U2AF participates in the binding of TAP (NXF1) to mRNA.

TAP/NXF1 is a conserved mRNA export receptor serving as a link between messenger ribonucleoproteins (mRNPs) and the nuclear pore complex. The mechanism by which TAP recognizes its export substrate is unclear. We show here that TAP is added to spliced mRNP in human cells. We identified a distinct region of TAP that targets it to mRNP. Using yeast two-hybrid screens and in vitro binding studies, we found that this region coincides with a direct binding site for U2AF35, the small subunit of the splicing factor U2AF. This interaction is evolutionarily conserved across metazoa, indicating its significance. We further found in human cells that the exogenously expressed large U2AF subunit, U2AF65, accumulates in spliced mRNP, leading to the recruitment of U2AF35 and TAP. Similarly to TAP, U2AF65 stimulated directly the nuclear export and expression of an mRNA that is otherwise retained in the nucleus. Together with our finding that U2AF is continuously exported from the nucleus, these data suggest that U2AF participates in nuclear export, by facilitating TAP's addition to its mRNA substrates.

TAP/NXF1 is a conserved mRNA export receptor serving as a link between messenger ribonucleoproteins (mRNPs) and the nuclear pore complex. The mechanism by which TAP recognizes its export substrate is unclear. We show here that TAP is added to spliced mRNP in human cells. We identified a distinct region of TAP that targets it to mRNP. Using yeast two-hybrid screens and in vitro binding studies, we found that this region coincides with a direct binding site for U2AF35, the small subunit of the splicing factor U2AF. This interaction is evolutionarily conserved across metazoa, indicating its significance. We further found in human cells that the exogenously expressed large U2AF subunit, U2AF65, accumulates in spliced mRNP, leading to the recruitment of U2AF35 and TAP. Similarly to TAP, U2AF65 stimulated directly the nuclear export and expression of an mRNA that is otherwise retained in the nucleus. Together with our finding that U2AF is continuously exported from the nucleus, these data suggest that U2AF participates in nuclear export, by facilitating TAP's addition to its mRNA substrates.
TAP (also called NXF1)/Mex67p proteins define a conserved mRNA export pathway in eukaryotes (1)(2)(3)(4). The Saccharomyces cerevisiae Mex67p and Caenorhabditis elegans Ce-NXF-1 are essential for mRNA export (4,5), and the vertebrate TAP mediates the export of mRNA, but not the other classes of RNA (3,6,7). The Schizosaccharomyces pombe spMex67p is nonessential, but its mRNA export function is required in the absence of the essential Rae1p (8,9). TAP belongs to the NXF family of proteins that share the ability to bind both RNA and the components of the nuclear pore complex (10). The proposed substrate binding domain of TAP (aa 1   comprises an RNA binding domain and leucine-rich repeat domain and exhibits general RNA binding and high affinity binding to the viral CTE RNA (2,3,11). This region is also required for the direct binding to REF (also called Aly) proteins (1,12), which are the metazoan homologues of yeast mRNA export factor Yra1p (13).
At all metabolic steps, mRNA is part of ribonucleoprotein complexes (mRNP). The primary transcripts associate with a subset of heterogeneous nuclear ribonucleoproteins (hnRNPs) (reviewed in Ref. 14) and with the splicing machinery (15,16), leading to their nuclear retention before splicing. Splicing changes the mRNP composition by promoting the assembly of export-competent mRNP (17,18) and by depositing "splicing signature" proteins onto ligated exons (19). This protein complex is composed of the splicing-associated factors SRm160, DEK, and RNP S1, as well as REF and the RNA-binding protein Y14 (19,20). Hence, REF was proposed to provide a link to export, possibly by recruiting TAP (18,21,22). As shown in human cells, TAP is added efficiently to the endogenous postsplicing mRNP complexes (23). In contrast, TAP is not added to the complexes produced by splicing in vitro (18,19) or in microinjected oocyte nuclei (2). In this work, we addressed the assembly of TAP with the endogenous mRNP in human cells. We report that TAP is targeted to the spliced complexes, through a region that directly binds to U2AF35, the small subunit of the splicing factor U2AF. We further found that U2AF can recruit TAP to mRNP and is able to promote the nuclear export of mRNA in vivo.
Cell Fractionation-Subcellular fractionation was performed using the conditions adapted from Refs. 29, 31, and 32. HeLa cells were extracted with hypotonic buffer (33) containing 0.004% digitonin (CYT fraction). Nuclei were extracted with 10 mM Tris-Cl, pH 7.8, 100 mM NaCl, 2.5 mM MgSO 4 , 0.5% Triton X-100 (N fraction). The remaining material was sonicated in the same buffer by five 3-s bursts, using VibraCell ultrasonic processor and a microtip (Sonics and Materials, Inc.), set at 40% amplitude (NUP fraction). For 293 cell extraction, the above protocol was modified: hypotonic lysis was performed in the presence of 1% Triton X-100 (CYT), and the nuclear N and NUP fractions were prepared in the presence of 400 mM NaCl. Prior to use, all extracts were centrifuged at 14,000 ϫ g for 10 min. All manipulations were carried out at 4°C, and the buffers contained 0.1 unit/ml RNasin. Total RNA was isolated using the RNasol protocol (Tel-Test, Inc.), and reverse transcription-PCR was performed with a Titan kit (Roche Molecular Biochemicals). The primer pair for unspliced human glyceraldehyde phosphate dehydrogenase (GAPDH) transcript spanning nucleotides 3393-3425 and 3536 -3568 of the gene (accession no. J04038) and amplified a 176-bp fragment. The primer pair for the spliced GAPDH transcript spanning nucleotides 3377-3401 and 3988 -4012 and amplified a 325-bp fragment.
mRNP Capturing Assay-mRNP capturing was performed using oli-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 25 magnetic beads (Dynal). Extracts from 10 6 cells were adjusted to 5 ml in 15 mM HEPES, pH 7.7, 50 mM KCl, 200 mM NaCl, 2 mM EDTA, 0.2% Triton X-100 (293 cells) or 10 mM Tris-Cl, pH 7.8, 50 mM NaCl, 2.5 mM MgSO 4 , 0.5% Triton X-100 (HeLa cells), and incubated with 20 l of beads for 30 min at room temperature. As controls, some extracts were treated with 0.1 mg/ml RNase A for 10 min at room temperature, prior to binding to the oligo(dT) beads, or the empty beads (M280, Dynal) were used. After four washes in the same buffer, the mRNPs were eluted with water at 37°C, or by boiling in 1ϫ Laemmli sample buffer.
Recombinant DNA-U2AF mammalian expression vectors contain the coding regions inserted into pCMV-GFPsg25 (34) that provided a GFP tag. Yeast two-hybrid plasmids were constructed as described (5). The C. elegans U2AF35 cDNA (35), the eukaryotic expression plasmids for HA and GFP-tagged TAP proteins (24), and the bacterial expression plasmids for glutathione S-transferase (GST)-U2AF35 and GST-U2AF65 (28) were described. The expression plasmids for REF proteins and the p15 expression plasmid pEGFPN3zz-p15 were provided by E. Izaurralde. All recombinant DNAs were sequenced on both strands.

RESULTS
TAP Is Added to mRNP Complexes after Splicing-We studied the association of TAP with mRNP in HeLa cell extracts. The cytoplasmic fraction (CYT) was obtained by digitonin extraction in low salt buffer, and the soluble nuclear fraction (N) was extracted with 100 mM NaCl and 0.5% Triton X-100. Because the bulk of hnRNPs associate with the insoluble nuclear components, the remaining nuclei were solubilized by sonication yielding the NUP extract, according to Ref. 29. The distribution of spliced and unspliced transcripts in these extracts was examined by reverse transcription-PCR for GAPDH transcripts. Fig. 1A shows that the CYT and N fractions were enriched in spliced mRNA, whereas NUP was enriched in pre-mRNA, as expected (29). The NUP fraction also contained small amounts of spliced GAPDH transcripts that were detectable upon prolonged amplification (data not shown). We further analyzed the ␤-actin transcripts in human 293 cells by Northern blots, and found a comparable distribution of spliced and unspliced mRNAs in the subcellular fractions (see also Fig. 7C). Together, these data indicate that the observed distribution patterns are not likely the artifacts of a particular cell line or detection method.
From the HeLa fractions, poly(A)-containing mRNPs were captured to oligo(dT) beads and analyzed on immunoblots. Fig.  1B shows that TAP associated with mRNPs only in the CYT and N fractions. Using anti-hnRNP A1 antibodies, we confirmed that hnRNPs were enriched in NUP (29), in agreement with the notion that the general hnRNPs contain mostly pre-mRNA (23). We next showed that non-hnRNP proteins such as Ran GTPase and transportin were excluded from the mRNPs, as expected (32,36), confirming the specificity of the assay (Fig. 1B). Because capturing of TAP was sensitive to RNase pretreatment, and the bound complexes coeluted with poly(A) RNA (Fig. 1B), TAP is a bona fide mRNP component. This association was resistant to 400 mM NaCl (see also Fig. 2) and 0.5 mg/ml heparin (data not shown) similarly to that of tightly bound hnRNP proteins (29). In contrast, hGLE-2, the human orthologue of Rae1p (37,38), was not detected in mRNPs (Fig. 1B), suggesting that the stable TAP-mRNP association is specific and is not a generic property of mRNA export factors.
mRNP Assembly Requires a Distinct Region of TAP-We hypothesized that the binding of TAP after splicing is facilitated by protein factors. To obtain a probe for such factors, we mapped the mRNP association region of TAP. TAP and its deletion mutants were transiently expressed in human cells as GFP fusion proteins ( Fig. 2A, load), and tested for mRNP association in soluble nuclear extracts ( Fig. 2A, mRNP). As shown in Fig. 2A (left panel), like the endogenous TAP, the transiently expressed TAP-GFP (aa 61-619) associated with oligo(dT)-selected fraction in a poly(A) RNA-dependent manner. No selection occurred from RNase pretreated extracts or when using empty beads (left panel). Similar results were obtained using TAP tagged with hemagglutinin epitope peptide (HA) (see also Fig. 5), whereas no association was detected for the GFP moiety alone (data not shown). To exclude that this association was an in vitro artifact, we added recombinant GST-TAP61-372 to the extract prior to oligo(dT) capturing ( Fig. 2A, right panel). Although this protein contains the complete RNA-binding domain of TAP, it did not access the mRNPs that were assembled in vivo and did not bind oligo(dT) on its own. To identify the minimal binding region, N-and C-terminal deletions of TAP (see Fig. 2, panel A and summary in panel B) were tested. TAP61-140 was the smallest region that associated efficiently, whereas further truncation to aa 61-120 reduced this association. We concluded that TAP is targeted to mRNP via a region spanning aa 61-140.
The mRNP Assembly Region of TAP Selects U2AF35 in the Yeast Two-hybrid Screen-To search for proteins that interact with the mRNP assembly region, we performed a yeast twohybrid screen of a HeLa complementary DNA (cDNA) library using TAP62-140 fused to the DNA-binding domain of LexA as bait. Using interaction-mating screens, we found that 2 of 24 primary clones also interacted with the complete human TAP, its human homologue TAPX2/NXF2 (10,24), and the C. elegans orthologue Ce-NXF-1 (4,5). No interactions occurred with the C-terminal TAP412-628, the human immunodeficiency virus-1 Rev nuclear export signal, or the empty plasmid, further indicating specificity of the observed interactions.
Sequence analysis showed that these cDNAs encode a truncated U2AF35 protein (aa 69 -240), the small subunit of the heterodimeric splicing factor U2AF. Using liquid culture as- says (Table I), we found that the C. elegans orthologue Ce-U2AF35 (35) also interacted with the complete human TAP, TAP62-140, TAPX2, and Ce-NXF-1. The interacting regions of TAP and TAPX2 are 57% similar, whereas that of Ce-NXF-1 does not share detectable homology. Hence, this interaction is conserved between divergent proteins across metazoan species.
Both U2AF35 and U2AF65, the large subunit of U2AF, contain a serine-arginine rich domain (RS) characteristic of a family of SR-related proteins. To exclude that the TAP proteins interacted with generic RS domains, we show that Ce-U2AF35, which naturally lacks RS, interacted efficiently with all TAP proteins tested (Table I). In addition, the RS deletion in hU2AF35 (aa 69 -178) impaired but did not abolish the interaction (Table I), whereas the isolated RS domain of hU2AF35 (data not shown) or the hU2AF65 (Table I) did not interact. Thus, the RS domain did not represent the primary TAP interaction site, and the interaction is specific for U2AF35.
The mRNP Assembly Region of TAP Coincides with the Direct Binding Site of U2AF35-We tested the interaction between purified recombinant TAP and U2AF35 proteins in vitro (Fig. 3A). TAP proteins were produced in E. coli as GST fusion proteins, and the U2AF35 protein was expressed in baculovirus. The same amounts of purified GST-TAP proteins and the GST-U2AF65 were immobilized on glutathione beads and pulldown assays were performed in 300 mM salt and 1 mg/ml bovine serum albumin to ensure that only strong interactions are detected. Fig. 3A shows that the GST moiety alone did not bind U2AF35, whereas GST-U2AF65 bound efficiently, confirming the specificity of the assay. We found that serial deletion mutants of GST-TAP61-372 including GST-TAP61-141 bound to U2AF35 efficiently, whereas the C-terminal truncation to aa 121 severely impaired the binding, and GST-TAP540 -619 and 195-372 bound poorly. Comparison of the input and the bound samples (Fig. 3A) confirmed that a signif-icant fraction of U2AF35 was bound. These data show that TAP has a direct binding site for U2AF35, which is located between aa 61 and 141, and coincides with the mRNP association region (aa 61-140, Fig. 2).
We next examined the ability of TAP61-141 to bind RNA. As shown previously, TAP61-187 binds RNA directly under low stringency conditions (39). We therefore examined the general RNA binding of a series of TAP peptides to a Bluescript-transcribed RNA (4) using gel shift assays. As shown in Fig. 3B, a core RNA binding site was found between aa 61 and 121. Similar results were obtained using different RNA probes (data not shown).
To study whether RNA and U2AF35 can bind to TAP simultaneously, we performed the pull-down assays in the presence of SRV-1 CTE RNA (3) or human histone H4 mRNA (2). We used GST-TAP61-198 to ensure that the core RNA binding site was present in the context of the complete RNA binding domain (aa 119 -198; Ref. 11). Both RNAs bound to GST-TAP61-198 efficiently, but did not compete for TAP-U2AF35 binding even at the highest concentration tested (4 M, Fig. 3C). As negative control, we used GST-TAP61-102, which showed only weak binding to both RNA and U2AF35, as expected (see Fig. 3, A and B). Hence, the mRNP association region of TAP comprises independent direct binding sites for U2AF35 and RNA.
U2AF Associates with Both Unspliced and Spliced mRNP-U2AF is known to act early in splicing and does not associate with the ligated exons in vitro (for recent reviews, see Refs. 40 and 41). Although TAP is added after splicing (Fig. 1), it is targeted to mRNP via its U2AF35 binding region (Figs. 2 and 3), suggesting the participation of U2AF at this step. One possibility is that the pre-mRNP-bound U2AF interacts with TAP transiently and facilitates TAP's addition to the spliced exons in a manner that is coupled to intron removal. Alternatively, a fraction of U2AF may persist on spliced mRNA, thereby providing the entry sites for TAP.
To examine whether U2AF can associate with spliced mRNPs in vivo, we immunoprecipitated the endogenous U2AFcontaining complexes from HeLa cell extracts. Analysis of these complexes by immunoblots for hnRNP A1 (Fig. 4A) and by Northern blots for ␤-actin and histone H4 transcripts (Fig. 4B) confirmed that anti-U2AF35, but not the control antibodies, efficiently precipitated mRNPs from the NUP extract. Fig. 4B further shows that U2AF35 associated with the spliced as well as the unspliced ␤-actin transcripts (lanes 4 and 7), but the pre-mRNA was enriched in the U2AF35-containing complexes compared with the unprecipitated fraction (lane 4 versus lane 2). The U2AF35 antibodies also precipitated histone H4 mRNA (Fig. 4B, lane 4 versus lane 2). Comparison of the ␤-actin transcripts that coprecipitated with hnRNP A1 (Fig. 4B, lane 6) and U2AF35 (lane 7) showed that the proportion of the unspliced RNA is higher in the U2AF-containing complexes.
Further analysis of the oligo(dT)-purified complexes confirmed that U2AF-containing mRNPs were found predominantly in the pre-mRNA-enriched NUP fraction. In addition, such complexes were present in the N and CYT fractions, both of which are enriched in spliced mRNA ( Fig. 4C; see also Fig. 1A). Because the cytoplasmic levels of U2AF-containing complexes were low, they could be the result of some leakage during fractionation. However, we found that the exogenously expressed U2AF accumulates efficiently with the spliced mRNP in the cytoplasm (see Fig. 5B), confirming that it can maintain stable association with mRNPs in this compartment.
Because U2AF can bind to oligo(dT) directly (42), the assays shown in Fig. 4C were performed from diluted extracts and in low salt buffer, to reduce the possible direct binding and to ensure that only mRNP-associated U2AF were detected. To U2AF Facilitates TAP-mRNA Assembly further exclude this potential artifact, we show that the majority of U2AF-containing complexes coeluted with poly(A) RNA (Fig. 4C, eluate). As additional controls, partially purified human U2AF dimer (hU2AF) was captured in the absence of cell extract (Fig. 4C, hU2AF) or the extracts were treated with RNase A prior to capturing (Fig. 4C, ϩ RNase). In both controls, U2AF35 did not copurify with poly(A) RNA, confirming that our procedure detected only mRNP-associated U2AF. Similar results were obtained using the U2AF65 antibodies (data not shown). Together, these data demonstrate that a fraction of U2AF associates with spliced mRNPs in vivo, whereas the majority of U2AF-containing complexes are unspliced. In addition, a naturally intronless and non-polyadenylated histone mRNA also exhibited association with U2AF.
We next used DEK as a marker of the "splicing signature" (19) to probe its potential association with the human complexes. Western blot analysis of crude HeLa extracts showed The TAP/NXF cDNAs were inserted into the bait plasmid pEG202 in-frame with LexA, whereas U2AF35 cDNAs were cloned into the prey plasmid pJG4 -5 in-frame with the activation domain. Two-hybrid interactions were measured by liquid ␤-galactosidase assay (58), normalized to the values obtained between empty pJG4 -5 and the respective LexA fusions, and are shown as value ranges: 1-2 (Ϫ), 2-5 (ϩ), 5-15 (ϩϩ), 15-40 (ϩϩϩ The TAPX2 region that matches aa 61-140 of TAP was assembled from two published sequences: aa 87-148 of TAPX2 (24), followed by aa 41-57 of the partial TAPX2 coding region (GenBank™ accession no. CAB75659).

FIG. 3. TAP directly binds to U2AF35 in vitro.
A, the indicated recombinant proteins were produced in E. coli and used in GST pulldown assay with purified baculovirus-produced U2AF35 (28). GST-U2AF65 protein (28) and the empty GST moiety served as positive and negative controls, respectively. Prior to binding, all proteins were treated with micrococcal nuclease to remove the residual polynucleotides. U2AF35 (30 ng/l) and 0.2 g of immobilized GST proteins were incubated in 10 mM phosphate, pH 7.4, 300 mM NaCl, 0.2% Tween 20, 1 mg/ml bovine serum albumin for 15 min at room temperature. The bound U2AF35 was detected on immunoblots in GST pull-downs; load, an 1:10 aliquot of the input U2AF35. The results of two independent experiments are shown. B, 20 nM 32 P-labeled RNA was transcribed from a Bluescript vector and used in mobility shift assay (4) with 5-10 ng/l GST-tagged proteins (2). The positions of free (F) and bound (B) RNA are shown to the right. C, pull-down assays were performed as described in panel A using the indicated GST-TAP proteins. The reactions contained 4 and 0.4 M in vitro transcribed CTE or histone H4 RNA (2, 3), as indicated on the top. The complexes were analyzed on immunoblots for U2AF35, and on ethidium bromide-stained gels for RNA.

FIG. 4. Association of U2AF with the endogenous human mRNP.
A, 2 g of anti-U2AF35 or normal rabbit immunoglobulins (control) were bound to protein A-Sepharose beads and added to HeLa cell extracts (N and NUP). After 10 min at 4°C, immunoprecipitates (bound) were collected and analyzed on immunoblots for hnRNP A1 and DEK proteins as indicated. Load, 1:100 aliquot of the input NUP extract. B, U2AF35 complexes were immunopurified from NUP extracts as in panel A (lanes 4 and 7). hnRNP complexes were isolated from the same extracts as described (29) using 1 g of purified hnRNP A1 monoclonal antibody (lane 6). Poly(A) RNA (lane 1) or total RNA (lanes 2-7) was extracted from the immunoprecipitates (bound); 1:10 aliquots of the supernatants after precipitation (unbound) and 1:10 of the input extract (lane 1). The ␤-actin and the histone H4 transcripts were detected on Northern blots. C, mRNP complexes were captured to oligo(dT) beads from HeLa extracts (CYT, N, and NUP) and analyzed on immunoblots for U2AF35 and DEK. As control, 10 g of partially purified, RNase A-treated human U2AF (hU2AF) (57) was captured under the same conditions, but in the absence of extract. load, samples prior to capture (1:150 of the input); eluate, water-eluted oligo(dT)captured complexes; after elution, beads after elution with water; ϩRNase, captured after digestion with RNase A. Similar results were obtained in several independent experiments. that DEK was present in the insoluble nuclear fraction NUP (Fig. 4C, load) as well as in the chromatin fraction (29) (data not shown), consistent with its association with the nuclear matrix components (43,44). DEK was not detectable in oligo(dT)-captured mRNP (Fig. 4C) and was not found in U2AF35-containing complexes (Fig. 4A). Thus, the putative DEK complexes may be unstable or insoluble, or they may be transient and thus difficult to detect under these conditions, compared with the in vitro splicing system. In support of this idea, recent data (45) show that some "splicing signature" factors such as REF are associated with mRNA transiently, whereas the others such as Y14 are more persistent.
U2AF65 Recruits U2AF35 and TAP to mRNP Complexes-To address the effect of U2AF on the assembly of TAP with mRNP, we coexpressed the HA-tagged TAP with the U2AF subunits in human 293 cells and tested their mRNP association in soluble nuclear (Fig. 5, A, C, and D) and cytoplasmic (Fig. 5B) extracts. We found that untagged (Fig. 5, A  and B) or GFP-tagged U2AF65 (Fig. 5C) accumulated efficiently in mRNP. The expression of U2AF65 led to concomitant, dose-dependent recruitment of the endogenous U2AF35 and HA-TAP to these complexes (Fig. 5A). Similar results were obtained using the N and the CYT fractions (data not shown). Among other recruited factors, we identified the SAP130, -145, and -155 proteins (Fig. 5B), which are parts of the U2 snRNPassociated SF3b protein complex (46,47). It is likely that SAP155 was recruited via direct binding to U2AF, because it is able to directly bind U2AF in vitro (48). The endogenous DEK was not recruited (Fig. 5C). Under the same conditions, the expression of U2AF65 did not alter the splicing and nuclear retention of ␤-actin transcripts, indicating that there was no redistribution of general pre-mRNP to the soluble nuclear extract (see Fig. 7C); therefore, U2AF likely accumulated on spliced mRNP.
The expression of U2AF35 alone did not lead to the accumulation of U2AF and HA-TAP in mRNPs (Fig. 5C). Because U2AF65 provides the major RNA-binding determinant of U2AF and U2AF35 by itself binds poorly to RNA (49,50), these results suggested that U2AF bound to mRNA directly via U2AF65 and recruited TAP via direct binding to the U2AF35 subunit. The exogenous expression of U2AF65 also led to the overall depletion of HA-TAP, but not of DEK, from the soluble nuclear extracts (Fig. 5C, load) because of TAP's redistribution to the insoluble NUP fraction (data not shown). In cells expressing high levels of U2AF65, a trace of this protein associated with oligo(dT) in an RNase-independent manner (Fig. 5A). However, the association of HA-TAP with oligo(dT) was completely abolished by RNase pretreatment, even at high U2AF65 levels (Fig. 5A), confirming that HA-TAP was recruited to the mRNP-associated, but not to the free, U2AF.
We next tested whether the increased levels of REF proteins affected TAP's association with mRNP. As representatives, we used the mouse REF1-II and REF2-II proteins, which were GFP-tagged and expressed in 293 cells to levels similar to those of U2AF65-GFP. Although REF, like U2AF, has the potential to bridge TAP and mRNA (12), HA-TAP did not accumulate in mRNP in response to the REF1-II and REF2-II expression (Fig.  5D). Because the REF proteins fused to neutral tags such as GFP or GST have been shown to possess export properties (18,21), it is unlikely that this lack of accumulation is caused by the presence of GFP tag. These data suggest that the U2AF-mediated recruitment of TAP is specific and is not a generic property of factors that have dual affinity to TAP and to RNA. This is consistent with our finding that within TAP the integrity of the U2AF35 (aa 61-140) but not the REF (aa 1-372) binding region was required for the stable association between TAP and mRNP (Fig. 2). Together, these results indicate that TAP can be recruited to the mRNA-bound U2AF in vivo.
U2AF65 Stimulates the Export of Nuclear-retained mRNA-We asked whether the U2AF-mediated accumulation of TAP in mRNP complexes affected their nuclear export. TAP assembles with mRNP in vivo (Ref. 23 and this study), whereas the existing in vitro and Xenopus oocyte systems did not allow addressing its mRNA association. In cell culture, TAP has been shown (10,51) to stimulate the otherwise inefficient expression of DM138 pre-mRNA, which contains an intron-encoded CAT reporter gene (26). In this system, TAP is a limiting factor for the export of an mRNA that is otherwise retained in the nucleus, but not for efficiently exported mRNAs (10,51). However, the exogenous expression of TAP alone leads to a relatively small stimulation of CAT production, whereas the coexpression of TAP's p15 cofactor leads to further stimulation (10,51,52). Therefore, the DM138 assay is a system of choice with which to study the TAP's cofactors. Fig. 6A illustrates this assay, and shows transfections of human 293 cells with a fixed amount of pDM138 and increasing amounts of the HA-TAP plasmid. The presence of TAP led to a dose-dependent increase of CAT, but had little effect on expression of the cDNA-encoded luciferase mRNA (Fig. 6A), and coexpression of p15 further stimulates the CAT production (Fig. 6C), as expected (10, 51, were captured to oligo(dT) beads. Equivalent amounts of extracts were treated with RNase A before capturing. The complexes that coeluted with poly(A) RNA were analyzed by Western immunoblots as indicated. B, the proteins isolated from cytoplasmic mRNP from A were analyzed on Coomassie-stained SDS-PAGE. The molecular masses of marker proteins (in kilodaltons) are shown to the left. The bands of SAP proteins, indicated to the right, were identified using standard mass spectrometry microsequencing protocols (Beckman Research Institute) resulting in high confidence matches (SAP130, 6 peptides, 5% coverage; SAP145, 7 peptides, 10% coverage; SAP155, 20 peptides, 17% coverage). C, nuclear mRNP from cells transfected with 250 ng of HA-TAP plasmid in the absence or presence of 25 ng of pU2AF35 and 25 ng U2AF65-GFP were analyzed as described in A. Load, 1:100 aliquots of crude extracts. Asterisk indicates the endogenous U2AF65 protein. D, nuclear mRNP from cells transfected with HA-TAP together with the indicated amounts of U2AF65, REF1-II-GFP, and REF2-II-GFP were analyzed as described in A. The results of three independent transfection experiments are shown in panels A, C, and D.
U2AF Facilitates TAP-mRNA Assembly 52). We then performed the DM138 assay under the same conditions (see Fig. 5) that led to the U2AF-mediated accumulation of TAP in mRNP. When fixed amounts of DM138 and TAP were coexpressed with increasing amounts of U2AF65, we observed a dose-dependent stimulation of CAT, but not of the GFP expression, comparable with that obtained with TAP (Fig.  6B). However, the expression of U2AF65 alone had no effect on CAT production (Fig. 6D), and, when both TAP and p15 were present, coexpression of U2AF65 had no further effect (Fig.  6C). We concluded that, like p15, U2AF65 augments the TAPmediated stimulation of CAT production. However, this effect is only seen when the TAP effect is suboptimal, such as in the absence of cotransfected p15.
To further examine the effects of U2AF65 on nuclear export, we cotransfected DM138 in the absence or in the presence of exogenous TAP, HA-U2AF65, and p15, and analyzed the CAT transcripts on Northern blots. Because CAT pre-mRNA has inefficient splice sites, it was poorly spliced as expected (Fig.  7A, unbound). In the absence of exogenously expressed proteins, CAT pre-mRNA was mostly nuclear (Fig. 7, A and B), whereas the spliced mRNA was exported efficiently, as described previously (26). As expected, expression of TAP alone led to a significant increase of CAT pre-mRNA export, which was further stimulated upon coexpression of p15 (Fig. 7B). Notably, expression of U2AF65 alone also resulted in the accumulation of CAT pre-mRNA in the cytoplasm (Fig. 7, A and   B), and coexpression of TAP or TAP/p15 had no further effect (Fig. 7B). In some experiments (Fig. 7A), the presence of U2AF65 led to a shortening of CAT pre-mRNA. This is similar to the observed shortening of human immunodeficiency virus pre-mRNAs during Rev-mediated export (53). As proposed for Rev, this could be a result of the poly(A) tail shortening, but this observation was not further investigated. The effect of U2AF65 was specific for CAT pre-mRNA, because the nucleocytoplasmic distribution of the cDNA-encoded luciferase mRNA (Fig. 7A, LUC) or the ␤-actin (Fig. 7C) and GAPDH (data not shown) transcripts were not affected. Accordingly, U2AF65 did not stimulate the protein production from the efficiently exported transcripts such as GFP (Fig. 6B) and LUC (data not shown), similarly to TAP (10,51). Because the levels of nuclear CAT pre-mRNA were not increased significantly by U2AF65 (Fig. 7A), its cytoplasmic redistribution cannot be attributed to nuclear leakage caused by the elevated nuclear levels, but rather points to its increased export.
To examine whether this export stimulation was direct or, alternatively, was the result of the inhibition of nuclear retention, we immunopurified the U2AF-containing complexes using HA antibodies from the cells that coexpressed the DM138 and the HA-tagged U2AF65. Northern blot analysis showed that HA-U2AF65 associated with CAT pre-mRNA both in the nucleus and in the cytoplasm, whereas its cytoplasmic association with LUC and the U snRNPs was undetectable (Fig. 7A,  bound). Thus, a fraction of HA-U2AF65 is likely bound to CAT pre-mRNA during export, although its de novo association in the cytoplasm cannot be excluded. Although HA-U2AF65 bound detectably to the efficiently exported LUC mRNA in the nucleus, its export was not further increased (Fig. 7A, bound), supporting the notion that in this system the stimulation can be only revealed using the inefficiently exported reporters such as DM138 CAT. It is likely that the transcripts such as LUC are exported at maximal rates and therefore TAP/p15 and U2AF65 are not rate-limiting in this system.
To rule out that the effects of U2AF65 were caused by its general interference with spliceosomal retention, similar to the phenotype of the trans-dominant mutants of spliceosomal release factor HRH1 (54), we analyzed the efficiently spliced ␤-actin transcripts (Fig. 7C). U2AF65 did not affect their splicing significantly and, importantly, did not lead to the redistribution of pre-mRNA to the soluble CYT and N extracts, suggesting that there is no general interference with nuclear retention under these conditions. Similar results were obtained using the GAPDH transcripts (data not shown).
To further establish the specificity of U2AF65 effects, we analyzed the endogenous U snRNPs under the same conditions. U1, U2, U4, and U5 exhibited low cytoplasmic to nuclear ratios, similar to that of CAT pre-mRNA, but the expression of U2AF65 did not significantly affect these ratios (Fig. 7A, unbound). Although the U6 snRNP is believed to be only nuclear, it was also detectable in the cytoplasmic fraction, whereas in the presence of U2AF65 its cytoplasmic accumulation was significantly reduced. There was no significant general leakage under these conditions, because in the same extracts the CAT pre-mRNA, as well as the U6 snRNP in the presence of U2AF65, were exhibiting low cytoplasmic to nuclear ratios (Fig. 7A, unbound). However, we cannot exclude the leakage of this particular snRNP during fractionation. Formally, the presence of U6 snRNP in the cytoplasmic extracts can be also attributed to its impaired nuclear import in 293 cells, but these effects were not further investigated. Together, these data indicate that U2AF65 stimulates specifically the export of CAT pre-mRNA.
Collectively, our results demonstrated that, in cultured human cells, U2AF65 could overcome the nuclear retention of CAT pre-mRNA by directly stimulating its nuclear export. Because, under the same conditions U2AF65 recruits extra TAP to the general mRNP complexes (Fig. 5) and TAP acts additively with U2AF65 to stimulate the CAT expression (Fig. 6), it is likely that the effect of U2AF65 is mediated by U2AF-bound TAP. DISCUSSION Here, we report that U2AF can recruit TAP to mRNP complexes and directly promotes the export of mRNA, which is otherwise retained in the nucleus. The mRNP assembly of TAP does not require its high affinity CTE binding site (aa 61-372), supporting the previous observations that TAP uses different modes to interact with CTE and with the cellular mRNAs (2, 3, 10, 11). We found that TAP is targeted to the spliced mRNP via a region that directly binds to U2AF35 subunit, and this interaction is conserved across metazoan species. This novel link of U2AF to nuclear export is further supported by our finding that both U2AF subunits can shuttle between the nucleus and the cytoplasm (data not shown), an observation that has been recently confirmed by another group (55). Thus, both U2AF35 and U2AF65 belong to a "shuttling subset" of SR family proteins, with a proposed role in nuclear export (56). Although our mapping of the mRNP assembly region of TAP did not indicate the participation of REF, it may be involved in weaker interactions contributing to TAP's assembly, which were not detected by our assays. In addition, the TAP-REF recognition may be transient and not rate-limiting; therefore, it could be difficult to detect in vivo. In support of this idea, a recent study demonstrated that REF is removed from nuclear mRNPs upon their cytoplasmic translocation in microinjected Xenopus oocytes (45). Thus, similarly to U2AF, REF likely performs a transient role and does not provide a permanent link between TAP and mRNA. At present, it is not clear whether REF and U2AF are part of the same underlying mechanism of TAP's assembly or act independently.
Although the mRNP assembly region of TAP (aa 61-140) has affinity to RNA, it does not include the complete, conserved RNA binding domain (aa 119 -198) and only contains one of its two RNP motifs (RNP2, aa 122-127) (11). Because TAP/ Mex67p proteins have conserved affinity to RNA and can be cross-linked to poly(A) RNA in vivo, TAP likely contacts mRNA directly during export. It is therefore plausible that TAP transiently binds to protein factors in mRNP and further establishes the direct contact with RNA.
Using total human poly(A) mRNPs, we show that splicing removes the bulk of U2AF from pre-mRNP, whereas TAP associates with spliced mRNP only. For individual transcripts, we confirmed the expected association of U2AF with pre-mRNA in both the endogenous human mRNPs and in the complexes assembled in microinjected oocyte nuclei (data not shown). Some U2AF was also found on mRNAs that lack the functional 3Ј splice sites, such as the spliced ␤-actin mRNA and the naturally intronless histone H4 mRNA.
Regarding the metabolic steps at which the TAP-U2AF interaction takes place, one possibility is that TAP binds to U2AF in spliced (or intronless) mRNPs, thereby committing them to export. Alternatively, U2AF may transiently interact with TAP in pre-mRNPs and facilitate its addition to ligated exons in a manner that is coupled to splicing, similarly to the addition of "splicing signature" (19,20).
To test whether the Xenopus laevis oocytes system is suitable to study the export function of U2AF, we performed nuclear coinjection experiments with purified, recombinant U2AF dimer. Using pull-down experiments, we found that U2AF assembled efficiently with the intron-lariat and the precursor of adenovirus-derived Ad2 mRNA, as well as with H4 RNA, but not with U1 and U6 snRNAs, tRNA and the spliced Ad2 mRNA. When U2AF was coinjected, the splicing of Ad2 precursor was ϳ2-fold reduced, but the nucleocytoplasmic ratios of the Ad2 precursor and the splicing products, as well as these of the control RNAs were not affected (data not shown). These experiments suggested that U2AF is not rate-limiting for the nuclear export of mRNA in microinjected X. laevis oocytes, and therefore the cell culture assays were chosen to further probe its export function.
We found that, in cell culture, the exogenously expressed U2AF65 accumulated in spliced mRNPs, leading to the recruitment of U2AF35 and TAP, which is consistent with the direct binding of TAP to mRNP-associated U2AF. Individual mRNAs also associated with U2AF under these conditions, but the export of efficiently exported mRNAs was not further stimulated by U2AF65, which is in agreement with the finding that TAP is not limiting for the mRNAs that are exported efficiently (10,51). In contrast, the export of an otherwise nuclear-retained CAT pre-mRNA is greatly increased by coexpressed U2AF65. U2AF65 associated with CAT pre-mRNA during export, further suggesting a direct effect. We found a good correlation between the cytoplasmic CAT pre-mRNA levels and CAT expression, upon export stimulation by TAP/p15. In contrast, U2AF65 alone stimulated the export, but not the expression of CAT pre-mRNA, whereas TAP in the presence of U2AF65 did not further stimulate export, but led to an increase of CAT expression (Figs. 6 and 7), thus likely acting at a post-export step. Because TAP is a stable component of the endogenous, cytoplasmic mRNP complexes (Fig. 1) and persists on poly-some-associated mRNPs (data not shown), it is possible that TAP has a role after export.
Because U2AF stimulates the export of model, nuclear-retained mRNA, and cooperates with TAP to stimulate its expression, it may play a similar role for natural transcripts that are subject to nuclear retention. Additionally, this export phenotype revealed with CAT pre-mRNA might reflect a dedicated mechanism by which TAP is added to spliced mRNA. In this model, TAP is first targeted to pre-mRNA via transient interactions with splice site-bound U2AF and is deposited onto exons as a result of subsequent splicing. At elevated concentrations of U2AF65, this splicing requirement can be bypassed, allowing the unregulated deposition of TAP onto pre-mRNA (such as DM138 CAT) leading to export stimulation. The development of in vitro models that reflect the addition of TAP to spliced complexes will aid in the understanding of the underlying molecular mechanisms.