The yeast AP‐1‐like transcription factor, Yap1p, activates genes required for the response to oxidative stress. Yap1p is normally cytoplasmic and inactive, but will activate by nuclear translocation if cells are placed in an oxidative environment. Here we show that Yap1p is a target of the β‐karyopherin‐like nuclear exporter, Crm1p. Yap1p is constitutively nuclear in a crm1 mutant, and Crm1p binds to a nuclear export sequence (NES)‐like sequence in Yap1p in the presence of RanGTP. Recognition of Yap1p by Crm1p is inhibited by oxidation, and this inhibition requires at least one of the three cysteine residues flanking the NES. These results suggest that Yap1p localization is largely regulated at the level of nuclear export, and that the oxidation state affects the accessibility of the Yap1p NES to Crm1p directly. We also show that a mutation in RanGAP (rna1‐1) is synthetically lethal with crm1 mutants. Yap1p export is inhibited in both rna1‐1 and prp20 (RanGNRF) mutant strains, but Yap1p rapidly accumulates at the nuclear periphery after shifting rna1‐1, but not other mutant cells to the non‐permissive temperature. Thus, disassembly of export complexes in response to RanGTP hydrolysis may be required for release of substrate from a terminal binding site at the nuclear pore complex (NPC).
Transport of proteins across the nuclear envelope is highly selective and can be temporally regulated. For example, many transcription factors are maintained in an inactive state in the cytoplasm until a signal is received that promotes their translocation into the nucleus, either downstream of a particular signal transduction pathway or at a specific point in the cell cycle (Karin and Hunter, 1995). Until recently such regulation has been attributed to changes in the rate of nuclear import; however, it has now been demonstrated that proteins can also be actively exported from the nucleus to the cytoplasm (Fornerod et al., 1997; Fukuda et al., 1997; Ossareh‐Nazari et al., 1997; Stade et al., 1997) and that export may also be an important regulatory mechanism by which the steady‐state level of a protein within the nucleus can be controlled (Beals et al., 1997; Gorner et al., 1998; Roth et al., 1998; Toyoshima et al., 1998; Yang et al., 1998).
The master regulator of nuclear transport is a Ras‐like GTPase called Ran (Melchior et al., 1993; Moore and Blobel, 1993), which is thought to serve as a marker for the nuclear compartment (Izaurralde et al., 1997; Gürlich, 1998). Although Ran itself is present in both nucleus and cytoplasm, the proteins that regulate its nucleotide‐bound state are compartmentalized. The Ran GTPase activating protein (RanGAP/Rna1p; Becker et al., 1995; Bischoff et al., 1995) is cytosolic (Hopper et al., 1990), whereas the guanine nucleotide exchange factor (RCC1/Prp20p; Bischoff and Ponstingl, 1991) is exclusively nuclear (Clark et al., 1991). Thus, Ran should be primarily in the GTP‐bound form in the nucleus and in the GDP‐bound form in the cytoplasm. Indeed, mislocalization of RanGAP to the nucleus, mutation of PRP20 or RNA1, or introduction of GTPase or exchange‐defective Ran mutants all inhibit nuclear transport (Palacios et al., 1996; Weis et al., 1996; Gürlich et al., 1997; Izaurralde et al., 1997; Richards et al., 1997; Oki et al., 1998).
The targets of Ran are members of a family of proteins referred to as β‐karyopherins (also known as β‐importins), each of which has a binding site for the GTP‐bound form of Ran (Rexach and Blobel, 1995; Lounsbury et al., 1996). The karyopherins are soluble cargo receptors that shuttle across the nuclear pore complex (NPC) by binding to specific nucleoporins (Izaurralde and Adam, 1998). The best characterized is Kap95 which, along with an auxiliary protein termed karyopherin α, binds to and imports substrates that contain an SV40‐type nuclear localization sequence (NLS) (Gürlich et al., 1994, 1995a,b; Enenkel et al., 1995; Moroianu et al., 1995). The GTP‐bound form of Ran binds to Kap95 with high affinity and causes it to dissociate from the α subunit and the NLS‐bearing substrate (Rexach and Blobel, 1995). Because RanGTP is restricted to the nucleus, cargo should therefore be released preferentially within the nuclear interior, resulting in import against a concentration gradient. Other members of the karyopherin family have since been shown to import distinct classes of substrates, including ribosomal proteins (Rout, 1997), hnRNP proteins (Aitchison et al., 1996; Nakielny et al., 1996; Bonifaci et al., 1997; Fridell et al., 1997; Pemberton et al., 1997), snRNPs (Huber et al., 1998) and tRNA processing factors (Rosenblum et al., 1997).
Recently it was shown that three different members of the β‐karyopherin family mediate nuclear export rather than import. A sequence necessary and sufficient to promote rapid nuclear export has been identified on several proteins, including HIV Rev, TFIIIA and PKI (Fischer et al., 1995; Wen et al., 1995; Fridell et al., 1996; Figure 10). This nuclear export sequence (NES) consists of several leucine residues with uneven spacing between, designated L(X)2‐3L(X)2‐3LXL (Bogerd et al., 1996), and is recognized by Crm1p, a member of the β‐karyopherin family (Fornerod et al., 1997; Fukuda et al., 1997; Ossareh‐Nazari et al., 1997; Stade et al., 1997). Another β‐karyopherin‐like protein, CAS, is responsible for recycling karyopherin α back to the cytoplasm (Kutay et al., 1997). The region of karyopherin α that is recognized by CAS has yet to be determined. Finally, Exportin(tRNA) exports tRNA (Arts et al., 1998; Kutay et al., 1998). In contrast to the karyopherins that mediate import, Crm1p, CAS and Exportin(tRNA) bind to their respective substrates strongly only in the presence of RanGTP, and in each case a cooperative complex is formed between karyopherin, RanGTP and cargo (Fornerod et al., 1997; Kutay et al., 1997, 1998; Arts et al., 1998). Thus, assembly of export complexes would be favored in the nucleus, whereas dissociation would result when cytoplasmic RanGAP promotes hydrolysis of the bound RanGTP.
The elucidation of pathways that mediate nuclear exit suggests an additional mechanism by which the relative steady‐state level of proteins in the nuclear and cytoplasmic compartments can be maintained. Thus, what has previously been referred to as ‘retention’ of proteins within the cytoplasm could reflect an increase in the rate of nuclear export relative to nuclear import, rather than simply an inhibition of nuclear import. In fact, there are few instances in which cytoplasmic retention has been directly attributed to inhibition of nuclear import.
Well before the role of Crm1p in nuclear export was elucidated, a number of crm1 mutants were identified by virtue of their altered sensitivity to a variety of drugs, including staurosporine, caffeine and leptomycin B (Nishi et al., 1994; Turi et al., 1994; Kumada et al., 1996). The latter is now known to inhibit Crm1p directly (Fornerod et al., 1997; Fukuda et al., 1997; Ossareh‐Nazari et al., 1997; Kudo et al., 1998), but resistance to other drugs may be the result of deregulated transcriptional activity in crm1 mutants. Specifically, Schizosaccharomyces pombe crm1 mutants constitutively express genes activated by the AP‐1 transcription factor Pap1p (Toda et al., 1992). Pap1p, and the Saccharomyces cerevisiae ortholog, Yap1p, is activated by oxidative or toxic stress and mediates transcriptional activation of a number of target genes involved in the stress response (Schnell et al., 1992; Wu et al., 1993; Gounalaki and Thireos, 1994; Hirata et al., 1994; Kuge and Jones, 1994; Toone et al., 1998). Interestingly, Yap1p activity is regulated by differential localization; under normal conditions it is maintained in the cytoplasm but it becomes nuclear in response to oxidative stress (Kuge et al., 1997). The C‐terminal domain containing three critical cysteine residues is required for maintaining Yap1p in the cytoplasm (Kuge et al., 1997).
Here we show that under normal conditions Yap1p is restricted to the cytoplasm by virtue of rapid Crm1p‐mediated export, rather than by inhibition of import. We also show that Yap1p has an NES embedded within its cysteine‐rich domain (CRD) and that Yap1p binding to Crm1p is sensitive to oxidation. Mutational analysis suggests that the cysteine residues serve as redox sensors which regulate the availability of the NES, and thus the activity of the transcription factor. We also provide evidence that RanGAP activity is required to release Crm1p‐associated export cargo into the cytoplasm.
We identified the S.cerevisiae CRM1 gene in a screen for mutations that are lethal in combination with a temperature‐sensitive mutation in the yeast RanGAP, rna1‐1. We obtained three different crm1 alleles (crm1‐1, 1‐2 and 1‐3; Table I), as shown by complementation and genetic crosses to a marked CRM1 allele (see Materials and methods). Although CRM1 is essential (Toda et al., 1992), none of our alleles has an appreciable growth phenotype on its own. However, crm1‐3 is resistant to 10 mM caffeine (data not shown).
Maintenance of Yap1p in the cytoplasm requires Crm1p function
Schizosaccharomyces pombe crm1 mutants had been identified previously that constitutively express genes dependent on the AP‐1 like transcription factor, Pap1p (Toda et al., 1992; Kumada et al., 1996). In S.cerevisiae, the cellular localization of the Pap1p ortholog (Yap1p) is regulated. Under normal conditions Yap1p is cytosolic, but it becomes nuclear in response to oxidative stress (Kuge et al., 1997). Because Crm1p has recently been identified as a nuclear export factor (Fornerod et al., 1997; Fukuda et al., 1997; Neville et al., 1997; Ossareh‐Nazari et al., 1997; Stade et al., 1997), these observations suggested to us that the maintenance of Yap1p in the cytoplasm might rely on rapid Crm1p‐dependent export of Yap1p, rather than on inhibition of its nuclear import, as was previously proposed (Kuge et al., 1997). To test whether Yap1p localization is altered in our crm1 mutant alleles, we fused full‐length Yap1p to green fluorescent protein (GFP). The Yap1‐GFP fusion protein is functional because it is able to suppress the oxidation‐sensitive phenotype of yeast cells lacking a functional copy of the YAP1 gene (Figure 1A). We found Yap1‐GFP to be predominantly cytoplasmic in wild‐type strains, consistent with previous results (Kuge et al., 1997), but largely nuclear in all of our crm1 mutant strains (Figure 1B and data not shown). The extent of mislocalization was striking considering the lack of any growth phenotype in the crm1 mutants. We conclude that the cytoplasmic localization of Yap1p is dependent on the export factor Crm1p. Similar results were recently obtained for Pap1p (Toone et al., 1998).
Crm1p binds to Yap1p
To determine whether the effect of crm1 mutants on Yap1p localization is direct, we looked for an interaction between Crm1p and Yap1p. First, we used a bacterially expressed glutathione S‐transferase (GST)‐Yap1 fusion protein to precipitate proteins from cytosolic extracts of yeast strain LDY903, which carries a functional HA epitope‐tagged copy of CRM1 (see Materials and methods). GST‐Yap1p did not precipitate Crm1p under these conditions (Figure 2A, lane 6). However, when we added bacterially expressed, GTP‐loaded His6‐Gsp1pG21‐V (a yeast Ran mutant, defective for GTP hydrolysis), GST‐Yap1p precipitated a small amount of Crm1p (Figure 2A, lane 7). The amount was somewhat less than that precipitated by GST‐NESPKI, a known Crm1p substrate (Figure 2A, lane 5), and the significance of this result will be addressed below. In both cases, His6‐Gsp1pG21‐V‐GTP was also precipitated (Figure 2A, lower part of panel). Thus, Crm1p forms a complex with Yap1p and does so only in the presence of RanGTP. This behavior is what one would expect of an export complex formed by Crm1p (Fornerod et al., 1997).
We also tested the interaction between Crm1p and Yap1p in the yeast two‐hybrid system (Figure 2B). Full‐length YAP1‐B42AD produced a strong signal in combination with CRM1‐LexABD, as well as with KAP123‐LexABD, but also showed some background with an empty LexABD plasmid, most probably because of its potential for transactivation. We therefore tested a construct containing only the C‐terminal CRD of Yap1p, because this region had previously been shown to mediate the localization of Yap1p (Kuge et al., 1997) and should not confer transactivation. YAP1(CRD)‐B42AD retained a strong interaction with CRM1‐LexABD, but not with vector alone or other members of the β‐karyopherin family. Thus, Crm1p interacts with Yap1p through the CRD. We also note that the strong interaction between full‐length YAP1‐B42AD and KAP123‐LexABD raises the possibility that Kap123p binds to a different region of Yap1p, perhaps one required for import of Yap1p.
The yap1p CRD contains an NES
A canonical NES first described in the HIV‐1 Rev and cellular protein kinase inhibitor (PKI) proteins (Fischer et al., 1995; Wen et al., 1995) has recently been shown to mediate binding of export substrates to Crm1p (Fornerod et al., 1997; Fukuda et al., 1997; Ossareh‐Nazari et al., 1997; Stade et al., 1997). Inspection of the CRD of Yap1p revealed a sequence rich in hydrophobic residues (Y611SDIDVDGLCSEL), although not with the consensus spacing [L(X)2‐3L(X)2‐3 LXL; Bogerd et al., 1996; Figure 10]. To determine whether these hydrophobic residues mediate Yap1p export, we made various single and double point mutations to alanine in the context of the full length Yap1‐GFP fusion protein (Figure 3A). All of the fusion proteins were expressed at similar levels (Figure 3C). Yap1Y611‐A, I614‐A‐GFP showed normal cytoplasmic localization (Figure 3B, panels c and d). In contrast, Yap1L619‐A‐GFP was constitutively nuclear (Figure 3B, panels g and h), similar to the behavior of wild‐type Yap1‐GFP in a crm1 mutant background (Figure 1B). The same phenotype was observed with the Yap1I614‐A, V616‐A‐GFP double mutant (Figure 3B, panels e and f). Thus, L619 and V616 are both critical for export. We also note that Yap1C620‐T‐GFP was shown previously to be constitutively nuclear, whereas Yap1C620‐A‐GFP was localized normally (Kuge et al., 1997). This result could be explained if certain mutations at this position interfere with NES function rather than with regulation of Yap1p localization in response to oxidation, as discussed below.
To determine whether mutations that lead to nuclear accumulation of Yap1‐GFP also abrogate binding between Yap1p and Crm1p, we introduced the L619‐A and C620‐T mutations into the GST‐Yap1 fusion protein. Neither mutant fusion protein showed a detectable interaction with Crm1p under conditions in which wild‐type GST‐Yap1p interacts very efficiently (Figure 4A). As summarized in Table II, we also tested several other mutant fusion proteins. In all cases, the results correlate well with the localization of the various Yap1‐GFP mutant proteins.
We also tested the mutants in the two‐hybrid assay with CRM1‐LexABD. With the yap1(CRD)L619‐A‐B42AD, yap1(CRD)C620‐T‐B42AD and yap1(CRD)C629‐A‐B42AD mutants we observed no detectable signal, whereas all other mutants tested interacted normally (Figure 4B, Table II). Again, the two‐hybrid results correlated exactly with the localization of the corresponding GFP fusion protein. We conclude that Yap1p interacts with Crm1p through an NES‐like sequence within the CRD, and that this interaction is required for export of Yap1p and its steady‐state maintenance in the cytoplasm.
The interaction between Crm1p and Yap1p is regulated by oxidation
The results described above suggest that the steady state cytosolic localization of Yap1p is maintained by rapid export from the nucleus via Crm1p. We sought next to determine whether the accumulation of Yap1p in the nucleus in response to oxidative stress was triggered by inhibiting the binding of Crm1p to Yap1p. The three cysteine residues in and around the Yap1p NES have been shown to be critical for this aspect of Yap1p regulation. If all three are changed to amino acid residues which individually have no effect on Yap1p localization, Yap1p can no longer respond to oxidation and remains cytoplasmic (Kuge et al., 1997). An attractive scenario is that oxidation of the cysteine residues somehow masks the NES, thus preventing export of Yap1p and allowing it to accumulate in the nucleus.
In the experiments described above (Figure 2), the interaction between GST‐Yap1p and Crm1p was assayed in the absence of reducing agents, conditions that are likely to promote oxidation during the binding step. To determine whether binding might be strengthened under reducing conditions, we repeated the experiment in the presence or absence of 1 mM dithiothreitol (DTT) (Figure 5). We found that ∼20‐fold more Crm1p was precipitated by GST‐Yap1p in the presence of reducing agent (Figure 5, compare lanes 8 and 9). Binding remained completely dependent on added RanGTP. This effect was specific to Yap1p, because GST‐NESPKI precipitated somewhat less Crm1p in the presence of reducing agent, perhaps due to competition by endogenous Yap or other redox‐sensitive proteins.
To determine whether the cysteine residues in the Yap1p CRD regulate the interaction with Crm1p, these three residues were mutated to alanine or threonine, and the interaction of the mutant GST‐Yap1p fusion proteins with Crm1p was assessed in vitro (Figure 6). GST‐Yap1pC620‐A and GST‐Yap1pC598‐T, C629‐T mutant proteins precipitated Crm1p in a redox‐sensitive manner similar to wild‐type GST‐Yap1p, consistent with the observation that the localization of the corresponding Yap1‐GFP mutants responds to oxidation in a manner indistinguishable from wild‐type (Kuge et al., 1997). However, the triple GST‐Yap1pC598‐T, C620‐A, C620‐T mutant, which confers constitutive cytoplasmic localization to Yap1‐GFP (Kuge et al., 1997), bound strongly to Crm1p even in the absence of reducing agents, showing that the interaction is no longer sensitive to oxidation. These results suggest that oxidation of any one of the cysteine residues in the CRD activates Yap1p by masking the NES and preventing Crm1p‐mediated nuclear export.
Addition of oxidants, such as diamide, to the growth medium has been shown to cause translocation of Yap1p to the nucleus (Kuge et al., 1997). This allowed us to use the two‐hybrid assay to test whether the interaction between Yap1p and Crm1p was sensitive to oxidants in vivo (Figure 7). Treatment of cells with 0.75 mM diamide inhibited the interaction between wild‐type YAP1(CRD)‐B42AD and CRM1‐LexABD, as well as that between yap1(CRD)C620‐A‐B42AD or yap1(CRD)C598‐TC629‐T‐B42AD and CRM1‐LexABD. In contrast, the interaction of CRM1‐LexABD with the triple yap1(CRD)C598‐T, C620‐A, C629‐T‐B42AD mutant was completely resistant to diamide. The two‐hybrid results exactly parallel those obtained in the in vitro binding studies (Figure 6), and taken together they provide strong support for the idea that oxidative stress promotes the accumulation of Yap1p in the nucleus by inhibiting formation of an export complex with Crm1p.
Our results suggest that a single cysteine residue within the CRD is sufficient to confer redox sensitive binding to Crm1p, making it unlikely that masking of the NES requires disulfide linkages between two cysteine residues. However, Yap1p contains three other cysteine residues (C303, C310, C315) outside of the CRD which might participate in the response to oxidation. We therefore mutated these residues in the context of GST‐yap1C598‐T, C629‐T, leaving only a single cysteine residue (C620) remaining. A GST‐Yap1 fusion protein carrying the five cysteine mutations displayed DTT‐dependent interaction with Crm1p, similar to wild‐type GST‐Yap1p (Figure 6). No difference in binding was observed when the GST moiety was cleaved off, ruling out the possibility that cysteine residues within GST itself can contribute to redox sensitivity (data not shown). We conclude that a single cysteine residue within the CRD is necessary and sufficient to mediate the response to oxidative stress in the context of full‐length Yap1p.
Yap1‐GFP is concentrated at the nuclear periphery in an rna1‐1 mutant
The three crm1 alleles described here were obtained in a screen for synthetic lethality with the temperature‐sensitive RanGAP mutant, rna1‐1. RanGAP is thought to be required for dissociation of export factors in the cytoplasm and subsequent recycling to the nucleus (Bischoff and Gürlich, 1997; Floer et al., 1997; Gürlich et al., 1997; Lounsbury and Macara, 1997). Thus, the export defect of a crm1 mutant might be exacerbated by inefficient recycling, explaining the synthetic lethality between crm1 and rna1‐1 mutants. To determine whether the rna1‐1 mutation itself has any effect upon the localization of Yap1p, we transformed strains W303 (wild‐type) and LDY546 (rna1‐1) with YAP1‐GFP, and localized the fusion protein. At the permissive temperature, Yap1‐GFP was localized throughout the cell (Figure 8). However, within 15 min of shifting rna1‐1 to the non‐permissive temperature, Yap1‐GFP became predominantly nuclear. Importantly, we observed consistently a distinct nuclear rim staining pattern in many cells, suggesting that Yap1‐GFP was accumulating at or near the nuclear envelope. This pattern, as well as the speed with which nuclear accumulation occurred, suggests that dissociation of the export complex by RanGAP may be required to release substrate from the NPC, perhaps at a terminal step of export. Consistent with this idea, we found that the behavior of Yap1‐GFP in the prp20‐1 mutant differed from that in rna1‐1. Although Yap1p‐GFP also accumulated in the nucleus at the non‐permissive temperature in prp20‐1 mutant cells, it took longer (∼45 min) for significant nuclear accumulation to develop. Furthermore, we never observed rim staining similar to that seen with the rna1‐1 mutant (Figure 8). The behavior of the prp20‐1 mutant is consistent with the notion that gradual depletion of the nuclear pool of RanGTP should inhibit formation of export complexes, trapping Yap1‐GFP within the nucleus.
Yap1p has a potential NLS at its N‐terminus that could mediate its nuclear import, presumably by karyopherin α/β or a related protein. The dependence of karyopherin‐mediated nuclear import on Ran has been well documented (Moore and Blobel, 1993; Melchior et al., 1995), and is thought to reflect the ability of nuclear RanGTP to dissociate import complexes once they have crossed the NPC (Gürlich et al., 1996). In spite of this, Yap1p relocated from the cytoplasm to the nucleus in both rna1‐1 and prp20 mutants, suggesting that import of Yap1p was not affected at early time points. To investigate this question more closely, we took advantage of the yap1L619‐A mutant, which cannot be exported and is constitutively nuclear (Figure 3), to isolate the import step. We found that Yap1L619‐A‐GFP relocated from the nucleus to the cytoplasm after shifting rna1‐1 mutant cells to the non‐permissive temperature, but did so with somewhat slower kinetics than the block to nuclear export of wild‐type Yap1‐GFP (Figure 9). The behavior of Yap1L619‐A‐GFP is identical to that of an NLS‐GFP reporter protein examined in parallel. Furthermore, Yap1L619‐A‐GFP never showed accumulation at the nuclear rim. These results suggest that import of Yap1‐GFP is also blocked in the rna1‐1 mutant, but this occurs more slowly than inhibition of export. Furthermore, a functional NES is required to trap Yap1‐GFP at the nuclear rim in the rna1‐1 mutant.
We have shown that the yeast AP‐1‐like transcriptional activator, Yap1p, is a regulated substrate of the Crm1p export factor. Yap1p binds to Crm1p via an NES located in the Yap1p CRD. This interaction is inhibited under conditions of oxidative stress, allowing Yap1p to accumulate in the nucleus and activate transcription of genes required for the stress response. The three cysteine residues within the CRD are crucial for this inhibition and may serve as redox sensors that induce masking of the NES upon oxidation. These observations reveal a novel mechanism by which nuclear transport can be temporally regulated.
The Yap1p NES is unusual in that it contains mostly hydrophobic amino acids other than leucine and the spacing between these residues is not canonical (Figure 10). It does have several acidic residues in between the hydrophobic residues, a feature common to many NESs. This difference may simply reflect the fact that yeast NESs have diverged significantly from their metazoan counterparts. Leptomycin B binds to human and S.pombe Crm1p, and inhibits its interaction with the NES, but does not bind to S.cerevisiae Crm1p (Fornerod et al., 1997). If Leptomycin B acts as a competitor for the NES binding pocket as suspected, this result suggests that this region of Crm1p is significantly divergent in S.cerevisiae. Such a difference would also explain why Rev functions only poorly in yeast (Stutz and Rosbash, 1994). An equally plausible explanation is that proteins such as Yap1p, whose binding to Crm1p must be tightly regulated, have NESs that bind less strongly and are thus more sensitive to perturbation. Finally, it is possible that this region of Yap1p is not a bona fide NES, but instead mediates binding to another protein that contains a canonical NES. In this regard, we note that NES‐like sequences are very similar to leucine‐rich repeats that mediate other protein‐protein interactions. The answer to these questions should become clear as more yeast NESs are identified.
In this regard, it is important to note that our crm1 mutants have no growth phenotype, yet they profoundly mislocalize Yap1p. The essential substrate(s) of Crm1p, which are not yet known, could have more efficient NESs and thus not be mislocalized in our mutants. Alternatively, export of all Crm1p substrates may be very inefficient in the crm1 mutants, but this may be tolerated under normal conditions. If this is the case, then it is unlikely that negative regulation of proteins like Yap1p is the essential function of Crm1p because our crm1 mutants are clearly not able to maintain enough export to keep these substrates inactive. Another class of potential Crm1p substrates are proteins that mediate RNA export. Crm1p has been shown to mediate snRNA export in metazoan cells (Fornerod et al., 1997) and may also be required for export of other RNA species, including mRNA (Pasquinelli et al., 1997; Stade et al., 1997). While absolutely essential, such export may not be limiting under normal laboratory conditions, and thus would need to be only marginally effective to maintain viability.
How do the cysteine residues within the Yap1p CRD mask the NES? The simplest explanation would be that they form intrachain disulfide linkages with one another or with other cysteine residues within Yap1p, thus embedding the NES in a conformation that is not accessible to Crm1p. However, we find that a Yap1p mutant missing all but a single cysteine residue within the CRD still binds to Crm1p in a redox‐sensitive manner, suggesting that intrachain disulfide linkages are not required. A different possibility is that the cysteine residues within the Yap1p CRD form linkages with a third protein, as yet unidentified, which serves to mask the NES. This question is currently under investigation.
The behavior of the Yap1‐GFP fusion protein in an rna1‐1 mutant provided interesting insights into the requirements for RanGAP in transport. Although RanGAP is required for both import and export, its effect on export appears to be much more direct. Wild‐type Yap1‐GFP relocated from cytoplasm to nucleus very rapidly after shifting rna1‐1 mutant cells to the non‐permissive temperature, suggesting that export was blocked at a time when import was still occurring. Under these conditions, Yap1‐GFP was often concentrated at the nuclear rim. In contrast, although Yap1‐GFP also accumulated in the nucleus in a prp20 mutant strain, we never observed rim staining. The simplest interpretation of these results is that RanGTP is required to assemble export complexes that are then targeted to the NPC, whereas RanGAP is required to dissociate these complexes and release substrates, perhaps from the terminal binding site at the cytoplasmic face of the NPC. We envision that the inability to release export complexes would cause subsequent substrates to stall within the NPC, leading to rapid cessation of export and accumulation of Yap1p at the NPC and within the nucleus. Consistent with this view, the Yap1L619‐A‐GFP mutant, which cannot bind to Crm1p, normally accumulates in the nucleus but never concentrates at the nuclear rim.
RanGAP could be required directly to hydrolyze RanGTP bound to export complexes. Alternatively, RanGAP could function indirectly by releasing Ran from RanBP1. RanBP1 has been shown in vitro to remove RanGTP from the CAS/importin α/RanGTP export complex, thereby inducing disassembly of the complex (Bischoff and Gürlich, 1997). RanGAP activity may therefore only be required to recycle RanBP1 from the RanGTP‐bound state after one round of dissociation. This idea would reconcile our results with those showing that microinjection of a nonhydrolyzable Ran mutant does not inhibit nuclear export (Izaurralde et al., 1997; Richards et al., 1997).
A growing number of proteins have now been shown to undergo regulated nuclear export. In most cases, the mechanism by which export is regulated has not been fully elucidated, but so far many appear to involve phosphorylation (Beals et al., 1997; Engel et al., 1998; Yang et al., 1998). Direct regulation of Yap1p export by oxidation uncovers a novel mechanism by which cells can sense and respond to environmental changes. However, because we do not yet understand exactly how the cysteine residues within the Yap1p CRD function, we cannot rule out the possibility that phosphorylation is also involved in the oxidation response. In this regard, we note that the S.pombe ortholog, Pap1p, is constitutively cytoplasmic in strains carrying a mutation in the stress‐activated MAP kinase Sty1p (Toone et al., 1998). Thus, phosphorylation by Sty1p may contribute to the inhibition of export upon oxidative stress in some unknown fashion. Alternatively, phosphorylation may regulate Yap1p/Pap1p import by a distinct mechanism.
Given the number of examples already in the literature, it seems likely that regulation of nuclear export is at least as common as that of nuclear import, and that the steady‐state localization of many proteins is largely determined by the relative rates of the two processes. Indeed, it may be that apparent exclusion from one compartment or the other is less dependent on a physical barrier to movement across the NPC than on the relative rates of import versus export.
Materials and methods
Enzymes for molecular biology were purchased from Boehringer Mannheim (Indianapolis, IN), Pharmacia (Piscataway, NJ) and New England Biolabs (Beverly, MA). Lyticase was purchased from Enzogenetics (Corvallis, OR). 5‐fluoroorotic acid (5FOA) was obtained through the Genetics Society of America consortium. A yeast genomic library carried in pRS314 (C.Connelly and P.Hieter, unpublished) was furnished by Dr Phil Hieter, as were all of the pRS vectors (Sikorski and Hieter, 1989). Anti‐GFP antibody was a kind gift from Jason Kahana and Pam Silver.
Strains and microbial techniques
The yeast strains and plasmids used are listed in Table I, and only relevant genotypes are listed in the text. Unless otherwise designated, all yeast strains are derived from W303. Mutants originally present in other strain backgrounds were crossed to W303 at least three times prior to use. Yeast cell culture, media preparation and genetic manipulations were performed essentially according to Sherman et al. (1986). Yeast shuttle plasmids and linear fragments were introduced into yeast by lithium acetate transformation (Ito et al., 1983). Selection against Ura+ strains was accomplished by culture on solid synthetic media containing 1 mg/ml 5FOA (Boeke et al., 1987). Sectoring assays were performed on solid synthetic media containing 25% of the normal amount of adenine and histidine supplements (SCsec). DNA cloning was performed using standard techniques outlined in Sambrook et al. (1989).
Synthetic lethal screen
An ade2 ade3 sectoring screen (Bender and Pringle, 1991) was used to identify mutants that are lethal in combination with rna1‐1. Yeast strains LDY431 (MATa rna1‐1 ade2 ade3 ura3 leu2 his7) and LDY432 (MATα rna1‐1 ade2 ade3 ura3 leu2 trp1) were transformed with pLDB382 (RNA1 URA3 ADE3 CEN) and mutagenized with either EMS (to 80% lethality) or ultraviolet (UV) irradiation (to 50% lethality). Colonies (150 000 total) were screened for the inability to form white sectors indicative of loss of the ADE3 containing plasmid. Nonsectoring colonies were further screened for sensitivity to 5‐FOA, indicating that they could not survive loss of the plasmid‐linked URA3 marker. Of 187 candidates obtained from this analysis, 55 reverted to Sec+ FOAr after transformation with pLDB93 (RNA1 TRP1 CEN), and 15 of these segregated 2:2 in backcrosses to either LDY431 or LDY432. We refer to these mutants as rna1‐1 lethal, or rle. The mutants fell into four complementation groups, the largest being RNA1 itself. The second largest group contained three members, rle1‐1 to 3 (LDY924, −928 and −932) The RLE1 gene was cloned by complementation of the non‐sectoring, FOAs phenotype of LDY924 (rle1‐1 rna11 <RNA1 CEN URA3 ADE3) with a yeast genomic library in pRS314. Three overlapping clones were obtained, all of which contained the CRM1 gene. An XhoI‐SalI fragment containing the entire CRM1 gene cloned into pRS315 (pLDB392) complemented all three rle1 alleles. Allelism of rle1 with CRM1 was confirmed by crossing each of the alleles to LDY884 or LDY885 (CRM1::LEU2) and showing segregation opposite the marked CRM1 allele. A precise disruption of CRM1 with the bacterial KAN gene was made by PCR using the KAN‐MX2 cassette as a template with hybrid oligonucleotides LOL6 and LOL18 (Baudin et al., 1993). The resulting fragment was transformed into diploid yeast strain LDY436, yielding strain LDY880. Replacement of the endogenous CRM1 gene was checked by PCR on genomic DNA, using LOL3 and LOL20. Sporulation and tetrad dissection showed that viability segregated 2:2, and all viable spores were Kans.
PCR primers are described in Table III.
YAP1 (CRD)‐B42AD and mutations. Full‐length YAP1, or the CRD alone (amino acid residues 526‐650), was PCR‐amplified from YEp351‐YAP1 to introduce EcoRI and XhoI sites at the ends. The resulting fragments were cloned into EcoRI‐XhoI sites of pJG4‐5 to create in‐frame fusions to the B42 activation domain (pLDB438 and pLDB439). Site‐directed mutations in pLDB438 were generated by two step PCR mutagenesis. In the first step, complementary mutagenic oligonucleotides (Table III) were used in separate PCR reactions with each of two outside primers, homologous to pJG4‐5 vector sequences flanking the YAP1 insert (COL11 and COL12). The resultant PCR fragments overlap 30‐50 bp in the region containing the point mutations. The two PCR fragments were gel purified, combined and subjected to five cycles of annealing/extension in the absence of added primers, prior to addition of COL11 and COL12, and continued cycling. The final PCR product was co‐transformed into yeast strain SM13 with pJG4‐5 that had been linearized with EcoRI and XhoI. Intact plasmids generated by gap repair were rescued from Trp+ transformants. Multiple point mutations were generated sequentially, using the same procedure.
YAP1‐GFP and mutations. ‘SuperGlow’ GFP (Kahana et al., 1995) was PCR‐amplified from pJK19‐1 using hybrid primers LOL23 and LOL26 that are flanked by sequences homologous to the 3′ end of YAP1, to produce a fragment containing GFP fused in frame to the last codon of YAP1. YEp351‐YAP1 was linearized at the 3′ NdeI site of YAP1 and co‐transformed with the PCR product into yeast strain SM13 (yap1‐Δ2::hisG). The intact plasmid generated by gap repair (pLDB419) was rescued from a Leu+ transformant expressing YAP1‐GFP. Site‐directed mutations in pLDB419 (Table I) were generated by two‐step PCR mutagenesis followed by gap repair, as described above, using the same mutagenic primer pairs with outside primers COL13 and COL14. For co‐transformation, pLDB419 was linearized with PacI and NcoI.
GST and His6 fusion proteins. Primers COL30 and COL31 were used to PCR amplify YAP1 or yap1 mutants from the YAP1‐GFP plasmid series described above. The PCR products were digested with EcoRV and EcoRI and ligated into pGEX2TK between SmaI and EcoRI, or pGEX4T‐1 between BamHI (blunt ended using Klenow) and EcoRI to generate the GST‐Yap1 series (Table I). To make GST‐NESPKI, primers MN35 and MN36 were annealed and cloned into EcoRI‐XhoI sites of pGEX4T‐1 to generate pLDB455. Construction of his6‐gsp1G21‐V (pLDB450) will be described elsewhere (J.Booth et al., in preparation).
CRM1HA. Oligonucleotide LOL13 was used to introduce the HA epitope into the most 3′ NsiI site of pLDB393 (CRM1 URA3 CEN), using the method of Kunkel et al. (1987) to generate pLDB396. pLDB396 was transformed into diploid yeast strain LDY880 (CRM1/crm1::KAN) prior to sporulation and tetrad dissection. Ura+ KanR spores were recovered and grew normally, showing that the epitope‐tagged CRM1 gene is functional.
Yeast cell fractionation, GST‐YAP1 precipitation, protein binding assays and immunoblot analysis
Diploid yeast strain LDY903 (crm1::KAN/crm1::KAN <CRM1HA URA3 CEN>) was grown in rich medium to mid‐log phase. Cells were spheroplasted and lysed, and the lysate was subjected to low‐speed centrifugation as described by Bogerd et al. (1994) to yield a supernatant fraction (S1) that contains the bulk of Crm1pHA. The S1 fraction was concentrated to one‐quarter volume using a Centricon 30 unit. His6‐Gsp1pG21‐V was purified from Escherichia coli using Ni‐NTA affinity chromatography as recommended by the manufacturer (Qiagen, Santa Clarita, CA). Induction of GST fusion proteins in E.coli was performed as described (Smith and Johnson, 1988). Induced cells were washed with ice‐cold 0.9% NaCl, collected by centrifugation for 10 min at 5000 r.p.m. in a Sorvall GS‐3 rotor and frozen in liquid nitrogen. Pellets were thawed at 37°C, resuspended in Buffer A [2.3 M sucrose, 50 mM Tris‐HCl pH 7.5 and protease inhibitiors (PI; Bogerd et al., 1994)] and incubated at 4°C for 30 min. Ten milliliters of Buffer B (50 mM Tris‐HCl pH 7.5, 10 mM KCl; 1 mM EDTA; 1 mM DTT; PI) containing 3 mg of lysozyme were added, followed by incubation on ice for 1 h. One 1/100 volume of 10% taurodeoxycholic acid, 1/100 volume of 1 M MgCl2 and 0.25 mg DNase were then added to lysed cells, and after incubation for 10‐15 min on ice, the lysate was centrifuged at 13 000 r.p.m. for 45 min at 4°C in a Sorvall SS‐34 rotor. To purify the GST fusion proteins, crude lysates were incubated with glutathione‐Sepharose at 4°C for 1 h in ELB2 (0.2 M NaCl, 0.05 M HEPES pH 7.0, 5 mM MgCl2, 0.1% Tween‐20) containing 1 mM DTT and PI. The resin was then washed five times in ELB2 + PI and resuspended to 50% bead volume.
Binding assays were performed in 100 μl volumes containing 60 μl yeast S1 fraction, 20 μl of 4× ELB2 (plus 1 mM GTP and PI, ± 1 mM DTT, ± 20 μg His6::Gsp1G21‐V‐GTP) and 20 μl of beads adsorbed with the GST fusion proteins. The mixture was rotated for 2 h at 4°C, and the beads were washed six times using ELB2 plus 0.25 mM GTP and PI, with or without 1 mM DTT. Bound proteins were eluted in Laemmli buffer and heated for 10 min at 95°C prior to SDS‐PAGE. Immunoblotting was performed essentially as described (Belanger et al., 1994) using mAb 12CA5 (Harvard cell culture facility, Cambridge, MA), anti‐EK (Invitrogen, Carlsbad, CA) and anti‐GST (Molecular Probes, Eugene, OR) antibodies. Bound antibodies were detected using the ECL detection system (Amersham, Cleveland, OH). In every experiment, blots were stripped and reprobed with anti‐GST antibody. In all cases, equivalent amounts of GST fusion protein were present in each lane (data not shown).
For analysis of Yap1‐GFP expression, whole‐cell lysates were prepared by boiling in SDS sample buffer, as previously described (Yan and Melese, 1993). Lysates were subjected to SDS‐PAGE and transferred to nitrocellulose. Blots were probed as described above, using rabbit anti‐GFP at a dilution of 1:5000.
pKW442 [CRM1‐LexABD (Stade et al., 1997)] and other β‐karyopherin fusions (Neville et al., 1997) were transformed into yeast strain RFY206. LDB432 (YAP1‐B42AD) and pLDB436 (YAP1 (CRD)‐B42AD) were transformed into yeast strain EGY48. To test interactions, EGY48 and RFY206 transformants were mated and diploids were selected on synthetic medium (2% glucose) lacking uracyl, tryptophan and histidine. Diploids were tested for LacZ expression by plating on synthetic medium (1% raffinose/2% galactose) lacking uracyl, tryptophan and histidine, containing 80 μg/ml X‐gal. To assay the interaction between CRM1‐LexABD and the yap1(CRD)‐B42AD mutants, CRM1 and yap1(crd) constructs were co‐transformed, along with the reporter plasmid pSH18‐34, into EGY48. Ura+ Trp+ His+ transformants were tested on X‐gal‐containing plates as described above.
To analyze the distribution of GFP fusion proteins, cells were grown in synthetic complete (SC) medium lacking the appropriate amino acid to early log phase, and then shifted to rich medium for 6 h. Cells were washed with phosphate‐buffered saline (PBS) and GFP signals were analyzed in living cells. Alternatively, cells were fixed in 75% ethanol for 15 min on ice, washed with PBS and stained using 1 μg/ml DAPI (added directly to fixed cells without further washes). Samples were observed using a Nikon Optiphot fluorescence microscope.
We thank Pam Silver, Scott Moye‐Rowley, Karsten Weis, Brett Finlay, Anita Hopper and Michael Rosbash for providing strains, antibodies and plasmids. We are grateful to Melissa Moore, Michael Rosbash, Jim Haber and members of the Rosbash and Davis laboratories for critical reading of the manuscript and for many insightful discussions. We particularly thank Megan Neville for providing the two‐hybrid constructs and James Booth for providing the His6‐gsp1G21‐V plasmid construct.
- Copyright © 1998 European Molecular Biology Organization