Proteins Connecting the Nuclear Pore Complex with the Nuclear Interior

The yeast strains Saccharomyces uvarum (NCYC 74, ATCC 9080), considered a strain of Saccharomyces cerevisiae (Mortimer and Johnson, 1986), or S. cerevisiae (W303; Thomas and Rothstein, 1989) were used throughout the procedure. Enriched nuclei, NEs and heparin-extracted NEs were prepared as previously described (Rout and Kilmartin, 1990, 1994; Strambio-de-Castillia et al., 1995; Rout and Strambio-de-Castillia, 1998). Enriched NPCs were prepared from nuclei (fraction 7) exactly as described in Rout and Blobel (1993) and Rout and Strambio-de-Castillia (1998).

Proteins contained in 30 ml of the highly enriched NE fraction described above (fraction 10), were precipitated by mixing the sample with 9 vol of methanol and harvested by centrifugation. The methanol pellet was solubilized in 4 ml of 10 mM MES, pH 6.5, 100 mM DTT, 1% SDS at 90°C for 10 min. The resuspended proteins were mixed with 36 ml of 20 mM MES, pH 6.5, 7 M Urea, 1% (vol/vol) Triton X-100, 0.1% SDS, 1 mM DTT (buffer 7). 16 ml of a 1:1 suspension of the cation-exchange S–Sepharose resin (8 ml of resin bed) were loaded on a broad base, 50-ml column and washed three times with 20 ml of buffer 7. The NE sample was loaded onto the column and was allowed to absorb onto the resin by incubating for 1 h at 25°C with gentle rocking. After the binding step, the flow-through from the column was harvested and pooled with a 20-ml wash in buffer 7 (this pooled material was termed unbound fraction). The column was eluted two times with 30 ml each of 1 M NaCl in buffer 7. Proteins from both bound and unbound fractions were harvested by methanol precipitation. Aliquots were separated on SDS-PAGE. After electrophoresis, the fractionation pattern of known NPC components was analyzed by immunoblotting using mAb414 (Davis and Blobel, 1986). The unbound fraction (termed S-NE), was found to be selectively depleted of most nucleoporins recognized by mAb414. Proteins from this fraction were harvested by methanol precipitation, resuspended in PBS, and used to immunize mice.

The production of hybridomas from B-lymphocytes derived from mice spleens was as previously described (Galfre and Milstein, 1981; Rout and Kilmartin, 1990). Supernatants were screened by indirect IF microscopy of whole yeast cells (Rout and Kilmartin, 1990; Wente et al., 1992; Kilmartin et al., 1993). Positive supernatants were also screened by immunoblotting of enriched NEs. Cells from positive lines were cloned up to four times by limiting dilution using a standard protocol (Galfre and Milstein, 1981; Rout and Kilmartin, 1990).

A λgt11 S. cerevisiae genomic expression library (Clontech Laboratories Inc., Palo Alto, CA) was screened by immunoblotting using mAb148G11 following the specifications of the manufacturer. Three positive λ clones containing an identical insert of ∼1.8 kb were obtained and purified to homogeneity by four consecutive rounds of screening. The insert from one of the positive λ clones was subcloned into pBluescript SK(+/−), sequenced from both ends using the T3 and the T7 standard primers and found to contain a sequence identical to the coding region for AA 1,105– 1,700 of MLP1 (Myosin Like Protein 1; Kolling et al., 1993). DNA sequence comparisons were performed using the BLAST algorithm (Altschul et al., 1990). Deducted amino acid sequences were compared with sequences in the SGD (Saccharomyces Genome Database), GenBank and EMBL databases using the FASTA algorithm (Parson and Lipman, 1988). Amino acid sequence alignments were performed using FASTA and CLUSTAL W v. 1.6 (Higgins et al., 1994). Amino acid sequences were analyzed using Protean v. 1.08 (DNAStar Inc.) and MacStripe 1.3.1 (Lupas et al., 1991).

IEM analysis of isolated NEs was performed using a modification of a published procedure (Wray and Sealock, 1984; Rout and Kilmartin, 1990; Kraemer et al., 1995). For the IEM of isolated whole nuclei, the nuclear preparations were subjected to mild osmotic shock by diluting them with 9 vol of PVP solution (8% polyvinylpyrrolidone [PVP]; 20 mM potassium phosphate, pH 6.5; 0.75 mM MgCl₂). These broken nuclei were then transferred (100 μl of diluted sample per well) to microtiter wells pretreated as described for the IEM of isolated NEs, and centrifuged at 23,500 g for 30 min at 4°C. Nuclei pellets were washed twice with PVP solution at 25°C followed by one wash each for 5 min with the following three solutions: (a) 25% (vol/vol) M buffer (5% dried milk in bt-DMSO [10 mM Bis-Tris-Cl, pH 6.50, 0.1 mM MgCl₂, 20% DMSO]), 75% (vol/vol) PVP solution; (b) 50% (vol/vol) M buffer, 50% (vol/vol) PVP solution; and (c) 75% (vol/ vol) M buffer, and 25% (vol/vol) PVP solution. The nuclei pellets were washed once in M buffer for 5 min at 25°C. Subsequent steps were also as described above with the following exceptions: (a) 0.5× PBS-K, 1 mM MgCl₂ was substituted with bt-DMSO; and (b) after treatment with osmic acid, samples were postfixed with 1% tannic acid in 50 mM potassium phosphate, pH 7.0, for 30 min at 25°C.

To generate green fluorescent protein (GFP)-tagged proteins we used the plasmid pGFP-N-FUS encoding the GFP gene under the control of the MET25 promoter (Niedenthal et al., 1996). DNA sequences encoding the following Mlp1p fragments: (a) NT1 (AA 1–667; pGFPNT1); (b) NT2 (AA 668–1446; pGFPNT2); (c) CT (AA 1,447–1,875; pGFPCT); and (d) putative nuclear localization sequence (NLS), pNLS (AA 1,486–1,545; pGFPpNLS), were PCR-amplified and cloned in-frame downstream of the GFP gene using the unique cloning sites XbaI and XmaI. pGFPNT1, pGFPNT2, pGFPCT, pGFPpNLS, and the negative control pGFP-N-FUS were transformed into W303 cells. To induce the expression of the GFP-tagged Mlp1p fragments, cells were grown to mid-log phase in selective medium without methionine. GFP fusion proteins were visualized in living cells after mounting on microscope slides with 2% methyl cellulose, using a Zeiss Axiophot microscope. (Carl Zeiss) equipped with a Chroma no. 41014 GFP Filter (Chroma Technology Corp.). Photomicrographic recording was performed using a Sony DKC5000 digital photo camera (Morrel Instruments Inc.) interfaced with Adobe Photoshop v. 4.01 (Adobe Systems, Inc.)

All yeast strains were derived from W303 (Thomas and Rothstein, 1989; Table I). Gene replacement of MLP1 and MLP2 was accomplished using published methods (Rothstein, 1990; Aitchison et al., 1995a,b). MLP1 was replaced with URA3 by generating a PCR product containing the entire URA3 gene flanked by 75 nucleotides directly upstream and 75 directly downstream of the MLP1 coding region. A similar procedure was followed to disrupt MLP2 with HIS3. A haploid strain carrying disrupted copies of both MLP1 and MLP2 (CSDC09α) was constructed by mating CSDC03α and CSDC05a and subsequent sporulation. Haploid strains of opposite mating types carrying the double deletion of MLP1 and MLP2 were mated to generate a homozygous diploid strain (CSDC13a/α; mlp1&mlp2Δhd). In all cases correct integration and segregation of each individual disruptions were verified by PCR analysis of the genomic DNA. The expression of Mlp1p in wild-type and mutant strains was analyzed by immunoblotting of whole yeast cell lysates and by indirect IF microscopy using mAb148G11.

For protein A tagging of Mlp1p and Mlp2p, PCR products were generated that contained the coding region for four and a half IgG-binding repeats of protein A followed by a HIS3, URA3 cassette flanked by 75 nucleotides immediately upstream and 75 nucleotides immediately downstream of the stop codon (i.e., excluding the stop codon itself) of each gene (Aitchison et al., 1995a,b).

The in vivo import assay was performed as described (Shulga et al., 1996). In brief, wild-type W303 and mlp1&mlp2Δhd cells expressing GFP (Cody et al., 1993) fused to the SV40 large T antigen NLS were incubated in dextrose-free medium containing 10 mM each of 2-deoxy-d-glucose and sodium azide at 30°C for 45 min. At the end of the incubation, cells were washed once and incubated in dextrose-containing medium at 30°C to allow reimport of the substrate into the nucleus. The number of normal cells showing a clear accumulation of GFP-NLS in the nucleus (nuclear cells) and the number of cells in which the reporter was cytoplasmic were counted at each time point. At least 40 independent cells were scored per time point. At least four independent sets of cells were counted to construct the graph presented in Fig. 8. The results are presented as the percentage of cells presenting nuclear signal as a function of time. Linear regression lines were drawn through the linear portion of each curve using KaleidaGraph and the slope of these straight lines were used to estimate the relative import rates.

Passive diffusion assays were performed as follows. pGFP-LEU transformed wild-type and mutant cells were grown as described (Shulga et al., 1996). Cells were harvested by centrifugation and resuspended in one-fifth of the initial volume of selective medium. Resuspended cells were held at 4°C until use. The diffusion assay was started by centrifuging the cells, washing them once in sterile water, resuspending them in 1 vol of 10 mM sodium azide and 10 mM deoxyglucose in dextrose-free selective medium and finally placing them at 30°C. Aliquots were taken at each time point and scored as described above. At least 3 independent sets of 40 cells were scored at each time point. The relative passive equilibration rates were estimated as described for the import rates.

To achieve the overexpression of Mlp1p in yeast cells, the MLP1 gene was inserted into the pYES2 yeast expression plasmid (Invitrogen Corp.) downstream of the GAL1/10 inducible promoter. The unique BspHI site located at nucleotide position 4958 of pYES2 was disrupted using the Klenow fragment of DNA polymerase I to generate pYES2-no BspHI. DNA from the λ clone λPM-5620 (Olson et al., 1986; ATCC catalog number 70598; America Type Culture Collection) containing a yeast genomic fragment of ∼16,800 bp from chromosome XI (chromosome bp coordinates: 611843–628638), was used as the template to generate a ∼650 bp PCR product containing the 5′ region of MLP1 from nucleotide position −12 (relative to the first bp of the coding region) to nucleotide position +634 flanked by a BamHI site at the 5′ end. This PCR product was cut with BamHI and EcoRI (site located at nucleotide position +572 of the MLP1 coding region) and inserted into the BamHI and EcoRI sites of pYES2-no BspHI, to generate pYES2-570MLP1. Finally, an ∼5,900-bp BspHI fragment of λPM-5620 that contains the entire MLP1 coding region except the first 8 bp, was inserted into pYES2-570MLP1 that had been linearized with BspHI (site located at nucleotide position +8 of the MLP1 coding region) to generate pGALMLP1.

pGALMLP1 was transformed into W303 cells by electroporation and transformants were selected taking advantage of the URA3 selectable marker present on the plasmid. Only freshly transformed cells were used for each experiment. To induce the expression of MLP1, W303/ pGALMLP1 cells were grown to mid-logarithmic phase in selective medium containing 2% D(+)-raffinose (referred to as raffinose throughout the text), transferred to selective medium containing 1% raffinose and 2% galactose and incubated at 30°C for up to 4 h. For repression of MLP1 expression, W303/pGALMLP1 cells were grown in raffinose as above, transferred to selective medium containing 1% raffinose and 2% dextrose and incubated for 4 h at 30°C.

SDS-PAGE and immunoblotting were performed essentially as described (Rout and Blobel, 1993). The intensity of bands on immunoblots was quantified using the ImageQuant v.1.1 software in the PhosphorImager system (Molecular Dynamics), when a Cy5-conjugated anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories) was used. Occasionally, computer images of enhanced chemiluminescence signals on photographic films were quantified using the gel plotting macro of NIHImage v.1.60 (Research Services Branch, National Institutes of Health, Bethesda, MD).

Cells were prepared for indirect IF microscopy using the procedure of Kilmartin and Adams (1984) with the modifications of Wente et al. (1992) and Kilmartin et al. (1993). Double labeling with mAb148G11 and a polyclonal rabbit anti-Nup159p antibody (Del Priore et al., 1997) was visualized using Cy3-labeled polyclonal donkey anti–rabbit IgG (cross absorbed against rabbit IgG) and DTAF-labeled polyclonal donkey anti–mouse IgG (cross absorbed against mouse IgG; Jackson ImmunoResearch Laboratories). In all single labeling experiments, the bound antibody was visualized using Cy3-labeled polyclonal donkey anti–mouse or anti–rabbit IgG (Jackson ImmunoResearch Laboratories). The staining conditions were as described (Wente et al., 1992). Photomicrographic recording was performed using a Zeiss Axioplan2 microscope (Carl Zeiss) equipped with a Photometrics SenSys A2S digital photo camera (Photometrics) interfaced with IPLab Spectrum p v. 3.1.1c (Signal Analytics).

The growth competition experiment was performed following published procedures (Smith et al., 1996; Rout et al., 1997; Thatcher et al., 1998).

Mutations or deletions of the genes that encode for certain nucleoporins (for example, Nup133p, Nup120p, Nup145p, and Nup159p) can cause the NPCs to accumulate at one side of the NE, giving rise to tight clusters that are easily identified as spots or patches when cells from such strains are stained with a nucleoporin specific antibody (Doye et al., 1994; Wente and Blobel, 1994; Aitchison et al., 1995a; Heath et al., 1995; Kraemer et al., 1995; Li et al., 1995; Pemberton et al., 1995). It was reasoned that proteins only partially associated with the NPC, or localized to areas of the NE that are not in close contact with the NPC, would either fail to cluster or would only partially cluster with the NPCs in these strains. Isolated yeast NEs prepared as previously described (Strambio-de-Castillia et al., 1995) were used to produce a panel of 114 anti-NE mAbs that were screened by an assay based on this NPC clustering phenomenon. The indirect IF pattern generated by each of the individual mAbs on fixed whole wild-type yeast cells was compared with the staining pattern obtained on a yeast strain carrying a gene disruption in the NUP133 gene (Pemberton et al., 1995). The anti-NE antibody, mAb148G11, was identified by this screen as recognizing an antigen that appeared to only partially cluster in the NUP133 disrupted strain. This mAb recognized a protein that runs as a single band of ∼220 kD (apparent molecular mass; p220) on immunoblots (see Fig. 3).

Both wild-type yeast and NUP133 deleted cells were double stained with mAb148G11 and a rabbit polyclonal antibody against the nucleoporin Nup159p (Fig. 1). In wild-type cells Nup159p and p220 colocalize to a great extent although this colocalization is not absolute. In particular, some areas of the NE are devoid of one or the other signals and some cells show some nucleoplasmic staining with mAb148G11 (Fig. 1, wt, p220 and merge, e.g., the three top cells). This raises the possibility that p220 may also be found in the nuclear interior. In the clustering strain the difference in the localization of Nup159p and p220 is striking (Fig. 1, ΔNUP133). In these cells Nup159p clearly forms tight clusters localized at the nuclear periphery consistent with its localization at the NPC, while p220 appears to be localized in large patches or even in continuous rims at the nuclear periphery that only partially overlap with the NPC clusters. Furthermore, Nup159p signal is closer to the cytoplasm than the p220 signal (Fig. 1, ΔNUP133, merge). This is consistent with the localization of p220 by IEM (see below) and is in accord with results recently obtained in Xenopus oocytes (Shah et al., 1998). In addition, similarly to data obtained with different spindle associated markers (Rout and Kilmartin, 1990), these findings demonstrate that it is possible to obtain IF localization to subregions of the NE even in yeast. Similar results were obtained when the localization of Mlp1p and Nup159p were compared in a NUP120 knock-out strain (data not shown; Aitchison et al., 1995a).

The gene encoding p220, mAb148G11, was identified as MLP1 (Kolling et al., 1993). MLP1 encodes for a protein of 1875 AA with a predicted molecular mass of 218 kD cloned on the basis of its cross-reactivity with a mAb recognizing human platelet myosin (Kolling et al., 1993). This nonessential protein was hypothesized to have a nuclear function on the basis of its subcellular localization (see Discussion). The major structural features of Mlp1p, the position of the cloned fragment of the gene and of the epitope of mAb148G11, are shown in Fig. 2 A. The NH₂-terminal ∼80% of the protein is predicted to have a high α-helical content and contains the heptad-repeats pattern characteristic of coiled-coil proteins. The COOH-terminal ∼400 amino acid residues of the protein are predicted to form a globular tail rich in phenylalanine and proline residues (Fig. 2 A, P/F rich region). The sequence with the highest degree of similarity to MLP1 based on FASTA analysis (28% identical and 66% similar) was the uncharacterized yeast open reading frame (ORF), YIL149c (referred to as MLP2 in Fig. 2 B). YIL149c is expected to encode a protein of 1,680 AA with a predicted molecular mass of 195 kD. MLP1 and YIL149c belong to a duplicated chromosomal region of the yeast genome (Block 38; Wolfe and Shields, 1997) present both on Chromosome XI and on Chromosome IX. The similarity between Mlp1p and Yil149p extends over the whole amino acid sequence and is underscored by the similarity of the overall predicted secondary structure of the proteins. Yil149p also contains a phenylalanine and proline rich COOH terminus. Based on its similarity to Mlp1p and the fact that MLP1 and YIL149c appear to have arisen from a genome duplication event, we propose the name MLP2 for YIL149c (see Discussion). The complete yeast genomic database (Clayton et al., 1997) contains no other putative homologues of MLP1.

Mlp1p and Mlp2p are also similar to the vertebrate and Drosophila Tpr proteins (Frasch et al., 1988; Mitchell and Cooper, 1992; Byrd et al., 1994; Zimowska et al., 1997; Fig. 2 B). These proteins are associated with structures found on the nuclear side of the NPC and may be involved in facilitating nucleocytoplasmic transport (Cordes et al., 1997; Zimowska et al., 1997; Shah et al., 1998). The main difference between Tpr and Mlp1p is that the COOH-terminal tail (excluding the predicted coiled-coil region) of Tpr is relatively much longer, spanning the final 30% of the protein, and has a much more marked acidic character (Byrd et al., 1994; Zimowska et al., 1997). A recently identified S. pombe uncharacterized ORF (SPC162.08c) also shows significant similarities to MLP1, MLP2, and Tpr.

Results from the indirect IF analysis shown in Fig. 1 demonstrated that Mlp1p is localized at the NE in areas that are only partly occupied by NPCs, and that it cannot be only a NPC constituent, even a peripheral one (e.g., a nuclear basket component). To better understand its localization, the fractionation pattern of Mlp1p was followed during the preparation of NEs and NPCs (Fig. 3; Rout and Blobel, 1993; Strambio-de-Castillia et al., 1995). As expected, the majority of the Mlp1p fractionated with the nuclei and highly enriched NE fraction, in agreement with the indirect IF data (Fig. 1). After heparin-extraction of NEs, nearly all of the NE-associated Mlp1p pool (68% of the total) was found in the heparin supernatant (Fig. 3, NE Prep, fractions 13 and 14), demonstrating it is not strongly associated with the membrane. Only some of Mlp1p cofractionated with the enriched NPC preparation (Fig. 3, E-NPCs), suggesting a partial or weak association of this protein with NPCs, again consistent with the indirect IF results (Fig. 1). It is important to note at this point that Mlp1p represents the first NE-associated protein to display such a subcellular fractionation pattern (see Discussion).

The ultrastructural localization of Mlp1p was investigated by pre-embedding labeling IEM using mAb148G11 on both isolated NEs (Fig. 4 A, B, and B′) and isolated whole nuclei that had been subjected to mild osmotic shock to expose the nuclear interior (Fig. 4 C). It should be remembered that using this procedure only the position of the epitope of the antigen is being localized rather then the position of the protein as a whole. On the scale of the immunostained structures here this may not actually reflect the extent of the localization of the entire protein. A second important caveat of many immunolocalization techniques is that the dimensions of the antibody-gold conjugate have to be taken into account in determining the precise localization of the epitope. Nonetheless, on isolated NEs labeled with mAb148G11, gold particles were found almost exclusively on the nuclear side of the NE often though not always in the vicinities of NPCs. The average of the distance between gold particles and the nearest NPC was 66 ± 19 nm on the y axis and 41 ± 34 nm on the x axis (n = 50; Fig. 5, B and C). On numerous occasions the gold appeared to be associated with fibrillar structures stretching from the nuclear side of the NE towards the nucleoplasm (Fig. 4 B′, arrow). This was in marked contrast to the labeling of Nup159p, a nucleoporin known to be associated with the cytoplasmic fibrils attached to the outer ring of the NPC (Fig. 5 A). In this case, and consistent with published results (Kraemer et al., 1995), the gold particles were found significantly closer to the NPC (Y = 33 ± 13 nm; X = 8 ± 8 nm; n = 50; Fig. 5 C). The results obtained with the Nup159p control demonstrate that the absence of Mlp1p signal on the cytoplasmic side of the NE can not be due either to a lack of accessibility or to gross alterations or damage of the NE.

The distribution of the Nup159p control in the isolated whole nuclei was unchanged with respect to the NEs (Y = 37 ± 11 nm; X = 12 ± 14 nm; n = 14; Fig. 5 C, Nuclei). While the majority of the Mlp1p signal in whole nuclei was found in the immediate vicinity of the NE, a significant fraction was found at a considerable distance into the nuclear interior, sometimes as much as ∼300 nm from the mid-plane of the NE (Fig. 5 C, Nuclei). This resulted in the average distribution of Mlp1p being further from the mid-plane of the NE than in isolated NEs (Y = 84 ± 56 nm; X = 33 ± 16 nm; n = 50; using the nearest NPC as a referent). The shorter distance in the case of isolated NEs could be due to the collapse of filaments during the NE isolation procedure (see Discussion). Gold was found even further into the nucleoplasm but it was difficult to establish whether this signal was significantly above background (Fig. 4 C). No obvious structure was observed in association with gold localized in the interior of the nucleus, although fibrils could again be seen associated with gold particles found near the NE. Interestingly, in both isolated NEs and isolated whole nuclei the Mlp1p signal appeared to extend a maximum of ∼120 nm from the cylindrical axis of the NPCs, which corresponds to the proposed minimum in vivo inter-NPC distance (Winey et al., 1997). The results of the IEM localization studies are consistent with the indirect IF and immunoblot results presented above (Figs. 1 and 3), and suggest that Mlp1p is associated with filaments localized at an interface between the nuclear interior and the NPC.

To determine which region of Mlp1p is responsible for its nuclear localization, the coding sequence of the protein was roughly divided in three thirds (NH₂-terminal 1, NT1; NH₂-terminal 2, NT2; and COOH-terminal, CT; Fig. 2 A) and each third was GFP-tagged. In addition, a short sequence localized at the non-coiled-coil COOH terminus was selected because of its high lysine and arginine content and hence its similarities to known NLSs, and was also GFP-tagged (putative NLS, pNLS; Fig. 2 A). The intracellular localization of GFP was determined in living cells expressing either GFP alone or the GFP-tagged fragments of Mlp1p (Fig. 6). As expected, untagged GFP had a predominantly cytoplasmic distribution even though it was not excluded from the nucleus. A similar distribution was observed with cells expressing GFP fused to the both of the NH₂-terminal thirds of Mlp1p and to the putative NLS. Interestingly however, the COOH terminus of Mlp1p was able to direct the targeting of GFP to the nucleus but not to the NPCs. This suggests that it contains a bona fide NLS that has no obvious homologies to other NLSs, and is not sufficient for NPC association.

To determine the subcellular localization of Mlp2p, the gene was genomically tagged with an in-frame COOH-terminal fusion of the IgG binding domains of protein A (pA; Aitchison et al., 1995a). As a control, MLP1 was similarly tagged. Highly enriched NEs fractions were prepared from the Mlp1p- and Mlp2p-tagged strains (Strambio-de-Castillia et al., 1995) and the fractionation pattern of the proteins was assessed on immunoblots. As expected, the fractionation pattern of Mlp1p-pA was indistinguishable from the one observed in Fig. 3 (data not shown). Similarly, Mlp2p-pA cofractionated with the highly enriched NE fraction (Fig. 7 A, fraction 10) but was almost entirely stripped off by the heparin treatment (Fig. 7 A, fractions 13 and 14). The tagged strains were used to determine the subcellular localization of the proteins by indirect IF microscopy (Fig. 7 B). Both proteins were found localized predominantly at patches found at the nuclear periphery similar to that previously observed for Mlp1p (Fig. 1). The localization of Mlp1p and Mlp2p in tagged strains was also determined by preembedding IEM on isolated NEs. Again these two proteins displayed a very similar localization pattern to that observed for Mlp1p (data not shown). These observations are consistent with the hypothesis that Mlp1p and Mlp2p are functional homologues in accordance with their structural similarities and their genetic interaction (see below).

The entire coding regions of both MLP1 and MLP2 were individually disrupted in the diploid yeast strain W303 by integrative transformation of the URA3 and HIS3 genes respectively. Each heterozygous diploid strain (mlp1:: URA3/+ and mlp2::HIS3/+) was sporulated and tetrads were dissected. In both cases four viable spores from most tetrads were observed, demonstrating that neither of these genes is essential and confirming and extending published results (Kolling et al., 1993). Immunostaining with mAb148G11 revealed that the punctate IF staining pattern and the ∼220-kD immunoblot band recognized by both antibodies was absent in mlp1::URA3 cells but was present in mlp2::HIS3 cells (data not shown). This confirmed that mAb148G11 binds specifically to Mlp1p and demonstrated that mAb148G11 does not cross-react with Mlp2p. Segregants of opposite mating types carrying the individual disruptions as confirmed by both phenotypic and genotypic analyses were mated and sporulated. Spores were isolated that carried both selectable markers demonstrating that the double knock-out of MLP1 and MLP2 is viable.

To assess the degree of selective disadvantage conferred by individual and double disruptions in the MLP1 and MLP2 genes, the mlp1Δ, mlp2Δ, and mlp1&mlp2Δ strains were each grown competitively with their wild-type counterpart in rich medium (Smith et al., 1996; Rout et al., 1997; Thatcher et al., 1998). While mlp1Δ and mlp2Δ competed successfully with wild-type, the strain harboring the double deletion lost ground rapidly even though it was initially added in twofold excess and appeared to be eliminated from the population after 30 generations (data not shown). These results demonstrated that mlp1&mlp2Δ had a fitness defect relative to the parental stock equal to 24% (selection coefficient 0.235 ± 0.021; Thatcher et al., 1998) and was effectively non-viable outside the protected laboratory environment. These results also indicated that MLP1 and MLP2 are homologous.

To investigate the possibility that Mlp1p and Mlp2p may be involved in transport of molecules in and out of the nucleus, an in vivo import assay was performed as described by Shulga et al. (1996). This assay allows the detection of kinetic defects in the import rates of a NLS-GFP reporter (Fig. 8 A). In this assay logarithmically growing yeast cells constitutively expressing NLS-GFP are harvested and poisoned in order to block the production of energy. Under these conditions the active import of the NLS-GFP reporter into the nucleus is dramatically reduced resulting in the equilibration of the GFP signal between the nucleus and the cytoplasm by passive diffusion across the NPC. When the metabolic inhibitors are removed and the cells are allowed to recover, NLS-GFP is once again rapidly imported in the nucleus. The relative rates of accumulation of the mutant strain are compared with the ones found with wild-type, revealing any defect in the mutant's efficiency of nuclear import. The steady state distribution of NLS-GFP was indistinguishable in homozygous diploid cells carrying a double deletion of MLP1 and MLP2 as compared with wild-type (data not shown). Nevertheless, mlp1&mlp2Δhd (homozygous diploid) cells displayed a markedly slower relative accumulation rate of NLS-GFP into the nucleus with respect to their wild-type counterpart, 13.5 ± 0.2%/min (wild-type) versus 8.9 ± 0.1%/min (mlp1&mlp2Δhd). Significantly, when double mutant cells were allowed to recover for extended periods of time (up to double the time required for wild-type), the initial equilibrium distribution of reporter protein was regained, again indicating that the efficiency of import and not its steady state balance was affected by the absence of Mlp1p and Mlp2p. The rates of passive equilibration of the NLS-GFP reporter during the incubation with the metabolic inhibitors were also measured in both wild-type and mutant cells (Fig. 8 B). In this case, mlp1&mlp2Δhd displayed a significant increase in the relative passive nuclear egress rates of the NLS-GFP reporter as compared with wild-type, −8.75 ± 1.0%/min (wild-type) versus −13.1 ± 0.6%/ min (mlp1&mlp2Δhd). A model that could explain both the import and the diffusion assay results is presented in Fig. 8 C. The basic assumption of this model is that the rates of passive diffusion of the reporter across the NPC are constant, both in the presence and in the absence of metabolic inhibitors (i.e., they do not require NTP hydrolysis). Consequently, during recovery, after the removal of the metabolic inhibitors (− Inhibitor), the rate of active import becomes greater than the diffusion rate and the net effect is accumulation of the transport substrate into the nucleus. If the rate of import is lower in mlp1&mlp2Δhd cells with respect to wild-type the prediction is that mutant cells will take a longer period of time to reach the steady state levels of nuclear accumulation of the substrate, and this is exactly what is observed. On the other hand, during the incubation with the inhibitors (+ Inhibitor), the rate of active import is drastically reduced in both wild-type and mutant cells. However, in the mutant cells this residual import is even further compromised as compared with the wild-type cells. As the diffusion rate remains constant this will appear as an overall faster egress rate from the nucleus in the mutant cells.

The involvement of Mlp1p and Mlp2p in active nuclear export was investigated using two steady state assays (data not shown). In the first assay, the subcellular distribution of poly(A)⁺ RNA was analyzed by in situ hybridization using digoxigenin-labeled oligo(dT)₃₀ as a probe (Wente and Blobel, 1993). In the second assay, the steady state distribution of a nuclear export sequence (NES)-GFP was studied by direct fluorescent microscopy (Stade et al., 1997). In both cases no effect on export was detected at steady state and the double mutant cells appeared indistinguishable from wild-type.

When Mlp1p was expressed from a high copy number 2 μm plasmid, the anti-Mlp1p antibody used by Botstein and coworkers recognized intensely staining dots and sometimes rings localized adjacent to the nucleus (Kolling et al., 1993). This localization does not correspond to the native localization of Mlp1p discussed in this manuscript (see above). Mlp1p was overexpressed in W303 cells in order to investigate whether this discrepancy could be accounted for by different levels of expression of this protein. The entire coding region of MLP1 was subcloned in a 2 μm–based yeast expression vector under the control of the GAL inducible promoter (see Materials and Methods). Using this plasmid (pGALMLP1), it was possible to overexpress Mlp1p at least ∼100-fold over wild-type levels as demonstrated by semi-quantitative immunoblotting performed with mAb148G11 (Fig. 9 A). Strikingly, overexpression of Mlp1p at these levels was not lethal as demonstrated by the ability of cells carrying pGALMLP1 to grow for days on galactose (data not shown). Cells containing this construct were either induced with galactose for various periods of time or repressed with dextrose for 4 h and analyzed by indirect IF microscopy using mAb148G11 to reveal the localization of Mlp1p (Fig. 9 B). As expected, cells in which the expression of the exogenous copy of MLP1 was repressed with dextrose (Fig. 9 B, 4 h [Glu]) showed a staining pattern very similar to the one observed in wild-type cells. In contrast, induced cells showed an increase of the Mlp1p-specific signal as a function of the induction time. Initially, small dots (1–4 per cell) could be seen at the nuclear periphery and a clear punctate rim pattern could still be distinguished in most cells (Fig. 9 B, 0 h [Gal]). Subsequently, these dots appeared to coalesce and generally gave rise to one prominent circular patch per cell and the nuclear rim pattern became increasingly more diffuse (Fig. 9 B, 0.5–2 h [Gal]). Finally, the large patch grew to occupy most of the nuclear interior (Fig. 9 B, 4 h [Gal]). This staining pattern closely resembled the one described by Kolling et al. (1993), thus reconciling the apparent difference. Careful examination of the overexpressing cells showed that the intense nuclear DAPI signal was either much reduced of excluded from the overexpression spots. It often appeared that the growing Mlp1p spheroids had pushed the chromatin aside, leaving clear indentations in the chromatin or in some cases even compressing it between two spheroids (Fig. 9 C). Cells overexpressing Mlp1p were observed by thin-section transmission EM to reveal whether any novel structure could be detected (data not shown). As a comparison, cells that had been grown in dextrose to repress the expression of the non-chromosomal copy Mlp1p were also analyzed using the same technique. Extensive electron-dense fibrillogranular networks that extended from different areas of the NE towards the nuclear interior were observed in cells grown in galactose but were absent in cells grown in dextrose. To determine if the electron-dense networks present in induced cells were indeed formed of large accumulations of Mlp1p, isolated nuclei from cells grown in both galactose (Fig. 10) and dextrose (data not shown) were immunostained with mAb148G11 and prepared for IEM using the same pre-embedding labeling technique used above. As expected, the dense fibrillogranular networks present in induced cells specifically stained with mAb148G11 demonstrating that they contain large quantities of Mlp1p and that the Mlp1p spheroids are highly permeable to large macromolecules such as the antibody-gold conjugate. In cells grown in dextrose, Mlp1p appeared to have a localization that was indistinguishable from wild-type.

The clustering of yeast NPCs within characteristic patches in the NE induced by the deletion of certain nucleoporins has previously been used to confirm the association of various proteins with NPCs (Grandi et al., 1995; Aitchison et al., 1996; Simos et al., 1996; Siniossoglou et al., 1996; Rout et al., 1997). Here we used this clustering assay to screen a bank of monoclonal antibodies raised against NE proteins, and isolate those which recognize NE components not associated with the NPC from those recognizing NPC components and nuclear transport factors. We expected to find mainly integral nuclear membrane proteins, analogous to the LAPs or LBR proteins in vertebrate NEs (Senior and Gerace, 1988; Worman et al., 1988, 1990; Foisner and Gerace, 1993). Instead, a monoclonal antibody was found which identified a protein, Mlp1p, associated with the nuclear periphery of the NPC and NE and extending into the nuclear interior. Sequence comparisons of the predicted amino acid sequence of Mlp1p with the entire GenBank database revealed numerous proteins with strong similarities. The most significant was a second yeast ORF predicted to have arisen as a chromosomal duplication event (Wolfe and Shields, 1997). We showed this to encode a redundant homologue of Mlp1p which we therefore named Mlp2p. Next, the vertebrate protein Tpr, its probable Drosophila homologue, and an uncharacterized S. pombe ORF showed strong similarity to Mlp1p and Mlp2p along their entire lengths. Though sequence similarity alone is not necessarily definitive, given the similarities in their localization and possible roles in nuclear transport, we conclude that Mlp1p and Mlp2p are the yeast homologues of Tpr (see below; Cordes et al., 1997; Zimowska et al., 1997; Bangs et al., 1998; Cordes et al., 1998; Shah et al., 1998). This represents the first identification and characterization of Tpr homologues in yeast, setting the stage for studies that will hopefully elucidate the functions of this protein in a genetically and molecularly tractable system.

The Tpr homologues are predicted to consist of a coiled-coil NH₂ terminus occupying most of the primary sequence. It seems likely that the putative coiled-coil domain forms an extended structure that may in turn be involved in organizing higher order homopolymers (for example, filaments). In contrast, the COOH terminus may interact with heterologous factors and anchor these polymeric structures to the NPCs or to the nuclear interior (see also below). Despite its name, based on its cross reactivity with a mAb against myosin (most likely accounted for by the large coiled-coil NH₂-terminal domain; Kolling et al., 1993), Mlp1p does not belong to the myosin family due to the different organization of the structural domains along the primary sequence of the protein.

IF microscopy of the NPC clustering strains shows a partial coclustering of Mlp1p with nucleoporin markers, indicating a significant fraction of Mlp1p is associated with the NPCs. Our IEM experiments localizing Mlp1p and Mlp2p to the nuclear face of the nuclear envelope in yeast agree with the recent results obtained with vertebrate Tpr (Cordes et al., 1997). Our IEM localization also suggests that, like Tpr, Mlp1p and Mlp2p form extensive filamentous structures radiating into the interior of the nucleus from foci at the NPC periphery, perhaps attached at the distal ends of the NPC baskets or fishtraps (Ris and Malecki, 1993; Cordes et al., 1997; Zimowska et al., 1997). Also like Tpr, Mlp1p penetrates deep into the yeast nucleus, as far as half of the nuclear radius, and is thus potentially in contact with at least 80% of the chromatin. Interestingly, a significant fraction of Mlp1p failed to cocluster with NPCs, remaining instead as numerous punctate foci distributed around the nuclear periphery. However, virtually all of Mlp1p cofractionated with the NE fraction, and the majority of Mlp1p signal was in close proximity to the NE inner membrane by IEM. Taken together, this suggests that there is also a portion of Mlp1p not associated with NPCs but distributed around the interporous regions of the NE inner face. This may differ from Tpr, which was reported to be absent from these interporous regions, although other studies have found structures believed to be composed of Tpr interconnecting between NPCs over the inner NE surface (Ris and Malecki, 1993; Cordes et al., 1997). The punctate IF staining pattern of Mlp1p in regions of NE devoid of nucleoporin signal (and hence NPCs) may indicate that it can also organize from non- NPC-associated foci distributed around the nuclear face of the inner nuclear membrane. In an alternative scenario, structures formed of Mlp1p would absolutely depend on NPCs for their nucleation at the NE but would be subsequently able to spread all around the nuclear rim in a non-NPC-dependent manner.

It has long been observed that the NPCs are structurally continuous with the nuclear interior (Monneron and Bernhard, 1969; Franke and Falk, 1971). Interconnecting open channels have been observed radiating from the nuclear interior towards NPCs (Fawcett, 1981; Berezney et al., 1995), and the movement of proteins and RNAs along distinct intranuclear pathways can be studied both in vivo and in vitro (Lawrence et al., 1989; Xing et al., 1993; Zachar et al., 1993; Misteli et al., 1997). These and other similar observations have led to speculations that efficient exchange of material between the nuclear periphery and the nuclear interior could occur along a filamentous network of tracks. (Blobel, 1985; Lawrence et al., 1989, 1993; Meier and Blobel, 1992, 1994; Xing et al., 1993; Berezney et al., 1995; Ris, 1997; Pederson, 1998). Indeed, one model proposes the existence of a nucleoskeleton (defined as a nonchromatin intranuclear structural framework) composed of a filamentous network organized from the nuclear periphery and playing a major role in the direction of nuclear trafficking to and from the NPCs (Berezney et al., 1995). Until recently, the existence of tracks had remained unproven, as no clear candidates for track components had emerged.

A case can be made for the Mlp1p/Tpr family of proteins meeting some of the characteristics expected of nucleoskeletal track components. As one would expect for track proteins they are localized further away from the NPCs than the most peripherally described structures of the NPCs (the nuclear baskets and cytoplasmic filaments) and penetrate most of the volume occupied by chromatin (as discussed above; Ris and Malecki, 1993; Cordes et al., 1997; Ris, 1997). Their peripheral association with the NPC is emphasized by the partial fractionation of Mlp1p with isolated NPCs, though an independent tight association with the nuclear periphery is suggested by its cofractionation with enriched NEs and continuous distribution around the NE even in strains where the NPCs have clustered. Furthermore, such components would not necessarily be stoichiometric with respect to NPC proteins, and not coassemble with them during NPC assembly. Instead, they could carry an NLS to be imported separately into the nucleus. Thus, the domain necessary for NPC association could be separated from that for nuclear localization, as indeed we showed for Mlp1p and was also recently shown for vertebrate Tpr (Bangs et al., 1998; Cordes et al., 1998). The Tpr family seems to form filamentous structures of considerable length (unsurprising considering their predicted coiled-coil structure) that are likely polymeric (Cordes et al., 1997). Interestingly, the most distal of the intranuclear signal appeared to collapse towards the inner nuclear membrane during the preparation of isolated NEs. Preliminary data indicate an extreme version of this collapse occurs when NEs are isolated from cells overexpressing Mlp1p (data not shown; see below). Thus it would seem, as expected of a track component, that interaction with intact chromatin is required to maintain the normal distribution of Mlp1p within the nucleus and even opens the possibility of a role for this protein in the maintenance of the nuclear architecture. As lamins are absent from yeast, it may even substitute in part for their function. The ability of Mlp1p to self-assemble and form regular polymers that could account for such structural functions remains to be demonstrated and will be the subject of future studies.

As with Tpr, the copious labeling of the COOH-terminal epitope recognized by our antibody indicates that this part of the protein is free, perhaps to interact with chromatin or transport factors (Cordes et al., 1997). A possible interaction has been found between Tpr and the vertebrate karyopherin β1 nuclear import factor (Shah et al., 1998). Here we show that Mlp1p and Mlp2p are required for the efficient nuclear import of a substrate of the homologous yeast karyopherin β1 plus karyopherin α. Although this represents the first in vivo evidence for a role of this protein family in nuclear transport, it must still be regarded with caution given the potential for pleiotropic effects in this experiment (see Results). Nonetheless, together such data support the idea of a direct role for the Tpr family in nuclear transport by active transport of substrates between binding sites positioned opportunely along the filaments. However, Tpr also coincides with clear channels extending from the NPC into the nuclear interior (Zimowska et al., 1997). Overexpression of Mlp1p forms spheroidal fibrillogranular structures which can occupy large portions of the nucleus. Overexpression aggregates such as these usually only display some of the characteristics of the protein under normal conditions, and other characteristics may be anomalous. However, contrary to many other proteinaceous aggregates (e.g., Capsey et al., 1990) the overexpression spheroids of Mlp1p are surprisingly open to large molecules, as indicated by the apparently unhindered access and accumulation of antibody conjugated ∼10 nm gold (added before fixation, embedding and sectioning for IEM) throughout them. In addition, it is interesting that the Mlp1p spheroids displaced chromatin as they grew to form large chromatin free regions within the nucleus. One could imagine a polymeric cylinder of Mlp1p with the properties of these spheroids projecting from the NPC into the nucleus. Such a structure would maintain a chromatin free channel that remains highly permeable to even large proteinaceous particles. These data are consistent with binding tracks for transport factors, but also raise another possible and less direct role for this protein family; they may maintain efficient nucleocytoplasmic transport by holding open diffusion channels, for the unhindered intranuclear movement of transport substrates.

It came as a considerable surprise that the deletion of both yeast Tpr homologues was not lethal. The function of both proteins must therefore not be essential in yeast, or other functionally redundant proteins must exist. Either way, this result holds important consequences for the increasing number of studies underway on the vertebrate Tpr proteins. Thus despite a reported association with transport factors (Shah et al., 1998) and a deleterious effect of vertebrate Tpr overexpression on mRNA export (Bangs et al., 1998), Tpr homologues are not necessary to support basic nuclear transport. In vivo experiments should be conducted to determine if vertebrate cells have a greater requirement than yeast for Tpr, and what factors (such as cell size) might contribute to any differences. Further work is also required on the Mlp1p/Mlp2p deficient strain to find conditions where at least one of these proteins is required for viability. Such experiments may lead to interacting proteins and processes; yeast studies have been key in showing that many nucleocytoplasmic transport operatives are redundant with numerous subtly overlapping functions, and the Tpr family are proving no exception (Rout et al., 1997; Wozniak et al., 1998). The behavior of these proteins should also be studied in living yeast cells (such as by using GFP-tagged Mlp1p), taking advantage of available conditional mutant strains, to avoid some of the problems previously associated with studying potential nucleoskeletal proteins biochemically (Singer and Green, 1997; Pederson, 1998).

In conclusion, we believe that Mlp1p and Mlp2p, together with their multicellular eukaryotic counterparts, satisfy all the criteria expected of non-nucleoporin nucleoskeletal components. We further suggest they represent the strongest candidates to date for the molecular components of the long hypothesized nuclear tracks connecting the NPCs with the nuclear interior. The characterization of the homologues of Tpr in a genetically tractable organism complements the studies of these proteins in vertebrate systems and opens new avenues in the study of nuclear cell biology.

GFP

green fluorescent protein

IEM

immunoelectron microscopy

IF

immunofluorescence

MLP

myosin-like protein

NE

nuclear envelope

NES

nuclear export sequence

NPC

nuclear pore complex

NUP

nucleoporin

PVP

polyvinylpyrrolidone