Nuclear Import and the Evolution of a Multifunctional RNA-binding Protein

DF5 was used as the parent strain for all new strains constructed (Finley et al., 1987). The Lhp1–PrA strains and the Kap108–PrA strain in the wild-type background were described previously (Rosenblum et al., 1997). The Kap95-PrA strain was the generous gift of M.P. Rout (Rockefeller University, New York, NY) and J.D. Aitchison (University of Alberta, Alberta, Canada) (Aitchison et al., 1996). The Kap108-PrA/Δlhp1 strain was generated by direct integration of a PCR product (as described in Aitchison et al., 1995) directly into a haploid Δlhp1 strain (provided by S. Wolin, Yale University, New Haven, CT). Proper integration was assessed by PCR and Western blotting as described previously (Rosenblum et al., 1997). Yeast strains with and without plasmids were grown at 30°C in dropout and YPD media, respectively (Ausubel et al., 1997).

The green fluorescent protein (GFP) constructs were assembled in pYX242 (Novagen, Madison, WI). pYX242 is a 2-μ plasmid with the triose phosphate isomerase promoter and LEU2 marker. The GFP used, a yeast codon-optimized, fluorescence optimized mutant (yEGFP3), was the kind gift of B. Cormack (Stanford University, Stanford, CA) (Cormack et al., 1997). All restriction enzymes were purchased from New England Biolabs (Beverly, MA), Pfu was purchased from Stratagene (La Jolla, CA). yEGFP3 was cloned into pYX242 using the enzymes HindIII and SalI to generate the plasmid pYX242–GFP. LHP1 was amplified from S. cerevisiae genomic DNA (Promega, Madison, WI) and cloned into pYX242–GFP using primer-encoded BamHI and ApaI sites to generate pYXLhp1–GFP. Restriction of pYXLhp1–GFP with HindIII, filling in the overhangs with T4 DNA polymerase and religating generated the fragment 1 construct, pYXfrag1–GFP. Fragments 2 and 3 were cloned with primer-encoded EcoRI/BamHI and BamHI/HindIII sites, respectively to generate pYXfrag2–GFP and pYXfrag3–GFP. The human La was amplified from a human fetal bone marrow cDNA library (Clontech, Palo Alto, CA) using primer-encoded BamHI and HindIII sites, generating pYXhsLa–GFP. Drosophila and S. pombe Las were amplified from cDNAs kindly provided by S. Wolin and cloned into pYX242–GFP using primer-encoded BamHI and HindIII sites to generate pYXdmLa–GFP and pYXspLa–GFP. pYXspLa35–GFP and pYXspLa67–GFP, which contain the last 35 and 67 aa of Sla1, respectively, fused to GFP, and were cloned from Sla1 cDNA using primer-encoded EcoRI and HindIII sites. The construct encoding the last 35 aa of Drosophila La fused to GFP, pYXdmLa35–GFP, was generated by restriction of pYXdmLa–GFP with EcoRI and religating. Yeast strains were transformed with plasmids via electroporation (Ausubel et al., 1997).

To express human La in bacteria for in vitro import assays, the hsLa– GFP cassette from pYXhsLa–GFP was amplified and cloned into pET21b (Novagen) using primer-encoded BamHI and SalI sites. The Escherichia coli strains XL1-blue and BL21(DE3) (Stratagene) were used for general cloning and protein expression, respectively. Bacterial strains were transformed by heat shock (Ausubel et al., 1997).

Total yeast extracts were isolated as previously described (Rosenblum et al., 1997). After electrophoresis through a 7.5% polyacrylamide gel, proteins were transferred to nitrocellulose (0.4 μ, Schleicher & Schuell, Keene, NH). Blots were visualized with amido black (0.1% amido black in 40% methanol, 10% acetic acid) before blocking in 5% milk in TBS (10 mM Tris, 150 mM NaCl, pH 7.5) for 15 min. Blots were incubated with polyclonal anti-GFP antibodies (Clontech) at 1:1,500 dilution in 5% nonfat dry milk in TBS for 2 h. After incubation, blots were rinsed with water, washed twice with 0.05% NP-40 in TBS, and then once with TBS. The blots were incubated with horseradish peroxidase-coupled donkey anti– rabbit antibodies (Amersham, Arlington Heights, IL) at 1:2,000 dilution in 5% milk in TBS for 1 h. After this incubation, blots were washed in a similar manner to before. Enzyme-linked antibodies were detected using luminol-based chemiluminescence (Schneppenheim et al., 1991).

For the purification of protein A fusion proteins from yeast, cytosol was prepared from strains essentially as described (Rout and Blobel, 1993). For a typical purification, cytosol from ∼2 g (wet weight) of cells in a final volume of 50 ml was incubated overnight at 4°C with 25 μl of IgG Sepharose (made from coupling CNBr-activated Sepharose 4B [Pharmacia Biotech, Piscataway, NJ] to rabbit IgG [Cappel, Malvern, PA] following standard protocols). The Sepharose was isolated and washed at least six times with 1 ml TB (20 mM Hepes, pH 7.5, 110 mM KOAC, 2 mM MgCl₂, 0.1% Tween 20), before elution with acid (0.5% HOAc, 150 mM NaCl) or MgCl₂ (various concentrations in 20 mM Hepes, pH 7.5, 0.05% Tween 20). Protein samples were concentrated by the method of Wessel and Flugge (1984). Proteins were identified using a combination of MALDI-TOF mass spectrometry and Edman sequencing (Fernandez et al., 1994; Gharahdaghi et al., 1996). Peptide masses were compared with several databases (http://prospector.ucsf.edu or http://prowl.rockefeller.edu/cgi-bin/ProFound).

For the purification of hsLa–GFP from E. coli, BL21(DE3) cells harboring pEThLa–GFP were grown in 30 ml 2× YT at 37°C to an OD₆₀₀ of 0.7. This starter culture was then diluted 1:20 in fresh 2× YT and grown to an OD₆₀₀ of 0.7 at 37°C. Recombinant protein expression was induced by addition of isopropyl B-p-thiogalactopyranoside to 0.5 mM. Cells were allowed to grow for an additional 3 h at 30°C. Cells were harvested, resuspended in 30 ml of 50 mM KH₂PO₄, 300 mM NaCl, 30 mM beta-mercaptoethanol, pH 8.0, and lysed by passing twice through a French pressure cell at 900 pounds per square inch. After sedimentation of cell debris, the fusion protein was purified by virtue of the plasmid-derived hexahistidine sequence using 1.5 ml Talon resin (Clontech) following the protocols supplied by the manufacturer.

Sequences were aligned using ClustalW 1.6 (Megalign Program; DNAStar, Inc., Madison, WI). Phylogenetic analysis was performed using Phylip (PHYLIP v 3.5c; distributed by J. Felsenstein, University of Washington, Seattle, WA). Both programs were used with standard settings.

GFP was visualized in live yeast after overnight growth in selective medium or on selective agar. 4 μl of culture or one colony resuspended in 4 μl of water was placed on a slide. All micrographs were taken using a 63× oil objective on an Axiophot microscope (Carl Zeiss, Thornwood, NY) and Zeiss filter set number 9 (excitation, 450–490 nm; beamsplitter, 510 nm; emission, long-pass 520 nm) with coincident differential interference contrast (Nomarski) optics. Images were transferred directly to Adobe Photoshop 3.0.5m (Adobe Systems, San Jose, CA) using a Sony DKS-5000 digital photo camera (Tokyo, Japan).

Import assays were performed in digitonin-permeabilized HeLa cells essentially as described (Moore and Blobel, 1992). Each assay contained 1 mM GTP and a combination of the following factors: 0.5 μg of Kapα2, 0.5 μg of Kapβ1, 4 μg of Ran, 60 ng of p10, 1 μg of human La-GFP, and 500 μg of HeLa S-100 cytosol.

Previously, we demonstrated that Kap108p could be isolated from yeast cytosol in a complex whose main components were Lhp1p and rpL11p (Rosenblum et al., 1997) (rpL11p was previously known as Rpl16p, for new S. cerevisiae ribosomal protein nomenclature see Mager et al., 1997). At the time we could not differentiate between two coexisting, dimeric complexes of Kap108p and a single trimeric complex of all three members. To address this issue, we generated a yeast strain that contained a genomically expressed fusion between protein A and Kap108p (Kap108– PrA) but was deleted for LHP1. We then isolated Kap108–PrA by virtue of the affinity for IgG-Sepharose of protein A. The results are shown in Fig. 1,a. In the first lane, for comparison, cytosolic Kap108–PrA was purified from cells wild type for LHP1 and the bound proteins were eluted with acid. The second panel of Fig. 1,a shows the result of a similar experiment done in parallel with cytosol from the LHP1-deletion strain. rpL11p eluted from Kap108–PrA in the LHP1 deletion strain at the same concentrations of MgCl₂ that dissociate this complex isolated from a strain containing Lhp1p. As such, it is evident that Lhp1p does not mediate the rpL11p/Kap108p interaction. We have also detected other, smaller ribosomal proteins in complex with Kap108p (Rosenblum et al., 1997). In Fig. 1 a, right, these smaller ribosomal proteins, between ∼11 and 17 kD, appear to elute from the column earlier than rpL11p, peaking in the 100 mM MgCl₂ fraction. Therefore, we cannot exclude the possibility that these ribosomal proteins bind to Kap108p via rpL11p.

In a similar fashion, we determined whether the Kap108p/ Lhp1p interaction might be mediated by rpL11p. To do so, we purified Lhp1p, by virtue of an Lhp1p–protein A fusion protein, from strains with and without Kap108p. As shown in Fig. 1,b, the profile of proteins associated with Lhp1p is quite different from that of proteins associated with Kap108p. The profiles of proteins bound to Lhp1– PrA are identical between strains with (Fig. 1,b, left) and without (Fig. 1,b, right) Kap108p with the exception of a prominent band of ∼105 kD. As expected, more Lhp1–PrA and associated factors are present in the cytosol of the KAP108-deficient strain. We identified two Lhp1p-associated proteins. The 50-kD protein eluting mostly at 100 mM MgCl₂ was identified by sequencing three peptides, xxxANLIAAGAxxL, xLLCGAAIGTIDAD, and xxxAQLEGxV (Fernandez et al., 1994). The peptides uniquely identified Pur5p, inosine monophosphate dehydrogenase, the enzyme that catalyzes the first dedicated step of GMP biosynthesis. The 105-kD protein eluting mostly in the 250 mM MgCl₂ fraction and present only in strains containing Kap108p was analyzed by mass spectrometry (Gharahdaghi et al., 1996). A database search was performed using the observed masses of eleven tryptic peptides generated from this protein (from 964.1 to 2,132 D), and a mass tolerance of 0.5 D. This search uniquely identified this protein as Kap108p, with the peptides covering 14% of the Kap108p sequence. That the interaction between Lhp1–PrA and Kap108p is direct is strongly suggested by the final two lanes of Fig. 1 b, left. In these two lanes, the only predominant proteins remaining on the column were Lhp1–PrA and Kap108p. Further evidence that the Lhp1–PrA/Kap108p interaction is not bridged by rpL11p is suggested by the absence of a distinct band at 22 kD eluting with or after Kap108p from Lhp1–PrA.

As both Lhp1p and rpL11p could bind directly to Kap108p, we determined whether these two substrates bind to independent sites on Kap108p or if their binding sites functionally overlap. This experiment was done in the cellular milieu to avoid nonspecific interactions between these three highly charged proteins. To do so, we constructed a plasmid to overexpress an Lhp1–GFP fusion protein in S. cerevisiae. When Kap108–PrA was purified from a strain harboring this plasmid, a large quantity of Lhp1–GFP copurified, but no endogenous Lhp1p or rpL11p could be seen by Coomassie blue staining (Fig. 1 c). Since Lhp1–GFP could compete with both Lhp1p and rpL11p for binding to Kap108–PrA, it is evident that the binding sites for Lhp1p and rpL11p on Kap108p overlap functionally. This functional overlap could be due to identical or overlapping binding sites or to steric interactions between the ligands.

To further characterize the nuclear import of Lhp1p via Kap108p we next delimited a subdomain of Lhp1p that was capable of conferring nuclear localization via Kap108p. To do so, we constructed plasmids for the in vivo expression of Lhp1p and three fragments of Lhp1p, each fused to GFP. A schematic of these constructs is shown in Fig. 2,a. These fragments overlap by at least 25 amino acids to minimize the likelihood that an NLS would be missed because it had been split between two fragments. That these fragments were properly expressed as full-length products was confirmed by Western blotting with anti-GFP antibodies (Fig. 2,b). Consistent with our previous observations with Lhp1–PrA, in wild-type cells, Lhp1–GFP is restricted to the nucleus, whereas in the KAP108 deletion, Lhp1–GFP is unable to target the nucleus (Fig. 2,c, first row) (Rosenblum et al., 1997). As such, the GFP tag, like the PrA tag we have previously used, did not affect the nuclear targeting of Lhp1p. Fragment 1, aa 1–136, contains the La domain which is highly conserved throughout the La family (Van Horn et al., 1997). As a GFP fusion, fragment 1 was unable to accumulate in nuclei of wild-type cells (Fig. 2,c, second row). Fragment 2 spans aa 112–224, whereas the RRM spans aa 124–209. This fragment, which was not as highly expressed as the other fragments (Fig. 2,b), did accumulate in nuclei of wild-type cells (Fig. 2,c, third row). In contrast, when fragment 2 was expressed in cells deleted for KAP108, no accumulation of GFP was seen in nuclei (Fig. 2,c, third row). Fragment 3 is composed of the COOH terminus of Lhp1p, aa 188–275. This fragment is composed of just under 50% of charged residues, and encompasses a region that corresponds to the NLS of human La that has recently been defined by microinjection of in vitro translated hsLa mutants into Xenopus laevis oocytes followed by dissection (Simons et al., 1996). Interestingly, this region was also able to direct nuclear accumulation, yet this accumulation was insensitive to the presence of Kap108p (Fig. 2,c, fourth row). That fragments 2 and 3 contained discrete NLSs was suggested by their different dependence on Kap108p for nuclear localization. To demonstrate that each fragment had its own NLS, a construct was made that encoded the portion of fragment 3 that did not overlap with fragment 2. This shorter fragment demonstrated localization identical to that of fragment 3, in both wild-type and KAP108-deleted cells (data not shown). As full-length Lhp1p is unable to accumulate in the nucleus in the absence of Kap108p (Fig. 2 c, first row), it appears that fragment 3 contains not the endogenous NLS of Lhp1p, but a secondary NLS unmasked by its expression out of context of Lhp1p.

In an attempt to further analyze the fragment 2-type NLS, we first selected a 20-aa portion of this fragment that appeared similar to a stretch of rpL11p. This portion, from lysine 152 to phenylalanine 171 of Lhp1p was highly charged, with 10 charged amino acids, eight of them basic. However, this portion, when fused to GFP, was unable to target the nucleus (data not shown). Next, two proteins were identified that appeared to have significant homology to fragment 2, Snp1p, and the unknown open reading frame YOL041c. Although both of these proteins were nuclear as GFP fusions in wild-type cells, neither lost this localization when expressed in Δkap108 cells, suggesting that they are not substrates of Kap108p (data not shown).

It should be noted that although fragment 2 was the only construct examined that was able to confer Kap108p-dependent nuclear import, it did not contain as strong of a signal for nuclear import as full-length Lhp1p. Whereas no Lhp1–GFP could be seen in the cytoplasm of wild-type cells, fragment 2–GFP was observable in the cytoplasm of wild-type cells. This is the case even though fragment 2 overlapped the Kap108p-independent fragments by 25 NH₂-terminal and 37 COOH-terminal amino acids. The most likely explanation for the inefficient import of fragment 2–GFP is that fragment 2, although 113-aa long, does not contain complete structural information for interaction with Kap108p. Another possibility is that fragment 2 is actively exported from the nucleus, perhaps by virtue of binding to RNA.

As the Kap108p-dependent NLS of Lhp1p did not coincide with the NLS of human La mapped using the Xenopus assay, we undertook an analysis of intracellular targeting of heterologous La proteins in S. cerevisiae. In the course of evolution, four major branches of La proteins have arisen (Fig. 3,a). The clearest diversion in primary structure among these branches occurs between branches I and II and branches III and IV. In diverging from branches I and II, the branch III and IV proteins grew in primary sequence by ∼30%. For example, spLa is 298-aa long whereas dmLa is 390-aa long. The added sequence is most apparent in the COOH-terminal half of the larger proteins. scLa and spLa are 36% identical, spLa and dmLa are 28% identical, and dmLa and hsLa are 34% identical. GFP fusion constructs were generated for each of these proteins. Each construct was transformed into wild-type cells and cells deleted for KAP108. As shown in Fig. 3,b (left column), human, Drosophila, and S. pombe La proteins were targeted to the nucleus when expressed in S. cerevisiae as fusions with GFP. Significantly, all of the fusion proteins were entirely restricted to the nucleus in wild-type cells, indicating that they were substrates of an S. cerevisiae Kap. Surprisingly, this Kap is Kap108p only for Lhp1p, the endogenous S. cerevisiae La protein (Fig. 3 b, right column). Even Sla1, the S. pombe La protein that is 36% identical to Lhp1p, is targeted to the nucleus in the absence of Kap108p. This targeting of Sla1 occurs even though it, like Lhp1p, diverges markedly from the COOH-terminal domain shared by the La proteins of higher eukaryotes (Van Horn et al., 1997). The NLS of the human La resides in precisely this COOH-terminal domain (Simons et al., 1996).

Two sequences similar to the consensus bipartite NLS recognized by Kapα/Kapβ1 have been identified in hsLa and termed box A and box B (Simons et al., 1996). Box A perfectly corresponds to the consensus, with 2 NH₂-terminal basic residues, a 10-aa spacer, and a COOH-terminal cluster of three basic residues, whereas box B does not, having a 12-aa spacer and only two basic residues on the COOH-terminal side. Interestingly, box A was shown to be unable to confer nuclear localization on hsLa whereas box B was shown to be the hsLa NLS (Simons et al., 1996). Although we were initially drawn to the differences between box B and the consensus, the ability of human La to localize to the nucleus in the absence of Kap108p led us to consider that human La is brought to the nucleus in an α-Kap– dependent manner. We tested for this biochemically in yeast by introducing our hsLa–GFP plasmid into a strain that carried a genomic protein A fusion with the S. cerevisiae Kapβ1, Kap95p (Kap95–PrA). We then purified cytosolic Kap95p–PrA from this strain by immunoaffinity chromatography. As shown in Fig. 4,a, Kap60p, the S. cerevisiae Kapα, copurifies with Kap95–PrA from an untransformed strain. As Kap95p/Kap60p appears to be the major route to the nucleus, transporting perhaps hundreds of proteins, individual import substrates cannot generally be visualized by Coomassie blue staining (Aitchison et al., 1996; Makkerh et al., 1996). This inability to clearly see substrates most likely results from the stoichiometry of the interaction—each substrate is probably two to three orders of magnitude less abundant than the isolated Kap60p/ Kap95p dimer. Surprisingly, when Kap95–PrA is isolated from cytosol of a strain carrying the human La–GFP fusion, a clear band corresponding to the fusion can be seen by Coomassie blue staining. The ability to visualize hsLa– GFP by Coomassie staining most likely results from the high expression of this construct, making the fusion protein one of the main substrates of this pathway. That the fusion product is observable is even more remarkable when the trimeric nature of the isolated complex is taken into account; purified Kap95–PrA is bound to Kap60p, which in turn is bound to human La–GFP. The identity of the hsLa–GFP band was confirmed by Western blot analysis with anti-GFP antibodies (Fig. 4,b). Significantly, in a separate experiment performed in parallel, no hsLa–GFP copurified with Kap108–PrA (Fig. 4 b).

The foregoing experiments strongly suggested that human La is targeted to the S. cerevisiae nucleus via the Kap60p/ Kap95p pathway. To address the targeting of La in human cells, we subjected recombinant hsLa–GFP to the permeabilized HeLa cell assay in the presence of recombinant human factors. As shown in Fig. 5 (first panel), the fusion protein, when added alone to permeabilized HeLa cells, did not localize to the nucleus. In the presence of HeLa cytosol, however, hsLa–GFP was rapidly and efficiently targeted to the nuclei of HeLa cells (Fig. 5, second panel). In the presence of Kapα2 and Kapβ1, a portion of hsLa–GFP could be seen to accumulate around the nuclear rim (Fig. 5, third panel). Furthermore, with the addition of Kapα2, Kapβ1, Ran and p10, the efficient nuclear import of HeLa cytosol was completely reconstituted (Fig. 5, bottom panel). As is the case with other Kapα/Kapβ1 substrates, hsLa–GFP, in the presence of either Kapα2 or Kapβ1, with or without Ran and p10, did not localize to the nucleus (data not shown). The ability to reconstitute hsLa– GFP nuclear import with human Kapα/Kapβ1 and the similarity of the previously mapped hsLa NLS to the consensus bipartite NLS, combined with our experiments in S. cerevisiae (Figs. 2 and 3) provide firm evidence that La is targeted to the nucleus via the Kapα/Kapβ1 pathway in human cells.

As we had detected a division in the route to the nucleus taken by La proteins of different species, we sought a possible evolutionary explanation. We had mapped the Lhp1p NLS to the central region, overlapping its RRM, whereas the human La NLS resides at the COOH terminus. In Fig. 6,a, we have listed the COOH-terminal 35 amino acids of each of the La proteins cloned to date, highlighting basic residues which are critical for the interaction with Kapα (Conti et al., 1998). Although this stretch from each protein contains several basic residues, scLa and spLa only have one basic residue on the COOH-terminal end of this region, whereas bipartite NLSs generally have three basic residues in their COOH-terminal cluster (Makkerh et al., 1996; Conti et al., 1998). The La proteins from the multicellular eukaryotes have clear bipartite NLSs at their COOH termini and are ∼100 amino acids larger than those from the yeasts (Fig. 6,a, right column). To asses the ability of this region to act as an NLS in the branch II and branch III proteins, we constructed reporter constructs of the last 35 aa of both spLa and dmLa. When expressed in S. cerevisiae, the last 35 aa of dmLa concentrated in the nucleus, with no observable GFP signal in the cytoplasm (Fig. 6,b). This suggests that the last 35 aa of dmLa can mediate its nuclear import, presumably by Kap60p/ Kap95p in S. cerevisiae as a bipartite consensus region begins at Lys-358, 35 aa from the COOH terminus of dmLa. In contrast, the last 35 aa of spLa did not concentrate in the nucleus, possibly due to the insufficiency of the single COOH-terminal basic residue (Fig. 6 b). However, a larger COOH-terminal construct of Sla1, beginning with Met-232 could concentrate in the nucleus (data not shown). This region of Sla1 contains a putative bipartite NLS beginning at Lys-255, which is 44 aa from the COOH terminus.

We have previously described Kap108p as a protein essential for the nuclear localization of Lhp1p (Rosenblum et al., 1997). By the experiments described above (Fig. 1) we have now extended this finding to indicate that Kap108p imports Lhp1p without an adapter protein. It appears that of the entire karyopherin superfamily, only Kapβ1, or Kap95p in S. cerevisiae, needs an adapter. Significantly, rpL11p and Lhp1p seem to compete for binding to Kap108p in vivo.

We have looked for mislocalization of rpL11p–GFP fusions in KAP108-deletion strains, but such fusions were only mislocalized in KAP123-deletion strains (data not shown). As nuclear import of ribosomal proteins appears to be a distributed, overlapping cellular task, we cannot rule out the possibility that Kap108p plays a role in the import of rpL11p (Rout et al., 1997). In addition, our observations, particularly with regard to the in vivo competition between Lhp1p and rpL11p, are also consistent with a regulatory role for the interaction between rpL11p and Kap108p. Not only is RPL11 one of the most highly transcribed genes in S. cerevisiae, its expression is also highly regulated (Neuman-Silberberg et al., 1995; Velculescu et al., 1997). There could, therefore, be feedback between expression of rpL11p and nuclear import of Lhp1p.

The beta karyopherin superfamily in S. cerevisiae likely contains as many as 14 members. Since S. cerevisiae is the only eukaryote whose genome has been sequenced so far, the number of karyopherins in other organisms is currently unknown. Direct homologues of Kap95p, Cse1p, Kap104p, Kap121p, Crm1p, and Los1p have been identified in higher eukaryotes and expressed sequence tags for homologues of other yeast Kaps have been detected (Pemberton et al., 1998; Wozniak et al., 1998). In addition, splicing variants of the human homologues of Kap104p have been identified (Siomi et al., 1997) (Bonifaci, N., and G. Blobel, unpublished data). As such, it is possible that humans have many more than 14 Kaps. No direct homologues of Kap108p have been identified, and our finding that Kapα/Kapβ1 can import hsLa into human cells suggests that no new Kaps for the import of La proteins will be needed. However, two Kaps, Xenopus RanBP7 and human RanBP8, appear to be equally related to both Kap108p and Nmd5p (Kap119p), a protein that we have recently shown to be a Kap (Albertini et al., 1998). The functional relevance of this homology between RanBP7/ RanBP8 and Kap108p/Nmd5p (Kap119p) is not known, since import substrates for RanBP7 and RanBP8 have not been determined. Also, Kap108p and Nmd5p (Kap119p), although the most similar pair of all the yeast Kaps (26% identical) appear to have no obvious functional overlap (Albertini et al., 1998).

The maintenance of the Kap108p import pathway in S. cerevisiae may have more to do with the early stage of evolution of Lhp1p than with a need for a specific pathway for this protein. Lhp1p is clearly a structural and functional homologue of human La (Yoo and Wolin, 1994; Van Horn et al., 1997; Fan et al., 1998). Even so, we have shown here that the nuclear import of La proteins in yeast and humans are not mediated by directly homologous Kaps. Import of spLa, dmLa, and hsLa in S. cerevisiae takes place even in the absence of Kap108p. In addition, we have been unable to observe an interaction between hsLa and Kap108p (Fig. 4 b). As such it appears that S. pombe, Drosophila, and humans should not need a Kap108p homologue for La import. Although we have been able to directly assess the Kap requirement in human cells, we have not yet done so in S. pombe or Drosophila, so it remains possible that additional Kaps mediate La import in these organisms. Although La seems to use different Kaps for import in S. cerevisiae and S. pombe, it is possible that other proteins will be identified that diverge differently. For this reason, although we believe our results indicate that spLa and dmLa are transported to the nucleus by an α-Kap–dependent pathway, it remains possible that other mechanisms exist for the transport of these proteins in S. pombe and Drosophila. It is clear that this is not the case for humans, as the hsLa NLS has been thoroughly characterized and shown to be similar to the consensus bipartite NLS (Simons et al., 1996). Three scenarios exist for the conservation of nuclear import pathways: (a) directly homologous Kaps can transport directly homologous substrates; (b) directly homologous Kaps can transport nonhomologous substrates, as is the case for the yeast Kap104p and the human Kapβ2 (Siomi et al., 1998); and (c) nonhomologous Kaps can transport directly homologous substrates, as we have described here. It remains to be seen how many examples of each of these scenarios will be found.

La proteins have been cloned from nine divergent species and are highly conserved (Fig. 3 a). In addition to their sequence conservation, functional conservation of La's role in tRNA processing has been rigorously demonstrated (Van Horn et al., 1997; Fan et al., 1998). However, La proteins in higher eukaryotes have been shown to have many functions other than facilitating tRNA processing. These functions occur in both the nucleoplasm and cytoplasm. Other nucleoplasmic functions of La proteins include facilitating initiation, termination, and recycling of RNA polymerase III transcripts (Gottlieb and Steitz, 1989a,b; Fan et al., 1997; Lin-Marq and Clarkson, 1998). Although human La, like homologous La proteins, is nuclear (or nucleolar depending on cell cycle stage [Deng et al., 1981]) at steady state, potentially cytoplasmic functions have been described for this protein. These include: stimulating cap-independent IRES-dependent translation for a variety of viruses (Svitkin et al., 1994; Ali and Siddiqui, 1997); unwinding double-stranded RNA and RNA/DNA duplexes (Bachmann et al., 1990; Huhn et al., 1997) and protecting the highly regulated and structurally unusual histone mRNA from cell cycle–dependent degradation (McLaren et al., 1997). It remains to be seen whether the S. cerevisiae La protein has any of these functions. It has been shown that the inability to translate certain IRES-dependent viral transcripts in S. cerevisiae extracts is not due to inactivity of Lhp1p. Instead, a small RNA inhibitor of this function of La has been identified in S. cerevisiae (Das et al., 1994, 1996).

The evolution of La, a multidomain, multifunctional protein most likely occurred through a number of discrete steps. It is possible that evolution of a Kapα/Kapβ1-dependent NLS allowed La to attain additional functionality, perhaps by decreasing the mass of La used as a signal for nuclear localization (see Fig. 7 for a schematic representation of the four La proteins studied here). As shown above, even a large, 113-aa domain encompassing a Kap108p-dependent NLS only weakly concentrated fragment 2 in the nucleus (Fig. 2). NLSs for α-Kap–independent β-Kaps appear to be larger than NLSs recognized by Kapα. The NLS of Nab2p for Kap104p-mediated import has been mapped to a region 50- or 110-aa long, and the hnRNP A1 NLS for Kapβ2-mediated import is at least 38-aa long (Siomi and Dreyfuss, 1995; Siomi et al., 1998; Truant et al., 1998). In addition, Kap111p/Mtr10p, which imports Npl3p, uses an NLS on the order of 130-aa long (Pemberton et al., 1997; Senger et al., 1998). In contrast, a 27-aa domain of human La (Simons et al., 1996) and a 35-aa domain of Drosophila La (Fig. 6,b) strongly concentrate fusion proteins to the nucleus via α-Kap–dependent pathways. In fact, α-Kap–dependent NLSs are often shorter than 12 aa and can even be as short as 5 aa (Dingwall and Laskey, 1991). In addition, the Kap108p-dependent NLS of Lhp1p includes the RNA-binding domain, and it is therefore possible that RNA binding and Kap108p binding of Lhp1p are mutually exclusive (many NLSs of DNA- or RNA-binding proteins have been mapped to sites overlapping or adjacent to DNA- or RNA-binding domains; LaCasse and Lefebvre, 1995). An example of this exclusivity of binding La has been described for the human protein. A monoclonal antibody, mAb SW3, which binds within the RRM of hsLa, is unable to bind to RNA-associated La (Pruijn et al., 1995). It could be envisioned that directionality for Lhp1p import could be conveyed in this manner. Newly synthesized, cytoplasmic Lhp1p could be recognized in the cytoplasm by Kap108p. Kap108p could then direct Lhp1p through the NPC by virtue of its interaction with nucleoporins (Rosenblum et al., 1997). Once in the nucleus, the Kap108p/Lhp1p complex could be dissembled in the presence of a high concentration of oligouridylate-containing RNA polymerase III transcripts in the nucleus (Senger et al., 1998). The overlap between the Lhp1p NLS and its RNA recognition motif may also affect cytoplasmic functions of La proteins. As shown in Fig. 1 b, a large proportion of cytoplasmic Lhp1p is stably bound to Kap108p. Given that this binding appears to be mediated by a large region that includes the RNA-binding domain of Lhp1p, it appears unlikely that this domain would be accessible for cytoplasmic functions in S. cerevisiae. As such, there may have been an evolutionary advantage to have the NLS of the more complex La proteins in a region not involved in direct RNA binding.

It is possible that using different domains for an NLS would require different modes of regulation for nuclear import/export. One possible mode for regulation of hsLa is by phosphorylation. Phosphorylation of the major casein kinase II site of La, serine 366, has recently been shown to impair the RNA pol III transcription factor activity of La in vitro (Fan et al., 1997). As a corollary to this observation, it would be expected that the unphosphorylated form of La would be present in the nucleus where it performs its transcription-related functions. Our observation that bacterially expressed, and therefore unphosphorylated, hsLa–GFP is actively transported into HeLa cell nuclei is consistent with this expectation. Further, the proximity of Ser-366 to the NLS of La (aa 382–408) raises the possibility that La activity may be coupled to its localization. Such feedback would appear to be particularly important for proteins like La that have nuclear and cytoplasmic functions.

A scheme for the evolution of La can be generated, taking into account known domains of La and our results regarding the divergence of its nuclear targeting (see Fig. 7 for a schematic of relevant domains of representative La proteins). In such a scheme, the common ancestor of the modern La proteins (Fig. 3 a) was imported via an α-Kap– independent β-Kap. The progenitor Kap recognized the progenitor La via a large, complex domain. Branch II diverged from branch I in part by acquiring a strong α-Kap– dependent NLS in the COOH-terminal domain of the protein. In the case of the S. pombe La, the NLS probably begins 44 aa from the COOH terminus. The second major evolutionary divergence (between branch II and branch III) occurred with the acquisition of an even more COOH-terminal NLS, the dmLa NLS, for example, begins within 35 aa of the COOH terminus. This second divergence coincided with a 30% increase in the size of La. Further specialization of the central domain, which was no longer required for nuclear import, led to the addition of an atypical RRM and an ATP-binding motif in the proteins of branch IV (Birney et al., 1993). In the case of human La, the NLS is contained within 27 aa of its COOH terminus (Simons et al., 1996).

Lhp1p also appears to have a weak Kap108p-independent NLS that can be unmasked by removing the first 187 amino acids of the protein. This NLS does not appear to function in the context of full-length Lhp1p (Fig. 2 c, top right panel), but is likely to be relevant evolutionarily, as a corresponding domain of Sla1 does appear to be responsible for nuclear import.

Recently a molecular phylogenetic analysis of the karyopherin superfamily was used to suggest an evolutionary framework for this class of proteins (Malik et al., 1997). This analysis demonstrated that the Arm motifs of α-Kaps and the HEAT motifs of β-Kaps are related. As a result, it is likely that an α-Kap–independent progenitor gave rise to both the α-Kaps and the β-Kaps. The β-Kaps that resulted consisted of both α-Kap–dependent and –independent types (Malik et al., 1997). If this were indeed the case, it might be predicted that such an evolutionary divide might be observed in comparing nuclear transport pathways used for a conserved protein in divergent organisms. It is possible that we have uncovered such a division. An additional possibility is that the evolution of nuclear transport mechanisms of La proteins was intimately related to the specialization of α-Kaps. S. cerevisiae only has one α-Kap whereas humans have at least six α-Kaps, each with a unique substrate specificity and expression pattern (Pemberton et al., 1998). In this light it is possible that the development of additional α-Kaps, which occurred after the divergence of S. cerevisiae from humans (Malik et al., 1997), was a key element in the development of α-Kap– dependent NLSs in La proteins. Given the difference in size between NLSs of α-Kap–dependent and –independent pathways, it would be surprising if other classes of proteins did not recapitulate similar evolutionary strategies. Whether Kap108p-like proteins evolved to specialize a role involving ribosomal proteins will be determined with the continued characterization of this function in S. cerevisiae and the cloning and characterization of Kaps in higher eukaryotes.

We thank E. Ellison for valuable technical assistance and members of the laboratory for numerous discussions. In addition, we would like to thank M. Rout and B. Cravatt for insightful comments that improved the manuscript. We are grateful to B. Cormack (yEGFP3 cDNA), M. Matunis (human Kapα2), N. Yaseen (human Ran), S. Wolin (scLa and dmLa cDNA), M. Rout, and J. Aitchison (Kap95-PrA strain and KAP123-deletion strain) for providing valuable reagents, and J. Fernandez and F. Gharahdaghi (Rockefeller University Protein/DNA Technology Center) for protein sequencing and mass spectral analysis.

aa

amino acid

GFP

green fluorescent protein

IRES

internal ribosome entry site

Kap

karyopherin

NLS

nuclear localization sequence

NPC

nuclear pore complex

RRM

RNA recognition motif