Two membrane proteins were identified through their genetic interaction with the nucleoporin Nup84p and shown to participate in nuclear envelope morphogenesis in yeast. One component is a known sporulation factor Spo7p, and the other, Nem1p, a novel protein whose C‐terminal domain is conserved during eukaryotic evolution. Spo7p and Nem1p localize to the nuclear/ER membrane and behave biochemically as integral membrane proteins. Nem1p binds to Spo7p via its conserved C‐terminal domain. Although cells without Spo7p or Nem1p are viable, they exhibit a drastically altered nuclear morphology with long, pore‐containing double nuclear membrane extensions. These protrusions emanate from a core nucleus which contains the DNA, and penetrate deeply into the cytoplasm. Interestingly, not only Spo7− and Nem1−, but also several nucleoporin mutants are defective in sporulation. Thus, Spo7p and Nem1p, which exhibit a strong genetic link to nucleoporins of the Nup84p complex, fulfil an essential role in formation of a spherical nucleus and meiotic division.
In eukaryotic cells, the nuclear envelope separates the nuclear from the cytoplasmic compartment. Nuclear pore complexes (NPCs) which are embedded in the nuclear membrane regulate transport between these two compartments (for a review see Panté and Aebi, 1995; Doye and Hurt, 1997). Based on electron microscopy studies, the NPC has been proposed to consist of a central spoke assembly with 8‐fold symmetry to which a cytoplasmic and nuclear ring are attached. From these rings, filaments emanate towards the cytoplasm and nucleus, respectively. Although the yeast NPC is slightly smaller than its vertebrate counterpart, the overall architecture of the NPC is conserved in all eukaryotic cells (Rout and Blobel, 1993; Yang et al., 1998). During the past years, many NPC‐associated proteins (nucleoporins) have been identified and characterized, particularly in the yeast Saccharomyces cerevisiae, where ∼30 nucleoporins are known to date, and many of them are organized in distinct NPC subcomplexes (for a review, see Doye and Hurt, 1997). In contrast to the increasing knowledge of the composition of NPCs, very little is known about how and where this complex organelle assembles. It is believed that a fusion event between the outer and inner nuclear membrane generates a hole in the envelope, which is then followed by the ordered assembly of individual nucleoporins or their respective subcomplexes. Indeed, intermediates during NPC biogenesis recently have been visualized in growing nuclear envelopes from Xenopus egg extracts (Goldberg et al., 1997). However, the role of transmembrane proteins in the biogenesis and assembly of the NPCs is not clear, since deletion of the only known integral pore membrane protein in yeast, Pom152p, does not affect cell growth (Wozniak et al., 1994). Thus, other components of the nuclear envelope must participate in the fusion and anchoring of NPCs within the nuclear pore membrane. It has been suggested that Ndc1p, an essential membrane protein required for the insertion of the spindle pole body (SPB) into the nuclear envelope, might play a similar role during NPC insertion and assembly (Winey et al., 1993).
The nuclear envelope is composed of two distinct membranes enclosing the perinuclear space (for a review, see Goldberg and Allen, 1995; Gant and Wilson, 1997). The outer nuclear membrane is continuous with the endoplasmic reticulum (ER) and thus also functions in protein translocation and secretion. The inner nuclear membrane in higher eukaryotes is covered by the nuclear lamina, which faces the nucleoplasm and participates in the organization of the chromatin and nucleoskeleton. However, the yeast nuclear envelope, which in contrast to higher eukaryotes does not break down during nuclear division, appears not to have an underlying nuclear lamina.
During interphase, the surface of the nuclear envelope in yeast cells increases (Jordan et al., 1977; Winey et al., 1997), but the mechanism of nuclear membrane biogenesis and how this is coordinated with NPC assembly is not known. There are several examples where the biogenesis and organization of the nuclear membrane in yeast is impaired; in one case, overexpression of membrane proteins such as HMG‐CoA reductase isoform 1 (HMG1) (Wright et al., 1988), cytochrome b5 (Vergeres et al., 1993) or chicken lamin B receptor (Smith and Blobel, 1994) caused a striking proliferation of the nuclear membrane, forming a stack of paired membranes which are arranged in layers around the nucleus. This structure, which was called ‘karmellae’ and corresponds to a proliferated ER membrane, is free of NPCs. The link between the overproduction of a membrane protein and the generation of karmellae is not clear, although it was reported that a luminal domain of HMG1 is responsible for the membrane proliferation (Parrish et al., 1995). Drastic nuclear membrane perturbations, which are completely different from ‘karmellae’ membrane stacks, have been observed in nucleoporin mutants. The structural abnormalities of the nuclear envelope concern both NPC and nuclear membrane organization. These include grape‐like arrangements of both NPCs and nuclear membrane domains, for instance in nup145 mutants (Wente and Blobel, 1994), an outer nuclear membrane seal over the NPCs, as in the case of nup116 mutants (Wente and Blobel, 1993), or cells with long nuclear envelope projections, as in nup1 mutants (Bogerd et al., 1994) or nup170pom152 double mutants (Aitchison et al., 1995).
Drastic structural defects of the nuclear envelope were also observed in nup85 mutants, and to a lesser extent in nup84‐disrupted cells (Goldstein et al., 1996; Siniossoglou et al., 1996). In particular, the nuclear envelope of nup85‐disrupted cells has protrusions which come in contact with each other, thereby engulfing part of the cytoplasmic compartment. At these contact zones, the nuclear pores clustered extensively, forming grape‐ or blister‐like NPC/nuclear membrane arrangements. Mutations in MTR7/ACC1, which encodes acetyl‐CoA‐carboxylase, a key enzyme in fatty acid biosynthesis, also cause an expansion of the perinuclear space due to the wide separation of outer and inner nuclear membrane, and protuberances extending from the inner membrane into the intermembrane space (Schneiter et al., 1996). This was taken as evidence for a specific need for lipids with very long fatty acid chains to stabilize the insertion of NPCs at the pore‐membrane interface and allow for a correct nuclear membrane biogenesis. Finally, the integral nuclear/ER membrane protein Snl1p was found recently in a genetic screen for high copy suppressors of a temperature‐sensitive mutant overproducing the Nup116p C‐terminal domain (Ho et al., 1998). It was suggested that Snl1p plays a role in stabilizing NPC structure and function because its overproduction can rescue the nuclear membrane herniations of nup116 mutants.
We previously have identified the Nup84p nucleoporin complex consisting of six subunits, with an essential role in nuclear envelope/NPC organization and mRNA export (Siniossoglou et al., 1996). Strikingly, this complex contains Sec13p, a COPII coat protein involved in vesicular transport from the ER to the Golgi (Pryer et al., 1993). This suggested a link between distinct steps in nuclear membrane and pore biogenesis, and vesicular transport. Furthermore, the Nup84p complex plays a crucial role in correct NPC distribution within the nuclear membrane (see above). In this study, we extend these studies of nuclear envelope organization in yeast and describe a novel complex consisting of two nuclear/ER membrane proteins, Spo7p and Nem1p, which were found in a genetic screen with a nup84 disruption allele. Both Spo7p and Nem1p are essential for the maintenance of normal spherical nuclear morphology, but do not participate in nucleocytoplasmic transport reactions. Surprisingly, we found that not only the Spo7p‐Nem1p complex, but also a subset of nucleoporins, including Nup84p, are essential for sporulation, suggesting a so far unrecognized link between this process and nuclear envelope/NPC organization.
A protein required for sporulation and a novel conserved protein genetically interact with the nucleoporin Nup84p
Nup84p is a member of a large nucleoporin complex (Siniossoglou et al., 1996). Yeast cells lacking NUP84 show a moderate defect in mRNA export as compared with nup85, nup120 and nup145 mutants, but are strongly impaired in nuclear pore distribution. This prompted us to use the nup84 disruption strain in a synthetic lethal screen in order to isolate novel factors involved in nuclear membrane and pore biogenesis (see Materials and methods). Among the 25 000 mutagenized colonies, nine mutants were synthetically lethal (sl) with the nup84::HIS3 disruption allele. The complementing genes of five of these sl mutants were identified. Sl243 and sl384 were complemented by the nucleoporin gene NSP1 (Hurt, 1988) and sl273 by another nucleoporin gene, NUP116 (Wente et al., 1992; Wimmer et al., 1992). Sl269 was complemented by SPO7 (DDBJ/EMBL/GenBank accession No. 349744; SGD accession No. YAL009w) and sl235 by a novel gene (DDBJ/EMBL/GenBank accession No. 500822; SGD accession No. YHR004c) designated NEM1 (for nuclear envelope morphology; see also below) (Figure 1A). NEM1 encodes a protein of 446 amino acids with a predicted mol. wt of 50 kDa. SPO7 encodes a protein of 30 kDa and was identified many years ago as a gene required for sporulation in yeast (Esposito and Esposito, 1969, 1974).
Although Spo7p and Nem1p are not homologous, they share some common features in their protein sequences. Both are very basic (isoelectric point of 10.1 for Spo7p and of 9.1 for Nem1p) and contain in their primary sequence extended stretches of hydrophobic amino acids, suggesting that they might be membrane proteins (see Figure 3A). A database search revealed no Spo7p homologues in other organisms so far. However, several open reading frames (ORFs) in different organisms, including Schizosaccharomyces pombe, Caenorhabditis elegans and human, exhibit high homology within the 200 C‐terminal amino acids of Nem1p (Figure 1C). Furthermore, three additional uncharacterized ORFs (YLR019w, YLL010c and YPL063w) in S.cerevisiae are also homologous to the Nem1p C‐terminal domain (Figure 1C; only two are shown). This suggests that the C‐terminal domain of Nem1p defines a novel protein family which is evolutionarily conserved. The identity between the C‐terminal domain of Nem1p and its highly related counterparts ranges between 34 and 37%, with 100% conserved blocks of 8‐10 residues (Figure 1C). Within their N‐terminal domains, the various members of this family do not show an overall sequence similarity, although individual members exhibit homology within this part of the protein (e.g. the N‐terminal domain of YLR019w and YLL010c).
NEM1 and SPO7 genetically interact with other NPC components
NEM1 or SPO7 are not essential for cell viability; however, nem1::HIS3 and spo7::HIS3 mutants exhibit a slightly reduced growth rate at higher temperatures (37°C) (Figure 1B) and grow more slowly at lower temperatures (16°C) (data not shown). This non‐essential phenotype enabled us to test genetically the functional interactions of NEM1 and SPO7 with other components of the NPC (Table I). As expected, cells without NEM1 or SPO7 are not viable in a genetic background of the nup84::HIS3 gene disruption, i.e. they are synthetically lethal. A complex pattern of genetic interactions was observed for the nem1::HIS3 and spo7::HIS3 mutants in relation to other members of the Nup84p complex. Whereas the combination spo7::HIS3 seh1::HIS3 yielded progeny which grow extremely slowly at 30°C when compared with the corresponding single mutant strains and are thermosensitive at 37°C, the nem1::HIS3 seh1::HIS3 pair gave viable cells with a normal growth rate (Table I). The opposite pattern was obtained for the nup85Δ allele when tested in combination with spo7::HIS3 and nem1::HIS3. Interestingly, spo7:: HIS3, but not nem1::HIS3, was synthetically lethal with the nup188::HIS3 disruption allele (Table I). NUP188 was often found in sl screens with nucleoporin mutants (Nehrbass et al., 1996; Zabel et al., 1996; Teixeira et al., 1997) and is linked predominantly to structural components of the NPC including Pom152p, the only known transmembrane nucleoporin in yeast (Wozniak et al., 1994; Nehrbass et al., 1996). However, no genetic interaction was detected between disruption alleles of POM152 and SPO7 (Table I). Finally, no synthetic lethal relationship was found between SPO7 and NEM1, since the combination of their disruption alleles yielded viable haploid progeny.
In conclusion, genetic tests revealed that NEM1 and SPO7 genetically interact with several members of the NPC implicated in NPC/nuclear envelope biogenesis. This prompted us to analyse whether Spo7p and Nem1p are also NPC proteins.
Spo7p and Nem1p are nuclear envelope/ER membrane proteins
To analyse the functional relationship between Spo7p, Nem1p and nucleoporins, we determined their subcellular location by tagging both proteins with the green fluorescent protein (GFP). Nem1p‐GFP and Spo7p‐GFP fusion constructs were expressed, under their authentic promoters, in the nem1::HIS3 and spo7:HIS3 mutants, respectively. Nem1p‐GFP and Spo7p‐GFP were able to complement the lethal phenotype of the corresponding sl mutants, sl235 and sl269, respectively, and were therefore functional (data not shown).
Although the expression of Spo7p‐GFP was low (as judged by Western blotting using anti‐GFP antibodies), a weak nuclear envelope staining was observed in the fluorescence microscope, which was significantly enhanced when Spo7p‐GFP was expressed from a high copy plasmid (Figure 2A). Interestingly, this ring‐like staining was not punctate but smooth, which is more typical of a nuclear membrane/ER distribution (Preuss et al., 1991). Furthermore, a discontinuous labelling close to the plasma membrane was observed which could be the underlying ER. Indeed, when a double staining experiment was performed using antibodies against Sec61p, an ER membrane protein, both signals overlapped (data not shown). Finally, Spo7p‐GFP did not cluster in the nup133 deletion mutant, where bona fide nucleoporins cluster (Doye et al., 1994; Pemberton et al., 1995) (data not shown). We therefore conclude that Spo7p localizes to both the nuclear membrane and the ER, but is not associated with NPCs. However, we cannot exclude that a minor fraction of Spo7p interacts with nuclear pores (see also Discussion).
The expression of NEM1‐GFP under its authentic promoter and from a centromeric plasmid was too low to detect a distinct GFP fluorescence signal in the cells. We therefore constructed an N‐terminal GFP‐NEM1 fusion under the control of the NOP1 promoter and expressed from a centromeric plasmid. This fusion protein could complement sl235 and, when localized in a nem1::HIS3‐disrupted diploid, exhibited a weak nuclear envelope and peripheral ER staining (Figure 2B). This suggested that Nem1p also localizes to the nuclear/ER membrane.
Hydropathy plot analysis indicated that Nem1p and Spo7p contain potential membrane‐spanning sequences (Figure 3A). Interestingly, the predicted transmembrane domains of Spo7p and Nem1p are two closely spaced hydrophobic stretches, separated by a lysine and proline in the case of Nem1p and a longer, 14 residue hydrophilic sequence in the case of Spo7p (Figure 3A). In order to test whether they are integral membrane proteins, Nem1p and Spo7p were tagged with protein A (see Materials and methods). This yielded functional fusion proteins able to complement the corresponding sl mutants (data not shown). When the spheroplasts were lysed in a buffer devoid of detergent, neither Nem1p‐ProtA nor Spo7p‐ProtA were recovered in the soluble supernatant where most of the cellular proteins partition (data not shown). However, the same lysis buffer containing 1% Triton X‐100 efficiently extracted both Nem1p‐ProtA and Spo7p‐ProtA. To show whether Nem1p and Spo7p are membrane associated or integral membrane proteins, a crude membrane fraction was prepared and extracted with a variety of reagents (Figure 3B). Following ultracentrifugation, the insoluble pellet (P) and soluble supernatant (S) were analysed by Western blotting using anti‐ProtA antibodies. For control reasons, we also followed Sec61p, an integral ER membrane protein, and Nup85p, a component of the NPC. Detergent‐free buffer or buffer containing 1 M salt did not release Nem1p‐ProtA or Spo7p‐ProtA. Addition of 1% Triton X‐100 solubilized most of Nem1p‐ProtA and Spo7p‐ProtA. On the other hand, incubation of the membrane fraction in pH 11.5 (0.1 M sodium carbonate) buffer did not extract Nem1p‐ProtA and Spo7p‐ProtA. As expected, the integral membrane protein Sec61p, but not Nup85p, was resistant to pH 11.5 extraction. Taken together, these results indicate that Nem1p and Spo7p behave like integral membrane proteins.
NEM1 and SPO7 are essential for correct nuclear morphology
Cells lacking the NEM1 or SPO7 gene do not exhibit defects in nucleocytoplasmic transport pathways, as tested for nuclear uptake of a nuclear localization signal (NLS)‐containing import substrate and nuclear export of mRNA (data not shown). We therefore looked for alterations in nuclear pore and nuclear membrane organization in these mutants by following the location of a nuclear pore reporter Nup49p‐GFP (Belgareh and Doye, 1997; Bucci and Wente, 1997) (Figure 4A). This revealed a strikingly different pattern of Nup49p‐GFP distribution. Instead of an exclusive punctate nuclear envelope staining typical for wild‐type cells, Nup49p‐GFP was found in many foci scattered throughout the cytoplasm of spo7 and nem1 mutants. Often these foci appeared to be aligned on a ribbon‐like array, which either surrounded the vacuole or was underlying the plasma membrane. To exclude that only Nup49p‐GFP mislocalized in nem1 and spo7 mutants, we purified Nup49p‐ProtA from wild‐type and spo7‐deleted cells. In both cases, Nup49p‐ProtA co‐purified with the other three members of the Nup49p complex (Grandi et al., 1993), showing that its assembly is not affected (data not shown). Furthermore, when the distribution of another nucleoporin, Seh1p‐GFP (Siniossoglou et al., 1996), was analysed, the same abnormal NPC staining was observed (data not shown). This all suggested that the overall NPC distribution is severely altered in nem1 and spo7 mutants. To test whether this is paralleled by an abnormal nuclear morphology, we analysed the localization of an intranuclear protein tagged with GFP, Pus1p‐GFP (Hellmuth et al., 1998). A strikingly different nuclear morphology was observed in Spo7− and Nem1− cells as compared with wild‐type cells (Figure 4B). Instead of being round, nuclei are irregularly shaped and elongated, and exhibit long and thin projections. The nucleus in a non‐dividing cell often consists of two lobes interconnected by a long nuclear membrane extension (Figure 4B, arrow).
Finally, Spo7p‐GFP served as a marker to depict the nuclear envelope boundary in Nem1− cells (Figure 4C). This revealed that the nuclear envelope is no longer spherical, but exhibits long protrusions extending into the cytoplasm and frequently surrounding the vacuole. We also saw repeatedly that single cells contain two nuclear compartments connected by a thin membranous filament. However, no irregular DNA staining was noticed in Spo7− and Nem1− cells (Figure 5). In single cells which appeared to contain two nuclear compartments, the DNA, depicted by Hoechst 33258 staining, either in vivo (Figure 5A) or in fixed cells (Figure 5B), was round and compact, and always restricted to one nuclear compartment. Thus, nem1 and spo7 mutants are not bi‐ or multinucleate, which is consistent with the fact that the cell growth and, accordingly, also nuclear division is not inhibited in these mutants. In conclusion, the nuclear membrane organization is drastically altered in yeast mutants lacking SPO7 or NEM1. The nucleus is no longer round, but exhibits long extensions which contain NPCs and an intranuclear content, but not DNA.
To analyse the nuclear envelope morphology of the nem1::HIS3 and spo7::HIS3 mutants at an ultrastructural level, we performed thin section electron microscopy (Figure 6). We could often see in non‐dividing cells either a single nucleus with a long nuclear envelope protrusion (Figure 6E) or two nuclear compartments interconnected by a long and thin double nuclear membrane extension (Figure 6A, B, D and F). NPCs were seen distributed all over the nuclear membrane of these irregularly shaped nuclei (Figure 6A and B). Accumulation of additional membranes adjacent to the nuclear envelope and the plasma membrane could be also seen (Figure 6E). In conclusion, the ultrastructural analysis confirmed the data from the in vivo localization of the various GFP reporters and demonstrated that the nuclear morphology is severely impaired in the spo7 and nem1 mutants.
Purification of ProtA‐tagged Nem1p reveals association with Spo7p
Tagging of Nem1p with ProtA allowed its efficient one‐step affinity purification by IgG‐Sepharose chromatography from a whole‐cell lysate. As revealed by SDS‐PAGE and Coomassie staining, the purified Nem1p‐ProtA eluate contained two major bands (data not shown). The larger was identified as Nem1p‐ProtA (by Western blotting) and the lower band migrating at ∼30 kDa as Spo7p (by mass spectrometry analysis; see Materials and methods). To confirm this result, a nem1::HIS3 spo7::HIS3 strain co‐expressing Nem1p‐ProtA and a Myc epitope‐tagged version of Spo7p (Spo7p‐Myc), both expressed from low‐copy plasmids, was constructed (see Materials and methods). When Nem1p‐ProtA was affinity purified, Spo7p‐Myc co‐enriched in the eluate in a stoichiometric ratio, as seen by Coomassie staining and Western blotting using anti‐Myc antibodies (Figure 7A). Therefore, Nem1p and Spo7p form a complex in vivo. Next, we asked whether nucleoporins, in particular Nup84p, are present in the purified Nem1p‐Spo7p complex. However, we could not detect any Nup84p, Nup85p or Sec13p co‐purifying with Nem1p‐ProtA by Western blot analysis (data not shown). Thus, no stable physical interaction between the Nem1p‐Spo7p complex and the Nup84p complex could be shown. However, this does not exclude that both complexes transiently interact in vivo (see Discussion).
As mentioned above, Nem1p contains an N‐terminal domain with no homology to known proteins and a conserved C‐terminal domain related to several proteins from yeast to human. To find out which domain is involved in the interaction of Nem1p with Spo7p, we generated three Nem1p truncation mutants, one lacking the putative membrane‐spanning sequence (Nem1ΔHp; Ile87‐Ile126), one lacking the conserved C‐terminal domain (Nem1ΔCp; Lys250‐Asn446) and one lacking the N‐terminal domain (Nem1ΔNp; Phe3‐Phe236). All truncated proteins were tagged with protein A at their C‐termini and expressed under the control of the authentic NEM1 promoter from low copy plasmids (Figure 7B). Surprisingly, the deletion of the hydrophobic sequence did not impair the Nem1p function, because Nem1ΔHp was able to complement the sl235 mutant (data not shown). However, sl235 was not complemented by either Nem1ΔCp or Nem1ΔNp (data not shown). Next, we transformed the nem1::HIS3 spo7::HIS3 mutant, harbouring the Spo7p‐Myc fusion protein, with the different Nem1p‐ProtA truncation constructs. After affinity purification from whole‐cell extracts as described before, the corresponding eluates were analysed by Western blotting using anti‐Myc antibodies (Figure 7C). Interestingly, the deletion of the conserved Nem1p C‐terminal domain (Nem1ΔCp‐ProtA) completely abolished the interaction with Spo7p‐Myc, whereas the C‐domain alone (Nem1ΔNp‐ProtA) was still able to bind to Spo7p‐Myc (Figure 7C). In conclusion, our data show that Spo7p and Nem1p form a stable complex in yeast. The conserved C‐terminal domain of Nem1p alone is sufficient and necessary for the physical interaction with Spo7p.
Membrane association of Nem1p is mediated both by its hydrophobic sequence and interaction with Spo7p
The unexpected finding that the putative transmembrane sequence within Nem1p is dispensable for its function raised the question of whether this hydrophobic sequence is involved at all in membrane association. Therefore, we analysed in more detail the targeting of Nem1p to the membrane. Cells disrupted for NEM1 and expressing either full‐length Nem1p or one of the truncation constructs described above were converted to spheroplasts, lysed in detergent‐free buffer and the extracts were centrifuged at 74 000 g for 30 min (see Materials and methods). Samples from the supernatant and pellet were analysed by Western blotting using anti‐ProtA antibodies. Full‐length Nem1p was found exclusively in the crude membrane pellet, as expected. Surprisingly, all three Nem1p mutant constructs were also recovered in this fraction, suggesting that they are still membrane associated (Figure 8A, Spo7+).
This result indicated that, besides the hydrophobic sequence, the C‐terminal domain of Nem1p may also contribute to membrane association, probably by directly binding to Spo7p which is also a membrane protein. To analyse this possibility, the same extraction procedure was repeated as described above, but now in a spo7::HIS3‐disrupted strain expressing the various Nem1p deletion constructs (Figure 8A, Spo7−). Full‐length Nem1p and Nem1ΔCp were still found in the pellet. On the contrary, about half of Nem1ΔHp and almost all of the C‐terminal domain (Nem1ΔNp) became soluble. Finally, in contrast to the full‐length protein, the Nem1ΔHp mutant is completely extractable by alkaline treatment (Figure 8B). These data suggest that (i) the hydrophobic sequence within Nem1p mediates the integral insertion into the nuclear membrane and (ii) deletion of the putative membrane‐spanning sequence from Nem1p still allows membrane association via binding to Spo7p, but this interaction is only peripheral, i.e. sensitive to alkaline extraction. Finally, a Nem1ΔHp‐GFP fusion protein, expressed from a high copy‐number plasmid, was localized in nem1::HIS3 cells. A diffuse cytoplasmic staining was observed (Figure 8C, left panel), most probably due to the limiting amount of Spo7p which could act as its membrane receptor. Indeed, when Spo7p was co‐overexpressed in the same strain, a distinct nuclear membrane/ER staining was seen (Figure 8C, right panel). This shows that overproduced Nem1ΔHp‐GFP can be targeted to the nuclear membrane via its interaction with overproduced Spo7p.
Taken together, these data demonstrate the presence of two different domains that determine the membrane association of Nem1p; a hydrophobic sequence that mediates direct interaction with the lipid bilayer, and the conserved C‐terminal domain that binds to another membrane protein, Spo7p. This most probably also explains why the deletion of the hydrophobic domain is dispensable for the function of Nem1p.
A role for nuclear envelope and NPC components in sporulation
SPO7 was identified initially in a genetic screen for genes involved in sporulation in yeast (Esposito and Esposito, 1969). Since lack of Spo7p or Nem1p in cells causes severe nuclear envelope abnormalities during vegetative growth, we tested whether Nem1p is also required for sporulation. Diploid strains homozygous for the deletion of SPO7 or NEM1 were tested for their ability to sporulate in liquid sporulation medium (see Materials and methods). This revealed that not only Spo7p, but also Nem1p is cessential for sporulation (Figure 9). This sporulation defect was complemented by plasmid‐borne SPO7 and NEM1 genes, when transformed into the disrupted diploid strains (Figure 9). Inhibition of sporulation must occur quite early in these diploids since neither by DNA nor Pus1p‐GFP staining could we detect proper nuclear division during meiosis (Figure 9B). We also found a strict correlation between the requirement for the different Nem1p domains for sporulation and vegetative growth (Figure 9A): Nem1ΔHp is able to complement the sporulation defect of the nem1 diploid mutant, in contrast to the Nem1ΔCp and the Nem1ΔNp truncations.
The genetic interaction of SPO7 and NEM1 with NUP84 prompted us to test whether a diploid nup84‐disrupted strain is also defective in meiosis. Strikingly, Nup84p is essential for sporulation (Figure 9A). Furthermore, a sporulation defect was also found for the other members of the Nup84p complex. Diploids homozygous for nup85Δ deletion are unable to sporulate, and diploids homozygous for seh1 disruption sporulate with only a 10% efficiency as compared with a corresponding wild‐type strain (data not shown). Furthermore, diploid homozygous for nup133 deletion are unable to sporulate. In contrast, diploids homozygous for nup2 (data not shown) or nup188 deletion (Figure 9A) are not affected in sporulation. These results show that nucleoporins which affect NPC distribution or nuclear envelope morphology are required for sporulation in yeast (see also Discussion).
How mutations in certain nuclear pore proteins can affect the overall morphology of the nuclear membrane is unknown. We have identified genetic interactions between the nucleoporin Nup84p that is required for NPC distribution (Siniossoglou et al., 1996) and a complex of two integral nuclear membrane/ER proteins, Spo7p and Nem1p, which participate in shaping the yeast cell nucleus. Furthermore, our work also points to an as yet unrecognized link between nuclear membrane morphogenesis, nuclear pore distribution and sporulation.
Very little is known about why a nucleus is normally round‐shaped in eukaryotic cells and which factors control this morphology. It is assumed that the densely packed chromatin inside the nucleus is involved in nuclear morphogenesis. Furthermore, the nuclear lamina which forms a fibrous nucleoskeletal meshwork underlying the nuclear envelope in higher eukaryotic cells was suggested to provide structural support to the nucleus and to serve as a site of chromatin attachment (Gant and Wilson, 1997), thereby also participating in nuclear morphogenesis. In S.cerevisiae, a nuclear lamina, which could participate in attaching the chromatin to the nuclear membrane, does not exist. It is possible that in yeast the spherical organization of the nucleus is not only mediated by intranuclear contacts between the inner nuclear membrane and the chromatin (or chromatin‐attached proteins), but also by interactions between the outer nuclear membrane and an underlying cytoskeletal structure. Integral components of the nuclear membrane, such as Spo7p and Nem1p, or peripheral nuclear pore proteins could participate in such an interaction.
In the absence of Spo7p and Nem1p, the nucleus loses its spherical organization and long double nuclear membrane projections appear instead, which arise from a main DNA‐containing nuclear body and penetrate deeply into the cytoplasm. An intriguing model for the function of the Spo7p‐Nem1p complex would be that they negatively control nuclear membrane proliferation by inhibiting outgrowth of the nuclear membrane at distinct sites. Amplification or proliferation of the nuclear membrane can happen in yeast by overexpression of membrane proteins (Wright et al., 1988; Smith and Blobel, 1994) or by inhibiting very long chain fatty acid biosynthesis (Schneiter et al., 1996).
It is intriguing that despite Spo7p and Nem1p being organized in a complex, they exhibit a different pattern of genetic interactions with nucleoporin mutants. Furthermore, there is no synthetic lethality between the spo7 and nem1 null alleles. Thus, SPO7 and NEM1 seem to perform different functions in nuclear envelope biogenesis. Nevertheless, both of them are synthetically lethal with NUP84. An interesting possibility could be that the Spo7p‐Nem1p complex physically interacts with Nup84p. However, Spo7p‐GFP did not cluster in a nup133 mutant, suggesting that at least the major pool of the complex does not associate with the NPC. Furthermore, we could not detect biochemically any interaction between Spo7p‐Nem1p and any of the components of the Nup84p subcomplex. Therefore, if such an interaction occurs, it has to be very transient or involve only a minor pool of the Spo7p‐Nem1p complex.
The putative transmembrane domains in both Spo7p and Nem1p are two closely spaced hydrophobic stretches (see Figure 3A). Therefore, it is possible that the hydrophobic domain inserts as a hairpin‐like structure into the membrane, with a short luminal loop in the case of Spo7p. As a consequence, both the N‐ and C‐terminal domains of Spo7p and Nem1p should face the cytoplasm. Such an unusual topology has been described previously for VIP21‐caveolin (Monier et al., 1995). The possibility that the Spo7p‐Nem1p interaction takes place on the cytoplasmic and not in the luminal side of the nuclear membrane and ER is consistent with the observation that a truncated Nem1p lacking its membrane‐spanning domain can still interact with Spo7p on the nuclear membrane/ER and is functional.
Nem1p belongs to an evolutionarily conserved protein family. Furthermore, within a given organism (e.g. S.cerevisiae), Nem1p has several homologues which have in common a highly related C‐terminal domain of ∼200 amino acid residues. Accordingly, this domain could perform a common function or have a similar folding. Interestingly, overexpression of the different Nem1p yeast homologues (see Figure 1C) could not rescue the nuclear envelope defects observed in the nem1::HIS3 mutant (S.Siniossoglou, unpublished data). Furthermore, when the conserved C‐terminal domain of Nem1p was replaced by the homologous C‐terminal region of YLL010c, the resulting hybrid protein was not able to complement sl235 and the nem1 mutant phenotype (S.Siniossoglou, unpublished data). This raises the interesting possibility that different Nem1p homologues in an organism are involved in different aspects of membrane organization. In agreement with this speculation is the finding that the tagged GFP version of the yeast Nem1p homologue YLL010c is located at the plasma membrane, but not at the nuclear membrane (S.Siniossoglou, unpublished data). It is also worth mentioning in this context that two human homologues of Nem1p, OS4 and HYA22, recently were implicated in cancer development (Ishikawa et al., 1997; Su et al., 1997). One intriguing possibility thus could be that oncogenic cell progression is linked to aspects of altered membrane morphogenesis.
It is striking that nuclear division proceeds normally in SPO7‐ and NEM1‐deleted cells (no inhibition of cell growth at 30°C), despite the drastically altered nuclear envelope morphology. Although it is difficult to imagine how the irregularly shaped nuclei with their long extensions are segregated correctly between mother and daughter cells, there is another example in yeast which shows that this is possible. In the case of karmellae membranes, which are closely attached to the nuclear membrane, these extra membrane stacks remain in the mother cells and are not transmitted during mitosis, whereas the daughter cell inherits an unelaborated nucleus (Wright et al., 1988). A similar mechanism could be true for the nuclear envelope segregation in nem1 or spo7 mutants. The nuclear envelope protrusions may remain in the mother cell during mitotic division and only the main nuclear body containing the DNA segregates between the mother and daughter cell.
It was reported 24 years ago that a spo7‐1 homozygous diploid is unable to undergo pre‐meiotic DNA replication, one of the earliest events during meiotic division (Esposito and Esposito, 1974). In contrast to the majority of sporulation genes in S.cerevisiae, SPO7 and NEM1 are constitutively expressed (our data and Whyte et al., 1990) and perform an important function during vegetative growth. Therefore, it is likely that the same function performed during vegetative growth is also needed during sporulation. This is supported by the observation that the ability of the different nem1 mutants to rescue the sl235 mutant correlates with their ability to complement the sporulation defect in the nem1::HIS3 diploids. Our analysis also showed that some nucleoporins, including members of the Nup84p complex, are required for sporulation. It is a novel finding that nucleoporins participate in this differentiation process. However, not every structural nucleoporin is essential for sporulation. NUP188, which is linked to many NPC components, is dispensable for sporulation, although nup188 mutants are impaired in nuclear envelope organization (Zabel et al., 1996).
Why cannot sporulation proceed in some nucleoporin and nem1 or spo7 mutants? It may be that a spherical nuclear envelope organization and/or a random distribution of NPCs are a prerequisite for meiotic nuclear division. It is known that the complicated and highly coordinated meiotic division steps take place within the confines of a single nuclear envelope, adopting a very specific three‐dimensional configuration (Moens et al., 1974). This complex process probably cannot be achieved in nem1 and spo7 mutants which contain these nuclear membrane protrusions. Alternatively, there could be a check point control before entering meiosis, which measures the stage of the nuclear membrane organization including NPC distribution. Therefore, an abnormal nuclear membrane structure as seen in spo7 or nem1 mutants or clustered NPCs, as is the case of the nup84 mutant, may be sensed and signal to the sporulation pathway not to proceed with meiosis. Interesting in this context is that late sec mutants also block sporulation, but in this case it appears that they are required for spore wall formation (Neiman, 1998).
The fact that Spo7− or Nem1− cells are not viable during mitotic growth if they carry a disrupted NUP84 gene may be explained by a synergistic impairment in nuclear membrane organization and nuclear pore distribution. It is known that nup84::HIS3‐disrupted cells exhibit morphological defects in the nuclear membrane which resemble those described for Nup85− cells (e.g. NPC clustering and grape‐like NPC structures; Siniossoglou et al., 1996). It is conceivable that the combination of the different nuclear envelope/NPC abnormalities present in Spo7−/Nem1− cells (this work) and the nup84::HIS3 disruption mutant (Siniossoglou et al., 1996) are no longer tolerated by the cell and thus cause cellular death, e.g. during nuclear division or segregation of NPCs into the daughter cell nucleus.
In summary, through a genetic approach, we identified a novel complex of nuclear envelope/ER membrane proteins which participate in nuclear envelope biogenesis and organization. Genetic links point to an unexpected relationship between NPC distribution, nuclear envelope morphogenesis and sporulation. It will be our future aim to unravel the molecular mechanisms of these morphogenetic links.
Materials and methods
Yeast strains, microbiological techniques, plasmids and DNA manipulations
The yeast strains used in this work are listed in Table II. Standard DNA manipulations (restriction analysis, end filling, ligation, PCR amplification and DNA sequencing) were performed according to Maniatis et al. (1982). Microbiological techniques (growth and transformation of yeast and Escherichia coli strains, plasmid recovery, mating and tetrad analysis) were done as described previously (Sherman et al., 1986). The following plasmids were used: pUN100, ARS1/CEN4 plasmid with the LEU2 marker (Elledge and Davis, 1988); pRS314 and pRS316, ARS4/CEN6 plasmids with the TRP1 and URA3 markers, respectively (Sikorski and Hieter, 1989); YCplac111, ARS1/CEN4 vector with the LEU2 marker (Gietz and Sugino, 1988); YEplac181, 2μ vector with the LEU2 marker (Gietz and Sugino, 1988); YEplac122, 2μ vector with the TRP1 marker (Gietz and Sugino, 1988); pASZ11, ARS/CEN vector with the ADE2 marker (Stotz and Linder, 1990); YDp‐H, pUC9‐based plasmid with the HIS3 marker; pHT4467‐URA3‐ADE3 (Segref et al., 1997); pHT4467‐URA3‐ADE3‐NUP84, the NUP84 gene as a SmaI‐SalI fragment was inserted into the SmaI‐SalI polylinker site of pHT4467‐URA3‐ADE3; pASZ11‐ADE2‐NUP49‐GFP, a SacI‐BamHI fragment containing the NUP49‐GFP fusion (Belgareh and Doye, 1997) was inserted into the SacI‐BamHI site of pASZ11‐ADE2; pASZ11‐ADE2‐PUS1‐GFP (Hellmuth et al., 1998).
Synthetic lethal screen with a nup84::HIS3 mutant and cloning of SPO7 or NEM1
To perform a synthetic lethal screen with the nup84::HIS3 disruption allele, the red/white colony‐sectoring assay was employed (Kranz and Holm, 1990; Wimmer et al., 1992). A screening strain was constructed by mating the nup84::HIS3 deletion mutant (Siniossoglou et al., 1996) with CH1305 (Kranz and Holm, 1990), selecting after tetrad dissection the ade2 ade3 nup84::HIS3 haploid progeny and transforming it with the pHT4467‐URA3‐ADE3‐NUP84. UV mutagenesis was done as described earlier (Wimmer et al., 1992). Approximately 25 000 colonies were analysed for a red, non‐sectoring phenotype at 30°C on YPD plates containing 4% glucose. After three successive rounds of selection, nine mutagenized colonies that were synthetically lethal with nup84::HIS3 but not with NUP84 were isolated (sl 154, 235, 243, 252, 269, 273, 338, 343 and 384). Out of these, sl243 and sl384 were complemented by the NSP1 gene (Hurt, 1988) and sl273 by NUP116 (Wente et al., 1992; Wimmer et al., 1992). sl235 and sl269 were transformed with a yeast genomic library inserted in the pUN100 vector (Bergès et al., 1994). The complementing plasmid was recovered from transformants that regained both the red/white sectoring phenotype and growth on plates containing 5‐fluoro‐orotic acid (5‐FOA). Restriction analysis demonstrated that three transformants from sl235 contained overlapping genomic inserts that did not correspond to the NUP84 gene. DNA sequencing showed that the smallest insert contained a 9 kb fragment from chromosome VIII in the ORC10‐ERG11 intergenic region. From the four different ORFs found in this genomic region, YHR004c (now called NEM1) was finally shown to complement the sl235 mutant. Similar analysis demonstrated that two red/white sectoring transformants from sl269 contained a 12 kb insert from chromosome I in the MDM10‐VPS8 intergenic region and that the complementing activity resided in the SPO7 gene.
Gene disruption of SPO7 and NEM1
For the deletion of the SPO7 gene, a pRS314‐SPO7 construct containing a unique BamHI site before the stop codon of SPO7 (see later) was digested with ClaI‐BamHI, and the entire SPO7 ORF, except for the 43 first amino acids, was removed. A filled‐in HIS3 gene was then inserted, and the resulting spo7::HIS3 allele was excised as an ApaI‐EcoRI fragment and used to transform the diploid strain RS453. Another spo7 null strain was made by transforming a haploid RS453 strain with the LEU2 gene flanked at the 5′ end with 36 bp of sequence from bp −36 to +1 of SPO7 and at the 3′ end with 42 bp of the untranslated region (UTR) sequence. For the deletion of NEM1, a pRS316‐NEM1 construct containing a unique BamHI site before the stop codon of NEM1 (see later) was digested with BamHI‐EcoRI. This resulted in the removal of the DNA sequence coding for the entire NEM1 ORF, plus 80 nucleotides upstream of the ATG. Into these blunt‐ended sites, the HIS3 gene was inserted as a BamHI filled‐in fragment. The nem1::HIS3 allele was excised as an ApaI‐SacI DNA fragment and transformed into a diploid RS453 strain. In all cases, the heterozygous His+ or Leu+ transformants were analysed for correct integration at the SPO7 and NEM1 locus, respectively, by PCR. Correct integrants were sporulated on yeast extract‐peptone‐acetate (YPA) plates and tetrads were dissected on YPD.
Construction of SPO7 and NEM1 fusion genes and NEM1 truncation mutants
To epitope tag SPO7 and NEM1, a unique BamHI site was introduced just before the stop codon by PCR, resulting in the addition of six nucleotides (5′‐GGATCCTGAAAG‐3′ for SPO7 and 5′‐AACGGATCCTGACAA‐3′ for NEM1). A BamHI fragment encoding either two IgG‐binding domains from Staphylococcus aureus protein A (380 bp), the triple Myc tag (140 bp) or the S65T/V163A variant of the GFP (720 bp) was inserted in‐frame into the BamHI site generated at the stop codon of SPO7 or NEM1. Most fusion proteins were expressed under the control of the SPO7 or NEM1 promoter from centromeric vectors (YCplac111‐LEU2‐SPO7‐ProtA, pRS314‐TRP1‐SPO7‐Myc, pRS314‐TRP1‐SPO7‐GFP, pUN100‐LEU2‐NEM1‐ProtA, pUN100‐LEU2‐NEM1‐GFP) or 2μ vectors (YEplac122‐TRP1‐SPO7‐GFP). All these fusion proteins were functional since they could complement the growth defect of spo7::HIS3 or nem1::HIS3 cells at 37°C and rescue the synthetic lethal phenotype of the sl269 or sl235 mutant, respectively.
The deletion of the hydrophobic sequence from Nem1p (Ile87‐Ile126) was done as follows: a 0.85 kb fragment coding for the promoter and the 86 N‐terminal residues of Nem1p followed by an XbaI restriction site has been generated by PCR. A second 1.6 kb PCR fragment was generated, starting with an XbaI site at residue 127 and extending to residue 446 followed by the protein A tag and the 3′ UTR sequence of NEM1. The two PCR fragments were digested by SphI‐XbaI and XbaI‐KpnI, respectively, and ligated into a SphI‐KpnI‐digested YCplac111‐LEU2 vector. The sequence of the Nem1ΔHp‐ProtA fusion in the area of the deletion is S(86)‐SR‐E(127). Due to the generation of an XbaI site, the two residues SR were introduced into the sequence. For the construction of Nem1ΔHp‐GFP, the protein A tag was replaced by a BamHI GFP fragment and the resulting fusion was subcloned into the high copy (2μ) vector YEplac181‐LEU2 as a KpnI‐SphI fragment. For the deletion of the conserved C‐terminal domain of Nem1p, two PCR fragments, one coding for the promoter and sequence of Nem1p until Q(249) followed by a BamHI site, and a second one coding for a BamHI site followed by the NEM1 stop codon and its 3′ UTR, were digested with SphI‐BamHI and BamHI‐SacI, respectively, and ligated into a SphI‐SacI‐digested pUN100‐LEU2‐NEM1‐ProtA constuct. The resulting construct coded for a ProtA‐tagged Nem1p fusion protein lacking the 197 C‐terminal residues (K250‐N446). To construct an NEM1 mutant expressing only the conserved C‐terminal domain, two PCR fragments, one coding for the promoter and the two first residues of Nem1p followed by a 3′ SacI site and a second PCR coding for a 5′ SacI site followed by residues P(237) to N(446) of Nem1p and the ProtA tag, were digested by SphI‐SacI and SacI‐KpnI, respectively, and ligated into an SphI‐KpnI‐digested YCplac111‐LEU2 vector. The sequence of this Nem1ΔNp‐ProtA construct in the area of the deletion was M(1)‐N(2)‐EL‐P(237) (the SacI site resulted in the generation of the two additional residues EL).
For the construction of the N‐terminally tagged Nem1p‐GFP fusion, a 240 bp fragment of NEM1, starting after its ATG codon and having a 5′ PstI site introduced by PCR, was digested with PstI‐HindIII and ligated with a HindIII‐KpnI fragment, containing the rest of the NEM1 ORF and its 3′ UTR, into a PstI‐KpnI‐digested ARS/CEN vector with the ADE2 marker harbouring the PNOP1‐GFP cassette. The resulting Nem1p‐GFP was expressed under the control of the NOP1 promoter and could complement the lethal phenotype of the sl235 mutant.
Affinity purification and extractions
Affinity purification of Nem1p‐ProtA was performed as described in Siniossoglou et al. (1998). To determine whether Spo7p and Nem1p are membrane proteins, spheroplasts were prepared from early log phase cultures of strains Spo7p‐ProtA or Nem1p‐ProtA (Table II). One gram of spheroplasts was lysed with a Dounce homogenizer in 16 ml of 150 mM NaCl, 20 mM Tris‐HCl, pH 8.0, and centrifuged at 15 000 r.p.m. for 10 min at 4°C with a Beckman JA‐25.50 rotor. The insoluble pellet was resuspended in 16 ml of either buffer alone (150 mM NaCl, 20 mM Tris, pH 8.0) or buffer containing 1% Triton X‐100 or 1 M NaCl. Alkaline extraction was performed resuspending the membrane pellet in 0.1 M sodium carbonate (pH 11.5). Extracts were incubated for 30 min on ice and then separated into a soluble and pellet fraction by ultracentrifugation at 407 500 g for 60 min at 4°C (SW60 rotor, Beckman). Equivalent amounts from both fractions were dissolved in SDS sample buffer, heated at 50°C for 3 min, analysed by SDS‐PAGE, blotted onto nitrocellulose and probed with antibodies as described below. All extractions were performed at 4°C in the presence of a protease inhibitor cocktail. To determine the membrane association of the different Nem1p truncation mutants, full‐length Nem1p‐ProtA or Nem1ΔHp‐ProtA, Nem1ΔCp‐ProtA and Nem1ΔNp‐ProtA were expressed either in a nem1::HIS3 or spo7::HIS3 strain. Spheroplasts from the corresponding mutants were prepared as described above and lysed with a Dounce homogenizer in 150 mM KCl, 20 mM Tris‐HCl pH 8.0, 5 mM MgCl2. Extracts were incubated on ice for 30 min and then centrifuged at 74 000 g for 20 min with a JA‐25.50 rotor at 4°C. Equivalent amounts from the pellet and the supernatant were analysed by Western blotting. The antibodies were used at the following dilutions: anti‐ProtA, 1:4000; anti‐Myc mAb 1:400; anti‐Sec61p, 1:2000; anti‐Nup85p, 1:500; anti‐hexokinase 1:1000.
Mass spectrometric analysis
Spo7p was identified using techniques and strategies of analysis as previously described (Shevchenko et al., 1996). The protein band was ‘in‐gel’ reduced, S‐alkylated, and degraded using trypsin. A small aliquot of the supernatant was deposited together with 10% formic acid on a microcrystalline matrix layer and analysed by MALDI‐time of flight mass spectrometry. For peptide sequencing, the remaining material was extracted on a microcolumn, concentrated and desalted, eluted into a nanoelectrospray needle (Wilm and Mann, 1996) and analysed in a triple quadrupole mass spectrometer (API III, Perkin‐Elmer Sciex, Toronto, Canada). All database searches using peptide mass maps and peptide sequence tags were performed with the PeptideSearch program in a non‐redundant database currently containing >300 000 entries. Six tryptic peptides were obtained from Spo7p (sequence coverage 20.8%). Additional MS/MS experiments with the unseparated peptide mixtures on a nanoelectrospray triple quadrupole mass spectrometer revealed the sequence of two peptides belonging to SPO7: LVLNPR and NLLILEDDLR.
Localization of GFP fusion proteins and electron microscopy
The in vivo location of Spo7p‐GFP, Nem1p‐GFP, Pus1p‐GFP, Nup49p‐GFP and Seh1p‐GFP was analysed in strains expressing the corresponding GFP fusion proteins plus the ADE2 gene (plasmid pASZ11‐ADE2) according to Segref et al. (1997). The cells were examined in the fluorescein channel of a Zeiss Axioskop fluorescence microscope. Pictures were taken with a Xillix Microimager CCD camera. Digital pictures were processed by modules of the software Openlab (Improvision, Coventry, UK). Thin‐section electron microscopy was done as described in Doye et al. (1994). Where indicated, instead of osmium tetroxide, 3% potassium permanganate was used to stain for intracellular membrane structures.
SDS‐PAGE and Western blot analysis were performed according to Maniatis et al. (1982). Diploid yeast strains were sporulated in liquid YPA medium at 30°C for 3 days. For DNA staining of cells expressing Nup49p‐GFP, cells were washed with water and stained with Hoechst 332358 (Fluka) for 3 min. For DNA staining of cells expressing Pus1p‐GFP, cells were fixed and spheroplasted as described in Wimmer et al. (1992) and stained with Hoechst 332358. For DNA staining of wild‐type, Spo7− and Nem1− diploids during sporulation, samples were fixed with ethanol for 10 min, washed with phosphate‐buffered saline, spheroplasted with cytohelicase and stained with Hoechst 332358 for 3 min.
The excellent help of Andrea Hellwig from the institute of Dr W.Huttner (University of Heidelberg, Neurobiology) in electron microscopy is acknowledged. We also thank Dr Gunnar von Heijne (Stockholm, Sweden) for help with sequence analysis of Spo7p and Nem1p, and Dr R.Schekman for the anti‐Sec61p antibody. The critical reading of the manuscript by Professor F.Wieland, Drs V.Doye and G.Simos is gratefully acknowledged. J.R. acknowledges support from the Fonds der Chemischen Industrie (FCI). E.C.H. was the recipient of a grant from the Deutsche Forschungsgemeinschaft (SFB352).
- Copyright © 1998 European Molecular Biology Organization