Misassembled nuclear pore complexes (NPCs) are removed by sealing off the surrounding nuclear envelope (NE), which is conducted by the endosomal sorting complexes required for transport (ESCRT) machinery. Recruitment of ESCRT proteins to the NE is mediated by the interaction between the ESCRT member Chm7 and the inner nuclear membrane protein Heh1, which belongs to the conserved LEM family. Increased ESCRT recruitment results in excessive membrane scission at damage sites but its regulation remains poorly understood. Here, we show that Hub1-mediated alternative splicing of HEH1 pre-mRNA, resulting in production of its shorter form Heh1-S, is critical for the integrity of the NE in Saccharomyces cerevisiae. ESCRT-III mutants lacking Hub1 or Heh1-S display severe growth defects and accumulate improperly assembled NPCs. This depends on the interaction of Chm7 with the conserved MSC domain, which is only present in the longer variant Heh1-L. Heh1 variants assemble into heterodimers, and we demonstrate that a unique splice segment in Heh1-S suppresses growth defects associated with the uncontrolled interaction between Heh1-L and Chm7. Together, our findings reveal that Hub1-mediated splicing generates Heh1-S to regulate ESCRT recruitment to the NE.
In eukaryotic cells, the double lipid bilayer membrane forming the nuclear envelope (NE) physically separates the nucleoplasm from the cytoplasm. Translocation of macromolecules in and out of the nucleus is mediated by nuclear pore complexes (NPCs), which are composed of multiple copies of ∼30 different protein components called nucleoporins (Alber et al., 2007; Cronshaw et al., 2002; D'Angelo and Hetzer, 2008; Rout et al., 2000; Strambio-De-Castillia et al., 2010). Notably, disruption of the precise structure of NPCs has been linked to various diseases, including cancer (Nofrini et al., 2016; Sakuma and D'Angelo, 2017).
The endosomal sorting complexes required for transport (ESCRT) components have been previously shown to be involved in NPC assembly and turnover (Lee et al., 2020a; Toyama et al., 2019; Webster et al., 2014, 2016). The ESCRT machinery comprises five evolutionarily conserved complexes, ESCRT-0, I, II, III and the AAA ATPase Vps4. Together, they orchestrate membrane remodeling and scission in diverse cellular processes including receptor sorting, virus budding, cytokinesis and plasma membrane wound repair (Alonso Y Adell et al., 2016; Hurley, 2015; Vietri et al., 2020a). In addition, the ESCRT machinery has been implicated in nuclear functions. For instance, the ESCRT machinery coordinates resealing of NE fragments in humans and fission yeast after cell division (Lee et al., 2020b; Olmos et al., 2015; Pieper et al., 2020; Vietri et al., 2015). Moreover, yeast mutants defective in either ESCRT-III or Vps4 accumulate misassembled NPCs at specific NE areas coined the ‘storage of improperly assembled nuclear pore complexes compartment’ (SINC; Webster et al., 2014). While ESCRT-dependent surveillance is crucial for maintaining nuclear compartmentalization, uncontrolled recruitment to the NE can be detrimental for cell viability (Gu et al., 2017; Thaller et al., 2019; Willan et al., 2019), suggesting that this process needs to be tightly regulated. However, how excessive ESCRT recruitment to NE is prevented remains unknown.
The ESCRT machinery is recruited to the NE through interaction of the ESCRT-III adaptor CHMP7 (Chm7 in budding yeast) with the inner nuclear envelope protein LEM2 (also known as LEMD2 in mammals) (Gu et al., 2017; Halfmann et al., 2019; Lee et al., 2020b; Pieper et al., 2020; Thaller et al., 2019; von Appen et al., 2020; Webster et al., 2016). LEM2 and its paralog MAN1 (also known as LEMD3) belong to the conserved LEM (for Lap2, emerin, MAN1)-domain protein family. They comprise a LEM and a winged-helix MSC domain (for MAN1-Src1 C-terminal) at their N- and C-terminus, respectively, which face the nucleoplasm and are separated by two transmembrane domains (TM; Brachner and Foisner, 2011). While the LEM domain has been the focus of many studies, several important functions have been recently reported for the MSC domain. For instance, LEM2 anchors telomeres and induces transcriptional silencing through its MSC domain; it further recruits CHMP7 to seal NE holes (Barrales et al., 2016; Gu et al., 2017; Halfmann et al., 2019; Hirano et al., 2018; Pieper et al., 2020; Thaller et al., 2019; von Appen et al., 2020). In budding yeast, the LEM2 homolog Heh1 (also known as Src1) also interacts with Chm7 (Thaller et al., 2019; Webster et al., 2016). Interestingly, HEH1 is a rare example of alternative splicing in Saccharomyces cerevisiae, resulting in two variants – a long form of 834 amino acids (aa), denoted Heh1-L, and a short form of 687 aa (Heh1-S; Fig. S1A). Alternative splicing causes a shift in the open reading frame of Heh1-S, resulting in the appearance of 49 unique residues that replace the second TM and the C-terminal MSC domain (Grund et al., 2008; Rodríguez-Navarro et al., 2002; Fig. S1A). This splicing reaction is assisted by Hub1 (also known as UBL5), which belongs to the family of ubiquitin-like proteins but does not form conjugates (Karaduman et al., 2017; Lüders et al., 2003; Mishra et al., 2011; Wilkinson et al., 2004; Yashiroda and Tanaka, 2004). Cells lacking Hub1 express only Heh1-L (Karaduman et al., 2017; Mishra et al., 2011). While Heh1 has been described to contribute to NE integrity, DNA tethering and chromatin silencing (Chan et al., 2011; Grund et al., 2008; Mekhail et al., 2008; Webster et al., 2016), the physiological relevance of Hub1-mediated HEH1 splicing remains largely unexplored.
Here, we examine the link between Hub1-mediated splicing and NE integrity. We demonstrate that Heh1-S is critical for normal cell growth when ESCRT-III subunits are absent. Lack of Heh1-S causes a severe growth phenotype due to a toxic gain-of-function of Heh1-L, which results from increased interaction with Chm7. Moreover, we find that Heh1-S and Heh1-L form heterodimers, and that the unique splice segment in Heh1-S is critical for the suppression of cellular toxicity. We propose that Hub1-dependent alternative splicing modulates the binding of Chm7 to Heh1-L, thus avoiding excessive recruitment of Chm7 to the NE.
ESCRT-III deficiency is toxic in cells lacking Hub1
When overexpressed, Hub1 enhances splicing efficiency and activates several cryptic introns (Karaduman et al., 2017). In particular, Hub1 overexpression induces mis-splicing of Did4 (also known as Vps2), which together with Snf7 (also known as Vps32), Vps20 and Vps24 are the core components of ESCRT-III (Babst et al., 2002; Karaduman et al., 2017). Mutations in Hub1 further exacerbate the growth defects associated with the deletion of the ESCRT-III subunit VPS24 (Costanzo et al., 2016). Moreover, we identified several members of the ESCRT-III complex in a genetic screen for mutants displaying synthetic growth defects in combination with hub1Δ (our unpublished results). Together, these findings suggest a functional link between Hub1 and the ESCRT machinery. To further examine this relationship, we crossed a hub1Δ strain with a mutant lacking the ESCRT-III subunit Did4. Consistent with previous studies, we found that lack of Hub1 alone does not affect growth at 37°C. However, additional deletion of HUB1 enhances the temperature sensitivity of the did4Δ mutant (Fig. 1A). Similarly, we observed a synthetic growth phenotype for cells lacking Hub1 and Snf7. In contrast, no synthetic genetic interaction was seen for hub1Δ in combination with a deletion of the ESCRT-II component VPS25 (Fig. 1B). These data suggest that Hub1 becomes essential when the ESCRT-III but not ESCRT-II complex is absent.
Heh1-S is essential for normal growth of ESCRT-III mutants
While Hub1 promotes the generation of the shorter spliced variant Heh1-S, its absence causes nearly exclusive expression of the long form Heh1-L (Karaduman et al., 2017; Mishra et al., 2011; Fig. S1A–C). Elevated temperature is also known to impair splicing (Cuenca-Bono et al., 2011; Hossain et al., 2011; Kawashima et al., 2014; Meyer et al., 2011), and consistent with this, we observe that Heh1-S is less efficiently produced at higher temperatures even in wild-type (WT) cells (Fig. S1B,C). To discern whether the synthetic growth defect of hub1Δ did4Δ cells is due to defective HEH1 splicing, we generated yeast strains that exclusively expressed either Heh1-L or Heh1-S from the endogenous locus (HEH1-L and HEH1-S strains, respectively; Fig. 1C). Both intron-less variants display similar expression levels and NE localization, and do not affect nuclear morphology (Fig. 1C; Fig. S1D). Using a functional assay for NPC integrity (Yewdell et al., 2011), we further confirmed that both alleles fully complement the heh1Δ phenotype (Fig. S1E,F). These results suggest that both spliced versions are functional and act redundantly in proper NPC assembly.
Next, we used the HEH1-L and HEH1-S strains to study whether concomitant deletion of HEH1-S affects growth in strains lacking ESCRT proteins. Similar to the hub1Δ mutant (Fig. 1A,B), the absence of Heh1-S produced a severe growth defect in ESCRT-III mutants (Fig. 1D,E; Fig. S2A,B). Interestingly, whereas the absence of Heh1-L caused no detectable phenotype at 30°C, the growth defect seen for the single ESCRT-III mutants at 37°C was partially suppressed in the corresponding double mutants (Fig. 1D,E; Fig. S2A,B). Conversely, combining HEH1-L with mutants lacking ESCRT-0, ESCRT-I, ESCRT-II or ESCRT accessory genes did not aggravate the growth defect (Fig. S2), suggesting that this genetic interaction is specific for ESCRT-III.
Deletion of ESCRT-III components results in clustering of NPCs at the SINC due to disruption of the pathway that removes defective NPC assembly intermediates (Webster et al., 2014, 2016). To assess whether lack of Heh1-S affects NPC clustering, we analyzed SINC formation in did4Δ and HEH1-L did4Δ strains using a C-terminal GFP fusion of the nucleoporin Nup192 (Nup192GFP). In agreement with previous studies (Webster et al., 2014, 2016), we found a significant increase of Nup192 accumulation at the SINC in the ESCRT-III mutant compared to WT and HEH1-L cells (10% versus 3–4%; Fig. 1F). However, Nup192 accumulation was further increased in did4Δ cells expressing only Heh1-L (more than 30%; Fig. 1F). We also observed that Nup192 dots mislocalized to the cytoplasm in some HEH1-L did4Δ cells (Fig. 1F, arrowheads), as reported for mutants affecting NPC assembly (Zhang et al., 2018). Together, these results imply that the exclusive presence of Heh1-L contributes to the accumulation of NPCs at the SINC in cells deficient in the ESCRT-III machinery.
Phenotypes associated with Heh1-S deficiency depend on the MSC domain of Heh1-L
Heh1-L contains two conserved domains exposed to the nucleoplasm – the LEM and the MSC domain (Fig. 1C, left panel). In Schizosaccharomyces pombe, these domains perform distinct functions for the Heh1-L homolog, Lem2 (Barrales et al., 2016). We sought to investigate which domain of Heh1 is responsible for the synthetic growth defect of ESCRT-III mutants expressing only Heh1-L. To this end, we generated multiple truncated versions of Heh1 lacking individual nucleoplasmic or transmembrane domains (ΔLEM, ΔTM1, ΔTM2 or ΔMSC) or combinations thereof (ΔTM1-TM2 and ΔTM2-MSC) and expressed them as plasmid-borne N-terminal GFP fusions under the control of the inducible GAL1 promoter in strains lacking endogenous HEH1 (Fig. 2A). Most of the truncated versions were expressed at similar levels as Heh1-L and localized to the NE, with the exception of those lacking the first TM domain, which displayed reduced expression and had a mainly nucleoplasmic localization (Fig. S3A,B).
While heh1Δ did4Δ cells display normal growth, overexpression of full-length HEH1-L causes a severe growth defect. Notably, deleting the MSC domain nearly completely rescued this growth defect (Fig. 2B). A comparable outcome was found for constructs in which Heh1-L is not membrane-bound (ΔTM1 and ΔTM1-TM2) or the MSC domain might not be properly exposed (ΔTM2; Fig. 2B). In contrast, lack of the LEM domain did not suppress the growth defect (Fig. 2B). Similar results were obtained for heh1Δ did4Δ and heh1Δ did2Δ cells grown at 37°C, although, under this condition, the suppression of the Heh1-L-mediated phenotype was partially weaker (Fig. S3C,D). In S. pombe, a region of Lem2 adjacent to the first TM mediates the interaction with another nuclear membrane protein, Bqt4 (Hirano et al., 2018; Hu et al., 2019). However, expression of mutants lacking the equivalent region in Heh1-L or other N-terminal regions affected the growth of heh1Δ did4Δ cells to the same extent or even stronger as the full-length protein (Fig. S3E). These findings suggest that the toxicity of Heh1-L does not require N-terminal domains. Recently, it was shown that cytosolic exposure of the MSC domain of Heh1-L is sufficient to recruit the ESCRT machinery to the NE (Thaller et al., 2019). We therefore assessed the phenotype of two distinct Heh1-L mutants that lack a functional nuclear localization signal (NLS), resulting in the cytoplasmic exposure of the MSC domain (ΔNLS and R176A; Lokareddy et al., 2015; Fig. 2A). Notably, these mutants displayed an even more severe growth defect than WT Heh1-L when expressed in heh1Δ did4Δ cells (Fig. 2C; Fig. S3F). Together, these findings imply that the MSC domain is responsible for the slow growth phenotype of ESCRT-III mutants expressing only Heh1-L.
LEM-containing proteins are highly conserved from yeast to humans (Brachner and Foisner, 2011). To identify conserved residues within the MSC domain that are potentially critical for its function, we performed sequence alignment of proteins from different species. In silico analysis of this region revealed several conserved residues (Fig. 3A), which we mutated into alanine (A) to assess their functional relevance. Using immunoblotting and live-cell imaging, we verified the proper expression and localization of these MSC mutants (Fig. 3B; Fig. S4A,B). Most of them displayed a growth phenotype similar to WT Heh1-L when expressed in heh1Δ did4Δ cells. However, specific tryptophan (W) to alanine mutants were found to partially suppress the phenotype of Heh1-L, especially when combined with each other (e.g. W786A and W817A; Fig. 3C; Fig. S4C). Interestingly, these conserved tryptophan residues form a hydrophobic surface in the MSC of human MAN1 but are not present in other winged-helix domains (Caputo et al., 2006). From these data, we conclude that the presence of a functional MSC domain of Heh1-L triggers the growth defect observed in ESCRT-III mutant cells.
The growth defect of cells lacking Heh1-S and ESCRT-III is caused by Chm7
Heh1-L recruits the ESCRT machinery to the NE through interaction with the chimeric ESCRT-II/III protein Chm7 (Bauer et al., 2015; Gu et al., 2017; Thaller et al., 2019; Webster et al., 2016). Chm7 exists in an inactive or active (open state) conformation due to the presence of auto-inhibitory helices (Lata et al., 2008; Webster et al., 2016). Recruitment to the NE likely requires Chm7 being present in the open state (Thaller et al., 2019). In other species, recruitment of Chm7 homologs further requires the presence of the MSC domain of Lem2, which is the homolog of Heh1-L (Gu et al., 2017; Pieper et al., 2020; Thaller et al., 2019; von Appen et al., 2020). We therefore assessed whether the constitutively active version of Chm7 lacking the inhibitory helices (chm7OPEN) is critical for binding to the MSC domain of Heh1-L. Using yeast two-hybrid (Y2H) and co-immunoprecipitation (co-IP) assays, we found that the MSC domain of Heh1-L interacts with chm7OPEN but not full-length Chm7 (Fig. 4A,B; Fig. S5A). Moreover, as expected, Heh1-S that lacks the MSC domain showed no binding to Chm7 (Fig. 4B; Fig. S5A). To test whether the residues critical for the toxicity of Heh1-L when overexpressed in ESCRT-III mutants are also important for binding to Chm7, we examined the association of Heh1-L and chm7OPEN in those mutants. Indeed, W786A and W817A mutations diminished the binding of Chm7 to the MSC of Heh1-L (Fig. 4C,D). We further found that a mutation in which the MSC is not properly exposed (ΔTM2) completely abolished the interaction (Fig. 4D). These results demonstrate that the MSC domain preferentially binds to chm7OPEN and that this interaction involves the residues W786 and W817 of Heh1-L.
Since Chm7 is required for SINC formation and interacts with Heh1-L (Thaller et al., 2019; Webster et al., 2016), we speculated that Chm7 might directly contribute to the growth phenotype of ESCRT-III mutants lacking Heh1-S. To test this hypothesis, we examined the growth of cells expressing exclusively Heh1-L in ESCRT-III mutants (either did4Δ or did2Δ) in presence and absence of Chm7. Remarkably, the severe growth phenotype of Heh1-L overexpression was nearly fully suppressed when Chm7 was absent (Fig. 4E; Fig. S5B–D). Consistent with our findings, overexpression of Heh1-S did not cause any growth defect (Fig. 4E; Fig. S5B–D). Finally, we tested whether the synthetic interaction of HUB1 and DID4 also depends on CHM7. Indeed, deleting CHM7 in cells lacking Hub1 and Did4 completely rescued the growth phenotype (Fig. 4F). Altogether, these data suggest that uncontrolled association between Chm7 and Heh1-L is detrimental in cells lacking the ESCRT-III complex.
Heh1-S impairs the interaction between Chm7 and Heh1-L
Encouraged by our finding that Heh1-L binds specifically to the open conformation of Chm7, we wondered whether exclusive expression of Heh1-L in chm7OPEN cells results in a growth defect analogous to mutants lacking ESCRT-III members (Fig. 1D,E). While we observed no defect under normal conditions (Fig. 5A, left panel), we found that the presence of chm7OPEN and Heh1-L, but not Heh1-S, caused significant sensitivity towards DMSO (Fig. 5A, right panel), which has been reported to affect the structure and properties of biological membranes (Gurtovenko and Anwar, 2007; Notman et al., 2006). Conversely, deletion of CHM7 improved growth in the presence of DMSO (Fig. 5A). Intriguingly, the severe growth phenotype of HEH1-L chm7OPEN cells upon DMSO treatment was rescued when plasmid-borne HEH1-S was concomitantly expressed from its endogenous promoter (Fig. 5B).
We next explored the mechanism by which Heh1-S rescues the growth defect of HEH1-L cells expressing chm7OPEN or lacking ESCRT-III members. By performing co-IP experiments in the presence or absence of Heh1-S, we found that co-expression of Heh1-S reduces the interaction between Chm7 and Heh1-L (Fig. 5C, compare lanes 2 and 3). It has previously been reported that Chm7 is recruited to the NE by Heh1-L and forms a distinct focus (Webster et al., 2016). Using live-cell imaging, we analyzed whether the presence of Heh1-S might affect Chm7 accumulation at the NE. In agreement with our biochemical data, we found that co-expression of Heh1-S reduced the number of Chm7 foci at the NE from ∼20% to ∼10% in HEH1-L cells (Fig. 5D,E). Moreover, many of these Chm7 foci colocalized with enrichment of Heh1-L (dots) but were largely decreased in cells expressing Heh1-S (Fig. 5D,E). Together, these findings imply that Heh1-S interferes with the binding of Chm7 to Heh1-L and prevents the toxicity of constitutively active Chm7, especially under conditions that affect membrane organization.
A unique sequence element of Heh1-S is critical to control the interaction of Heh1-L and Chm7
In higher eukaryotes, LEM-containing proteins have been shown to oligomerize (Berk et al., 2014; Mansharamani and Wilson, 2005). We therefore examined whether Heh1-S can bind to Heh1-L, which might explain its ability to prevent the interaction of Heh1-L with Chm7. By expressing different epitope-tagged versions of Heh1-L and Heh1-S, we found that both of them can form homo- and hetero-dimers (Fig. S6A). Using Y2H experiments, we mapped the interaction between Heh1-L and Heh1-S to individual domains (Fig. 6A,B). The heterodimerization interface comprises a region of the lumen domain (LD) shared between both forms (residues 476 to 638; Fig. 6B; Fig. S6B), which was further confirmed by co-IP experiments (Fig. 6C). Together, these findings demonstrate that Heh1-S and Heh1-L interact through their lumen domain.
We next explored whether Heh1-S directly competes with Chm7 for Heh1-L association; however, we observed no difference in Heh1-L and Heh1-S heterodimer formation in the presence of chm7OPEN (Fig. 5C; compare co-immunoprecipitated Heh1-S6HA in lane 3 and 4). Furthermore, Heh1-L requires its MSC domain to bind Chm7 (Fig. 4A,C), whereas the LD is needed for Heh1-S interaction (Fig. 6B). This suggests that Heh1-S presents a distinct feature that is necessary to impair Chm7 recruitment to the NE. Indeed, Heh1-S contains 49 unique residues at its C-terminal (Grund et al., 2008; Rodríguez-Navarro et al., 2002), which are highly conserved among budding yeast species (Fig. S6D). Notably, we found that a Heh1-S mutant lacking these unique residues was unable to suppress the sensitivity of HEH1-L chm7OPEN cells towards DMSO (Heh1-S 1-638aa; Fig. 6D). Interestingly, this unique fragment was not required for heterodimerization (Fig. 6C, compare lane 4 with lanes 5 and 6). Conversely, complementation was independent of the LEM domain (ΔLEM; Fig. 6D). We could exclude that lack of complementation was due to altered protein stability, since all examined mutants were expressed to similar levels compared to full-length Heh1-S (Fig. S6C). Furthermore, expression of N-terminal truncations of Heh1-S also rescued the growth defect of HEH1-L chm7OPEN cells (Fig. S6E). Additionally, Heh1-S lacking the dimerization domain (Heh1-S Δ552-638) failed to suppress the phenotype of HEH1-L chm7OPEN cells in the presence of DMSO (Fig. S6F).
In order to obtain mechanistic insights into the regulation of Chm7 recruitment, we examined whether the HEH1-L chm7OPEN phenotype is affected by Heh2, a paralog of Heh1 (King et al., 2006). However, the absence of HEH2 did not decrease the suppression of the DMSO sensitivity of HEH1-L chm7OPEN cells by Heh1-S (Fig. S6G). In addition, we tested the role of inner nuclear membrane protein Nur1, which forms a complex with Heh1-L (Mekhail et al., 2008) and, in S. japonicus, is required to recruit the ESCRT machinery to the NE (Pieper et al., 2020). Indeed, deletion of NUR1 suppressed the phenotype of HEH1-L chm7OPEN cells independently of Heh1-S (Fig. S6G). Hence, we speculated that Heh1-S might disrupt the interaction between Nur1 and Heh1-L to block Chm7 recruitment. However, using co-IP experiments, we could neither detect an interaction between Heh1-S and Nur1 nor a decrease in binding of Nur1 to Heh1-L (Fig. S6H), implying that Heh1-S acts in an Heh2- and Nur1-independent manner. Taken together, these results suggest that the unique residues of Heh1-S modulate the interaction between Heh1-L and Chm7, and imply that suppression of the growth defect cannot be explained only by heterodimerization. We thus conclude that the spliced form Heh1-S modulates the interaction of Heh1-L with Chm7, and prevents excessive recruitment of Chm7 to the NE.
The ubiquitin relative Hub1 assists pre-mRNA splicing in yeast and human cells (Ammon et al., 2014; Mishra et al., 2011; Wilkinson et al., 2004). Although Hub1 upregulation is correlated with cadmium tolerance in budding yeast (Chanarat and Svasti, 2020), the physiological role of Hub1-mediated splicing remains poorly understood. Here, we unveil that Hub1 modulates ESCRT recruitment to the NE by regulating the alternative splicing of the inner nuclear membrane protein Heh1.
Whereas Heh1-L recruits the ESCRT member Chm7 to seal NE holes, persistent interaction between them is toxic for cells (Thaller et al., 2019; Vietri et al., 2020b; Webster et al., 2016; Willan et al., 2019), indicating that this process needs to be tightly regulated. This association is resolved by the ESCRT-III/Vps4 complex, since depletion of ESCRT-III members causes Chm7 accumulation at the NE in yeast and human cells (Gu et al., 2017; Pieper et al., 2020; Webster et al., 2016; Willan et al., 2019). In addition, lack of the AAA ATPase Vps4 in fission yeast causes a growth defect accompanied by NE defects, which can be suppressed by deleting Chm7 or blocking its recruitment to the NE (Gu et al., 2017; Pieper et al., 2020). In agreement with these findings, we provide several lines of evidence that uncontrolled Heh1-L-mediated recruitment of Chm7 is the underlying cause of the growth defect in ESCRT-III mutants. First, masking or complete removal of the MSC domain suppresses the synthetic growth phenotype caused by Heh1-L overexpression in ESCRT-III mutants (Fig. 2B; Fig. S3). Second, a similar outcome is seen when Heh1-L is expressed with mutations within its MSC domain that specifically disrupt the interaction with Chm7 (Figs 3C and 4D). Finally, this synthetic growth phenotype strictly depends on the presence of Chm7 (Fig. 4E; Fig. S5C–E).
In S. cerevisiae, Chm7 is constitutively exported from the nucleus by the major exportin Crm1, suggesting the existence of an active mechanism to physically separate Chm7 and Heh1-L (Thaller et al., 2019). In agreement with this finding, we found that overexpression of nuclear import-deficient Heh1-L mutants in the ESCRT-III did4Δ mutant caused a stronger phenotype than expression of WT Heh1-L (Fig. 2C; Fig. S3). Moreover, CHMP7, the human homolog of Chm7, also contains a functional nuclear export signal, indicating a similar regulation in mammalian cells (Vietri et al., 2020a,b). Our findings imply an additional layer of regulation that involves the alternative splicing of Heh1. Cells lacking Hub1, which are compromised in producing the spliced variant Heh1-S, display impaired growth when combined with ESCRT-III mutants (did2Δ or did4Δ) and accumulate misassembled NPCs at the SINC (Fig. 1; Fig. S2). Importantly, this phenotype seems to arise from a toxic gain-of-function of Heh1-L and not the mere absence of Heh1-S, since no synthetic growth phenotype was detected in heh1Δ cells. Furthermore, deletion of CHM7 also rescued the synthetic growth phenotype of hub1Δ did4Δ cells (Fig. 4F), further supporting the idea that this defect is due to the lack of HEH1 splicing. Importantly, we found that the interaction between Chm7 and Heh1-L is impaired by Heh1-S expression (Fig. 5C), implying that it prevents or counteracts excessive Chm7 recruitment. Intriguingly, this behavior is highly reminiscent of the regulation of human seven in absentia homolog 1 (Siah-1), an ubiquitin ligase implicated in cellular stress responses (Qi et al., 2013). Alternative splicing generates the shorter variant Siah-1S. The short version forms heterodimers with Siah-1, which seems to be critical for complex formation with Siah-interacting protein (Mei et al., 2007).
A recent study has shown that human LEM2 condensates into a liquid-like phase through intrinsically disordered regions (IDR), which are present at the N-terminal domain, and contribute to NE reassembly and post-mitotic nucleocytoplasmic compartmentalization (von Appen et al., 2020). Interestingly, IDRs are also present in the N-terminal domain of Heh1 (Fig. S7), suggesting that they may play a similar role in yeast. However, overexpression of Heh1-L lacking the LEM domain or the N-terminal region impairs growth of ESCRT-III mutants to the same extent as the full-length protein (Fig. 2B; Fig. S3C–E). Conversely, the growth defect of HEH1-L chm7OPEN cells can still be rescued when expressing LEM- or N-terminal truncated versions of Heh1-S (Fig. 6D; Fig. S6E). While this does not exclude the possibility that other functions of Heh1 may be influenced by the presence of the N-terminal IDRs, these data suggest that at least Chm7 recruitment and its regulation by Heh1-S is controlled in a manner independent of N-terminal motifs.
The specific role of Heh1-S in modulating the recruitment of ESCRT-III to the NE implies the presence of distinct features not present in Heh1-L. Besides the absence of the second transmembrane domain and the MSC domain, Heh1-S contains 49 unique residues at its C-terminus due to the frameshift generated by Hub1-mediated alternative splicing of HEH1 (Grund et al., 2008; Mishra et al., 2011; Rodríguez-Navarro et al., 2002). Interestingly, while not involved in heterodimerization with Heh1-L, the presence of these residues in Heh1-S is critical to rescue DMSO toxicity in HEH1-L chm7OPEN cells (Fig. 6D). Hence, we propose a model in which the repair of membrane damage at the NE is fine-tuned by the two splice products of HEH1. NE ruptures or misassembled NPCs promote the interaction between Heh1-L and Chm7, which in turn recruits ESCRT-III proteins and Vps4 to repair damage through membrane scission. In wild-type cells, Hub1-spliced Heh1-S modulates the interaction between Chm7 and Heh1-L (Fig. 7, left panel). However, in cells lacking Heh1-S, Chm7 recruitment is not regulated (Fig. 7, right panel), leading to its continuous recruitment. This most likely results in excessive activity and deleterious phenotypes, such as DNA damage, membrane fenestration or ‘karmallae’ formation, as reported in other species (Gu et al., 2017; Vietri et al., 2020a,b).
To date, alternative splicing of HEH1 has been only demonstrated for the subphyla Saccharomycotina and Ascomycota (Hurtig et al., 2020). Nevertheless, the importance of restraining ESCRT activity at the NE suggests that the role of Heh1-S here might be fulfilled by a different factor in other species.
The mechanism by which Heh1-S regulates the recruitment of Chm7 still remains elusive. We speculate that the unique residues of Heh1-S might recruit a yet to be identified factor to avoid excessive NE remodeling. Since the phosphatidic acid-binding activity of Chm7 is necessary for its function at the NE in yeast and humans (Olmos et al., 2016; Thaller et al., 2020 preprint), Heh1-S might control the membrane composition at NE ruptures to avoid persistent recruitment of Chm7. Interestingly, in S. pombe Lem2 acts as a barrier to control the membrane flow between the nuclear envelope and endoplasmic reticulum; nonetheless the MSC domain is dispensable for this function (Hirano et al., 2020; Kume et al., 2019). Although additional experiments are required to determine whether Heh1-S modulates membrane composition at the NE, such a possible function would explain how Heh1-S recovers the growth defect of ESCRT-III mutants at elevated temperatures (Fig. 1, Fig. S2). Supporting this idea, overexpression of ole1+, which encodes the sole Δ9 fatty acid desaturase, can partially suppress the phenotypes of cells lacking Chm7 homolog in S. japonicus (Lee et al., 2020b).
Our data provide a novel example for the key role that alternative splicing plays in regulating a specific cellular function, such as NE integrity. Given the strong conservation of LEM domain proteins, our results may provide insights into an ancient molecular mechanism evolved to avoid excessive NE remodeling.
MATERIALS AND METHODS
Yeast methods and molecular biology
S. cerevisiae strains and plasmids are listed in Tables S1 and S2, respectively. Standard protocols for transformation, mating, sporulation and tetrad dissection were used for yeast manipulations. Yeast growth assays were performed by spotting five-fold serial dilutions of the indicated strains on solid agar plates. Yeast strains isogenic to W303 or DF5 were used for genetic studies, and PJ69-7a for yeast two-hybrid assays. Cells were grown at 30°C (if not indicated otherwise) in YPD medium (2% glucose) or synthetic complete medium lacking individual amino acids to maintain selection for transformed plasmids. Protein tagging and the construction of deletion mutants were conducted by a PCR-based strategy (Janke et al., 2004; Knop et al., 1999) and confirmed by immunoblotting and PCR, respectively. For galactose induction, cells were cultured in synthetic complete medium lacking the specific amino acids plus 2% raffinose, and galactose was added to a mid-log phase yeast cells to a final concentration of 2%. To generate untagged HEH1-L and HEH1-S strains, the HEH1 genomic region after the intron (+2046 to +2631) was cloned into the integrative plasmid YIplac128. This construct was amplified together with the LEU2 marker and integrated into the endogenous HEH1 locus at positions +1916 or +1920 to generate HEH1-L or HEH1-S cells, respectively. The HEH1 locus was subsequently sequenced to confirm proper integration and intron removal in the obtained strains.
Cloning methods used standard protocols or the Gibson Assembly Master Mix (NEB). HEH1 and CHM7 were cloned into pGADT7 or pGBKT7 vectors (Clontech™) for yeast two-hybrid experiments. Plasmids with point mutations or deletions were constructed by a PCR-based site-directed mutagenesis approach. Maps and primer DNA sequences are available upon request.
Total protein extracts from 2×107 cells (OD600=1) were prepared by trichloroacetic acid (TCA) precipitation (Knop et al., 1999). Proteins solubilized in HU loading buffer [8 M urea, 5% SDS, 200 mM Tris-HCl pH 6.8, 20 mM dithiothreitol (DTT) and Bromophenol Blue 1.5 mM] were resolved on NuPAGE 4–12% gradient gels (Invitrogen), and analyzed by standard immunoblotting techniques. Dpm1 antibody was used as loading control.
Polyclonal Hub1 antibodies were raised in rabbits and have been described previously (Mishra et al., 2011; 1:5000). Monoclonal (3F10; 1:1000) antibody directed against the HA epitope and monoclonal (B-2; 1:1000) antibody against GFP were purchased from Roche and Santa Cruz Biotechnology (Dallas, TX), respectively. Monoclonal Dpm1 (5C5; 1:2000) antibody was purchased from Thermo Fisher Scientific.
Live-cell imaging and analysis
Imaging was performed on a Zeiss AxioObserver Z1 confocal spinning-disk microscope equipped with an Evolve 512 (Photometrics) EMM-CCD camera through a Zeiss Alpha Plan/Apo 100×/1.46 oil DIC M27 objective lens. Optical section images were obtained at focus intervals of 0.2 μm. Subsequent processing and analyses of the images were performed in Fiji/ImageJ (Schindelin et al., 2012). DAPI staining was performed by adding 2.5 µg ml−1 DAPI to mid log-grown cells, incubating for 30 min, and washing with PBS before subjecting to imaging. Measurement of the diameter of the nucleus was performed using Fiji/ImageJ in cells expressing Heh1-LGFP or Heh1-SGFP; each data point represents the average length of at least 35 cells from the same image. For SINC quantification, cells were grown overnight in complete minimal medium (synthetic complete medium), diluted in fresh medium to an OD600=0.3 and allowed to grow for 4 h (∼2 divisions) before imaging. For quantification of Cmh7 accumulation at the NE, CHM7GFP HEH1-Lmars cells carrying the empty vector or a plasmid bearing HEH1-S6HA with the endogenous promoter were grown overnight in minimal selective medium (synthetic complete medium lacking leucine), diluted in fresh medium to an OD600=0.15 and allowed to grow for 6 h (∼3 divisions) before imaging. Statistical analyses of the mean values were performed using R statistical language, and different groups are marked with letters at the 0.05 significance level from a Tukey's post hoc test.
Cell lysates were prepared by resuspending the pellets from 150–200 OD600 units in 800 μl lysis buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 10% glycerol, 1 mM EDTA pH 8, 0.5% NP-40, 1× complete EDTA-free protease inhibitor cocktail (Roche), 2 mM PMSF, 20 mM N-ethylmaleimide (NEM)]. Cells were lysed by bead beating (Precellys 24, Bertin instruments) with zirconia/silica beads (BioSpec Inc.) and lysates were cleared by centrifugation (800 g, 5 min). Clarified extracts were incubated with anti-HA–Sepharose (Roche) or GFP-Trap (Chromotek) for 1.5 h at 4°C, beads were washed four times with lysis buffer and two times with wash buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA pH 8). Proteins were eluted by boiling with HU loading buffer and analyzed by immunoblotting.
Protein sequence alignment
Prediction of intrinsically disordered regions
Predictions were performed using the Protein DisOrder prediction System (PrDOS; Ishida and Kinoshita, 2007), with a prediction false positive rate of 2%.
We thank F. Wilfling, C.-W. Lee and members of the Jentsch and Braun labs for fruitful discussions and critical comments on the manuscript; F. Wilfling and C.-W. Lee for strains; and K. Straßer for technical assistance. We dedicate this paper to the memory of Stefan Jentsch, a great scientist and mentor.
Conceptualization: M.C., S.B., S.J.; Methodology: M.C., L.M.C.; Validation: M.C.; Formal analysis: M.C.; Investigation: M.C., L.M.C.; Data curation: M.C.; Writing - original draft: M.C.; Writing - review & editing: M.C., L.M.C., B.P., S.B.; Visualization: M.C.; Supervision: S.B., S.J.; Project administration: S.B., S.J.; Funding acquisition: S.B., S.J.
Research in the S.J. laboratory was supported by Max Planck Society, Deutsche Forschungsgemeinschaft, Center for Integrated Protein Science Munich (CIPSM), Louis-Jeantet Foundation and a European Research Council (ERC) Advanced Grant. Research in the S.B. laboratory was supported by grants from the Deutsche Forschungsgemeinschaft (BR 3511/4-1). S.B. is a member of the Collaborative Research Center 1064 funded by the Deutsche Forschungsgemeinschaft and acknowledges infrastructural support.
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.250688.reviewer-comments.pdf
The authors declare no competing or financial interests.