Nuclear protein import proceeds through the nuclear pore complex (NPC). Importin‐β mediates translocation via direct interaction with NPC components and carries importin‐α with the NLS substrate from the cytoplasm into the nucleus. The import reaction is terminated by the direct binding of nuclear RanGTP to importin‐β which dissociates the importin heterodimer. Here, we analyse the sites of interaction on importin‐β for its multiple partners. Ran and importin‐α respectively require residues 1–364 and 331–876 of importin‐β for binding. Thus, RanGTP‐mediated release of importin‐α from importin‐β is likely to be an active displacement rather than due to simple competition between Ran and importin‐α for a common binding site. Importin‐β has at least two non‐overlapping sites of interaction with the NPC, which could potentially be used sequentially during translocation. Our data also suggest that termination of import involves a transient release of importin‐β from the NPC. Importin‐β fragments which bind to the NPC, but not to Ran, resist this release mechanism. As would be predicted from this, these importin‐β mutants are very efficient inhibitors of NLS‐dependent protein import. Surprisingly, however, they also inhibit M9 signal‐mediated nuclear import as well as nuclear export of mRNA, U snRNA, and the NES‐containing Rev protein. This suggests that mediators of these various transport events share binding sites on the NPC and/or that mechanisms exist to coordinate translocation through the NPC via different nucleocytoplasmic transport pathways.
Nuclear pore complexes (NPCs) are the sites of exchange of macromolecules between nucleus and cytoplasm (Feldherr et al., 1984). NPCs allow diffusion of small molecules up to ∼40–60 kDa and can accommodate active transport of particles as large as 25 nm in diameter or several million Daltons in molecular weight (reviewed in Bonner, 1978). Active transport across the NPC occurs in both directions, is generally saturable, and thus carrier‐mediated. A number of kinetically distinct pathways have been characterized by competition experiments or by the signals that direct transport. Import of most nuclear proteins is triggered by nuclear localization signals (NLSs) consisting of one or more clusters of basic amino acids (for review, see Dingwall and Laskey, 1991). In the following, we will use the term ‘NLS’ in its strict sense, referring to this basic type of signal only. The M9 domain of hnRNP A1 is different in sequence and confers import into the nucleus as well as nuclear export (Michael et al., 1995). Nuclear export of proteins like HIV Rev is by virtue of a short, leucine‐rich nuclear export signal (NES) (Fischer et al., 1995; Wen et al., 1995). The term ‘NES’ is used here to describe this Rev‐type signal only.
It is probable that the export of RNAs out of the nucleus is directed by associated proteins carrying NESs, M9 domains or perhaps still unidentified export signals (for reviews, see Gerace, 1995; Izaurralde and Mattaj, 1995; Görlich and Mattaj, 1996). For example, HIV Rev achieves export of viral RNAs containing Rev response elements. The best candidates to mediate export of cellular mRNAs are hnRNP A1 and its relatives (Piñol‐Roma and Dreyfuss, 1993). The NES‐containing TFIIIA contributes to nuclear export of 5S RNA (Guddat et al., 1990), and the cap binding complex (CBC) is involved in export of capped U snRNAs (Izaurralde et al., 1995). It is well established that the different classes of RNAs leave the nucleus via distinct routes. For example, export of tRNA is competed by excess tRNA but not by mRNAs, U snRNAs, 5S RNA or by rRNA (Zasloff, 1983; Jarmolowski et al., 1994; Pokrywka and Goldfarb, 1995).
The import of NLS‐containing proteins is the best‐characterized nucleocytoplasmic transport pathway. Following the observation that import into the nuclei of digitonin‐permeabilized cells depends on the re‐addition of cytosol or cytosolic fractions (Adam et al., 1990; Moore and Blobel, 1992; Adam and Adam, 1994), four soluble factors implicated in nuclear protein import were purified and the corresponding genes have been identified: the GTPase Ran/TC4 (Melchior et al., 1993a; Moore and Blobel, 1993), importin‐α (Görlich et al., 1994; Imamoto et al., 1995b), importin‐β (Chi et al., 1995; Görlich et al., 1995a; Imamoto et al., 1995a; Radu et al., 1995) and NTF2 (Moore and Blobel, 1994; Paschal and Gerace, 1995). Importin has also been called karyopherin and NTF2 is also called pp15 or p10.
For import, the NLS‐protein binds in the cytoplasm to the importin‐α/β heterodimer (Görlich et al., 1995a; Imamoto et al., 1995c). Importin‐α provides the NLS binding site (Adam and Gerace, 1991; Görlich et al., 1995a; Weis et al., 1995). Importin‐α in turn interacts with importin‐β via the N‐terminal importin‐β binding domain (IBB domain) (Görlich et al., 1996a; Weis et al., 1996). Binding to importin‐β with this IBB domain is sufficient for efficient nuclear entry and the IBB domain can therefore be regarded as the nuclear targeting signal of importin‐α. The trimeric NLS–importin‐α/β complex docks via importin‐β to the cytoplasmic periphery of the NPC (Görlich et al., 1995b; Moroianu et al., 1995) and is subsequently translocated through the NPC as a single entity. Translocation is finally terminated at the nucleoplasmic side of the nuclear pore by the disassembly of the NLS–importin‐α/β complex and the release of the NLS‐protein and importin‐α into the nucleoplasm. The importin heterodimer is dissociated by direct binding of RanGTP to importin‐β (Rexach and Blobel, 1995; Görlich et al., 1996b). The inactivation of the Ran binding site in importin‐β prevents termination, but not translocation, i.e. transport intermediates accumulate in the nuclear baskets of the NPC and are not released into the nucleoplasm (Görlich et al., 1996b). This indicates that RanGTP normally binds to importin‐β only in the nucleus. The two subunits of importin are probably returned separately from the nucleus back to the cytoplasm.
NLS‐mediated nuclear protein import requires cytoplasmic RanGDP and, as discussed above, seems to depend on nuclear RanGTP (Görlich et al., 1996b). This obligatory gradient in Ran's nucleotide binding state is generated by the nuclear GDP/GTP exchange factor for Ran, RCC1, and cytoplasmic RanGAP1 (Rna1p), the Ran‐specific GTPase activating protein (Ohtsubo et al., 1989; Hopper et al., 1990; Bischoff and Ponstingl, 1991; Melchior et al., 1993b; Bischoff et al., 1994, 1995a; Corbett et al., 1995). Low cytoplasmic RanGTP levels allow importin‐α to bind importin‐β, the high nuclear RanGTP concentration favours importin heterodimer dissociation in the nucleus. The RanGTP gradient thereby determines from which compartment the importin heterodimer can transport an NLS substrate.
In addition to the role of Ran in defining nuclear and cytoplasmic identity, Ran's GTP cycle probably also drives the translocation through the NPC. To achieve this, Ran would have to interact with components distinct from importin‐β. Nuclear protein import requires cytoplasmic RanGDP (Görlich et al., 1996b) and GTP hydrolysis by Ran (Melchior et al., 1993a; Moore and Blobel, 1993). This, together with the observation that RanGDP binds to the NPC, suggests that translocation involves GDP/GTP exchange and GTP hydrolysis by Ran that is bound to the NPC. However, it is still unclear how the GTP cycle on NPC‐bound Ran would translate into a directed movement.
Importin‐β consists of 876 amino acids and interacts with the NPC, importin‐α and Ran. We have determined the domains required for these interactions: residues 1–364 are needed for high‐affinity Ran binding, and residues 331–876 for importin‐α binding. Release of importin‐α upon RanGTP binding to importin‐β (Rexach and Blobel, 1995; Görlich et al., 1996b) is therefore probably explained by a conformational switch in importin‐β. The NPC binding of importin‐β survives deletion from either terminus and involves more than one interaction site. The 45–462 importin‐β fragment binds to the NPC but does not interact with either importin‐α or Ran. It has only a small effect on import substrate docking to the cytoplasmic side of the NPC, but blocks sites used at later import steps very efficiently. It therefore appears to bind irreversibly to that subset of sites at the NPC from which importin‐β is normally released as an importin‐β–RanGTP complex. This is consistent with a model for termination in which RanGTP binding to importin‐β not only disassembles the importin heterodimer, but also releases importin‐β from the NPC before re‐export can occur. The 45–462 importin‐β fragment efficiently blocks nucleocytoplasmic transport of several substrates, including hnRNP A1 M9‐mediated import and the export of the NES‐containing Rev protein, mRNA and U snRNA, but it only weakly blocks export of tRNA. This could indicate that importin‐β binds to sites at the NPC which are shared with mediators of other transport pathways, i.e. that different nucleocytoplasmic transport pathways converge at the level of NPC binding. Alternatively, or in addition, it is also possible that a nuclear pore complex has to be cleared of one cargo before the next can pass the central transporter. The cross‐competition of importin‐β derivatives with other pathways could thus also be due to this ‘traffic control’.
Importin‐β function requires intact N‐ and C‐termini
Importin‐β is a key component of nuclear protein import. In order to support the import reaction it has to interact with at least three components: importin‐α, the nuclear pore complex and Ran. A more detailed molecular dissection of the import reaction requires knowledge of which part of importin‐β accounts for which interaction. To approach the problem, we first tested bacterially expressed recombinant importin‐β fragments for their ability to support complete import. Permeabilized HeLa cells were incubated with Ran, an energy regenerating system, fluorescent nucleoplasmin as the import substrate, importin‐α, and full‐length or truncated forms of importin‐β. The reactions were stopped by fixation and import analysed by confocal fluorescence microscopy. The upper panel of Figure 1 shows that full‐length importin‐β supports import. No nuclear accumulation of nucleoplasmin occurs if full‐length importin‐β was replaced by truncated fragments lacking either the N‐ or the C‐terminus.
The deletion of the first 44 residues of importin‐β arrests import intermediates at the nuclear envelope and gives rise to the punctate, ring‐like pattern typical for nuclear pore complexes. This confirms our previous report that 45–876 importin‐β (ΔN44) is able to bind to importin‐α and to the NPC (Görlich et al., 1996b). It should be noted that deletion of only the N‐terminal 32 or 10 amino acids results in a similar phenotype. However, the block is not as tight as for the deletion of the first 44 amino acids (not shown, but see below). An importin‐β fragment lacking the N‐terminal 330 residues still allowed importin‐α binding and some NPC binding (Figure 1). However, efficient targeting of nucleoplasmin to the NPC requires a ∼10‐fold higher concentration of this fragment compared with wild‐type protein. This lower targeting efficiency is mainly due to a reduced affinity for the NPC (not shown). Further N‐terminal deletions could not be analysed as the resulting fragments did not fold and were insoluble when expressed in Escherichia coli.
The phenotypes of C‐terminal deletions were fundamentally different. If importin‐β lacked the C‐terminal 106 residues, the fluorescent signal in the confocal section showed only non‐specific binding to the cytoplasmic remnants of the permeabilized cells (Figure 1). The fragment thus fails to target the import substrate to the NPC. This could be due to a failure to interact with either importin‐α or the NPC, or both.
The importin‐α binding site
To distinguish between the possibilities, we tested directly the importin‐α binding of importin‐β fragments (Figure 2). First, lysates were prepared from E.coli expressing the indicated importin‐β derivatives (lanes 1–6). The fragments were then allowed to bind to an immobilized IBB domain (the importin‐β binding domain of importin‐α) or, as a control, to nickel agarose to which all the fragments bind by virtue of their his‐tag.
Wild‐type importin‐β binds efficiently to the IBB domain, as reported previously. In spite of the very basic nature of the IBB domain, this interaction is highly specific as only importin‐β and virtually none of the E.coli proteins was recovered in the bound fractions. Importin‐α binding was still retained if the N‐terminal 255 or 330 amino acids of importin‐β were deleted. Importin‐β 331–876, however, binds less strongly than the wild‐type protein. If the ionic strength was raised from 350 mM NaCl (as in Figure 2) to 650 mM, binding of the wild‐type or 256–876 importin‐β remained unaffected, but binding of 331–876 was severely reduced (not shown).
Even the importin‐β derivative with the smallest C‐terminal deletion tested, importin‐β 1–771, fails to bind importin‐α (Figure 2, compare lanes 4, 10 and 16). This explains the failure of this fragment to target import substrates to the NPC (Figure 1, bottom panel). Thus, it is approximately residues 331–876 of importin‐β that are required for binding to importin‐α.
Characterization of the NPC binding domain in importin‐β
To set up an assay for NPC binding of importin‐β derivatives, full‐length and truncated forms of importin‐β were expressed in E.coli, purified and modified with fluorescein 5′‐maleimide such that, on average, a single fluorescein was attached per molecule. The fluorescent signal from these modified fragments is thus a direct measure of their local concentration. We could therefore use confocal fluorescence microscopy to compare the efficiency with which different importin‐β fragments bind to NPCs of permeabilized cells. We first noted that the capacity of the NPC to bind importin‐β became considerably higher if Ran and an energy regenerating system were added (not shown). There are probably two contributions to this effect. First, translocation allows importin‐β to occupy additional sites further inside the NPC. Second, translocation likely clears endogenous (unlabelled) importin‐β from NPCs, making the sites available for the fluorescent proteins. In Figure 3, the NPC binding of some importin‐β fragments under these conditions is shown. The shortest C‐terminally truncated fragment with detectable NPC binding is importin‐β 1–306. However, under more stringent conditions, i.e. a lower concentration of the fluorescent proteins, the gradual decrease in NPC binding from importin‐β 1–618 to the 1–306 fragment was evident (not shown). In conclusion, it is approximately residues 1–618 of importin‐β that are required for full NPC binding activity. Some binding was still observed for the N‐terminal 306 residues (Figure 3), but also for amino acids 331–876 (Figure 1). This suggests that at least two separate sites of NPC interaction exist on importin‐β.
Ran interacts with the N‐terminus of importin‐β
To determine which part of importin‐β accounts for Ran binding, we separated a mixture of C‐terminal importin‐β truncations by SDS gel electrophoresis and transferred the fragments onto nitrocellulose. The amounts of the fragments were normalized by Western blotting with an antibody raised against a peptide present in all fragments (Figure 4). Ran interaction was then detected in an overlay blot with Ran [γ‐32P]GTP as a probe. As seen from Figure 4, all importin‐β fragments down to 1–364 showed significant Ran interaction. However, the signal is essentially lost by the deletion of a further 21 amino acids. Thus, the C‐terminal border of the Ran binding domain appears to be between residues 343 and 364.
Probing the blot with Ran [γ‐32P]GTP in the presence of a large excess of RanBP1 (Coutavas et al., 1993; Bischoff et al., 1995b; Schlenstedt et al., 1995) resulted in an enhancement of the specific signal, as reported before (Lounsbury et al., 1996). Thus, RanBP1 and importin‐β do not compete for the same binding site on Ran. The amplification of the signal is probably due to the stabilization of RanGTP by RanBP1 and to cooperativity that favours the formation of the trimeric RanBP1–RanGTP–importin‐β complex. This trimeric complex is also detectable by gel filtration and by affinity chromatography on immobilized RanBP1 (not shown). As expected, RanGTP binding to the importin‐β fragments is competed by importin‐β itself. It should be noted that the N‐terminus of importin‐β is absolutely crucial for interaction with Ran. No Ran binding is detectable in the overlay assay if as little as the 10 N‐terminal amino acids are deleted (not shown, but see Görlich et al., 1996b and below).
The overlay blot measures two effects. First, renaturation of the blotted proteins and second, the presence of the Ran binding domain. To quantify Ran binding more precisely and independently of refolding, we employed an enzymatic assay: binding of importin‐β protects RanGTP against GTPase activation by RanGAP1(Rna1p) (Floer and Blobel, 1996). The proportion of protected RanGTP can be measured and we have calculated the binding constants from the dose dependence of protection. Table I shows relative affinities for the importin‐β fragments. The affinity of the wild‐type protein was set to 100. The numbers confirm residues 1–364 as the Ran binding domain. There is a sharp transition at its C‐terminal border. The Ran affinity of importin‐β 1–343 is three orders of magnitude lower than that of 1–364. That the 1–771 and 1–618 fragments have a slightly lower affinity for Ran than importin‐β 1–462 is probably explained by a poorly folded C‐terminus of the longer fragments influencing the folding of the Ran binding domain.
Deletion of as few as 10 amino acids from the importin‐β N‐terminus results in a 10‐fold drop in affinity for Ran, consistent with our observation that the import activity of this mutant is severely affected. Inhibition of importin‐β transport activity is complete if the first 44 amino acids are deleted (Figure 1).
Importin‐β is released from a subset of NPC binding sites as a complex with RanGTP
The Ran binding site and the nuclear pore binding site overlap in importin‐β. This might indicate that RanGTP binding and binding to the NPC influence one another. However, at least three different classes of importin‐β binding sites at the NPC might exist, that show different effects of RanGTP. First, as importin‐β does not have to bind Ran on its way into the nucleus (Görlich et al., 1996b), there have to be sites at which normally no interaction between Ran and importin‐β must occur. Second, the termination sites are the sites where importin‐β interaction with RanGTP releases importin‐α. At termination, importin‐β could either remain NPC‐bound and return directly from the termination site to the cytoplasm, or alternatively, importin‐β could be detached from the NPC before re‐export. These two alternatives are tested below. Third, if we assume that cytoplasmic RanGAP1 is one factor required for the disassembly of the importin‐β–RanGTP complex, then the importin‐β–RanGTP complex has to be returned to the cytoplasm and has to bind to the NPC on its way out of the nucleus.
Different effects of the Ran–importin‐β interaction on importin‐β complexes with isolated nucleoporins have indeed been observed: RanGTP releases importin‐β from the yeast nucleoporin Nup1p (Rexach and Blobel, 1995). Likewise, a similar effect has been seen for the p62–importin‐β interaction in gel filtration experiments (F.R. Bischoff and V.Cordes, unpublished results). On the other hand, importin‐β can simultaneously bind RanGTP and certain nuclear pore proteins, because a number of nucleoporins are detectable on overlay blots using a complex between Ran [γ‐32P]GTP and importin‐β as a probe (Lounsbury et al., 1996; our unpublished data). However, in all these studies it is unclear if binding sites found in isolated or partially denatured nucleoporins are accessible and functional in intact nuclear pore complexes.
Looking at intact nuclear pore complexes, we wanted to know if the termination reaction would transiently release importin‐β from the NPC. Mutant forms of importin‐β which do not bind Ran, should, in this case, resist the normal release mechanism and bind to the termination sites irreversibly. We have tested this in two sets of experiments, the first of which is shown in Figure 5. Permeabilized cells were incubated in Xenopus egg extract together with an energy regenerating system, and NPC binding of three fluorescent importin‐β derivatives was tested. The fluorescent proteins were added to a concentration of 1 μM and had to compete with ∼3 μM unlabelled, endogenous wild‐type importin‐β present in the extract.
Without further addition, the binding of the importin‐β 1–462 fragment was reduced compared with that of the wild‐type protein, probably because the fragment does not contain the entire NPC binding site. The binding of importin‐β 45–462 (which does not bind Ran) was somewhat heterogeneous; on average, however, it was ∼2‐fold stronger than that of the full‐length protein.
These differences became striking in the presence of 10 μM RanQ69L, a Ran mutant which is GTPase‐deficient and therefore fixed in its GTP‐bound form (Klebe et al., 1995). This concentration of RanQ69L dissociates the importin heterodimer in the extract completely (not shown, but see Görlich et al., 1996b), i.e. most importin‐β should be complexed with RanGTP. RanQ69L addition caused only a slight decrease in NPC binding of wild‐type importin‐β, consistent with the assumption that the importin‐β–RanGTP complex can bind to at least some sites at the NPC (perhaps those on the export route). Binding of 1–462 was severely reduced by addition of the Ran mutant, demonstrating that RanGTP can release importin‐β from at least some sites at the NPC. In contrast, binding of 45–462 was enhanced, probably because the fragment could now occupy sites from which RanQ69L GTP has cleared endogenous importin‐β and related transport factors.
Taken together, two possible conclusions are suggested from this experiment. First, importin‐β is released as an importin‐β–RanGTP complex from a subset of sites at the NPC, probably the termination sites. Second, the importin‐β 45–462 fragment, as a consequence of its lack of response to Ran, binds to these sites much more stably.
Inhibition of NLS import by importin‐β dominant–negatives
Both the 1–462 and 45–462 importin‐β fragments bind to the NPC but not to importin‐α. They should therefore inhibit protein import in a dominant‐negative way. Figure 6 shows the dose dependence of the inhibition. Import of fluorescent BSA–NLS conjugate into the nuclei of permeabilized cells was measured in the presence of an energy regenerating system and a Xenopus egg extract which contains ∼3 μM endogenous importin‐β. A 50% inhibition was observed with a 2 μM concentration of the 1–462 fragment. Importin‐β 45–462 is a far more potent inhibitor, a concentration of only 0.1 μM being required for 50% inhibition. The import block was complete at 1 μM (see Figures 6 and 7), a finding consistent with the assumption that importin‐β 45–462 binds to sites at the NPC from which it cannot be cleared. Confocal microscopy revealed that the initial docking of the import substrate to nuclear pores was not yet affected at this concentration of the 45–462 fragment (Figure 7). This is consistent with the assumption that the 45–462 fragment binds irreversibly only to sites far inside the NPC where termination normally occurs and where RanGTP interaction is required to detach importin‐β from the NPC (see also Görlich et al., 1996b).
Importin‐β dominant‐negatives block major nucleocytoplasmic transport pathways
Next we wanted to know if the importin‐β dominant‐negatives inhibit only the importin‐dependent pathway or if there is cross‐competition between different pathways. Figure 8 shows nuclear import of a fusion protein composed of the nucleoplasmin core and an M9 domain, the nuclear targeting signal of hnRNP A1 (Michael et al., 1995; see Introduction). The panels are views on unfixed import mixtures, i.e. nuclear and cytoplasmic concentrations of the fluorescent fusion protein are directly shown. Without an inhibitor, a clear nuclear accumulation is observed. If, however, the dominant‐negative importin‐β 45–462 was added, import was completely abolished and the probe remained excluded from the nuclei. It should be noted that wild‐type importin‐β at higher concentrations also competitively inhibits M9‐mediated import (not shown).
To test whether export out of the nucleus would also be affected, we performed micro‐injection experiments into Xenopus oocytes. Initial experiment established that the importin‐β dominant‐negatives block nuclear export of the Rev protein with the same efficiency as they inhibit NLS‐ and M9‐dependent protein import (not shown). In Figure 9A we tested the effects on nuclear export of DHFR and histone mRNA, U1 and U5 snRNA, and a tRNA. As a control for nuclear integrity, the mixture also contained U6 snRNA, an RNA which is not exported from the nucleus. The RNAs were co‐injected into the nucleus together with buffer, wild‐type importin‐β, the 1–462 fragment or the 45–462 fragment. Full‐length importin‐β left tRNA export largely unaffected, but significantly inhibited the export of mRNA and U snRNA. When importin‐β is injected into the nucleus, it is rapidly exported (not shown) and importin‐β re‐export to the cytoplasm probably competes with mediators of mRNA and U snRNA export for binding sites at the NPC. The injection of importin‐β 1–462 at this concentration had less effect on RNA export than full‐length importin‐β, probably because the affinity of this fragment for the NPC is lower than that of wild‐type importin‐β. Co‐injection of 45–462, however, caused a complete inhibition of mRNA and U snRNA export. The concentration needed to block export is similar to that required to prevent NLS‐dependent protein import into oocyte nuclei (not shown). At this concentration (5.8 μM in the injected sample) importin‐β 45–462 also reduced tRNA export to some extent.
To determine whether a higher concentration of the importin‐β mutants would have a greater effect on tRNA export, we injected importin‐β 1–462 and, as a control 1–256, at 50 μM. Importin‐β 1–256 does not interact with Ran, importin‐α or the NPC, and is therefore a good negative control. The 45–462 fragment at this concentration was toxic to the oocytes, and could therefore not be used. While 1–256 had no effect on RNA export, even at this high concentration (Figure 9B, compare lanes 4–6 with lanes 10–12) the 1–462 fragment blocked mRNA, U snRNA and tRNA export completely (lanes 7–9). Thus, the dominant‐negative importin‐β mutants have a general impact on many distinct nucleocytoplasmic transport pathways. For the block of tRNA export by high concentrations of importin‐β 1–462 it is not yet clear if the inhibition is due to a competition for NPC binding or because importin‐β 1–462 titrates free RanGTP, for which it has a high affinity. We are currently testing these possibilities.
We have mapped domains in importin‐β which interact with Ran, importin‐α and with the nuclear pore complex. A region of the protein encompassing approximately amino acids 331–876 accounts for interaction with importin‐α and residues 1–364 account for RanGTP binding, whereas NPC binding appears complex and can be mediated by at least two non‐overlapping sites on importin‐β.
RanGTP dissociates importin‐α from importin‐β. This could be explained in two ways. First, Ran and importin‐α could compete for the same binding site at importin‐β. In this model, dissociation of the importin heterodimer would have to occur spontaneously and RanGTP would prevent re‐binding of importin‐α. Second, RanGTP binding to importin‐β could cause a conformational change in importin‐β that releases the alpha subunit. In this model, RanGTP could force the importin heterodimer to dissociate. Our observation of largely non‐overlapping sequence requirements for Ran and importin‐α binding suggests that Ran binding to importin‐β causes a conformational change that inactivates the importin‐α binding site located elsewhere on the molecule. Such an active displacement can of course be much more efficient than simple competition for re‐binding after spontaneous dissociation. A previous study of the binding sites on importin‐β for RanGTP and importin‐α, based on the use of a single mutant β‐protein and on competition assays with short peptides derived from β, reached the opposite conclusion as to the relative orientation of the binding sites on β for the two proteins (Moroianu et al., 1996). Although we have no good explanation for this discrepancy, we feel the methods used in our study, and the larger numbers of different mutants and binding assays employed, allow us to define the regions required for the specific protein–protein interactions with some confidence.
Importin‐β mediates translocation into the nucleus and makes contact with the nuclear pore complex. This translocation involves binding to a number of intermediate sites on the NPC (Görlich et al., 1996b; Panté and Aebi, 1996). As discussed previously (Görlich and Mattaj, 1996) a directed transfer from one site to the next, without intermediate detachment from the nuclear pore complex, would require more than one NPC binding site within importin‐β. Our observation that the N‐terminal 306 residues are sufficient for NPC binding and that importin‐β still binds the NPC without the first 330 amino acids indicates that distinct NPC binding sites on importin‐β exist. These sites could potentially be used independently and possibly alternately to move from one NPC site to the next. We are currently using the different importin‐β fragments to determine which parts of the NPC binding sites are required for the actual translocation reaction.
Ran has probably two distinct functions in nuclear protein import. First, Ran's GTP cycle probably drives translocation of the NLS–importin‐α/β complex into the nucleus. This appears to involve binding of RanGDP to the NPC, followed by nucleotide exchange and GTP hydrolysis, but not binding of RanGTP to importin‐β (Görlich et al., 1996b). Second, Ran regulates the interaction between importin‐α and importin‐β. It is the binding of nuclear RanGTP to importin‐β that finally terminates the import reaction and disassembles the importin heterodimer. There are indications to suggest that importin‐β might return to the cytoplasm as a complex with RanGTP. First, importin‐β is probably recycled to the cytoplasm without the alpha subunit. Re‐export of importin‐β complexed with RanGTP would explain why. Second, although the importin‐β–RanGTP complex forms in the nucleus, its disassembly is likely to be a cytoplasmic event as it requires two cytoplasmic proteins, RanGAP1 and RanBP1, plus an auxiliary factor (F.R.Bischoff, unpublished results). In addition, our observation that RanGTP causes the release of importin‐β from only a subset of sites at the NPC implies that the importin‐β–RanGTP complex could be a species that moves through the NPC.
Here, we provide evidence that termination, i.e. dissociation of the importin heterodimer on the nuclear side of the NPC, also involves at least a transient release of importin‐β from the nuclear pore complex. All importin‐β fragments which have retained nuclear pore binding but do not interact with either Ran or importin‐α, inhibit protein import in a dominant‐negative way. Those fragments, however, which have lost Ran binding (e.g. importin‐β 45–876 and 45–462) are by far the most potent inhibitors. Since mutants that do not bind RanGTP can be visualized on the nuclear face of the NPC (Görlich et al., 1996b) they probably bind to the sites of termination such that release by the normal mechanism via interaction with nuclear RanGTP is not possible.
The importin‐β dominant‐negatives not only block NLS‐dependent import, but with the same efficiency they also block other pathways, namely M9 import, NES‐dependent nuclear export, and export of mRNAs, U snRNAs and, though less efficiently, tRNAs. Aside from the export of U snRNAs (Görlich et al., 1996c; Izaurralde et al., 1997) none of these pathways is affected by inhibition of NLS‐mediated protein import with saturating amounts of NLS conjugates or IBB domain (Görlich et al., 1996c; Pollard et al., 1996; Izaurralde et al., 1997). Thus, each is mediated by a saturable receptor distinct from importin α/β. The fact that the importin‐β mutants block these transport events in spite of this may be explained in two ways. First, although different transporter molecules bind different substrates they might take a similar path through the NPC and function according to similar principles. This is supported by the recent identification of the M9 import receptor, transportin, which is distantly related to importin‐β (Pollard et al., 1996). The cross‐competition could indicate that at least some of the intermediate binding sites on the NPC are shared by the distinct transport mediators. Second, the cross‐inhibition might be a reflection of how the NPC coordinates trafficking between nucleus and cytoplasm. For example, when a large RNP is being exported, then incoming particles may have to be excluded until the central transporter of the NPC has been cleared. The information that a cargo is not yet released from a NPC might be transmitted to more distant parts of the NPC, preventing further transport substrates from entering the same NPC.
Importin‐β is at the heart of the NLS‐mediated import pathway. As such, importin‐β has to interact with multiple factors involved in transport and to help coordinate these factors to orchestrate directed movement of NLS‐containing proteins to the nucleus from the cytoplasm. Our results cast light on how importin‐β achieves these multiple interactions, and the effects of the dominant mutants suggest that other transport mediators will share at least some of these properties.
Materials and methods
Import assays with permeabilized HeLa cells
The basic method was described previously (Görlich et al., 1994, 1996a,b). Unless indicated otherwise, reactions were performed at 23°C and contained an energy regenerating system, i.e. 10 mM creatine phosphate, 0.5 mM ATP, 0.5 mM GTP, 50 μg/ml creatine kinase. When using Xenopus egg extracts (LSS) salts were not adjusted. When using hypotonic HeLa cell extract the ionic conditions were: 20 mM HEPES–KOH, pH 7.5, 140 mM potassium acetate, 5 mM magnesium acetate, 1 mM dithiothreitol (DTT), 250 mM sucrose. The standard assay based on recombinant import factors contained: 20 mM HEPES–KOH, pH 7.5, 80 mM potassium acetate, 4 mM magnesium acetate, 1 mM DTT, 250 mM sucrose, 1.5 μM Ran (GDP form), 150 nM RanBP1, 150 nM Rna1p, 150 nM pp15 and 2 mg/ml nucleoplasmin core to block non‐specific binding of proteins. For the concentrations of the other components see the figure legends.
Oocyte injections and analysis of microinjected RNA by denaturing gel electrophoresis, autoradiography or phosphoimager analysis were performed as described (Jarmolowski et al., 1994). The histone H4 mRNA was generated by transcription of the mRNA coding portion of the S3 clone (Meier et al., 1989), which was amplified by PCR and inserted into the BamHI site of the pBS‐vector. The concentrations of the recombinant proteins in the injected samples are indicated in the figure legends.
Recombinant expression and protein purification
Preparation of the following proteins was as described: C‐terminally his‐tagged Xenopus importin‐α, nucleoplasmin, nucleoplasmin core (Görlich et al., 1994), untagged recombinant, human Ran, Schizosaccharomyces pombe Rna1p and murine RanBP1 (Bischoff et al., 1995b). For import assays, Ran was expressed with an N‐terminal his‐tag, and purified on nickel agarose as described (Görlich et al., 1994). The bound nucleotide was exchanged for GDP using the EDTA method and final purification achieved on Superdex 75 equilibrated in 50 mM potassium phosphate, pH 7.2, 0.5 mM magnesium acetate. S.pombe Rna1p was expressed as a fusion with a z domain on the N‐terminus and a his‐tag on the C‐terminus. Purification was with nickel agarose and activity verified with an enzymatic GAP assay. NTF2 was expressed untagged. It was bound from the E.coli lysate (post‐ribosomal supernatant) to Q Sepharose FF equilibrated in 50 mM Tris–HCl, pH 8.0, and eluted in the NaCl gradient at 230 mM NaCl. Final purification was on Superdex 75 where it eluted at a position expected for the homodimer. The M9 fusion was prepared as follows: a BamHI–HindIII DNA fragment that codes for residues 255–320 of human hnRNP A1 and includes the M9 domain was generated by PCR. The fragment was then fused to the C‐terminus of the nucleoplasmin core domain. Expression was from the pQE9 vector (Qiagen) with an N‐terminal his tag. Fluorescein labelling of the purified fusion protein was with fluorescein 5′‐maleimide.
Expression of importin‐β fragments
Fragments were generated by PCR using Pwo proof‐reading DNA polymerase and were cloned into the NcoI–BamHI sites of pQE60. The constructs were verified by DNA sequencing. Expression was in the BLR/Rep4 E.coli strain at 17°C for 4 h. 2 mM PMSF was added immediately before the culture was chilled on ice. After centrifugation, the bacterial pellet was resuspended in 50 mM Tris–HCl, pH 7.5, 20 mM imidazole, 200 mM NaCl, 5 mM mercaptoethanol, and lysis was performed by sonication. The lysate was cleared by ultracentrifugation in a Ti50.2 rotor at 50 000 r.p.m. for 3 h and loaded overnight onto nickel agarose equilibrated in sonication buffer. The column was washed in the same buffer and eluted with a gradient ending at 400 mM imidazole. Peak fractions were pooled and buffer was exchanged for 50 mM Tris–HCl, 250 mM sucrose. The concentration of the fragments was determined by photometry at 280 nm using the molar extinction coefficient calculated from the amino acid composition (Edelhoch, 1967). For fluorescence labelling, fluorescein 5′‐maleimide dissolved in dimethylformamide was added at an equimolar ratio to the protein and the mixture was incubated for 1 h on ice. The reaction was quenched with 50 mM mercaptoethanol and free label separated on a PD10 column equilibrated in 50 mM Tris–HCl, pH 7.5, 250 mM sucrose. Efficiency of labelling was close to 100% for all of the fragments.
Labelling of Ran with [γ‐32P]GTP and GTPase assays were as described (Bischoff et al., 1994, 1995b; Görlich et al., 1996b). The reaction was performed in 20 mM HEPES–KOH, pH 7.4, 100 mM NaCl, 1 mM MgCl2, 1 mM sodium azide, 0.05% hydrolysed gelatin. Ran was used at 60 pM, far below the Kd of the importin‐β–RanGTP complex. Ran was pre‐incubated for 30 min with the importin‐β fragments before 2 nM Rna1p was added. After 5 min, GTP hydrolysis was measured as released γ‐32P‐labelled phosphate. The dissociation constants of various importin‐β derivatives from RanGTP was interpolated from dilution series as the concentration that protects 50% RanGTP against GTPase activation by Rna1p. The affinity for RanGTP is then inversely proportional to Kd. For convenience, the affinity of the wild‐type protein was set to 100 in the listing shown in Table I. The values were found to be highly reproducible in three independent series of measurements and the averages are shown.
The basic method followed a published protocol (Lounsbury et al., 1996) with some modifications. Briefly, proteins were separated by SDS–PAGE and transferred onto nitrocellulose. The blot was then incubated for 1 h at 4°C in renaturation buffer containing 20 mM MOPS, pH 7.1, 100 mM sodium acetate, 5 mM magnesium acetate, 5 mM DTT, 0.5% bovine serum albumin, 0.05% Tween 20, and subsequently for 30 min at 25°C in renaturation buffer plus 100 μM unlabelled GTP (20 ml total volume). 10 min later, 100 μl 1 nM Ran [γ‐32P]GTP was added alone, or with 5 μM RanBP1 or with 5 μM importin‐β. After another 10 min the blot was washed five times in renaturation buffer and subjected to autoradiography at −70°C.
We thank Petra Schwarzmaier for technical help, and Drs M.Dabrowski, R.A.Laskey and T.Sommer for critical reading of the manuscript. We are grateful to Drs E.Darzynkiewicz and J.Stepinski for the gift of the 7‐methyl GpppG and gamma methyl GTP cap analogues, and D.Schümperli for the Histone H4 clone. This work was supported by the Deutsche Forschungsgemeinschaft (DFG).
- Copyright © 1997 European Molecular Biology Organization