Exportin‐t (Xpo‐t) is a vertebrate nuclear export receptor for tRNAs that binds tRNA cooperatively with GTP‐loaded Ran. Xpo‐t antibodies are shown to efficiently block tRNA export from Xenopus oocyte nuclei suggesting that it is responsible for at least the majority of tRNA export in these cells. We examine the mechanism by which Xpo‐t–RanGTP specifically exports mature tRNAs rather than other forms of nuclear RNA, including tRNA precursors. Chemical and enzymatic footprinting together with phosphate modification interference reveals an extensive interaction between the backbone of the TΨC and acceptor arms of tRNAPhe and Xpo‐t–RanGTP. Analysis of mutant or precursor tRNA forms demonstrates that, aside from these recognition elements, accurate 5′ and 3′ end‐processing of tRNA affects Xpo‐t–RanGTP interaction and nuclear export, while aminoacylation is not essential. Intron‐containing, end‐processed, pre‐tRNAs can be bound by Xpo‐t–RanGTP and are rapidly exported from the nucleus if Xpo‐t is present in excess. These results suggest that at least two mechanisms are involved in discrimination of pre‐tRNAs and mature tRNAs prior to nuclear export.
RNA transcription occurs largely within the nucleus of a eukaryotic cell. Since most RNAs either function in the cytoplasm, or have to pass through the cytoplasm during maturation, the vast majority of cellular RNAs must be exported from the nucleus to the cytoplasm (reviewed by Izaurralde and Mattaj, 1995; Izaurralde and Adam, 1998). A number of RNA‐binding proteins that are thought to be mediators of export of particular RNAs have been identified (Piñol‐Roma and Dreyfuss, 1993; Izaurralde et al., 1995, 1997; Lee et al., 1996; Segref et al., 1997; Grüter et al., 1998). In addition, the nucleocytoplasmic transport receptor CRM1/Xpo1 is known to function in the export of specific cellular RNAs from the nucleus (Fornerod et al., 1997a; Stade et al., 1997). The relationship between CRM1 and other known nucleocytoplasmic transport receptors, all members of the importin‐β family (Fornerod et al., 1997b; Görlich et al., 1997) has led to the expectation that RNA export from the nucleus will generally occur in a manner analogous to the importin‐β mediated nuclear import of proteins (Görlich, 1997).
In general it is thought that RNA‐binding proteins involved in export are recognized in the nucleus by export receptor molecules, together with the RanGTPase in its GTP‐bound form (Fornerod et al., 1997a; Kutay et al., 1997), and the resultant complex is translocated to the cytoplasm through a nuclear pore complex (Corbett and Silver, 1997; Cole and Hammell, 1998; Dahlberg and Lund, 1998; Izaurralde and Adam, 1998; Mattaj and Englmeier, 1998; Ohno et al., 1998 for review). However, the second characterized export receptor involved in the transport of cellular RNAs, the tRNA export receptor [exportin(tRNA), exportin‐t or Xpo‐t], binds directly to tRNA in the presence of RanGTP, and does not require an adaptor protein to mediate its interaction with the substrate RNA (Arts et al., 1998; Kutay et al., 1998). Xpo‐t was found to bind to many, and perhaps all, different tRNAs (Arts et al., 1998; Kutay et al., 1998).
Other tRNA‐binding proteins that recognize common features of multiple different tRNAs include the bacterial translation elongation factor EF‐Tu (Nissen et al., 1995) and its eukaryotic equivalent eEF1 alpha (Negrutskii and El'skaya, 1998); the CCA‐adding enzyme (tRNA nucleotidyl transferase) that creates mature tRNA 3′ ends (Shi et al., 1998); ribonuclease P, which creates the mature 5′ end on all tRNAs by endonucleolytic cleavage (Altman et al., 1993) and the ribosome (Nazarenko et al., 1994). These proteins recognize universally conserved structural and sequence features of tRNAs (Nazarenko et al., 1994; Liu et al., 1998; Shi et al., 1998, and references therein). Since Xpo‐t is unrelated to these proteins it was of interest to determine which features of mature tRNAs are recognized by the Xpo‐t–RanGTP complex and mediate its binding. Note that there are many more examples of proteins, like the aminoacyl‐tRNA synthetases or tRNA modifying enzymes (Edqvist et al., 1993; Giegé et al., 1993), that recognize specific tRNA species and discriminate between them.
A further aspect of the Xpo‐t–RanGTP interaction relevant to tRNA export is that tRNAs, like almost all cellular RNAs, are transcribed in the form of precursors. The maturation of all tRNAs involves removal of 5′ and 3′ extensions, addition of the mature CCA 3′ end and both base and sugar modifications. In addition, many pre‐tRNAs contain introns that must be removed by splicing. Almost all of these processing steps occur in the nucleus prior to tRNA export (Melton et al., 1980; De Robertis et al., 1981; Nishikura and De Robertis, 1981). This raises the question of how pre‐tRNAs and mature tRNAs are discriminated so that only the latter are efficiently exported and, in particular, what is the role of Xpo‐t in this process.
In this study three aspects of Xpo‐t function in tRNA export are investigated. First, how Xpo‐t, together with RanGTP, recognizes and binds to tRNAs. Secondly, the involvement of Xpo‐t in discriminating between pre‐tRNAs and mature tRNAs with regard to nuclear export. Thirdly, we investigate the quantitative importance of Xpo‐t for tRNA export from the nuclei of Xenopus oocytes.
It was previously established that Xpo‐t can bind to many different tRNAs (Arts et al., 1998; Kutay et al., 1998). This suggested that Xpo‐t recognizes conserved structural features rather than individual tRNA sequences. For this reason, Xpo‐t–tRNA interactions were analysed via enzymatic and chemical protection and interference analyses utilising reagents that modify the ribose–phosphate backbone of the tRNA rather than reagents that target nucleotide bases. As a substrate tRNA we chose Saccharomyces cerevisiae tRNAPhe. This tRNA has been extensively studied by structural, biochemical and mutagenic methods (Kim et al., 1974; Ladner et al., 1975; Romby et al., 1987; Nazarenko et al., 1994) and this background information facilitated interpretation of the analyses presented here.
Footprinting of tRNA in the Xpo‐t complex
Fe/EDTA results in cleavage of ribose sugar rings through the local production of hydroxyl radicals, and has been used to probe aspects of tRNAPhe structure (Latham and Cech, 1989). Extensive protection of tRNAPhe against Fe/EDTA cleavage was seen on addition of Xpo‐t and RanGTP, which bind cooperatively to the tRNA (Arts et al., 1998; Kutay et al., 1998). The positions of protected riboses was determined using both 5′‐ and 3′ end‐labelled tRNAPhe (Figure 1A). Equal cleavage was seen in the presence and absence of the proteins at the resolved positions between 7–16 and 19–46 (Figure 1A, lanes 4–6, 11–13; note that since the T1‐cleaved markers carry an intact ribose ring they migrate above the corresponding band in the experimental lanes). Riboses at positions 18, 47–49 and 72 exhibited mild protection against cleavage whereas the riboses between positions 50 and 71 were significantly protected in the presence of Xpo‐t–RanGTP (Figure 1A, compare lanes 1–4 with 5 and 6). In addition, better seen with 5′ end‐labelled RNA, positions 4–6 and 17 were also protected against Fe/EDTA attack by Xpo‐t–RanGTP (Figure 1A, lanes 11–13). For technical reasons, positions 1–3 and 73–76 cannot be examined by this method.
The next probing agent used was ribonuclease V1, which cuts accessible double‐stranded helical regions in RNA. RNase V1 cuts several positions in the stems of the anticodon and acceptor arms of free tRNAPhe (Figure 1B, lane 3; Lockard and Kumar, 1981). When preformed tRNAPhe–Xpo‐t–RanGTP complex is probed with V1 nuclease, the majority of the cleaved positions in the acceptor stem (positions 67–70, see also Figure 3) were protected to some extent against cleavage while cleavages at most of the anticodon stem positions (positions 28–30) as well as a site in the D arm (position 22) were enhanced and the cleavage at 41 was little affected (Figure 1B, lanes 1–4, Figure 3). This suggested that Xpo‐t–RanGTP would interact with the acceptor arm and that binding would result in a conformational change in the tRNA resulting in increased accessibility of parts of the anticodon and D arms to V1 nuclease. Note that, as for Fe/EDTA‐induced cleavage, the protection afforded by Xpo‐t–RanGTP was incomplete. Whether this is due to partial accessibility to the cleaving reagents in the presence of bound Xpo‐t–RanGTP or to some displacement of Xpo‐t–RanGTP in the cleavage conditions is not known.
Pb2+ ions have been shown to cleave tRNAPhe between positions D17 and G18 (Werner et al., 1976; Brown et al., 1983; Behlen et al., 1990). Since Fe/EDTA‐induced cleavage at position 17 was reduced in the presence of Xpo‐t–RanGTP (Figure 1A, lanes 11–13), the effect of these proteins on lead cleavage was tested. Addition of Xpo‐t–RanGTP essentially blocked lead cleavage (Figure 1B, lanes 8–10).
Phosphate ethylation‐binding interference
To obtain further information on likely points of contact between Xpo‐t–RanGTP and tRNAPhe, ethylnitrosourea (ENU) modified tRNA was prepared. ENU ethylates phosphate groups. In this experiment, free tRNAPhe was ethylated at a ratio of approximately one ethyl group per tRNA molecule (Romby et al., 1985) then bound to immobilized Xpo‐t–RanGTP (Arts et al., 1998). Bound and free RNA populations were separated, cleaved at ethylated phosphate positions by high pH treatment, and fractionated by denaturing gel electrophoresis. Bands resulting from phosphates whose ethylation interfered with Xpo‐t–RanGTP binding should be under‐represented in the bound (P) as compared with the input and non‐bound (IN,S) fractions (Figure 2). The experiment is carried out under conditions of tRNA excess (Arts et al., 1998) and there is thus no enrichment of bands in the non‐bound versus input lanes at interfering positions. The results are shown in Figure 2, lanes 4–6, with positions whose ethylation interfered significantly with Xpo‐t–RanGTP binding indicated on the left. The strongest interference with binding was seen on ethylation of phosphates at positions 49, 56, 57, 59 and 62, making some of these positions good candidates for close contact with Xpo‐t–RanGTP (see Discussion).
The data from Figures 1 and 2, as well as the results of RNase T1 protection experiments (data not shown) are summarized in Figure 3. Protection or interference is seen over much of the acceptor and TΨC arms, indicative of these structures being directly involved in Xpo‐t–RanGTP interaction. Two phosphates at the anticodon stem–extra arm junction interfere with Xpo‐t–RanGTP binding (positions 43 and 44; Figure 2) making interaction here also likely. Other changes seen in the anticodon arm are generally enhancements of nuclease cleavage on binding, suggesting that this region of the tRNA is not in contact with Xpo‐t. The results from the D arm are more varied, and will be discussed in detail below.
Changes in the acceptor arm and in tRNAPhe tertiary structure affect functional interaction with Xpo‐t
To obtain evidence that the interaction study presented above was related to Xpo‐t function, i.e. to tRNAPhe nuclear export, we examined various mutant forms of yeast tRNAPhe with respect to their nuclear export and interaction with Xpo‐t. As expected, wild‐type tRNAPhe bound specifically to Xpo‐t in the presence of RanGTP, but did not bind in the presence of RanGDP (Figure 4A, lanes 1–3). tRNAPhe also did not bind to importin‐β in the presence of RanGTP (Figure 4A, lane 4), and neither U1 nor U6 bound specifically to either Xpo‐t or importin‐β (Figure 4A, lanes 1–4). Yeast tRNAPhe was exported from the nucleus of Xenopus laevis oocytes after microinjection (Figure 4A, lanes 5–10) at a rate similar to the export of X.laevis tRNAPhe (Figure 5A, lanes 1–6). Increasing the nuclear concentration of Xpo‐t by microinjection into oocytes increases the rate of tRNA export from the nucleus (Arts et al., 1998; Kutay et al., 1998) and this effect was also seen with yeast tRNAPhe (Figure 4A, lanes 11–14).
The interaction data presented in Figures 1 suggested that a ‘minihelix’ consisting of only the acceptor arm and TΨC arm might be a substrate for Xpo‐t–RanGTP binding and export. However, this was not the case. The minihelix construct showed no detectable binding to Xpo‐t–RanGTP in vitro (Figure 4F, lanes 1–4) and was too unstable on microinjection to assess whether it could be exported from the nucleus (data not shown).
Since the acceptor arm and TΨC arm minihelix was not sufficient for binding, the effect of removal of specific structural features from the tRNA was tested. Removal of the entire acceptor arm prevented Xpo‐t–RanGTP binding (Figure 4B, lanes 1–4) and nuclear export, even in the presence of excess Xpo‐t (Figure 4B, lanes 5–14). A much smaller deletion, removal of the CCA 3′ end, consistent with previous data on tRNASer (Kutay et al., 1998), also had a drastic effect on Xpo‐t–RanGTP interaction (Figure 4C, lanes 1–4). However, since the CCA end was re‐added to this mutant RNA very rapidly following injection into oocyte nuclei (Figure 4C, compare the injected RNA in lane 5 with the uninjected sample in the IN lane), the effect of CCA deletion on export could not be assessed.
The loop of the TΨC arm is essential for tRNA tertiary folding, mainly due to interactions between this loop and the D loop (Kim et al., 1974; Ladner et al., 1975). Before examining the contribution of the TΨC arm to binding of Xpo‐t–RanGTP it was therefore essential to determine whether tertiary folding was important for Xpo‐t–RanGTP interaction. To this end, a C56G mutant of yeast tRNAPhe was used. This mutation removes the C56–G19 base pair and thereby disrupts tRNAPhe tertiary folding (Nazarenko et al., 1994). This mutation prevented Xpo‐t–RanGTP binding and nuclear export (Figure 4D, lanes 1–4, 5–14). Taken together, the above data demonstrate that the acceptor arm and interaction between the D‐ and TΨC‐loops of tRNAPhe are requirements for both Xpo‐t binding and export.
Discrimination between pre‐tRNA and tRNA
Pre‐tRNA processing, i.e. intron removal and mature 5′‐ and 3′ end‐formation, occurs in the nucleus of Xenopus oocytes (Melton et al., 1980; De Robertis et al., 1981). tRNAs carrying 5′‐ and 3′ end‐extensions do not leave the nucleus even in the presence of excess Xpo‐t (Melton et al., 1980; Arts et al., 1998) and do not detectably bind to Xpo‐t–RanGTP (data not shown). Melton et al. (1980) did however observe export of an intron‐containing, end‐matured version of yeast tRNATyr in Xenopus oocytes. This prompted us to determine whether yeast tRNAPhe with its intron would interact with Xpo‐t–RanGTP.
This pre‐tRNA bound to Xpo‐t–RanGTP as well as wild‐type tRNAPhe (Figure 4E, lanes 1–4). On injection into oocyte nuclei, however, very little export of the intron‐containing species was seen, whereas the small amount of mature tRNAPhe produced was exported at a rate similar to injected tRNAPhe (Figure 4E, lanes 5–10). When excess Xpo‐t was coinjected with the pre‐tRNA (Figure 4E, lanes 11–14) the intron‐containing tRNA was exported at a rate similar to wild‐type tRNAPhe with excess Xpo‐t (Figure 4A, lanes 11–14). The same result was obtained whether RNA and Xpo‐t were coinjected or Xpo‐t was injected 30 min after the RNA (data not shown). These results demonstrate that intron‐containing tRNAPhe binds well to Xpo‐t–RanGTP, and suggest strongly that the reason that intron‐containing tRNA is not normally exported is due to competition between Xpo‐t and other nuclear proteins that recognize intron‐containing tRNAs and retain them in the nucleus (see Discussion).
As mentioned above, tRNA precursors without introns but carrying both 5′ and 3′ extensions and lacking the post‐transcriptionally added CCA end are not exported and do not bind Xpo‐t. To study the ‘end’ requirements for export in more detail, we made a series of tRNA constructs as illustrated in Figure 5A–E. These were based on X.laevis pre‐tRNAPhe (Lin‐Marq and Clarkson, 1995) since this pre‐tRNA has extensions of defined sequence and lacks an intron. tRNAs carrying a 5′ extension and having either a mature CCA end or an immature 3′ end were not substrates for export even in the presence of excess Xpo‐t (Figure 5B and C), and these RNAs did not bind to Xpo‐t–RanGTP in vitro (data not shown). Mature tRNAPhe derived from the artificial precursors by processing in the injected oocytes was however efficiently exported (Figure 5B and C). Similarly, a 5′ end‐mature but 3′ end‐extended precursor was not exported until processed (Figure 5D). An artificial construct that was 5′ end‐mature, included the ‘mature’ CCA residues, but also carried a 3′ extension attached to the CCA, was neither exported nor matured to a significant extent (Figure 5E). The 3′ extended RNAs also did not bind Xpo‐t–RanGTP in vitro (data not shown). These experiments show that the possession of either a 5′‐ or a 3′ end‐extension interferes with Xpo‐t binding and nuclear export. An RNA that has the mature sequence but that does not end at the conserved CCA (Figure 5E) is not an export substrate. The requirement for correct ends for Xpo‐t binding is not absolute, as at least some tRNAs lacking the 3′ CCA or carrying a few extra nucleotides at their 5′ ends do bind detectably to Xpo‐t in vitro (Arts et al., 1998; Kutay et al., 1998).
Aminoacylation is not required for tRNAPhe export
Recently it has been discovered that tRNA aminoacylation occurs in the nucleus of Xenopus oocytes as well as in the cytoplasm (Lund and Dahlberg, 1998). Since the identity elements of yeast tRNAPhe are G20, G34, A35, A36 and A73 (Sampson et al., 1989) the intron‐containing version of tRNAPhe (Figure 4E) should not be aminoacylated. It therefore seemed unlikely that aminoacylation would be an absolute requirement for export. Further, many of the binding studies reported here and previously (Arts et al., 1998; Kutay et al., 1998) were carried out with non‐aminoacylated, and indeed non‐modified, in vitro transcripts, demonstrating that neither aminoacylation nor modification is a requirement for Xpo‐t–RanGTP binding.
To examine the relationship between aminoacylation and export more definitively, we made use of two mutant forms of tRNAPhe that are known to be extremely poor substrates for aminoacylation by yeast phenylalanyl‐tRNA synthetase (Sampson et al., 1992). One is a double mutant in which G34 and A35 are both mutated to U, the second a triple mutant with these two changes plus an A73U mutation (Figure 6A and B). Both displayed specific binding to Xpo‐t–RanGTP in vitro (Figure 6A and B, lanes 1–4). The binding of the double mutant was indistinguishable from wild‐type tRNAPhe (data not shown) while the triple mutant bound less strongly (Figure 6A and B). The double mutant was exported at a readily detectable rate, although more slowly than wild‐type tRNAPhe (Figure 6A, lanes 5–10 and Figure 4A). The weaker binding of the triple mutant to Xpo‐t–RanGTP correlated with a lower export rate (Figure 6B, lanes 5–10). Export of both mutants was significantly increased by microinjection of excess Xpo‐t (Figure 6A and B, lanes 11–14).
To demonstrate that the mutants were not subject to aminoacylation in Xenopus oocytes, we analysed microinjected wild‐type and mutant tRNAPhe on acid gels that separate aminoacylated and non‐aminoacylated tRNA species (Varshney et al., 1991). As markers we used mature tRNAPhe purified from yeast before and after aminoacylation with purified yeast phenylalanyl‐tRNA synthetase (a gift of D.Kern, UPR 9002, Strasbourg). The aminoacylated tRNA migrates more slowly (Figure 6C, lanes 1 and 2). After microinjection into the nucleus or cytoplasm of oocytes and 30 min incubation, aminoacylation of tRNAPhe was examined. In both cases, two bands were seen, representing aminoacylated and non‐aminoacylated tRNAPhe in both samples (Figure 6C, lanes 3 and 6). Note that the nuclear‐ and cytoplasmic‐injected tRNAs migrate differently, probably due to their being differentially modified in the nucleus and cytoplasm (Nishikura and De Robertis, 1981). To demonstrate further that the upper bands (open triangles) represented aminoacylated tRNAPhe, the control RNAs and the wild‐type tRNA injected into the nucleus were subjected to alkaline pH to deacylate the tRNAs. After this treatment, the aminoacylated control tRNAPhe and the microinjected wild‐type tRNAPhe co‐migrated with non‐aminoacylated tRNAPhe (Figure 6C, lanes 9–11). In contrast to the wild‐type, both mutant forms of tRNAPhe were present as single species after either cytoplasmic or nuclear injection, equivalent to the faster‐migrating non‐aminoacylated form of the wild‐type tRNA (Figure 6C, lanes 4, 5, 7 and 8). Thus, these mutant tRNAs are not subject to aminoacylation, although they are exported from the nucleus. The difference in the rate of export between the double mutant and wild‐type tRNAPhe suggests that aminoacylation may increase export efficiency (see also Lund and Dahlberg, 1998).
Xpo‐t antibodies drastically reduce tRNA export
As shown using HeLa cell tRNA (Kutay et al., 1998), Xpo‐t does not seem to be specific for a subset of tRNAs, but rather to bind most or all of the population of tRNAs. This is also true of the tRNAs present in Xenopus oocyte extracts (data not shown) and suggests that Xpo‐t can in principle be the export receptor for all nuclear tRNAs. We wished to inhibit Xpo‐t function and assay the effect on tRNA export. To this end a rabbit antiserum was raised against recombinant human Xpo‐t. This antiserum specifically recognizes both human Xpo‐t and, although more weakly, a single band corresponding to the expected size of Xenopus Xpo‐t on Western blots (data not shown). The antiserum was affinity‐purified and injected into the nucleus of Xenopus oocytes. After 30 min incubation, a mixture of four RNAs was injected into the same nuclei. U6Δss, a control for nuclear integrity and injection accuracy, U1ΔSm to assay U snRNA export, a pre‐mRNA derived from the fushi‐tarazu (ftz) gene that is efficiently spliced in oocytes and yeast tRNAPhe. On injection, ftz pre‐mRNA is spliced and the resultant mRNA exported to the cytoplasm (Figure 8, lanes 1–6). U6 RNA remains largely nuclear while U1ΔSm and tRNAPhe are exported at different rates (Figure 7, lanes 1–6). Note that the figure shown is of an experiment where overall RNA export rates were slower than average, the result of variation between batches of oocytes. On injection of affinity‐purified Xpo‐t antibodies neither pre‐mRNA splicing, mRNA export nor U snRNA export were affected. In contrast, tRNAPhe export was reduced by 99% after 60 min and by 78% after 120 min (Figure 7, lanes 7–10). The percentage inhibition was derived by quantitation of the bands seen in a phosphoimager followed by subtraction of the percentage of the negative control, U6 RNA, that had leaked into the cytoplasm. Antibodies purified from pre‐immune or non‐immune sera did not affect tRNA export (data not shown). The inhibitory effect was not specific for tRNAPhe. Xpo‐t antibodies had a similar quantitative effect on the export rate of tRNAiMet when it was injected alone and on the entire population of yeast tRNAs when they were labelled by CCA transfer and injected together.
Three aspects of tRNA export from the nucleus, and the role of Xpo‐t in this process, were investigated in this study. First, detailed biochemical probing of the interaction between Xpo‐t–RanGTP and tRNA was presented. Analysis of this interaction was then pursued making use of mutant and variant forms of tRNAs and pre‐tRNAs. These studies provide evidence for the functional relevance of the interaction data, and suggest mechanisms by which pre‐tRNAs and mature tRNAs are discriminated by the export machinery such that only the latter are transported to the cytoplasm. Finally, evidence that Xpo‐t is the major tRNA export receptor in Xenopus oocytes was presented.
Extensive interaction between Xpo‐t–RanGTP and tRNA
Xpo‐t binds tRNA cooperatively with RanGTP (Arts et al., 1998; Kutay et al., 1998). Binding of the Xpo‐t–RanGTP complex to yeast tRNAPhe was examined by protection and interference analysis using a variety of probing reagents. These data are displayed on a representation of the 3‐D structure of tRNAPhe in Figure 8. It is immediately obvious that Xpo‐t–RanGTP interacts mainly with the upper half of the tRNA, composed of the acceptor and TΨC arms. The interactions presented in this figure are incomplete since analysis of tRNA variants indicates that the mature 5′ and 3′ ends of the tRNA are also important for Xpo‐t–RanGTP binding (Kutay et al., 1998 and this study). Further, mutation of the conserved purine (A) residue that precedes the mature CCA 3′ end to U also decreased Xpo‐t–RanGTP interaction, probably by disrupting stacking of the ACCA end on the acceptor helix.
Remarkably, ribose groups located much of the way along both strands of the acceptor and TΨC stems as well as riboses located in the TΨC loop are protected against modification by Xpo‐t–RanGTP. Further, in the TΨC arm, phosphates along both strands of the stem and in the loop, when modified, interfere with Xpo‐t–RanGTP binding. These data, together with ribonuclease V1 protection studies, support the hypothesis that Xpo‐t–RanGTP makes extensive interactions with this part of the tRNA, and essentially envelopes the upper part of the tRNA structure.
The fact that ethylation of several individual phosphate groups, those at positions 56, 57, 59 and 62, abolishes binding suggests that Xpo‐t–RanGTP may directly contact the backbone at these positions. The strong effect of ethylation at phosphate 49 on Xpo‐t–RanGTP binding is, on the other hand, not likely to be due to direct interaction. This phosphate interacts with the 2′ hydroxyl of ribose 7 (Ladner et al., 1975). Modification of phosphate 49 will therefore alter the conformation of tRNAPhe, and this is the more likely cause of binding inhibition.
Other factors that recognize most or all tRNAs have related modes of recognition. CCA‐adding enzyme (Shi et al., 1998), elongation factor‐Tu in the GTP‐bound state (Nissen et al., 1995) and RNase P (McClain et al., 1987) all bind to this region of the tRNA, although it is not thought that these enzymes make such extensive interactions as the Xpo‐t–RanGTP complex. A further difference between the interactions is that a minihelix consisting solely of the acceptor and TΨC arms is not a substrate for Xpo‐t–RanGTP binding (Figure 4F) while all three of the above‐mentioned tRNA‐binding enzymes recognize the minihelix and can use it as a substrate (McClain et al., 1987; Rudinger et al., 1994; Li et al., 1997). These three enzymes are all ancient components of the protein synthetic machinery, and their ability to bind the minihelix may be a reflection of tRNA evolution. It has been proposed that minihelices are the oldest remaining aspect of tRNA structure, with the D loop, anticodon arm and extra arm being later additions (Schimmel et al., 1993). Xpo‐t, being a more recent addition to the tRNA‐binding protein family, might have evolved at a later date, after tRNA structure had become more complex, and thus has evolved to recognize additional structural features.
This reasoning is supported by analysis of the basis of the difference in binding behaviour between Xpo‐t–RanGTP and the enzymes. Xpo‐t–RanGTP requires aspects of tRNA structure for binding that are not present in the minihelix. This can be deduced from the loss of binding to a mutant tRNA where the TΨC loop–D loop interaction is disrupted and is further supported by the protection by Xpo‐t–RanGTP of the ribose of D17. Other changes in accessibility of the D arm on Xpo‐t–RanGTP binding are suggestive of some alteration of the structure of this region, including the TΨC loop–D loop interaction. The nuclease T1 cleavages after G18 and G19 are not affected on binding while cleavage after G20 is reduced. In contrast, V1 cleavage after G22 increases on Xpo‐t–RanGTP binding (Figure 3).
The non‐uniformity of these changes suggests they may not be due to Xpo‐t–RanGTP interaction with the affected residue. Rather, they may reflect a change in tRNA structure induced by Xpo‐t–RanGTP. Similarly, since Pb2+ cleavage after D17 requires tertiary interaction between the D and TΨC loops, with the TΨC loop forming the Pb2+‐binding site (Brown et al., 1985; Behlen et al., 1990) the loss of lead cleavage on Xpo‐t–RanGTP binding may reflect a structural change in either loop rather than direct protection against cleavage. Finally, the increase in cleavage by V1 of the 5′ side of the anticodon stem is again suggestive of some conformational change in this area that increases accessibility towards the enzyme.
In summary, Xpo‐t–RanGTP binds mainly to the upper half of the tRNA and probably also contacts the backbone of the D loop and the anticodon stem–extra arm junction (Figure 8). Other changes in accessibility to enzymes and modifiers are likely to be due to conformational alteration in the tRNA induced by Xpo‐t–RanGTP binding.
Discrimination between pre‐tRNA and tRNA
The enzymes responsible for pre‐tRNA processing are located in the oocyte nucleus (Melton et al., 1980; De Robertis et al., 1981). In general, pre‐tRNA processing is completed in the nucleus before export (Figures 4E and 5; Melton et al., 1980). How does the cell discriminate between pre‐tRNA and tRNA, and only export the latter? Our data provide evidence for at least two separate mechanisms. First, tRNAs carrying the natural 5′ or 3′ extensions do not bind Xpo‐t–RanGTP in vitro and are not substrates for export in vivo. Thus, end processing is clearly a prerequisite for efficient interaction of a tRNA with the export machinery. It appears that the exact mature 5′ and 3′ ends are not essential for the interaction of all tRNAs with Xpo‐t, since, for example, tRNASer lacking the CCA showed reduced, rather than no, binding to Xpo‐t in vitro (Kutay et al., 1998).
The presence of an intron, on the other hand, did not affect Xpo‐t interaction in vitro. Melton et al. (1980) observed that intron‐containing tRNATyr could be exported. Although intron‐containing tRNAPhe was only exported to a very minor extent under normal conditions (Figure 4E), this pre‐tRNA was efficiently exported either on coinjection of excess Xpo‐t (Figure 4E) or if Xpo‐t was injected in excess subsequent to injection of the pre‐tRNA (data not shown). This suggests strongly that there is a competition between Xpo‐t–RanGTP and other nuclear proteins for binding to intron‐containing tRNA, and that, in this case, discrimination between the mature and immature species with regard to export is simply the result of competition between proteins that retain the pre‐tRNA in the nucleus and those that export it. The most obvious possible competitors of Xpo‐t–RanGTP are the proteins that constitute the tRNA splicing machinery (Di Nicola Negri et al., 1997; Fabbri et al., 1998 and references therein).
Xpo‐t and redundancy in export pathways
It is likely, but not certain (Arts et al., 1998), that Xpo‐t is the vertebrate homologue of the S.cerevisiae Los1p protein. The LOS1 gene was discovered in two different genetic screens, one for mutations that affected tRNA processing (Hopper et al., 1980; Hurt et al., 1987) and one for mutations that produced synthetic lethality together with a gene encoding a defective nuclear pore complex protein, or nucleoporin (Simos et al., 1996a). Subsequent analysis demonstrated interactions between LOS1 and several genes encoding proteins involved in various aspects of tRNA metabolism (Simos et al., 1996a,b; Hellmuth et al., 1998). Los1p is a member of the importin‐β family of proteins (Fornerod et al., 1997b; Görlich et al., 1997) which includes all currently known nucleocytoplasmic transport receptors (Ohno et al., 1998). These data are all consistent with the hypothesis that Los1p is, like Xpo‐t, a tRNA export receptor. However, LOS1 is not an essential gene (Hopper et al., 1980; Hurt et al., 1987) suggesting either that tRNA export can still occur in the absence of a dedicated receptor or that there is a redundant pathway of tRNA export in yeast. The fact that no other member of the yeast importin‐β family is closely related to Los1p (Fornerod et al., 1997b; Görlich et al., 1997; Ohno et al., 1998) suggests that none of these genes is likely to encode a protein that is functionally identical to Los1p. Thus, if there is a redundant tRNA export mediator in yeast, it is not closely related to Los1p.
Our analysis shows that there is a strong correlation between Xpo‐t–RanGTP binding and tRNA nuclear export. That is, for the RNAs tested, those that bound tightly to Xpo‐t–RanGTP in vitro were exported efficiently from the nucleus in vivo, those that did not bind were not exported, and those that bound with intermediate affinity were exported slowly in the presence of endogenous levels of Xpo‐t. The only exception to this correlation was intron‐containing pre‐tRNA, as discussed above. This correlation between binding and export would be consistent with the hypothesis that Xpo‐t could be the major, or indeed the only, mediator of tRNA export in Xenopus oocyte nuclei.
Further evidence on this topic came from antibody inhibition experiments. Antibodies against Xpo‐t, when injected into oocyte nuclei, had a strong and specific inhibitory effect on export of tRNAPhe and tRNAiMet when they were injected individually, or on the whole population of yeast tRNAs when they were injected together (Figure 7; data not shown). Injection of lower concentrations of antibody resulted in reduced inhibition, but it proved technically impossible to further increase the antibody concentration. We cannot therefore say if the minor amount of tRNA export seen in the conditions used is due to incomplete inhibition of Xpo‐t due to insufficient antibody or if it is due to export by some other mechanism. However, even in the latter case, if there is a redundant mediator of tRNA export in Xenopus oocytes that is not recognized by the Xpo‐t antibodies, it functions very inefficiently. The binding and antibody inhibition data would therefore suggest that Xpo‐t and other related proteins recognized by the antibody are either the major or the only tRNA export receptors in Xenopus oocytes.
Materials and methods
Expression and purification of the histidine‐tagged recombinant proteins exportin‐t, zz‐exportin‐t, z‐importin‐β, RanGTP, RanGDP and RanQ69LGTP was performed as described previously (Arts et al., 1998).
End labelling of yeast tRNAPhe
Purified tRNAPhe from yeast (Sigma) was either 5′‐ or 3′ end‐labelled. For the 3′ end‐labelling (Vlassov et al., 1981), 3′ terminal nucleotides were removed by incubating 10 μg tRNAPhe for 10 min at 20°C in 25 μl containing 0.1 μg phosphodiesterase (Boehringer), 50 mM Tris–HCl pH 8.0, 10 mM MgCl2. The reaction was stopped by phenol–chloroform extraction and ethanol precipitation. Three micrograms of the pretreated tRNA was 3′ end‐labelled by incubating for 45 min at 37°C in 25 μl containing 0.5 mg/ml yeast CCA adding enzyme (Rether et al., 1974), 50 mM Tris–glycine pH 8.5, 10 mM MgCl2, 8 mM DTT, 50 μM rCTP and 60 μCi [α‐32P]ATP (3000 Ci/mmol, Amersham). For 5′ end‐labelling, the tRNA was dephosphorylated by bacterial alkaline phosphatase (Amersham) and subsequently 5′ labelled with [γ‐32P]ATP and T4 polynucleotide kinase (Gibco‐BRL) according to the manufacturer's instructions. Both end‐labelled tRNAs were purified from an 8% polyacrylamide–7 M urea denaturing gel.
RNA protection assays
In the enzymatic and lead‐induced cleavage assays 50 000 c.p.m. of either 3′ or 5′ end‐labelled tRNA was used per reaction. For the hydroxyl radical protection experiments double this amount was used. In all footprinting assays, the RNA was mixed in buffer A (50 mM HEPES–KOH pH 7.5, 5 mM Mg acetate, 200 mM K acetate) and incubated for 15 min at 20°C. Recombinant exportin‐t and GTP‐loaded Ran were then both added to a final concentration of 0.4 to 3 μM, or omitted. Complex formation was allowed by incubating the reaction at 20°C for 30 min. Fe(II)EDTA reactions were as described (Hüttenhofer and Noller, 1992; Lentzen et al., 1996) in 25 μl buffer A containing 4 mM EDTA, 10 mM DTT, 2 mM (NH4)2Fe(SO4)2, 0.02% H2O2 for 2 min at 20°C.
RNase V1 probing was performed in 10 μl buffer A containing 2 μg carrier 5S RNA and 0.05 units of RNase V1 (Pharmacia) for 5 min at 20°C. Lead(II) induced hydrolysis (Behlen et al., 1990) was in 10 μl buffer A with 1 mM Pb(II) acetate for 5 min at 20°C and was stopped by addition of 5 μl 50 mM EDTA. Incubation controls were done in parallel without probing reagent. All reactions were stopped by the addition of 50 μl 0.3 M NaAc pH 5.0 and phenol–chloroform extraction. The RNAs were ethanol precipitated with 2 μg of 5S RNA as carrier (if not already present) and the pellets were washed with 80% ethanol, dried, and resuspended in 6 μl RNA‐loading dye. Equal counts were loaded in each lane for 15% denaturing PAGE. As standards, an alkaline ladder and an RNase T1 ladder were also loaded onto the gel. For the alkaline ladder, 100 000 c.p.m. of 3′‐ or 5′ end‐labelled tRNA mixed with 1 μg carrier RNA was hydrolysed in 70 mM sodium carbonate pH 9.0 for 5 min at 90°C. For the T1 ladder, 50 000 c.p.m. mixed with carrier was first denatured in 20 mM sodium citrate pH 4.5, 1 mM EDTA, 7 M urea and gel‐loading dyes for 5 min at 50°C. On ice, 0.005 U RNase T1 was added and the reaction was performed for 5 min at 50°C, stopped and kept on ice until loading.
Ethylation interference assay
Phosphate ethylation by ethylnitrosourea (ENU) was performed under denaturing conditions as described (Romby et al., 1985). Briefly, 2×106 c.p.m. of 3′ end‐labelled yeast tRNAPhe was incubated for 2 min at 90°C in 25 μl containing 100 mM sodium cacodylate pH 8.0, 1 mM EDTA, and 20% ENU which was added to the reaction as a saturated solution in ethanol. In the control reaction ethanol was added to 20%. The reactions were stopped by the addition of 50 μl 0.3 M Na acetate pH 5.0 and ethanol precipitated. The pellets were resuspended in 10 μl of buffer B (50 mM HEPES–KOH pH 7.9, 200 mM NaCl, 5 mM MgCl2) of which 7 μl was used for the in vitro‐binding assay and the other 3 μl was kept as the input fraction. The binding reaction, the separation of the bound and unbound fractions and the RNA extractions were performed as descibed (Arts et al., 1998) and in the section ‘in vitro binding’, 2 μg 5S RNA were used as carrier RNA. The washed and dried RNA pellets were dissolved in 10 μl 0.1 M Tris–HCl pH 9.0 and incubated for 10 min at 50°C to hydrolyse the RNA backbone at ethylated phosphates. The reaction was stopped by the addition of 50 μl 0.3 M Na acetate pH 5.0, ethanol precipitated and the washed pellet resuspended in 6 μl RNA‐loading dye. Again, equal counts were loaded on a 15% polyacrylamide‐7 M urea denaturing gel.
In vitro‐binding assay
In vitro‐binding assays of internally labelled, in vitro transcribed tRNA variants using immoblized zz‐exportin‐t or z‐importin‐β were as described previously (Arts et al., 1998). The bound and unbound fractions were incubated for 30 min at 37°C in 300 μl homogenization mixture containing 50 mM Tris–HCl pH 7.5, 5 mM EDTA, 1.5% SDS, 300 mM NaCl and 1.5 mg/ml proteinase K. Samples were phenol–chloroform extracted and ethanol precipitated with 10 μg glycogen as a carrier. RNAs were analysed on a 10% polyacrylamide‐7 M urea denaturing gel.
An affinity column carrying 1.5 mg recombinant human exportin‐t was prepared, using 500 μl of Affigel‐15 beads (Bio‐Rad) that had been quickly washed twice with ice cold deionized water. The beads were rotated for 4 h at 4°C in a total volume of 4 ml Transport buffer (20 mM HEPES–KOH pH 7.5, 140 mM K acetate, 5 mM Mg acetate, 1 mM DTT and 250 mM sucrose) containing the protein. The beads were spun down at 1000 r.p.m. for 5 min at 4°C and the supernatant removed. Three millilitres of 0.1 M ethanolamine pH 8.0 were added to the beads, and the mixture was rotated overnight at 4°C. The blocked beads were washed with 20 ml PBS and rotated with 10 ml immune serum, from a rabbit injected with recombinant human exportin‐t, for 3 h at 4°C. The column was allowed to settle and the supernatant drained. The column was washed with 20 ml of PBS and eluted with 10 ml of 0.2 M glycine pH 2.2, 0.5 M NaCl. Fractions of 1 ml were collected in tubes containing 0.5 ml 1 M Tris–HCl pH 8.0. The flow‐through fraction was collected and the binding and elution cycle repeated twice more. The OD280 of the fractions was measured and the most concentrated fractions pooled. Protein was precipitated with 60% (final) ammonium sulfate by stirring for 2 h at 4°C and spun at 20 000 r.p.m. at 4°C for 30 min. The pellet was resuspended in 600 μl PBS and dialysed overnight against PBS (final protein concentration 4 mg/ml). This affinity purified fraction was concentrated a further 10× with centrifugal concentrators (Nanosep, Pall Filtron Corporation) and used for micro‐injection experiments in Xenopus oocytes.
Microinjections of RNA into X.laevis oocytes, incubations, extractions and analyses were performed as described previously (Jarmolowski et al., 1994). In experiments where recombinant exportin‐t was co‐injected with the RNA mixtures, 0.15 pmol exportin‐t in Transport buffer was injected per oocyte. In control injections Transport buffer was used. In the double nuclear injection experiments with the affinity purified antibody raised against human exportin‐t, ∼0.7 μg antibody was injected into each oocyte nucleus. Oocytes were then incubated for 30 min at 19°C before injection of RNA mix. Both injection mixtures contained dye to allow selection for double‐injected nuclei.
Cloning of tRNA genes
The tRNA genes described in this paper were obtained as follows: two oligonucleotides were synthesized; (i) a sense oligo containing at the 5′ end a cloning site (HindIII) and the T7 promoter sequence followed by the 5′ half of the desired tRNA sequence, (ii) an antisense oligo containing at its 5′ end a cloning site (BamHI or XbaI), a restriction site for linearization followed by the antisense sequence of the 3′ half of the desired tRNA. The oligos were designed so that they could base pair with one another at their 3′ ends for ∼15 nucleotides. The sense and antisense oligos (15 μM of each) in annealing buffer (50 mM Tris–HCl pH 7.9, 10 mM MgCl2, 100 mM NaCl, 1 mM DTT) were denatured for 5 min at 95°C and allowed to anneal by incubation for 3 min at 65°C then slowly cooled to room temperature. From this mixture, 37.5 fmoles annealed oligos were extended by incubating for 5 min at 37°C, 10 min at 20°C in a 30 μl reaction containing 2.5 U Klenow fragment (Gibco‐BRL) and 0.2 mM of each dNTP in the buffer recommended by the manufacturer. The double‐stranded fragment was digested with BamHI (or XbaI) and HindIII and cloned into the appropriate cloning sites of pUC19. The inserts were completely sequenced. Details of the oligonucleotides used are available on request ().
32P‐labelled RNAs were synthesized according to Jarmolowski et al. (1994). Templates which are described elsewhere are: T7‐U1ΔSm, T7‐U6Δss, T7‐ftz (Hamm et al., 1987; Rio, 1988; Hamm and Mattaj, 1989; Jarmolowski et al., 1994; Boelens et al., 1995).
The minihelix RNA was obtained as a ribo‐oligonucleotide that was fractionated on a 10% polyacrylamide denaturing gel, visualized by UV shadowing and extracted. The RNA was labelled with [γ‐32P]ATP and T4 polynucleotide kinase.
Analysis of aminoacylated tRNAs
Labelled tRNAs were injected into either the nucleus or the cytoplasm of Xenopus oocytes. After 30 min incubation, RNA was extracted from total oocytes. Samples were extracted under acid conditions, using buffer I (8 M Guanidine‐HCl, 0.1 M Na acetate pH 4.5, 0.5% Sarkosyl, 10 mM EDTA, 5 mM DTT), followed by phenol–chloroform extraction (equilibrated with 0.3 M Na acetate pH 4.5, 10 mM EDTA) and then precipitated with ethanol. The pellet was resuspended in buffer II (8 M urea, 0.3 M Na acetate pH 4.5, 0.2% Sarkosyl, 10 mM EDTA, 1 mM DTT) and extracted twice with phenol–chloroform prior to ethanol precipitation. The pellet was resuspended in 0.3 M Na acetate pH 4.5, 0.1 mM EDTA, and re‐precipitated with ethanol. To deacylate the tRNAs, extraction was performed at pH 7.9. The samples were then treated as described above and ethanol precipitated. Pellets were washed, dried and resuspended in acid dyes (0.4 M Na acetate pH 5.0, 50% urea, 30% glycerol, 0.05% bromophenol blue, 0.05% xylene cyanol). The samples were loaded on a 10% polyacrylamide (ratio of acrylamide to bisacrylamide, 19:1), 8 M urea, 0.1 M Na acetate pH 5.0, denaturing gel. The 20 cm long gel was run for 20 h at 9 V/cm at 4°C using 0.1 M Na acetate pH 5.0 as running buffer and dried before exposure.
We thank D.Kern for the gift of phenylalanyl‐tRNA synthetase. We are also grateful to Matthias Hentze, David Tollervey and the members of our laboratory, in particular Maarten Fornerod, for critical comments on the manuscript and discussion of the project and Elsebet Lund for pointing out that the methionyl tRNA T7 transcript carries a short 5′ extension. G.‐J.A. and S.K. hold a TMR Marie Curie and an NIH research training fellowship, respectively.
- Copyright © 1998 European Molecular Biology Organization