, amino acid residues 195 (Lys), 197 (Phe), and 199 (Phe) were substituted by asparagine (195), alanine (197), and alanine (199). These residues were predicted to form part of the RNP-1 consensus sequence of U2AF 65 RRM1 (10), a prediction then confirmed by NMR analysis (59). Mutation of these residues in RNP-1 is predicted to have very detrimental effects in RNA binding by RRM1 (reviewed by Ref. 60), which in turn makes critical contributions to the overall RNA binding affinity of U2AF 65 (10). Consistent with this, Fig. 5G shows that RNA binding was impaired in this mutant when compared with the wild type protein. Both the wild type construct, GFP-U2AF 65 , and the mutant, GFP-U2AF 65 RRM1mut, were analyzed in heterokaryon assays (Fig. 5, A  and B, D and E). As depicted in Fig. 5, D and E, this mutant retains the ability to shuttle. Incubation of cells at 4°C confirms that neither GFP-U2AF 65 nor GFP-U2AF 65 RRM1 chimeras leak passively to the cytoplasm (Fig. 5, C and F). A similar result was obtained using HA-tagged U2AF 65 ⌬RNP2,3, a deletion mutant that covers the RNP consensus 2 and 3 of U2AF, showing impaired RNA binding activity (data not shown). Taken together, these results strongly suggest that U2AF shuttling is independent from RNA binding.
RS Domain Requirements for Nucleocytoplasmic Transport of U2AF-The finding that the RS domain of shuttling SR splicing factors is necessary for export of these proteins to the cytoplasm (30) prompted us to investigate the RS domain requirements for nucleocytoplasmic transport of U2AF. GFP chimeras were constructed using either the RS domain alone (GFP-RS-U2AF 65 and GFP-RS-U2AF 35 ), or deletion mutants  A and D). As a control, the cells were immunostained with anti-hnRNP C antibodies (B and E). To exclude that nuclear export of U2AF 65 is due to diffusion, cells expressing these chimeras were treated with emetine for 3 h and then incubated at 4°C for another 4 h (C and F). Bar, 10 m. RNA binding properties of U2AF 65 wild type (WT) and mutant RRM1mut proteins were analyzed in a gel retardation assay (G). The apparent dissociation constant (K d ) was estimated from the protein concentration that results in retardation of 50% of the RNA. The results show that although the wild type protein has an apparent K d of 5 ϫ 10 Ϫ9 , almost no binding was detected at a 20-fold higher concentration for the RRM1mut protein. Titration over a wider range of concentrations indicated that the difference in apparent affinity between the two proteins was at least 100-fold. Similar differences were observed with different uridine-rich RNA substrates both in the presence and in the absence of the RS domain (data not shown). Molar concentrations of protein in the different lanes: lane 1, no protein; lanes 2 and 6, 10 Ϫ7 ; lanes 3 and 7, 3 ϫ 10 Ϫ8 ; lanes 4 and 8, 10 Ϫ9 ; lanes 5 and 9, 3 ϫ 10 Ϫ9 . spanning the RS domain of both U2AF subunits (HA-U2AF 65 ⌬RS and GFP-U2AF 35 ⌬RS). HeLa cells expressing GFP-RS-U2AF 65 or GFP-RS-U2AF 35 were fused to murine 3T3 cells in the presence of the protein synthesis inhibitor emetine. At ϳ4 h after fusion, both chimeras were readily detected in the mouse nuclei, confirming their ability to be exported from the HeLa nuclei (Fig. 6, A and B, D and E). However, when these cells are treated with emetine and incubated at 4°C, the two GFP fusion proteins accumulate in the cytoplasm (Fig. 6, C and  F). Thus, the U2AF RS domains fail to retain the GFP chimeras in the nucleus, and as a consequence of their small size, both proteins diffuse passively through the pores to the cytoplasm. In contrast, after deletion of the RS domain in either U2AF 65 or U2AF 35 , the proteins remain in the nucleus at 4°C but are exported at 37°C (Fig. 6, G-L). This implies that, unlike the situation observed in SR proteins, in U2AF, the RS domain is dispensable for nuclear export. However, because U2AF is a heterodimer and the region of interaction between U2AF 65 and U2AF 35 does not overlap with the RS domain, there is the possibility that the presence of this domain in only one of the two subunits is sufficient to direct the nucleocytoplasmic shuttling of the complex. To address this idea, two additional deletion mutants of U2AF 65 were analyzed (Fig. 7). HA-U2AF 65 ⌬84 -150 has a deletion of amino acids 84 -150, spanning the region of interaction with U2AF 35 (43), and HA-U2AF 65 ⌬1-150 has a deletion of amino acids 1-150, spanning both the region of interaction with U2AF 35 and the RS domain.
The HA-U2AF 65 ⌬84 -150 protein is exclusively localized in the nucleus and retains shuttling ability ( Fig. 7A and data not shown), whereas HA-U2AF 65 ⌬1-150 shows significant accumulation in the cytoplasm (Fig. 7B). Thus, when U2AF 65 cannot interact with U2AF 35 , its RS domain becomes important for nuclear import.
Consistent with the view that U2AF is imported to the nucleus as a heterodimer requiring only one RS domain, overexpression of either U2AF 65 ⌬RS or U2AF 35 ⌬RS results in significant accumulation in the cytoplasm (Fig. 8, A and E). As expected, these two deletion mutants become exclusively localized in the nucleus of cells that simultaneously overexpress the wild type forms of U2AF 35 or U2AF 65 , respectively (Fig. 8, B and F). In contrast, overexpressed U2AF 65 ⌬RS remains in the cytoplasm of cells that simultaneously overexpress either wild type U2AF 65 (Fig. 8C) or the splicing factor SC35, which contains an RS domain but does not interact with U2AF 65 (Fig.  8D). Furthermore, overexpressed U2AF 35 ⌬RS remains in the cytoplasm of cells that simultaneously overexpress the deletion mutant HA-U2AF 65 ⌬84 -150, which lacks the region of interaction with U2AF 35 (Fig. 8G) or U2AF 65 ⌬RS (Fig. 8H).
In conclusion, these data support the view that the presence of an RS domain in either U2AF subunit is sufficient to target the heterodimeric complex into the nucleus. Given that the RS domain has been implicated in both nuclear import and export of SR proteins, it is most likely that the same model applies to export of U2AF from the nucleus to the cytoplasm.

DISCUSSION
In this work we show that the U2AF splicing factor shuttles between the nucleus and the cytoplasm by a carrier-mediated pathway that is independent from mRNA traffic. Several other proteins associated with pre-mRNA processing exhibit shuttling activity, namely hnRNP A1, A2, D, E, I, and K proteins (56,58,61) and the SR proteins ASF/SF2, SRp20, and 9G8 (37). Shuttling of hnRNPs most likely contributes to export of mRNA to the cytoplasm (58,62,63), whereas SR proteins appear to leave the nucleus as a cargo bound to the mRNAs (64). In fact, the RS domain is necessary for export of SR proteins, but unlike the M9 signal of hnRNP A1 (58), it is not capable of promoting export of an unrelated protein (37). Rather, the RS domain has to be coupled to a RNA binding domain to give rise to a shuttling protein (37). Consistent with the view that SR proteins piggyback on mRNA to exit the nucleus, several members of this family have been shown to bind to exon sequences (18,19) and to accompany the mRNA molecules in transit to the nuclear pores (65). The association of SR proteins with spliced mRNA could therefore serve as a guide for mature RNAs to cross the nuclear pores (66). Another possibility to be considered is that SR proteins remain bound to the mRNAs in the cytoplasm and play a role as regulators of mRNA stability or translation. Since the transport of SR proteins into the nucleus is mediated by specific import receptors (35,36) and regulated by phosphorylation and transcriptional activity (37,30), shuttling could at the same time provide a mechanism to control nuclear availability of splicing factors in response to external signals. This would allow a rapid tuning of splicing activity. Consistent with this view, cytoplasmic accumulation of the shuttling protein hnRNP A1 has been shown to occur in response to cellular stress and to correlate with changes in alternative splicing (40).
In contrast with SR proteins, both U2AF subunits have been shown to bind only to intronic sequences of pre-mRNA during the early steps of spliceosome assembly (10 -13) and to be later replaced by the U5 snRNP (67). Thus, it is unlikely that U2AF exits the nucleus as a passive cargo bound to mature mRNA. In agreement with this view, our data indicate that transport of U2AF to the cytoplasm does not require binding to mRNA and continues in the absence of mRNA traffic.
What function could U2AF shuttling perform? Very recent data indicate that U2AF 35 interacts with the mRNA export factor TAP and associates with spliced messenger RNP complexes in the nucleus, suggesting that the factor participates in coupling of splicing with export of mRNA to the cytoplasm. 2 Another possibility is that U2AF plays an as yet undetermined role in the cytoplasm. Recent observations suggest that the supply of U2AF is regulated in the cell (68,69). In this context, shuttling could also represent a rapid means of controlling the level of U2AF available in the nucleus for spliceosome assembly. A mechanism of U2AF regulation by retention has been recently proposed by MacMorris et al. (70). The Caenorhabditis elegans U2AF 65 homologue gene gives rise to two alternatively spliced mRNAs, one encoding the U2AF protein and another containing an exon with premature stop codons. This alternative mRNA is retained in the nucleus and contains 10 repeats of the 3Ј splice site consensus sequence. MacMorris et al. (70) propose that U2AF is sequestered by these repeats and consequently becomes unavailable for splicing.
A unique feature of the splicing factor U2AF is that it is composed of two subunits that bind tightly to each other and form a stable complex (9). Copurification of both subunits suggests that at least 90% of the protein present in HeLa nuclear extracts is in a heterodimeric complex (14). The interaction between the two subunits has also been shown to be essential in vivo (71). The data presented here indicate that deletions of the RS domain of either U2AF subunit do not alter the distribution and shuttling capacity of the protein complex. Similar results are obtained after deletion of any of the other functional domains of U2AF 65 , suggesting that both nuclear import and export of this protein complex are mediated by redundant signals. Furthermore, when the two subunits cannot interact, each RS domain becomes important for nuclear import. Based on these results we propose that U2AF enters and leaves the nucleus as a heterodimer, with the RS domain of each subunit serving as a redundant nuclear import and export signal. In good agreement with this view, deletion of the RS domain in either U2AF subunit is viable in vivo, whereas deletion of the RS region in both subunits simultaneously is lethal (72).
In summary, the work presented here reveals a novel property of the general splicing factor U2AF, i.e. shuttling between the nucleus and the cytoplasm. The RS domain present in either of the two U2AF subunits is involved in nucleo-cytoplasmic transport and subnuclear targeting, similarly to what was previously established for SR proteins. Possibly, higher eukaryotes have evolved a common pathway to control the availability of RS-containing splicing factors in the nucleus. Future studies are needed to determine how shuttling of these factors may contribute to regulate splicing.