Complex formation between Tap and p15 affects binding to FG-repeat nucleoporins and nucleocytoplasmic shuttling.

Mammalian Tap-p15 and yeast Mex67p-Mtr2p are conserved and essential mRNA export factor complexes that transport mRNPs through the nuclear pore. Here, we report that the small subunit p15 affects the binding of the large subunit Tap to repeat nucleoporins. BIAcore measurements revealed that recombinant Tap binds with high affinity (K(d) in the nm range) to repeat nucleoporins and dissociates from them very slowly. In contrast, when recombinant Tap was bound to p15, the derived heterodimeric complex exhibited a significant lower affinity to FG-repeat nucleoporins (K(d) in the microm range). Furthermore, when recombinant Tap lacking the N-terminal nuclear localization sequences (TapDeltaNLS) was microinjected in mammalian cells, it did not shuttle; however, TapDeltaNLS with bound p15 efficiently shuttles between nucleus and cytoplasm. We conclude that heterodimerization of Tap and p15 is required for shuttling of the functional Tap-p15 mRNA exporter complex.

Transport of macromolecules in and out of the nucleus occurs through the nuclear pore complexes (NPCs). 1 Nuclear import and export require soluble transport receptors and nucleoporins (for review see Ref. 1 and 2). Import and export cargoes are recognized by importins or exportins, respectively, which belong to the importin-␤ type receptor family. The cargo-receptor complexes translocate through the NPC based on the direct binding of the importin-␤ type receptors to phenylalanine-glycine (FG-) repeat-containing nucleoporins (3)(4)(5)(6). Although the mechanism of translocation through the pore is still unknown, several models exist. In one model, importin-␤, which has different affinities to various FG-repeats, migrates through the NPC toward the nucleoplasm due to an affinity gradient (7). For transportin and NTF2 it was suggested that they translocate through the NPC by equally low affinities to the different nucleoporins (8 -10).
Human Tap was first identified as the putative mammalian orthologue of the Saccharomyces cerevisiae mRNA exporter Mex67p (11) and to be necessary for the export of a viral RNA-export element called CTE (12). As does Mex67p, Tap shuttles between nucleus and cytoplasm and binds to poly(A) ϩ RNA in vivo (13). Tap exhibits a pronounced domain organization. The middle domain of Tap binds to p15 (13,14), while the C-terminal domain binds to various FG-repeat-containing nucleoporins (13,14). The N-terminal domain of Tap contains a basic NLS (13,14), which is recognized by transportin (14) and an RNA-binding domain (13,15) that exhibits a canonical RNP fold (16). Furthermore, in vivo Tap binds to a series of intranuclear proteins including Aly (Yra1p in yeast), which are recruited to the mRNA during splicing and mark the mature and thus export-eligible mRNA (17)(18)(19)(20).
Recent results indicate that p15 is crucial for nuclear mRNA export by Tap. However, p15, which is also called NXT1, was shown to be involved in Crm1-dependent nuclear protein export (21). Hence, p15 may play a role in multiple nuclear export pathways. Nuclear export of mRNAs in mammalian cells is stimulated by co-expression of p15 and Tap (22)(23)(24), and the complex formation of Tap and p15 was shown to be required for the stimulatory activity (25). The Tap-p15 complex is the functional orthologue of the yeast mRNA exporter Mex67p-Mtr2p, which indicates that the mRNA export pathway is conserved from yeast to human (13). Interestingly, a mutation of MEX67 that abrogates its interaction with Mtr2p leads to mislocation of Mex67p into the cytoplasm and concomitant nuclear accumulation of poly(A) ϩ RNA. This suggested that complex formation of Mex67p-Mtr2p is required for both pore association and mRNA export. Furthermore, complex formation between Mex67p and Mtr2p is required for nucleoporin binding in vitro (26). These findings led to the conclusion that Mtr2p could function in mRNA export by altering the affinity of Mex67p to repeat nucleoporins (26). Binding of Mex67p alone to FG-repeat nucleoporins has been also shown, but in this case it was not tested whether Mtr2p influences this interaction (27). On the other hand, Tap itself possesses a distinct affinity for FGrepeat nucleoporins. In this report, we show that p15 significantly decreases the affinity of Tap to various FG-repeats and enables the Tap-p15 heterodimer to translocate efficiently through the nuclear pores in both directions.
Protein Expression and Purification-Tap⌬NLS, Tap⌬NLS-p15, and * This work was supported by Grant-in-Aid 12CE2007 for COE Research from the Japanese Ministry of Education, Science, Sports, and Technology and the Human Frontiers Science Program. 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 U.S.C. Section 1734 solely to indicate this fact.
Microinjection Experiments and in Vitro Import Assay-Labeling of proteins with fluorescent dye was achieved with the Alexa 546 labeling kit (Molecular Probes) according to the manufacturer's instruction. Unincorporated dye was removed by PD-10 gel filtration column chromatography (Amersham Biosciences, Inc.) at 4°C. By pull down experiments, we determined that the binding of the Alexa 546 labeled Tap⌬NLS to FG-repeats was reduced at most by 20% (data not shown). Thus, we conclude that the fluorescent labeling did not grossly interfere with the repeat binding ability of Tap. For microinjection and in vitro import assays, concentrations of the Alexa 546-labeled proteins were adjusted to 2 mg/ml in phosphate-buffered saline.
Formation of HeLa cell homokaryons by Sendai virus and microinjection assay were carried out as reported (28). Thirty min after microinjection, cells were fixed with formaldehyde and observed with a Zeiss Axiophot II fluorescence microscope.
For in vitro import assays Madin-Darby bovine kidney cells were used. Cytosolic extract was prepared from Ehrlich ascites tumor cells as reported previously (29). Cells grown on glass 8-well multitest slides (ICN Biomedicals) were treated with ice-cold transport buffer (20 mM HEPES-KOH, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium acetate, 1 mM EGTA, 2 mM DTT (pH 7.3) containing 1 g/ml each leupeptin, aprotinin, pepstatin, and 40 g/ml digitonin for 5 min at 4°C. After washing with transport buffer at room temperature, the cells were incubated at 30°C for 30 min with import substrate mixtures containing 4 g of the labeled proteins with or without cytosolic extract (30 g) and an ATP-regenerating system as indicated in the figure legend. The cells were then fixed with formaldehyde and observed with a Zeiss Axiophot II fluorescence microscope.
All pictures were taken with a Cool Snap HQ CCD camera (Roper Scientific), and digital images were recorded and processed with Improvision Open Lab software. The signals in nuclei were measured by using the histogram function of Adobe PhotoShop software.
Pull Down Assays-Pull down assays were performed as reported previously (13,30). Unbound proteins were precipitated with 10% trichloroacetic acid. After extensive washing, bound proteins were eluted by treating the resin with SDS-sample buffer. Aliquots corresponding to 50% of input of the bound and unbound fractions were loaded onto SDS-14% polyacrylamide gels and protein bands were separated by electrophoresis and visualized by Coomassie staining. BIAcore Binding Assays-Ligands (GST-fused nucleoporin repeats) were immobilized via anti-GST antibody covalently attached on CM5 research-grade sensor chips (BIAcore). Activation and blocking of sensor chips and immobilization of anti-GST antibody were done using the Amine-coupling and GST antibody kits (BIAcore) according to the manufacturer's instructions. Analysis of protein-protein interactions was carried out in transport buffer containing 0.2% Tween 20 at a flow rate of 20 l/min at 20°C using a BIAcore 2000 system. Parallel injections of analytes over a GST-immobilized flow cell were subtracted from the data as background. Apparent K d values for Tap⌬NLS/p15-repeat binding were derived from Scatchard plots RU/concentration versus RU and linear regression analysis by Kyplot software (31,32). The absence of mass transport limitation and/or rebinding was tested by injecting Tap⌬NLS on sensor chips harboring different amounts of each GSTrepeat at a flow rate of 20 l/min at 20°C. Since the sensorgrams were superimposable (data not shown), it can be concluded that the diffusion of the analyte from the bulk flow to the sensor chip surface is not kinetically limiting. Thus K d values for Tap⌬NLS-repeat bindings were obtained by using standard kinetic equations supplied within the BI-Aevaluation 3.1 software.  heads (A, C,  E, G). After incubation at 37°C for 30 min cells were fixed with formaldehyde, and the intracellular localization of the injected proteins was examined with a fluorescence microscope. The same cells were also viewed with phase contrast (right panels).

Tap⌬NLS-p15 Complex Enters the Nucleus without the Aid of Soluble
Factors-The C-terminal domain of Tap (aa 540 -619) binds directly to nucleoporins and confers to Tap the ability to translocate through the nuclear pore (13,(33)(34)(35). However, in vitro Tap requires its N-terminal NLS and transportin for nuclear import (14). To find out whether p15 can confer transportin-independent shuttling of Tap, fluorescently labeled Tap⌬NLS lacking the transportin binding site (14) either alone or complexed with p15 was tested in an in vitro nuclear import assay. This revealed that Tap⌬NLS can enter the nucleus only when bound to p15 (Fig. 1A). In contrast, free Tap⌬NLS is only inefficiently imported and has the tendency to accumulate at the nuclear periphery (Fig. 1B). Unexpectedly, nuclear import of Tap⌬NLS-p15 was still observed when cytosol (containing karyopherins and Ran) and an energy-regenerating system were omitted from the in vitro assay (Fig. 1C). Under these conditions, GST-GFP-M9 import substrate was not imported into the nucleus unless cytosol and an ATPregenerating system were added (data not shown), indicating that most of the soluble transport factors (i.e.; transportin and Ran) as well as energy are deprived. Under the same conditions, Tap⌬NLS alone did not enter the nucleus (Fig. 1D). These data indicate that upon complex formation with p15 Tap⌬NLS gains the ability to translocate through the nuclear pore and no longer requires transportin/RanGTP for nuclear import. In the light of this finding, the role of the N-terminal NLS of Tap remains unclear. Transportin could be required for the rapid re-import of free Tap and/or for nuclear import of newly synthesized Tap only. Another possibility is that transportin is required for efficient release of mRNP cargoes from Tap-p15, since the N-terminal NLS and the RNA-binding domain of Tap are adjacent to each other.
Tap⌬NLS Shuttles between the Nucleus and Cytoplasm Only in Association with p15-To analyze nucleocytoplasmic shuttling of Tap⌬NLS-p15 in vivo, fluorescently labeled Tap⌬NLS, either alone or in complex with p15 were microinjected into the cytoplasm or nucleus of HeLa cell homokaryons. When free Tap⌬NLS was injected into one of the several nuclei of a homokaryon, it was not exported (Fig. 2C). Similarly, Tap⌬NLS injected into the cytoplasm of a homokaryon was not imported into the nucleus (Fig. 2D). In contrast, the Tap⌬NLS-p15 complex, regardless of the injection site, was able to translocate through the nuclear pores in both directions and rapidly accumulated in the nuclei (Fig. 2, A and B). The ability of Tap-p15 to shuttle depends on the ability of Tap⌬NLS to bind to FGrepeat nucleoporins, since Tap⌬NLS lacking its C-terminal domain but bound to p15 (GST-Tap⌬NLS⌬C-p15), showed a reduced binding to FG-repeat Nups (see below) and lost the ability to shuttle between the nucleus and cytoplasm (Fig. 2, E  and F). The inability of GST-Tap⌬NLS⌬C-p15 to shuttle is not due to the larger complex size caused by the GST tag, since GST-Tap⌬NLS-p15 is still able to translocate through the nuclear pores (Fig. 2 G and H). Taken together, our results show that complex formation with p15 enables Tap to translocate through the nuclear pores in both directions.
Tap Exhibits Different Affinities to Various FG-repeats in the Presence or Absence of p15-It is well documented that Tap can interact directly with FG-repeat-containing nucleoporins CAN/ Nup214 and hCG1 via its C-terminal domain (aa 507-619) (13,33). Since p15 is necessary for shuttling of Tap in vivo, p15 could modulate the FG-repeat binding ability of Tap. To test this possibility, the binding of Tap⌬NLS alone or the Tap⌬NLS-p15 complex to nucleoporin repeats was analyzed in vitro. As reported earlier, purified Tap⌬NLS can bind to the FG-repeat domains of CAN and hCG1 ( Fig. 3A and B, lanes  2-4). In addition to these FG-nucleoporins, Tap⌬NLS also bound efficiently to the FXFG-repeats of p62 and GLFG-repeats of Nup98 (Fig. 3, C and D, lanes 2-4; see also Ref. 14). In contrast to the strong binding of Tap⌬NLS to FG-repeats, binding of Tap⌬NLS complexed to p15 was weaker since most of the complex was recovered in the unbound fraction (Fig. 3,  A-D, lanes 5-7 and 13-15). We conclude that complex formation with p15 reduces the affinity of Tap to nucleoporin repeats.
To obtain quantitative data on the binding of Tap⌬NLS ϩ/Ϫ p15 to repeat nucleoporins, we performed BIAcore measurements. Repeat sequences from the different nucleoporins were immobilized on a sensor chip and varying concentrations of Tap⌬NLS or Tap⌬NLS-p15 complex were injected. Free Tap⌬NLS bound very efficiently to all of the repeat sequences tested and dissociated extremely slowly (Fig. 4A). The K d value of free Tap to nucleoporin repeats lies in the nM range, showing that Tap alone binds very strongly to repeat nucleoporins (Fig.  4A, inset). In contrast, the Tap⌬NLS-p15 complex showed a weaker and transient interaction with the different nucleoporin repeats; accordingly, ϳ10 times more Tap⌬NLS-p15 com- plex had to be used in the Biacore binding assays. Strikingly, the sensorgram exhibited almost a rectangular shape indicative of an extremely high dissociation rate of Tap-p15 from the repeat sequences (Fig. 4B, left panels). Since the on/off rates for Tap⌬NLS-p15 from FG-repeats (except CAN4 repeats) were too fast to be reliably calculated by kinetic analysis, the apparent dissociation constants (K d ) were obtained by Scatchard analysis. The Tap⌬NLS-p15 complex binds to nucleoporin repeats with a K d in the M range (see Fig. 4B, right panels). The efficient binding of Tap to repeat sequences requires the Cdomain of Tap, since deletion of this domain significantly reduced the amount of Tap-p15 bound to repeats (Fig. 4C). However, a residual but significant binding of Tap-p15 to several repeat nucleoporins is evident (See also Fig. 2, E and F). This residual binding may result in nuclear rim localization observed in microinjected cells (see Fig. 2, E and F). The observed lower on-rate of Tap⌬NLS-p15 to CAN4 FG-repeats in comparison to other tested FG-Nups is not understood at present. This could be due to a lower amount of immobilized GST-CAN4 on the sensor chip or may be a peculiarity of the CAN4 FG-repeat construct. In summary, p15 decreases the affinity of Tap to nucleoporin repeats about 1,000-fold. These data indicate that p15 functions to modulate the interaction of Tap with FGrepeat nucleoporins by making this association more transient.

DISCUSSION
Previous work has revealed that p15 is required for nuclear mRNA export (see Introduction). Here, we report a biochemical function for p15. Our data show that binding of p15 to Tap affects the affinity of the Tap-p15 complex to nucleoporin repeats. This enables the Tap-p15 complex to shuttle between the nucleus and the cytoplasm. How could p15 perform its function to modulate the binding of Tap-p15 to repeat nucleoporins? Interestingly, the middle domain of Tap, which binds to p15, and p15 itself show homology to NTF2, which was shown to bind directly to FG-repeat nucleoporins (13,26,36). Furthermore, a Mex67p-Mtr2p complex that lacks the C-domain of Mex67p, is still able to bind to repeat nucleoporins (26). In accordance with these observations, recent structural data of the Tap-p15 heterodimer indicates that the middle domain of Tap acts synergistically with the C-terminal NPC-binding domain in binding to nucleoporin repeats as well as in shuttling (25). This analysis also showed that p15 contributes only indirectly to repeat binding of Tap . Sensor responses at equilibrium (R eq ) were determined for each protein concentration from each sensorgram, and R eq /concentration of Tap⌬NLS values were plotted as a function of R eq . The slopes of the curves obtained by linear transformation yields the dissociation constants K d (insets). C, GST-Nup98, GST-p62, GST-hCG1, and GST-CAN4 were immobilized on a CM5 sensor chip as in A. Purified Tap⌬NLS⌬C-p15 (5.98 M, gray boxes) and Tap⌬NLS-p15 (5.98 M, black boxes) were then injected over the sensor chip. Control injections over a sensor chip with immobilized GST were subtracted from the data as background, and the RU value at equilibration was obtained as R eq values. Data are presented as average values of three experiments Ϯ S.D. tural studies furthermore revealed a hydrophobic pocket in M-domain of Tap, which binds to a single phenylalanine residue of nucleoporin FG-repeats. Thus, the Tap-p15 heterodimer may move through the NPC by consecutive low affinity interactions with repeat nucleoporins.
It is not clear whether Tap functions as a monomer in vivo. If so, the measured in vitro affinities of the Tap monomer could be meaningful. Accordingly, Tap with bound mRNP cargo could be first recruited to the nucleoplasmic side of the NPC due to its strong affinity to repeat nucleoporins. p15 would then associate to trigger the release of Tap from these high affinity binding sites due to an increased dissociation rate. If p15 stays permanently bound to Tap during pore passage, Tap-p15 would efficiently pass through the NPC channel by hopping between different repeat nucleoporins due to its comparably low affinities as suggested for transportin and NTF2 (10). Alternatively, multiple association/dissociation cycles of p15 and Tap may lead to their translocation through the pore channel. It is also conceivable that the cargo-loading state affects the directionality of Tap-p15 translocation, since binding of cellular mRNA may control the affinity of Tap to repeat nucleoporins. However, it is also possible that the Tap-p15 heterodimer is the only functionally relevant unit in nuclear mRNA export. Therefore, the observed high affinity binding of Tap to nucleoporins could be an in vitro artifact due to a misfolded Tap M-domain. The observation that co-expression of p15 with Tap is required for efficient export of mRNA (22,23) supports this speculation. Further work is required to clarify whether a free pool of Tap functions in vivo in mRNA export.
Our data support a model, in which binding of p15 to the middle domain of Tap affects the conformation and/or accessibility of the two shuttling/NPC-binding domains within Tap, thereby affecting Tap's association with and dissociation from repeat nucleoporins. Since the C-terminal repeat binding domain (aa 540 -619) of Tap alone is able to mediate NPC translocation in vitro (35), it is likely that both the C-domain and the M-domain of Tap act in concert as NPC shuttling devices. The correct folding of the middle domain of Tap, which requires binding to p15, may affect the overall conformation of the Tap-p15 heterodimer and hence might be a prerequisite for efficient shuttling of this mRNA export factor through the nuclear pores.
In summary, we have shown that p15 is a crucial co-factor of Tap that affects the interaction of Tap with repeat-containing nucleoporins. Thus, the Tap-p15 complex gains the capability to translocate efficiently through the NPC, which is the basis for export of mRNA from the nucleus to the cytoplasm.