Factors promoting nuclear envelope assembly independent of the canonical ESCRT pathway

In eukaryotes, the nuclear envelope (NE) separates the nucleoplasm and cytoplasm. During cell division, the NE undergoes complex dynamics, necessitating mechanisms for controlled remodeling. In metazoans, the NE is disassembled, partially or completely, at the beginning of mitosis and reassembled after the completion of chromosome segregation. NE remodeling also occurs during interphase nuclear pore complex insertion in all eukaryotes (Grossman et al., 2012; Ungricht and Kutay, 2017) and for spindle pole body (SPB) insertion and/or removal in yeast (Cavanaugh and Jaspersen, 2017; Ding et al., 1997). Finally, pathological “rupture” of the NE, which can then undergo membrane repair, occurs in a variety of circumstances (Halfmann et al., 2019; Hatch and Hetzer, 2014; Shah et al., 2017). Failure to establish or maintain NE integrity compromises genome integrity, as occurs during aging, cellular senescence, cancer, or cell migration through confined spaces (Bakhoum et al., 2018; Spektor et al., 2017; Umbreit and Pellman, 2017; Ungricht and Kutay, 2017). Understanding the mechanisms underlying NE dynamics thus has broad implications for cell and organismal physiology.

Recent studies have suggested that the endosomal sorting complex required for transport (ESCRT) machinery plays a major role in many of the above described aspects of NE dynamics (Frost et al., 2012; Radulovic et al., 2018). ESCRT is best known for mediating scission of membranes that bud away from the cytoplasm (Christ et al., 2017; McCullough et al., 2018; Schöneberg et al., 2017; Scourfield and Martin-Serrano, 2017). Using process-specific cues, the ESCRT-III proteins are recruited to different sites of action and assemble into polymers. The polymers are then remodeled by the ATPases Associated with diverse cellular Activities (AAA ATPase) Vps4 to promote membrane scission. In several cellular processes mediated by ESCRT, ESCRT-III proteins are recruited both by the ESCRT-I/II complexes and/or the Bro1-domain containing protein Alix (Henne et al., 2011; Schöneberg et al., 2017). At the NE, the only known adaptor protein for ESCRT-III recruitment is CHMP7 (Chm7 in budding yeast or Cmp7 in fission yeast; Vietri et al., 2015; Webster et al., 2016; Gu et al., 2017; Pieper et al., 2020), which contains both an ESCRT-II–like domain and an ESCRT-III–like domain (Bauer et al., 2015). CHMP7 itself can be recruited to the NE by a direct physical interaction with the inner nuclear membrane protein LEM2 (Heh1 in budding yeast and Lem2 in fission yeasts; Appen et al., 2019; Pieper et al., 2020). This interaction is normally prevented by spatial separation of CHMP7 and LEM2 because of active export of CHMP7 from the nucleus (Thaller et al., 2019; Vietri et al., 2019). Disruption of the NE during mitosis or pathological NE rupture enables LEM2 and CHMP7 to form a complex that activates ESCRT-III polymerization, providing one simple regulatory mechanism to control ESCRT activity at the NE. Intriguingly, near-complete RNAi-mediated depletion of LEM2, CHMP7, or ESCRT-III proteins (Olmos et al., 2015; Vietri et al., 2015) resulted in only a partial delay in reestablishing NE integrity after mitosis or rupture. This finding could be explained by incomplete RNAi depletion or by the existence of additional mechanisms that regulate NE dynamics. Supporting the latter possibility, NE repair was also only mildly affected by CHMP7 knockdown, but was abolished when three different LEM (Lap2-Emerin-Man1)-domain proteins were depleted (Halfmann et al., 2019).

Here we use the fission yeast Schizosaccharomyces japonicus to study mechanisms for NE assembly after mitosis. S. japonicus is an unusual yeast in that it undergoes a semi-open mitosis, where the NE is broken during anaphase with complete but transient loss of NE integrity (Aoki et al., 2011; Yam et al., 2011). After mitosis, the S. japonicus NE is reassembled (hereafter “sealing”), a process that requires membrane fusion, as occurs in metazoan cells. In S. japonicus, we find an important but nonessential role for Cmp7 and ESCRT-III in NE sealing, which enabled the design of a genetic screen to identify alternative mechanisms. This screen has identified factors that bypass the canonical ESCRT-mediated pathway for NE sealing.

The following findings suggested that the canonical LEM2-CHMP7-ESCRT-III pathway for postmitotic NE reassembly is operative in S. japonicus. First, Cmp7 and ESCRT-III localize to putative NE sealing sites. In postmitotic S. japonicus cells, NE remodeling is required at the spindle pole after SPB removal and at the mitotic NE “tail” located in the regions of the daughter NE facing the mid-spindle (Ding et al., 1997; Yam et al., 2011). Lem2 and its interacting partner, Nur1, are enriched at both the SPB and the mitotic NE tail (Pieper et al., 2020). Using Nur1 as a marker for NE sealing sites, we found that immediately after anaphase, both Cmp7 and Vps32, the key subunit of the ESCRT-III polymer (Schöneberg et al., 2017; Teis et al., 2008; Wollert et al., 2009), localized to the SPB and the mitotic NE tail (Fig. 1 A). Second, consistent with findings in other organisms (Gu et al., 2017; Webster et al., 2016), the localization of Cmp7 to potential NE sealing sites depends on Lem2 (Fig. 1 B). Interestingly, S. japonicus Cmp7 often formed foci in close proximity to the SPB in interphase, a stage when the NE is intact, and this localization also depends on Lem2 (Fig. 1, A and B).

Cmp7 was also required to establish normal nuclear morphology and NE integrity after mitosis, as evidenced by compromised accumulation of an import reporter (GST-GFP-NLS-GFP, referred to as GFP-NLS; Yam et al., 2011) in cmp7Δ cells (Fig. 1, C and D). This defect occurred preferentially following postmitotic NE reassembly because premitotic (mononucleated) cmp7Δ cells that established NE integrity from a previous cell cycle generally maintained it throughout interphase (Fig. 1 E). NE rupture was never observed when WT cells were imaged for a comparable time interval. Moreover, vps32Δ cells exhibited a similar magnitude NE integrity defect as cmp7Δ cells, evident from the nuclear/cytoplasmic ratio of GFP-NLS, despite a much stronger growth defect, determined by colony size after ascus dissection (Fig. 1, F and G). Consistent with another recent report (Pieper et al., 2020), these data indicate that Cmp7 is the main adaptor for the ESCRT-dependent NE sealing in S. japonicus, as observed in other cell types.

We next asked if S. japonicus cells might exhibit a stronger requirement for Cmp7 in stressed conditions. We focused on cold temperature, because of its profound effect on membrane fluidity and its potential to alter membrane and protein dynamics (Ernst et al., 2016). Indeed, cmp7Δ cells displayed an aggravated NE integrity defect and failed to form colonies on plates at 19°C (Fig. S1, A and B). cmp7Δ cells were also supersensitive to Leptomycin B (LMB; Fig. S1 C), which compromises nucleocytoplasmic transport by inhibiting nuclear export (Kudo et al., 1999; Nishi et al., 1994). LMB sensitivity can be explained by additive effects from defective nucleocytoplasmic transport with partial compromise of NE integrity–like ineffective bailing of a leaky boat.

To characterize NE defects at higher resolution, we performed serial section electron microscopy to compare nuclei from WT and cmp7Δ cells grown at 19°C (Fig. 1, H and I). Duplication and insertion of the SPBs in cmp7Δ cells was similar to that of WT strains (Fig. S1 D), indicating that Cmp7 is not required for the membrane fusion events that accompany SPB insertion and bipolar spindle formation. However, we frequently observed defects in nuclear shape and structure in cmp7Δ cells. The most prominent defect was NE rupture of post-mitotic cmp7Δ mutant cells that was never observed in WT cells (Fig. 1, H and I; and Fig. S1, E and F). This severe defect of S. japonicus cmp7Δ strains contrasts with the near-normal phenotype of cmp7Δ Schizosaccharomyces pombe strains (Gu et al., 2017). This likely is explained by the increased need for ESCRT-III to seal the NE in an organism that undergoes mitotic NE disruption. These NE abnormalities have similarities to the defects previously reported for S. pombe vps4Δ strains, which include disorganized membranes and some large NE ruptures (Gu et al., 2017). However, S. japonicus cmp7Δ strains did not exhibit organized membrane stacks or “karmellae.”

Surprisingly, despite the above described defects, cmp7Δ cells were viable at 30°C, with 76.9% cells showing nuclear accumulation of GFP-NLS (Figs. 1 F and S2 B). Therefore, although our data indicate an important role for Cmp7/ESCRT-III in the postmitotic NE sealing in S. japonicus, they also indicate the existence of one or more Cmp7 and/or ESCRT-III independent NE sealing mechanism(s) under normal growth conditions.

We designed a genetic screen to identify gene products that function independently of Cmp7 to promote NE sealing. We reasoned that we could most specifically identify parallel pathways by isolating mutations that would bypass a cmp7 deletion, enhancing the establishment and maintenance of NE integrity in the complete absence of Cmp7. In principle, this approach could identify gain-of-function mutations in genes that promote Cmp7-independent NE sealing, loss-of-function mutations in negative regulators, or gene products that become toxic in the absence of Cmp7.

A three-tiered screen was performed to identify cmp7Δ bypass suppressors (Fig. 2 A). A cmp7Δ strain was mutagenized, and fast-growing colonies at 19°C were isolated (Fig. 2 B). After this initial selection, mutant strains were prioritized based on LMB resistance (Fig. 2 C). Finally, we directly assayed the degree of restoration of NE integrity in these mutant strains using fluorescence reporters engineered into the parental strain (Fig. 2 D). We then employed backcrosses, pooled segregant analysis (Birkeland et al., 2010), and whole genome sequencing to identify putative causal suppressor mutations (Table S1). We identified 13 different mutated genes from 33 suppressor strains with a single putative causal mutation. These genes encoded NE proteins as well as proteins not previously implicated in NE remodeling. We introduced several of the mutations into the cmp7Δ strain as a definitive test of whether the mutations were responsible for suppression of the defect of cmp7Δ cells. Indeed, these mutations were sufficient to enhance the growth of cmp7Δ cells at 19°C, albeit with some variable penetrance (Fig. 2 E). Together, these findings validate our approach to identify genes involved in NE sealing after mitosis. Analysis of the most frequently isolated and/or strong phenotype mutations is described below.

In all organisms examined to date, the localization of CHMP7/Chm7 to sites of NE sealing depends on LEM2/Heh1 (Gu et al., 2017; Webster et al., 2016; Fig. 1 B). It was therefore unexpected that we obtained 11 suppressor strains that carried two different nonsense mutations in lem2, both terminating the ORF at the third amino acid. Moreover, we obtained three independent isolates of a missense mutation in Nur1 (Table S1 and Fig. 3 A). Because the lem2 nonsense mutations we identified were likely to be null alleles, we determined if lem2Δ could suppress cmp7Δ. Indeed, lem2Δ suppressed the growth defect of cmp7Δ cells. Strikingly, the deletion of lem2 also rescued the strong growth defect of vps32Δ strains. Deletion of nur1 also suppressed the growth defects of either cmp7Δ or vps32Δ strains, albeit less robustly than lem2Δ (Fig. 3 B). These findings are difficult to reconcile with the simplest linear model, and suggest additional effects of Lem2 on the maintenance of NE integrity.

One interpretation of these genetic findings is that in the absence of Cmp7, Lem2, and Nur1 become altered in a way that is deleterious to cells. Accordingly, the growth defect of cmp7Δ cells was enhanced by a single extra copy of lem2 expressed from its endogenous promoter, but not by an extra copy of lem2-m29, the loss-of-function allele (Fig. 3 C). Interestingly, although an extra copy of WT nur1 did not significantly affect growth, an extra copy of nur1-m78 resulted in some suppression of the growth defects (Fig. 3 C), indicating that nur1-m78 is at least a partially dominant mutation. This can explain why nur1Δ cmp7Δ strains exhibited only mild suppression of the cmp7Δ growth defect (Fig. 3 B), whereas nur1-m78 cmp7Δ strains showed strong suppression (Fig. 2 E).

Next, we addressed the mechanism for the toxicity of Lem2 in strains lacking Cmp7. We noted that in cmp7Δ strains, Lem2 localization is significantly altered. In WT cells, Lem2 localizes to the SPB, NE, and the mitotic tail (Yam et al., 2011). Strikingly, in most cmp7Δ cells, Lem2 forms one or more aggregates on the NE (Fig. 3 D). As the Lem2–Nur1 complex is known to tether chromatin to the NE (Banday et al., 2016; Barrales et al., 2016; Mekhail et al., 2008), it was recently proposed that unresolved chromatin-NE tethering by these aggregates might generate toxicity because of impaired chromosome segregation (Champion et al., 2019; Chen, 2019; Pieper et al., 2020). We performed a series of experiments to test this idea. First, we examined chromosome segregation in cmp7Δ cells by assaying the fluorescence intensity of the kinetochore cluster in anaphase cells (Gregan et al., 2007; Joglekar et al., 2008). We observed an approximately fivefold increase in chromosome segregation errors in cmp7Δ cells. However, inconsistent with the model that Lem2 aggregates mediate chromosome mis-segregation, cmp7Δ lem2Δ double mutant strains showed an increase, not the predicted decrease, in chromosome mis-segregation relative to cmp7Δ single mutants (Fig. 3 E). Additionally, chromosome segregation defects in S. japonicus do not produce strong cold sensitivity: strains lacking a monopolin complex component (mde4Δ) had a more severe chromosome segregation defect than cmp7Δ strains (Choi et al., 2009; Gregan et al., 2007), but mde4Δ strains did not exhibit a severe growth defect at 30°C and were only modestly growth-impaired at 19°C (Figs. 3 F and S1 G). Additionally, we did not observe a synergistic growth phenotype in cmp7Δ mad2Δ or vps32Δ mad2Δ double mutant strains, lacking ESCRT proteins and the spindle assembly checkpoint protein, Mad2 (Fig. 3 G). Finally, cmp7Δ strains were not significantly sensitive to thiobendazole (TBZ), which induces chromosome mis-segregation due to impaired microtubule function (Fig. 3 F). In summary, our data suggest that, although Lem2 adversely affects the growth of ESCRT-deficient strains, this growth defect is not primarily due to compromised chromosome segregation.

Alternatively, the “toxicity” of Lem2 in cmp7Δ strains could be due to effects on NE integrity. Consistent with results from our initial screen (Fig. 2 D), the NE integrity of lem2Δ cmp7Δ and lem2Δ vps32Δ double mutant strains was significantly improved relative to cmp7Δ or vps32Δ single mutants. This was apparent by measuring the nuclear/cytoplasmic ratio of GFP-NLS (Fig. 3 H) as well as the time interval for the accumulation of the GFP-NLS reporter in daughter nuclei after mitosis (Fig. 3 I). Thus, our data reveal a surprising inhibitory role for Lem2/Nur1 in postmitotic NE sealing in ESCRT-III mutants, which occurs, at least in part, by interfering with NE integrity. Lem2 aggregates could compromise NE integrity directly or indirectly, by inhibiting ESCRT-independent NE sealing.

The genetic interaction between lem2Δ and cmp7Δ that we observe in S. japonicus contrasts with the genetic interactions reported in S. pombe. In S. pombe, loss-of-function mutations in lem2 or cmp7 behave similarly, suppressing the pleiotropic defects of cells lacking Vps4, the ATPase that disassembles ESCRT-III filaments (Gu et al., 2017). Consistent with the LEM2–CHMP7–ESCRT-III linear recruitment model (Thaller et al., 2019; Vietri et al., 2019; Webster et al., 2016), this finding was most simply interpreted by Lem2 generating toxicity in an ESCRT-III– and Cmp7-dependent manner. By contrast, in S. japonicus, we found that cmp7Δ and lem2Δ have different phenotypes, displaying an alleviating genetic interaction. Our finding, therefore, identifies an ESCRT-III–independent effect of Lem2 in S. japonicus.

One mutant, m131, results in a Glu to Val substitution at position 308 in the SJAG_03891 ORF, which encodes one of the two Bro1-domain–containing proteins in S. japonicus (Fig. 4 A). The alx1-m131 mutation is adjacent to the concave ESCRT-III–binding surface, based on the structure of human and budding yeast Bro1 domains (McCullough et al., 2018; Fig. 4 A). Bro1-domain–containing proteins have not previously been implicated in postmitotic NE sealing. We named SJAG_03891 alx1 and investigated its localization and role in NE integrity.

Both Alx1 and Alx1-m131 localized to SPB and mitotic tails in dividing cells (Fig. 4 B) and are present in cells at similar steady-state levels (Fig. 4 C). Normal localization and protein levels raised the possibility that alx1-m131 might be a dominant gain-of-function mutation, a category of suppressors expected to be identified in our screen. Indeed, we found that alx1-m131 suppressed the growth defect of the cmp7Δ strain even in the presence of WT alx1 (Fig. 4 D). We determined if alx1-m131 suppressed the defect of cmp7Δ cells by preventing the formation of the toxic Lem2/Nur1 aggregates. However, the formation of the Lem2/Nur1 aggregates was unaffected in alx1-m131 cmp7Δ cells (Fig. 4 B).

Next, we characterized the growth and NE integrity of a cmp7Δ lem2Δ alx1-m131 strain to that of a control cmp7Δ lem2Δ strain. If alx1-m131 suppression of the cmp7Δ defect is explained by disruption of the Lem2 aggregates, it is expected that alx1-m131 cmp7Δ lem2Δ and cmp7Δ lem2Δ strains would be indistinguishable, as Lem2 toxic aggregation will not occur in either strain. Alternatively, if alx1-m131 were to promote NE sealing independent of the Lem2 aggregates, alx1-m131 should promote growth and NE integrity whether or not Lem2 is present in cells. Indeed, we found that alx1-m131 cmp7Δ lem2Δ strains showed improved growth and NE integrity relative to cmp7Δ lem2Δ strains (Fig. 4, E and F). Together, these data indicate that alx1-m131 is a dominant gain-of-function mutation that promotes NE integrity independent of Cmp7 and independent of toxic Lem2 aggregates.

The alx1-m131 mutation could enhance a normal function of Alx1 in maintaining NE integrity, or it could repurpose Alx1 from another ESCRT-regulated process (e.g., a neomorphic mutation). To distinguish these models, we characterized the phenotype of cells lacking Alx1. alx1Δ strains had no measurable growth defect but did exhibited a subtle NE-sealing defect. However, cmp7Δ alx1Δ strains did exhibit aggravated growth and an aggravated NE integrity defect (Fig. 4, F and G; and Fig. S2 P). Therefore, Alx1 contributes to growth and NE integrity independently of Cmp7, a function that is enhanced by the alx1-m131 mutation.

We next asked if alx1-m131 promoted NE sealing by recruiting ESCRT-III to the NE. In cmp7Δ cells, both Alx1 and Alx1-m131 colocalized with some of the Nur1 aggregates (Fig. 4 B), but Vps32 was not detected in these foci in either strain (Fig. 4 H). This negative result is not conclusive because it could be due to limited detection sensitivity, given that the Vps32 signal at NE sealing sites is weak (Fig. 1 A). Nevertheless, an alternative interpretation is that alx1-m131, at least in part, suppresses the NE integrity defects of cmp7Δ cells via an ESCRT-independent mechanism. The possibility the Bro1-domain proteins could have ESCRT-independent effects has been raised in prior studies of these proteins in the context of virus budding (Bardens et al., 2011; Popov et al., 2009). To assess this possibility, we compared the growth of alx1Δ vps32Δ, alx1-m131 vps32Δ, and vps32Δ strains. Surprisingly, alx1Δ vps32Δ strains exhibited a stronger growth defect than vps32Δ (Fig. 4 I), whereas alx1-m131 improved the growth of vps32Δ cells. These data indicate an ESCRT-independent function of alx1-m131. Although the genetic data are most consistent with this function being related to NE sealing, absent more information on the underlying mechanism, we do not exclude effects related to other ESCRT functions.

In addition to dedicated membrane fusion machineries, such as ESCRT, membrane fusion can be enhanced by altering the biophysical properties of the lipid bilayer (Chernomordik and Kozlov, 2003; Kinugasa et al., 2019; Penfield et al., 2019). Interestingly, our screen identified several mutations in ole1, the gene encoding the only Δ9 fatty acid desaturase in S. japonicus (Stukey et al., 1989; Makarova et al., 2020). Like its homologues (Tatzer et al., 2002), Ole1 localizes to the ER membrane (Fig. 5 A). Unsaturated acyl chains are poorly packed within lipid bilayers, which is predicted to increase membrane fluidity (Harayama and Riezman, 2018) in a manner that may promote NE sealing. To determine if increased fatty acid desaturase activity would promote NE sealing in the absence of Cmp7, we overexpressed ole1 from the medium-strength Purg1 promoter (Watt et al., 2008). Indeed, Purg1-ole1 partially suppressed the growth defect of cmp7Δ strains (Fig. 5 B). Moreover, overexpression of Ole1 also resulted in a subtle but statistically significant rescue of the NE sealing defect in binucleated cmp7Δ cells (Fig. 5 C). Together, these results further support an important contribution of lipid composition for NE sealing in the absence of Cmp7, as was suggested for S. pombe cells lacking Lem2 (Kinugasa et al., 2019). Optimal membrane lipid composition might promote the function of membrane fusion proteins or possibly facilitate the spontaneous fusion of tethered membranes.

Here, we have taken a genetic and cell biological approach to identify factors required for postmitotic NE sealing. S. japonicus is an ideal system to address this problem, because it undergoes partial mitotic NE breakdown but retains facile genetics. Our results highlight a key role of ESCRT in NE membrane sealing, but also establish that NE sealing can occur in the absence of ESCRT-III because strains lacking Cmp7 or Vps32 are viable with only partial defects in NE integrity.

In principle, ESCRT-independent NE membrane fusion might be promoted by another dedicated membrane fusion machinery, or more speculatively, by membrane tethering that enables spontaneous membrane fusion (Chernomordik and Kozlov, 2003). Both classes of mechanism would be affected by membrane lipid composition, a conclusion that is supported by the finds on ole1 described in this study. Previous work in Xenopus egg extracts has suggested a requirement for ER SNAREs for NE membrane assembly (Baur et al., 2007; Wang et al., 2013), which could provide a backup mechanism for ESCRT. However, these studies primarily relied on the addition of blocking antibodies or dominant negative constructs, and evidence for a SNARE requirement for NE fusion is currently lacking in other systems. Moreover, we note that our screen did not identify SNARE proteins or their regulators. An alternative to unidentified membrane fusion proteins mediating ESCRT-independent NE sealing, proteins that enrich near the NE sealing site (e.g., Lem2, Nur1, Alx1) could promote membrane tethering to facilitate spontaneous membrane fusion. Although Lem2 was recently shown to undergo a liquid–liquid demixing phase transition that might enhance membrane tethering (Appen et al., 2019), we found that deletion of Lem2 improves rather than compromises membrane fusion of ESCRT mutants, indicating that Lem2 is not the major ESCRT-III–independent sealing factor. Instead, our results implicate Alx1, and possibly proteins sequestered in Lem2 aggregates, as candidate components of a Cmp7/ESCRT-III–independent NE sealing pathway.

In summary, we present the results of the first unbiased genetic approach to identify factors required for NE sealing. Because common mechanisms may be used for many aspects of NE dynamics, our results may be relevant to other processes such as NE repair after NE disruption (Radulovic et al., 2018).

Transformation of S. japonicus cells was performed as described (Aoki et al., 2010), with 2–10 µg purified DNA fragments per transformation. All transformants were verified by PCR at both the 5′ and 3′ insertion junctions. Suppressor mutants created in a WT background were verified by sequencing.

Table S2 lists the S. japonicus strains used in this study. DNA fragments for gene targeting and gene deletion were designed as previously described (Bähler et al., 1998) with the exception that 575-bp homologous sequences flanking the designated insertion sites were used for homologous recombination. These homologous sequences were either synthesized as double-stranded gene blocks or PCR-amplified from WT genomic DNA. Fusion to sequences encoding fluorescent proteins and/or selection markers was done by PCR. To introduce extra gene copies, the relevant ORF and 1-kb flanking regulatory sequences were cloned into the plasmid pSJK1 (Aoki et al., 2010) after the kan gene using Gibson Assembly. The gene cassette and the kan gene were then introduced at the ura4 locus. To delete vps32, a gene that is important for cell growth, kan-Pvps32-vps32-Tvps32 was integrated into the ura4 locus before the endogenous vps32 was deleted. This strain was then crossed to a WT strain to obtain vps32Δ cells without kan-Pvps32-vps32-Tvps32. To introduce the suppressor mutations nur1-m78 and alx1-m131 to their endogenous locus in a WT background, the endogenous nur1 and alx1 ORFs were replaced with Sp-ura4⁺ (the ura4⁺ gene from S. pombe; Aoki et al., 2010). A second round of transformation was then to replace Sp-ura4⁺ with the mutant nur1-m78 or alx1-m131 sequence. 1.5 g/liter 5-fluoroorotic acid in yeast extract with five supplements (YE5S) medium was used for counterselection (Furuya and Niki, 2009).

Cells are grown in YE5S unless otherwise noted. To assay cell growth at different temperatures or with different chemicals, cell cultures in log-phase growth were diluted to the same OD₆₀₀ (0.2–0.4) and serially diluted (fivefold increments) in 96-well plates. To test for sensitivity to microtubule depolymerization, TBZ (Sigma-Aldrich) was added to a final concentration of 10 µl/ml. In a pilot experiment, we found that WT S. japonicus was sensitive to LMB at 20 ng/ml in yeast peptone dextrose (YPD). 3 ng/ml LMB was used as a secondary screen for mutations affecting nuclear integrity based on preliminary experiments showing that this concentration of LMB inhibits the growth of cmp7Δ strains but not the WT controls.

For crosses, cells of the opposite mating type were mixed on sporulation agar and incubated at 30°C for 12–16 h. Cells were then dissected on a YE5S plate using a yeast dissection microscope (Nikon Instruments, Inc., and Micro Video Instruments, Inc.). Asci were incubated at 32°C for 6–8 h before spores were separated onto different spots. Dissected spores were incubated at 30°C for 3–4 d before further experiments or analysis. A complete S. japonicus ascus contains eight spores (Furuya and Niki, 2009). However, a mature ascus that had released one or two spores was almost indistinguishable from a complete ascus under a yeast dissection microscope, making it infeasible to only analyze asci with eight spores. Only asci that had more than six spores after dissection were analyzed in this study.

To analyze colony size after asci dissection, plates containing colonies germinated from spores were scanned, and unbiased colony size analysis was performed using the “Analyze Particles” function in ImageJ/FIJI. Table S2 lists all strains used in the crosses. For each cross, colonies from 14–35 asci were analyzed. Colonies were then restreaked, and their genotypes were identified by replica-plating and/or PCR. For each genotype, colony sizes were normalized to the average size of WT colonies on the same plate. Because of the prominent growth defect of vps32Δ cells, we used PY7579 (vps32Δ::natMX6 ura4-D3::kan-Pvps32-vps32-Tvps32) in crosses to determine the genetic interactions between vps32Δ and deletions of other genes. Only the sizes of colonies that did not have the extra copy of vps32 at the ura4 locus were plotted. Colonies carrying alx1-m131 were identified by Sanger sequencing after ascus dissection.

Cells for microscopy were collected from liquid culture in log-phase growth at 28–30°C unless otherwise noted. After centrifugation, cell pellets were washed into the Edinburgh Minimal Medium (EMM) and imaged in 35-mm glass-bottom dishes (MatTek Corporation). For time-lapse videos, cells were imaged in YE5S. For visualization of Cmp7-Halo (Los et al., 2008), the fluorescent dye JF549 (Promega; Grimm et al., 2015) was added to a final concentration of 500 nM 2 h before imaging. Images were collected using a 100×/1.4 NA lens on a Nikon Ti2 Microscope equipped with a Hamamatsu Flash4.0 V2+ sCMOS camera and Lumencor SpectraX fluorescence light source, or a 100×/1.4 NA lens on a DeltaVision microscope (Applied Precision Inc.) equipped with a Photometrics CoolSnap HQ camera. Acquisition software was either NIS-Elements or DeltaVision softWoRx. A z-focal plane series was acquired at 0.2–0.4 µm intervals spanning 8 µm. Maximum-intensity projections of five 0.4-µm z-slices or ten 0.2-µm z-slices from the middle of the indicated cells are shown unless otherwise noted. Image analysis and preparation were performed using ImageJ/FIJI. Figures were assembled using Adobe Illustrator.

The percent binucleated cells with nuclear GFP-NLS in both daughter nuclei was determined qualitatively. For Fig. 5 C, because of the effect size, the samples were scored double-blind. To determine chromosome mis-segregation rate, strains expressing Mis12-mCherry (kinetochore) GFP-Atb2 (tubulin) were imaged. The fluorescent intensity of the Mis12 cluster in each z-slice was measured and corrected for cellular background as previously described (Wu et al., 2008). The sum intensity was then calculated for each cluster. Chromosome mis-segregation was defined by having a cluster with a relative fluorescent intensity that was lower than 0.376, which was three SDs from the average in WT cells. To minimize the contribution from photobleaching during z-stack acquisition, only cells with two clusters close to the same focal plane (separated by <2.0 µm in the z-axis) were analyzed.

Single z-slices at the middle of the cells in the field were used to measure the nuclear/cytoplasmic ratio of GFP-NLS (Webster et al., 2016). For each cell, the nuclear and cytoplasmic intensity of GFP-NLS in a region of interest with an area of ∼1 µm² were measured. The ratio was calculated after background subtraction. 507–553 cells were measured for each strain.

WT and cmp7Δ cells were grown in YE5S media in a 19°C shaking water bath for 72 h. Cells were back-diluted every 24 h to maintain an OD₆₀₀ below 1.0 and harvested during logarithmic phase growth by vacuum filtration onto 0.45 µm Millipore filters. The cells were then frozen in a Leica EM-ICE high-pressure freezer at ∼2,050 bar, then transferred to the Leica EM AFS2 for freeze substitution under liquid nitrogen into 2% osmium tetroxide/0.2% uranyl acetate/5% water/acetone. We used the following procedure for freeze substitution: −90°C to −80°C over 60 h, −80°C to −60°C over 5 h, −60°C for 4 h, −60°C to −20°C over 8 h, held at −20°C for 4 h, and −20°C to 0°C over 5 h, and finally samples were held at 0°C for 3 h. Samples were then removed from the AFS and brought to room temperature. Samples went through four changes of acetone over 1 h and were removed from the planchettes. They were embedded in acetone/Spurr’s mixtures to a final concentration of 100% Spurr’s over several days in a stepwise procedure as previously described (McDonald, 1999). 60–80-nm serial thin sections were cut on a Leica UC6 ultramicrotome, stained with 4% uranyl acetate and Sato’s lead, and imaged on a Zeiss Merlin scanning electron microscope equipped with a scanning transmission electron microscope detector and Atlas software at a pixel size of 20 nm. Additional higher resolution SEM images were acquired at a pixel size of 6 nm, and images of select SPBs and nuclei were acquired on a FEI Tecnai Spirit transmission electron microscope. Original data underlying this manuscript can be accessed from the Stowers Original Data Repository at http://www.stowers.org/research/publications/libpb-1516.

The protocol for mutagenesis of S. japonicus was modified from a protocol for S. pombe (Ekwall and Thon, 2017). 10⁸ PY7509 (cmp7Δ::kanMX6 GFP-NLS-ura4⁺ Nup85-mCherry-ura4⁺) cells from a log-phase YE5S culture were washed into EMM and incubated with 2% ethyl methanesulfonate for 2 h at room temperature, giving ∼50% survival. The culture was then washed into 5% sterile sodium thiosulfate to quench the activity of ethyl methanesulfonate, and then washed three more times in EMM. Mutagenized cells were then plated onto YE5S plates with a density of 10⁵ cells per plate, and then incubated at 19°C for 9 d. 470 large colonies were selected, restreaked for single colonies, and replica plated onto YPD plates with 3 ng/ml LMB to test their growth at 30°C. The top 50 strains that showed significantly improved growth in the presence of LMB compared with the parent strain were selected for further experiments, including imaging to determine the percent of binucleated cells with nuclear GFP-NLS.

Backcrosses to a WT strain (h⁺ GFP-NLS-ura4⁺ nup85-mCherry-ura4⁺) were used to identify candidates where suppression resulted from mutation of a single genetic locus. 47 out of 50 crosses gave rise to cmp7Δ colonies that grew at 19°C.

To identify putative causal mutations, we performed pooled-segregant analysis followed by whole genome sequencing (Birkeland et al., 2010). After backcrossing, cmp7Δ spores with improved growth at 19°C were categorized as suppressed (S), and cmp7Δ progenies that were cold sensitive were categorized as non-suppressed (N). 8–12 colonies, each from a different ascus, were selected for each category to ensure statistical power (Abe et al., 2012; Iida et al., 2014). To achieve near-equal representation of each colony, spores were grown separately in 96-well deep-well plates for 12–16 h at 30°C before being combined equally based on OD₆₀₀ measurements. Genomic DNA was extracted from each pool using MasterPure yeast DNA extraction kit (Illumina) with RNase treatment at 37°C overnight. Any residual small nucleotides were removed by AMPure XP beads (Beckman Coulter). The quality of genomic DNA was examined using Genomic DNA Screen Tape on a TapeStation 2200 (Agilent Technologies). Sequencing libraries were made using a plexWell kit (seqWell, Inc.), and the quality of the libraries was determined by Bioanalyzer (Agilent Technologies).

A total of 96 libraries (47 pairs of S and N pools, plus WT and cmp7Δ parental cells) were sequenced using NextSeq (paired end 75 × 2), producing on average 10 million reads per sample. Sequencing reads were aligned to the S. japonicus reference genome (GCA_000149845.2, release 35) using bwa (https://github.com/lh3/bwa, version bwa-0.7.13) in the paired-end mode by ‘‘bwa mem.’’ For all libraries, 94% of the reference genome was covered by at least one read. Before variant discovery was performed, duplicated sequencing fragments were removed by “MarkDuplicates” using the PICARD software suite, and realigning was performed around genomic regions of indels using The Genome Analysis Toolkit (v3.7-0-gcfedb67). Variants for each of the 96 samples were discovered using “HaplotypeCaller” method in Genomic Variant Call Format mode and merged using “GenotypeGVCFs.” Variants were annotated using SnpSift version 4.3s (Cingolani et al., 2012) using the annotation of S. japonicus (downloaded on October 25, 2017 using snpEff download method). All variants with at least one supporting read were collected and summarized in a table for each pair of S and N strains. SNPs that had 100% penetrance in the S pool (at least 14 reads) and 0% in the N pool of the same mutant (at least 3 reads) were identified as putative causal mutations. Out of 47 suppressors sequenced using the above method, 33 of them contained one putative causal mutation. Several mutations including lem2-m29 and nur1-m78 were isolated multiple times from the screen. Given that we identified different background SNPs in different isolates with the same putative causal mutation, these mutations almost certainly arose multiple times, independently, rather than being the consequence of cell proliferation during mutagenesis.

Cells with an OD₆₀₀ below 1.0 were harvested and snap frozen for protein analysis. For each strain, 1 g of cell pellet was pulverized with a mortar and pestle in liquid nitrogen. The frozen powder was resuspended in lysis buffer (10 mM Hepes, pH 7.4, 200 mM NaCl, 3 mM MgCl₂, 300 mM sucrose, 1 mM EDTA, 1% Triton X-100, 1 mM DTT, and freshly added PMSF and complete protease inhibitor tablet [Roche]). After centrifugation at 2,000 g for 10 min, the lysate was incubated with mNeonGreen trap magnetic beads (Chromotek) at 4°C for 1 h. After a 3× wash, bound proteins were released with sample buffer. The samples were then loaded for SDS-PAGE and Western blotting. Anti-mNeonGreen antibody (32F6, Chromotek) was used at 1:1,000 dilution. Anti-mouse IgG (Sigma-Aldrich) was used at 1:20,000 dilution. Blots were reacted with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and imaged by a ChemiDoc Imager (BioRad). Lysates were stained with Coomassie blue to assess input protein concentration.

Plots were generated using GraphPad Prism. Comparisons of colony size were made using Brown–Forsythe and Welch ANOVA to account for different variances of strains, followed by Tukey’s multiple comparison test. Other comparisons were made using one-way ANOVA followed by Tukey’s multiple comparison test.

Fig. S1 shows the accentuation of the nuclear envelope integrity defect of cmp7Δ cells at 19°C. Fig. S2 shows representative images of various mutants expressing GFP-NLS Nup85-mCherry cultured at 30°C. Table S1 lists bypass suppressors of cmp7Δ identified in this study. Table S2 lists S. japonicus strains used in this study.

The authors would like to thank S. Oliferenko (The Francis Crick Institute, London, UK), J.-Q. Wu (The Ohio State University, Columbus, OH), and JapoNet (National Institute of Genetics, Shizuoka, Japan) for reagents, and the Nikon Imaging Center at Harvard Medical School, Center for Cancer Computational Biology at Dana-Farber Cancer Institute, and Melainia McClain in the Electron Microscopy Core at the Stowers Institute for technical support. We thank L.-L. Du, D. Moazed, S. Oliferenko, T. Rapoport, N. Rhind, and members of the Pellman laboratory for discussion.

I-J. Lee was supported by Leukemia and Lymphoma Society Postdoctoral Fellowship 5436-16. The Jaspersen laboratory is supported by the Stowers Institute for Medical Research and by R01GM121443. D. Pellman is a Howard Hughes Medical Institute investigator, and this work was supported by 5R37 GM061345-14.

The authors declare no competing financial interests.

Author contributions: I-J. Lee and D. Pellman conceptualized and designed the experiments. I-J. Lee, E. Stokasimov, and N. Dempsey performed the experiments. I-J. Lee and E. Stokasimov carried out the bypass suppressor screen. I-J. Lee and N. Dempsey analyzed microscopy data. I-J. Lee performed crosses, dissections, and colony size analysis. J.M. Varberg and S.L. Jaspersen performed and analyzed EM. E. Jacob analyzed whole-genome sequencing data. I-J. Lee and D. Pellman wrote the manuscript with input from all co-authors.