SUMOylation regulates Lem2 function in centromere clustering and silencing Open Access

Uniform segregation of genetic material between daughter cells is essential for cellular proliferation and survival. Centromeres play a critical role in this process by directing the formation of kinetochore complexes, onto which spindles from opposite poles of dividing cells can attach to separate duplicated chromosomes. Centromere-associated heterochromatin contributes to centromere function and further promotes genome stability by silencing repetitive genetic elements and suppressing recombination (Allshire and Madhani, 2018). In addition, increasing evidence is emerging that the physical organisation of centromeres within the cell is also important, with clustering of centromeres during interphase seen in many organisms, often in the vicinity of the nuclear periphery (Fransz et al., 2002; Weierich et al., 2003; Solovei et al., 2004; Kozubowski et al., 2013; Padeken et al., 2013). Although the mechanisms and functions of centromere clustering are yet to be fully elucidated, there is evidence that this spatial organisation can facilitate loading of centromeric proteins (Wu et al., 2022), enhance transcriptional silencing (Padeken et al., 2013) and promote genome stability through prevention of micronuclei formation (Jagannathan et al., 2018).

The fission yeast Schizosaccharomyces pombe has proven a powerful model for the study of nuclear organisation. During interphase, S. pombe chromosomes adopt the so-called Rabl conformation, whereby all three centromeres are clustered together and anchored at the nuclear periphery adjacent to the spindle pole body (SPB; the microtubule organisation centre in yeast) (Funabiki et al., 1993; Mizuguchi et al., 2015). A key factor in this clustering is Csi1, which appears to provide a physical link between kinetochore and SPB-associated proteins. Deletion of Csi1 results in severe centromere-clustering defects as well as defects in chromosome segregation during mitosis (Hou et al., 2012). In addition, the inner nuclear membrane protein Lem2 appears to function in parallel with Csi1 to help cluster centromeres, with cells lacking both Csi1 and Lem2 showing synthetic clustering defects (Barrales et al., 2016). Lem2 is localised throughout the nuclear envelope, but shows Csi1-dependent enrichment at the SPB (Ebrahimi et al., 2018). Lem2 shares similarity with the Lap2-emerin-Man1 (LEM) subfamily of animal cell lamina-associated proteins, and although S. pombe lacks nuclear lamina, Lem2 shares conserved functions of lamin-related proteins, including maintenance of nuclear envelope structure, peripheral tethering of chromatin and chromatin silencing (Hiraoka et al., 2011; Gonzalez et al., 2012; Banday et al., 2016; Barrales et al., 2016; Tange et al., 2016; Hirano et al., 2018).

Similar to what occurs in multicellular eukaryotes, centromeres in S. pombe comprise two distinct types of chromatin – a core domain enriched for the centromeric histone variant CENP-A (Cnp1 in S. pombe) and flanking regions of repressive heterochromatin associated with pericentromeric repeats [comprising inner most repeat (imr) and outer repeat (otr) sequences]. A complex network of redundant mechanisms contribute to pericentromeric heterochromatin formation and silencing, including both RNA interference (RNAi)-dependent and -independent pathways (Reyes-Turcu and Grewal, 2012; Marina et al., 2013; Allshire and Ekwall, 2015; Chalamcharla et al., 2015; Martienssen and Moazed, 2015; Tucker et al., 2016; Taglini et al., 2020). Lem2 has also been implicated in pericentromeric silencing, with deletion of Lem2 causing loss of silencing most prominently in the imr region (Banday et al., 2016; Barrales et al., 2016). Interestingly, this role of Lem2 in silencing appears to be independent of its role in tethering centromeres at the nuclear periphery – whereas the N-terminal HeH/LEM domain of Lem2 associates with centromeric chromatin and is required for chromatin tethering (Barrales et al., 2016; Tange et al., 2016), the C-terminal Man1-Src1p-C-terminal (MSC) domain of Lem2 is sufficient to mediate centromere silencing, possibly via recruitment of repressive chromatin factors (Banday et al., 2016; Barrales et al., 2016). A growing number of partner proteins have been implicated in contributing to the diverse functions of Lem2; for example, binding to the inner nuclear membrane protein Bqt4 helps mobilise Lem2 around the nuclear envelope and mediates telomere-related functions (Ebrahimi et al., 2018; Hirano et al., 2018), whereas interaction of Lem2 with the RNA surveillance factor Red1 was recently found to contribute to regulation of meiotic transcripts (Martin Caballero et al., 2022). However, how Lem2 is differentially targeted for roles in these varied processes is yet to be fully elucidated.

Lem2 is one of many centromere- and kinetochore-associated proteins found to be subject to SUMOylation in fission yeast (Kohler et al., 2015). SUMO is a small protein modifier that is similar in structure to ubiquitin, and like ubiquitin, can be covalently attached to lysine residues in substrate proteins through a cascade of E1 activating, E2 conjugating and E3 ligase enzymes. However, whereas a relatively large and diverse array of E3 ligases confer substrate specificity for ubiquitylation, the complement of E3 SUMO ligases is much smaller, reflecting a greater role for spatial control in substrate definition (Psakhye and Jentsch, 2012; Jentsch and Psakhye, 2013). Although a classic fate of ubiquitylated proteins is proteasome-mediated degradation, SUMOylation is more often associated with modulating protein–protein interactions, either negatively, for example by obscuring a binding surface (Pichler et al., 2005), or more commonly positively, via non-covalent interaction of SUMO with a SUMO-interacting motif (SIM). Individual SUMO–SIM interactions often act synergistically to enhance binding affinities (Hecker et al., 2006; Yau et al., 2021), and these interactions can also influence protein localisation (Mahajan et al., 1998; Matunis et al., 1998). In addition, crosstalk between SUMO and ubiquitin modifications can occur; for example, in some circumstances polySUMOylated proteins can be recognised by SUMO-targeted ubiquitin ligases (STUbLs), with subsequent ubiquitylation driving extraction or degradation of target proteins (Uzunova et al., 2007; Perry et al., 2008; Nie et al., 2012; Kohler et al., 2013, 2015; Nie and Boddy, 2015).

Like other post-translational modifications, SUMOylation is dynamic and reversible. SUMO deconjugation is performed by the conserved ULP/SENP family of SUMO-specific proteases, which includes six SENP family proteins in human, and Ulp1 and Ulp2 in yeast. Differences in substrate specificity between these enzymes often appear to be determined by their distinct subcellular localisations (Li and Hochstrasser, 2003; Hickey et al., 2012). In Saccharomyces cerevisiae, Ulp2 is localised throughout the nucleoplasm, but shows highest activity towards SUMO–SUMO linkages and therefore the shortening of polySUMO chains. In contrast, Ulp1 shows broad specificity, removing SUMO from substrate proteins as well as processing SUMO precursors; however, it is spatially restricted, being localised primarily to the inner surface of the nuclear pore complex. Loss of this localisation results in a major shift in substrates, indicating that the physical location of Ulp1 normally restricts its activity towards certain SUMOylated proteins while enabling cleavage of others (Li and Hochstrasser, 2003).

Centromere- and kinetochore-associated factors are reported to be enriched amongst SUMOylated proteins in several species (Azuma et al., 2003; Montpetit et al., 2006; Zhang et al., 2008; Mukhopadhyay et al., 2010; Ban et al., 2011; Wan et al., 2012; Li et al., 2016; Restuccia et al., 2016), and both SUMO E3 ligase and SUMO protease enzymes have been found to colocalise with kinetochores (Joseph et al., 2004; Agostinho et al., 2008; Ban et al., 2011; Cubenas-Potts et al., 2013; Suhandynata et al., 2019). Indeed, the budding yeast genes encoding SUMO and Ulp2 were first identified as high-copy suppressors of a temperature-sensitive mutation in Mif2 (Meluh and Koshland, 1995), the orthologue of the mammalian centromere protein CENP-C. It was subsequently shown that the human orthologue of Ulp2, SENP6, is required for proper centromere assembly, by preventing hyper-SUMOylation of multiple proteins within the constitutive centromere-associated network (CCAN) to enable their assembly at centromeres (Liebelt et al., 2019). Conversely, recent evidence suggests that SUMOylation of kinetochore protein Nuf2 is required to promote recruitment of the SIM-domain-containing centromere-associated protein CENP-E, which is essential for proper alignment of chromosomes in metaphase (Subramonian et al., 2021). Hence the role of SUMO at centromeres appears multifaceted and is yet to be fully defined.

In S. pombe there is a single SUMO isoform (Pmt3), and the majority of SUMOylation is directed by one E3 SUMO ligase, Pli1. Deletion of Pli1 is associated with centromere-related defects, including impaired silencing at pericentromeric (imr) regions and sensitivity to the microtubule-destabilising drug, thiabendazole (TBZ) (Xhemalce et al., 2004). Similar defects are seen in cells lacking either the SUMO protease Ulp1 (Han et al., 2010) or the nucleoporin Nup132, which is required to tether Ulp1 to the nuclear periphery (Han et al., 2010; Nie and Boddy, 2015). It has previously been proposed that depletion of Pli1, and hence a reduction in SUMOylation, accounts for the defects in all three backgrounds. This is because Nup132-tethered Ulp1 functions to antagonise Pli1 auto-SUMOylation, and so in nup132Δ cells, polySUMOylated Pli1 accumulates and is subject to STUbL-mediated degradation (Nie and Boddy, 2015). However, the release of Ulp1 from the nuclear periphery in nup132Δ cells results not only in a reduction in global SUMOylation (linked primarily to destabilisation of Pli1), but also an increase in accumulation of SUMOylated proteins at the nuclear periphery (Kramarz et al., 2020). Indeed, it has been recently demonstrated that impaired processing of stalled replication forks in nup132Δ cells can be attributed to loss of deSUMOylation activity at the nuclear periphery, with deletion of Pli1 rescuing the defects (Kramarz et al., 2020). Hence, questions remain about precisely how alteration of the SUMO landscape in nup132Δ cells impacts on centromere function.

Towards investigating the role of SUMOylation in centromere function in S. pombe, here, we have directly tested whether centromeric defects in nup132Δ cells are a result of either Pli1 destabilisation and hence reduced SUMOylation, or, conversely, enhanced SUMOylation at the nuclear periphery. We present multiple lines of evidence that localised hyper-SUMOylation is the primary driver of centromeric defects in this background and identify Lem2 as a key factor whose SUMOylation contributes to defects in centromere silencing. Unexpectedly, we find that in contrast to the detrimental effects on centromeric silencing, hyper-SUMOylation can enhance centromere clustering, helping to rescue clustering defects in cells lacking Csi1. Interestingly, this effect is again at least partially mediated through SUMOylation of Lem2. Our results reveal a previously unappreciated role for SUMOylation in promoting centromere clustering, and suggest that SUMOylation might provide a mechanism for coordination of the diverse functions of Lem2, possibly influencing the balance of Lem2 interactions with alternate partner proteins.

Defects in centromere silencing in nup132Δ cells were previously proposed to arise as a result of destabilisation of Pli1, given that loss of Nup132-dependent localisation of SUMO protease Ulp1 has been shown to lead to increased Pli1 auto-SUMOylation and hence STUbL-dependent degradation (Nie and Boddy, 2015). In order to test this model, we sought to specifically block the ubiquitin-mediated destabilisation of Pli1.

Although Pli1 has been shown to be subject to ubiquitin-dependent degradation (Nie and Boddy, 2015), the specific ubiquitylation sites have not been identified. To address this, and because such an analysis has not previously been reported, we performed global identification of protein ubiquitylation sites in fission yeast under physiological conditions, by adapting previously described strategies for enrichment of ubiquitylated peptides based on affinity purification of peptides bearing the diGly moiety left by trypsin digestion of ubiquitin (Udeshi et al., 2013). Whole-cell protein extracts were prepared under denaturing conditions, and sequentially digested with Lys-C and trypsin. The resulting peptides were fractionated and subjected to diGly-immunoprecipitation (IP), from which bound peptides were eluted and analysed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Fig. S1A). We performed eight independent analyses, identifying a total of 5116 ubiquitylation sites on 1801 proteins (Table S1). By comparison, the largest previous study of ubiquitylation sites in fission yeast, which employed tagged and overexpressed ubiquitin, identified 1200 ubiquitylation sites on 494 proteins (Beckley et al., 2015) (Fig. S1B). Hence, our approach was both more physiological and more sensitive, resulting in identification of a greater number of ubiquitylation sites on a wider range of proteins (representing ∼35% of the S. pombe proteome).

From our global analyses, we identified three ubiquitylated residues in Pli1 – K15, K169 and K214 (Fig. 1A). To test whether ubiquitylation of these residues promoted degradation of Pli1 in nup132Δ cells, we replaced endogenous Pli1 with a mutant version in which the three identified lysine residues were mutated to arginine (Pli1^K3R). Both wild-type and mutant Pli1 were C-terminally tagged with FLAG. Western blot analysis indicated that in wild-type cells, Pli1^K3R mutant protein levels were similar to wild-type Pli1. However, in nup132Δ cells, whereas wild-type Pli1 was destabilised, Pli1^K3R levels remained high (Fig. 1B). That mutation of these three residues was sufficient to stabilise Pli1 confirms that these sites are responsible for the ubiquitin-dependent degradation of Pli1 in the nup132Δ background.

We next tested whether the stabilising Pli1^K3R mutation was also sufficient to rescue the centromere-silencing defects observed in nup132Δ cells. In wild-type cells, a ura4⁺ reporter gene inserted into the heterochromatic inner-most repeat region of centromere one (imr:ura4⁺) is repressed, and cells consequently grow well on medium containing the counter-selective drug 5-fluoroorotic acid (5-FOA). As reported previously (Xhemalce et al., 2004; Nie and Boddy, 2015), deletion of either nup132⁺ or pli1⁺ results in reduced growth in the presence of 5-FOA, indicating increased expression of ura4⁺ and hence loss of silencing (Fig. 1C). If the silencing defect in nup132Δ cells were due to destabilisation of Pli1, we would expect it to be rescued by expression of the stabilised Pli1^K3R mutant. However, this was not the case – nup132Δ cells expressing Pli1^K3R–Flag displayed a defect in silencing equivalent to those expressing wild-type Pli1–Flag. In contrast, expression of Pli1^K3R-Flag in an otherwise wild-type background did not affect silencing. Similarly, both nup132Δ and pli1Δ cells showed sensitivity to the microtubule-stabilising drug thiabendazole (TBZ), consistent with defects in centromere function, and this TBZ sensitivity was also not rescued by the Pli1^K3R mutation (Fig. 1C). Therefore, in contrast to previous assumptions, the centromere defects in nup132Δ cells do not appear to be due to destabilisation of Pli1.

Given that the centromere defects in nup132Δ cells appeared not to relate to destabilisation of Pli1 and hence reduced global SUMOylation, we reasoned that they might rather relate to increased SUMOylation at the nuclear periphery due to loss of nuclear membrane-tethered Ulp1 deSUMOylase activity. This would be consistent with the known association of S. pombe centromeres with the nuclear periphery, adjacent to the SPB (Hou et al., 2013; Mizuguchi et al., 2015). To investigate this model, we first tested whether overexpression of Ulp1 was sufficient to rescue the centromere defects in nup132Δ cells. Interestingly, although Ulp1 overexpression had no effect in wild-type cells, it partially suppressed the TBZ sensitivity of nup132Δ cells, consistent with the defect resulting from hyper-SUMOylation (Fig. 2A). We were unable to test whether Ulp1 overexpression suppresses the defect in silencing at imr:ura4⁺ given that this experiment required growth in minimal medium, but silencing defects in nup132Δ cells were only observed in rich medium (Fig. S2). As a complementary approach, we also generated a plasmid expressing a ‘lysine-less’ SUMO (Pmt3^KallR), in which each of the lysine residues in Pmt3 are replaced by arginine, such that SUMO can attach to substrates, but cannot form lysine-linked polySUMO chains. We expressed wild-type and mutant SUMO in ‘mature’ form (terminating in the diGly motif), to avoid any impact of Ulp1 mis-localisation on SUMO maturation. Strikingly, although expression of lysine-less SUMO caused a slight increase in TBZ sensitivity in otherwise wild-type cells, it clearly suppressed the stronger TBZ sensitivity seen in nup132Δ cells, suggesting that increased polySUMOylation in particular contributes to the defect (Fig. 2B). As a further, more direct test of the model, we also fused Ulp1 to the nucleoporin Nup107 in order to more specifically restore deSUMOylation activity at the nuclear periphery. This again resulted in partial suppression of TBZ sensitivity in nup132Δ cells, as well as rescue of the silencing defect at imr:ura4⁺ (Fig. 2C). Although all of these approaches will likely have pleiotropic effects and thus full suppression might not necessarily be expected, the results are consistent with a model in which centromere defects in nup132Δ cells are primarily due to localised reduction in Ulp1 deSUMOylation activity at the nuclear periphery.

We next set out to identify candidate proteins whose SUMOylation could account for the defects in centromere function in nup132Δ cells. We searched for candidates meeting three key criteria: (1) having a known role in silencing specifically at the centromere imr region, similar to Nup132; (2) having a known localisation to the nuclear periphery; and (3) that are known to be subject to SUMOylation in S. pombe. Interestingly, applying these criteria we identified one clear candidate – the inner nuclear membrane protein Lem2. Lem2 guides nuclear membrane assembly, directly interacts with centromeric and telomeric chromatin to anchor it at the nuclear periphery, and crucially, plays a role in heterochromatin silencing (Hiraoka et al., 2011; Gonzalez et al., 2012; Banday et al., 2016; Barrales et al., 2016; Tange et al., 2016; Ebrahimi et al., 2018; Hirano et al., 2018; Pieper et al., 2020). Lem2 has been reported to associate with Nup132 (Iglesias et al., 2020), and strikingly, effects of Lem2 deletion on centromere silencing have also been reported to be nutrition dependent, with stronger defects observed in rich medium than in minimal medium (Tange et al., 2016). This mirrors our observations for nup132Δ cells (Fig. S2), strongly suggesting that these two proteins function in the same pathway for centromere silencing.

A previous proteomic study identified a cluster of seven SUMOylation sites in the N-terminus of Lem2 (Kohler et al., 2015) (Fig. 2C). To investigate whether silencing defects in nup132Δ cells might relate to SUMOylation of Lem2 at these sites, we replaced endogenous Lem2 with a mutant version in which each of the seven SUMOylated lysine residues were replaced with arginine (Lem2^K7R). Western blot analysis confirmed that SUMOylated Lem2 is detectable in wild-type cells, and that this SUMOylation is lost with the Lem2^K7R mutant (Fig. 3B). Expression of Lem2^K7R alone showed no defect in silencing of the imr:ura4⁺ reporter, as assessed by growth on FOA. Strikingly, however, when expressed in nup132Δ cells, Lem2^K7R rescued silencing, restoring growth on FOA to near wild-type levels (Fig. 3C). This was confirmed by real-time quantitative PCR (RT-qPCR) analysis of imr:ura4+ transcript levels, which were elevated in nup132Δ cells, consistent with loss of silencing, but reduced to nearer wild-type levels in nup132Δ Lem2^K7R cells (Fig. 3D). These observations suggest that centromere-silencing defects in nup132Δ cells are linked to Lem2 SUMOylation.

Interestingly, although expression of Lem2^K7R rescued the centromere-silencing defects in nup132Δ cells, it did not rescue TBZ sensitivity (Fig. 3C). Rather, TBZ sensitivity was exacerbated in nup132Δ Lem2^K7R cells, suggesting that other defects, possibly in a pathway functioning in parallel with Lem2, underlie the TBZ sensitivity caused by deletion of Nup132.

Previous studies have shown that although the N-terminus of Lem2 plays a dominant role in tethering chromatin to the nuclear periphery, the C-terminal MSC domain alone is sufficient for Lem2-mediated centromeric silencing (Banday et al., 2016; Barrales et al., 2016). Given that Lem2 SUMOylation sites lie in the N-terminal region, we expected that deletion of the Lem2 N-terminal domain, including the SUMOylation sites, would also be effective in rescuing the silencing defects in nup132Δ cells. However, interestingly, this was not the case; although deletion of the Lem2 N-terminal domain (Lem2ΔN; comprising MSC domain only) did not itself impair centromeric silencing, it also did not rescue the silencing defects in nup132Δ cells (Fig. 3E). Thus suppressing Lem2 SUMOylation rescues silencing defects in nup132Δ cells, but only when the N-terminal domain is present.

The observation that the normally dispensable N-terminal domain of Lem2 is required for centromeric silencing in nup132Δ cells argues against the possibility that mutation of Lem2 SUMO sites simply serves to stabilise Lem2 in nup132Δ cells by blocking STUbL-mediated degradation, as seen for Pli1. Consistent with this, and in contrast to our observations for Pli1, no difference in protein stability was observed between GFP-tagged wild-type Lem2 and Lem2^K7R in nup132Δ cells (Fig. S3A). Additionally, live-cell imaging of Lem2–GFP or Lem2^K7R–GFP revealed no apparent difference in Lem2 localisation in wild-type versus nup132Δ cells (Fig. S3B). We therefore suspect that SUMOylation might rather impact on Lem2 function by affecting protein-protein interactions.

As well as its function in silencing, Lem2 plays an important role in facilitating clustering of centromeres at the SPB (Barrales et al., 2016). Given the Lem2-dependent defect in pericentromeric silencing in nup132Δ cells, we next tested whether centromere clustering is also affected in this background. Live-cell imaging was performed on cells expressing GFP–Cnp1 to visualise centromere positioning, with Sid4–RFP acting as a marker of the SPB. Consistent with previous reports (Barrales et al., 2016), although 100% of wild-type cells displayed a single SPB-associated GFP–Cnp1 focus (representing the three clustered centromeres) adjacent to the SPB, deletion of lem2⁺ resulted in a greater proportion of cells with two or even three GFP–Cnp1 foci, indicating centromere-clustering defects (Fig. 4A,B). Interestingly, nup132Δ cells also exhibited a mild but significant declustering of centromeres. However, whereas the silencing defects in nup132Δ cells were rescued by expression of Lem2^K7R, the centromere-clustering defects were not (Fig. 4B, compare bars 3 and 5). Expression of Lem2^K7R alone was also not associated with any centromere-clustering defects. Consequently, we find no evidence that the mild centromere-clustering defects in nup132Δ cells are dependent on Lem2 SUMOylation. Interestingly, we also noticed that, unlike the silencing defects, the centromere-clustering defects in nup132Δ cells are independent of nutritional status, as they also occur in cells grown in minimal medium (Fig. 4D, bar 5), consistent with there being a different underlying cause.

It has been shown previously that Lem2 and Csi1 function in parallel pathways to promote centromere clustering (Barrales et al., 2016). Given that the clustering defects we observed in nup132Δ cells appear to be unexpectedly Lem2 independent, we questioned whether nup132⁺ deletion might instead affect the Csi1-dependent centromere-clustering pathway. To test this, we generated nup132Δ csi1Δ double mutant cells, with the expectation that if Nup132 and Csi1 act in parallel pathways, we should observe synthetic defects compared to the single mutants, similar to what has been seen for lem2Δ csi1Δ cells; and conversely, if Csi1 and Nup132 function in the same pathway, we would observe no further augmentation of clustering defects. Unexpectedly, and contradicting either hypothesis, we found that deletion of nup132⁺ substantially suppresses centromere clustering defects in csi1Δ cells, observing normal clustering in 85.0% of nup132Δ csi1Δ double-mutant cells, as compared to only 63.9% of csi1Δ single mutant cells (Fig. 4C).

We next sought to understand the mechanism by which deleting nup132⁺ rescues centromere-clustering defects in csi1Δ cells. We previously found evidence that the TBZ sensitivity of nup132Δ cells is related to increased SUMOylation at the nuclear periphery. To test whether hyper-SUMOylation also accounts for the alleviation of centromere-clustering defects in csi1Δ cells upon deletion of nup132⁺, we tested whether the alleviation is inhibited by manipulations that reduce polySUMO accumulation. Interestingly, even in wild-type cells we found that overexpression of Pmt3^KallR, which suppresses polySUMOylation, resulted in a small but significant defect in centromere clustering, consistent with a role for polySUMOylation in promoting clustering (Fig. 4D, compare bars 1 and 3). Moreover, although overexpression of either Pmt3^KallR or Ulp1 (increasing de-SUMOylation) had little effect on centromere clustering in csi1Δ or nup132Δ single-mutant backgrounds, these were sufficient to fully suppress the rescue of centromere clustering in csi1Δ nup132Δ double-mutant cells (Fig. 4D, compare bar 13 to bars 15 and 16). The rescue was also entirely suppressed by deletion of pli1⁺, repressing global SUMOylation (Fig. 4C, compare bars 2 and 6). Together these observations strongly suggest that increased polySUMOylation at the nuclear periphery in nup132Δ cells can rescue centromere-clustering defects caused by loss of Csi1.

One known function of polySUMOylation is the recruitment of STUbLs for extraction and/or degradation of target proteins (Uzunova et al., 2007; Perry et al., 2008; Nie et al., 2012; Kohler et al., 2013, 2015; Nie and Boddy, 2015). We therefore tested whether the rescue of centromere clustering in csi1Δ cells upon deletion of nup132⁺ is dependent on the STUbL Slx8 or on Ufd1 (a component of a ubiquitin-selective chaperone) (Nie et al., 2012; Kohler et al., 2015). However, neither the deletion of slx8⁺ nor expression of mutant Ufd1 (ufd1ΔCt^213-342, lacking the C-terminal domain that mediates interaction with SUMO; Kohler et al., 2013), prevented the rescue (Fig. S4, compare bars 2 and 3, and 5 and 6). Thus, the mechanism of rescue does not appear to involve STUbL-dependent protein removal, and rather, SUMO might promote centromere clustering by influencing protein–protein interactions and/or localisation.

We previously found that centromere-silencing defects in nup132Δ cells relate to Lem2 SUMOylation, whereas the clustering defects in nup132Δ cells are not a result of Lem2 SUMOylation. We therefore questioned whether the nup132Δ-dependent rescue of centromere-clustering defects caused by absence of Csi1 is dependent on Lem2 SUMOylation. Strikingly, we found that although expression of Lem2^K7R did not affect clustering in wild-type nor csi1Δ cells, it did partially suppress the rescue seen in nup132Δ csi1Δ cells, with clustering reduced from 85.0% to 76.5% upon expression of Lem2^K7R (Fig. 4C, compare bars 1, 2 and 8). This suggests that Lem2 is at least one substrate whose SUMOylation helps to promote centromere clustering in the absence of Csi1.

One function of Csi1 is to help stabilise Lem2 at the SPB (Ebrahimi et al., 2018). Given that SUMOylation can influence protein localisation, we next examined whether the rescue of clustering defects in nup132Δ csi1Δ cells might relate to enhanced localisation of Lem2 at the SPB. Whereas GFP-tagged Lem2 was found to reliably localise to the SPB (as indicated by SPB marker Sid4–RFP) in wild-type and nup132Δ cells, we observed a loss of Lem2 localisation at the SPB upon deletion of csi1⁺, consistent with previous findings (Ebrahimi et al., 2018). However, remarkably, whereas only 75.0% of csi1Δ cells show normal Lem2 localisation at the SPB, this was increased to 85.2% in nup132Δ csi1Δ cells (Fig. 5A,B). That Lem2 localisation is partially rescued in the nup132Δ background is consistent with this contributing to the rescue of centromere clustering. To confirm that this enhancement in Lem2 SPB localisation, like the suppression of centromere-clustering defects, is dependent on Lem2 SUMOylation, we analysed the subcellular localisation of Lem2^K7R–GFP in nup132Δ csi1Δ cells. Interestingly, we found a reduction in SPB localisation of Lem2^K7R–GFP as compared to wild-type Lem2-GFP in nup132Δ csi1Δ cells, closely mirroring the effects on centromere clustering (Fig. 5B). We also noted a small reduction in SPB localisation of Lem2^K7R–GFP as compared to wild-type Lem2–GFP in csi1Δ cells. Although not statistically significant, these changes are consistent with Lem2 SUMOylation contributing to its localisation to the SPB in the absence of Csi1 (expression levels of Lem2^K7R versus wild-type Lem2 were comparable in both backgrounds; Fig. S5). Together, these observations support a model in which deletion of nup132⁺ can rescue centromere-clustering defects in csi1Δ cells by increasing the SUMO-dependent localisation of Lem2 to the SPB and amplifying its contribution to centromere clustering.

Finally, we questioned whether the function of Lem2 SUMOylation in the rescue of centromere clustering in csi1Δ cells relates solely to Lem2 localisation or whether SUMOylation also enhances Lem2 function in centromere clustering independently of localisation. It has been shown previously that Bqt4 functions to mobilise Lem2 away from the SPB and around the nuclear envelope, and that although Lem2 localisation at the SPB is destabilised in csi1Δ cells, it is rescued in bqt4Δ csi1Δ double mutant cells (Ebrahimi et al., 2018). We therefore asked whether the enhanced localisation of Lem2 to the SPB in bqt4Δ csi1Δ cells would be sufficient to rescue centromere-clustering defects, as in nup132Δ csi1Δ cells. However, this was not the case; although deletion of bqt4⁺ alone had little effect on centromere clustering, clustering defects in bqt4Δ csi1Δ double-mutant cells were equivalent to those in csi1Δ single-mutant cells (Fig. 5C). This suggests that SUMOylation of Lem2 plays an important role in its function in clustering, beyond enhancing its localisation at the SPB.

The dissociation of the deSUMOylase Ulp1 from the nuclear envelope in the absence of Nup132 has been found to cause seemingly paradoxical effects. On the one hand, there are reduced levels of E3 SUMO ligase Pli1 (Nie and Boddy, 2015) and thus reduced global SUMOylation, yet on the other hand, there are toxic effects of persevering polySUMOylated proteins that accumulate at nuclear pore complexes (Kramarz et al., 2020). In contrast to previous assumptions, here, we show that centromere-related defects in nup132Δ cells are not explained by destabilisation of Pli1. Rather, we find that suppressing global SUMOylation, or Lem2 SUMOylation specifically, is sufficient to suppress centromeric defects, strongly supporting the model that enhanced SUMOylation at the nuclear periphery is the primary driver of centromeric phenotypes in nup132Δ cells. Unexpectedly, we also find that although hyper-SUMOylation at the nuclear periphery is detrimental to centromere silencing, it can enhance centromere clustering – clustering defects in csi1Δ cells are partially rescued by deletion of nup132⁺, whereas depleting SUMO suppresses this rescue. We show that Lem2 is again a key SUMO substrate in this context, given that specifically suppressing Lem2 SUMOylation is also sufficient to partially supress the rescue. Our results reveal opposing effects of hyper-SUMOylation in general, and Lem2 SUMOylation in particular, on different aspects of centromere function, and suggest a key role for SUMOylation in regulating the diverse activities of Lem2.

Lem2 has been shown to play multiple roles at the nuclear periphery, including organisation and silencing of heterochromatin domains, maintenance of nuclear membrane integrity and regulation of nuclear-exosome-mediated RNA degradation (Hiraoka et al., 2011; Gonzalez et al., 2012; Banday et al., 2016; Barrales et al., 2016; Tange et al., 2016; Ebrahimi et al., 2018; Hirano et al., 2018; Pieper et al., 2020; Martin Caballero et al., 2022). Interactions with many different partner proteins are thought to enable Lem2 to localise to distinct subcellular domains and contribute to different pathways. Interestingly, evidence indicates that several of these interactions occur through the same (C-terminal MSC) domain of Lem2 (Gu et al., 2017; Pieper et al., 2020; Martin Caballero et al., 2022), yet how competing interactions are regulated to coordinate distinct functions of Lem2 remains largely unknown. The data presented here suggest that SUMOylation of Lem2 both impairs its function in centromere silencing and simultaneously enhances its function in centromere clustering. As the most parsimonious explanation for these observations, we favour a model in which SUMOylation regulates the competition between Lem2-associating factors, enhancing interactions with centromere-clustering components at the cost of those that mediate silencing functions (Fig. 5D). Evidence of competition between alternative functions of Lem2 has been observed before; in the related fission yeast S. japonicus, the ESCRT-III/Vps4 machinery has been shown to remodel Lem2 heterochromatin attachments, and ESCRT-III/Vps4-mediated release from heterochromatin is required to free up Lem2 to perform its function in nuclear envelope sealing at the end of mitosis (Pieper et al., 2020). We propose that spatially controlled deSUMOylation of Lem2 represents another layer of regulation of Lem2 interactions, such that Lem2 SUMOylation acts as a regulatory switch between pathways.

Our model proposes a role for SUMOylation in regulating Lem2 function, yet mutation of the seven known SUMOylation sites within Lem2 (Lem2^K7R) in an otherwise wild-type background does not result in any of the centromere-related phenotypes normally associated with impaired Lem2 function. However, our genetic analyses suggest that this might reflect redundancy with other pathways, given that (1) expressing Lem2^K7R in nup132Δ cells results in increased TBZ sensitivity, and (2) expressing Lem2^K7R in nup132Δ csi1Δ cells causes more-severe centromere-clustering defects. That the role of Lem2 SUMOylation might be hidden by parallel pathways is consistent with previous findings that show that multiple elements of Lem2 function are masked by redundant mechanisms (Barrales et al., 2016).

We show here that hyper-SUMOylation contributes to centromere-silencing defects in nup132Δ cells, and that this effect is dependent on SUMOylation of Lem2. Mutation of Lem2 SUMOylation sites rescues the silencing defects without any obvious effect on Lem2 localisation or stability, suggesting that SUMOylation likely alters Lem2 protein–protein interactions. The mechanism by which Lem2 promotes heterochromatic silencing remains opaque; although there is evidence that Lem2 influences the balance of recruitment of opposing chromatin factors, including the histone deacetylase (HDAC) repressor complex SHREC and anti-silencing protein Epe1, no physical association of Lem2 with these factors has been detected (Banday et al., 2016; Barrales et al., 2016). Interestingly, recent evidence indicates that Lem2 can also influence post-transcriptional silencing by interacting with the RNA surveillance factor Red1 to regulate RNA degradation by the nuclear exosome (Martin Caballero et al., 2022). An intriguing possibility is that this post-transcriptional mechanism might contribute to Lem2-mediated silencing of the centromeric imr:ura4+ reporter; in future it will be interesting to explore whether Lem2 interaction with Red1 is affected by SUMOylation.

An unexpected finding was that suppression of Lem2 SUMOylation can rescue centromere-silencing defects in nup132Δ cells only when the N-terminus of Lem2 remains intact. This was surprising because the Lem2 N-terminal domain has been previously shown to be dispensable for centromeric silencing (Banday et al., 2016; Barrales et al., 2016). A possible explanation is that in nup132Δ cells there is hyper-SUMOylation, and therefore potentially misregulation, of one or more other proteins whose function can normally compensate for absence of the Lem2 N-terminal domain. In particular, the N-terminus of Lem2 has been shown to be important for Lem2 binding to centromeric chromatin, yet expression of the Lem2 C-terminus alone is sufficient to maintain centromeric silencing in an otherwise wild-type background (Barrales et al., 2016; Tange et al., 2016). We speculate that, in this scenario, interactions with other centromere-localised proteins might be sufficient to localise Lem2 within sufficient proximity to centromeric chromatin to perform its silencing function. However, in nup132Δ cells, hyper-SUMOylation of one or more of these proteins might impair Lem2 interactions, such that there is increased dependency on N-terminus-mediated chromatin binding of Lem2 to allow the C-terminus to perform its function in silencing.

Csi1 and Lem2 have been shown to function in parallel pathways for centromere clustering (Barrales et al., 2016), but dissecting the relative contributions of these pathways is complicated by the fact that Csi1 also plays a role in stabilising Lem2 localisation at the SPB (Ebrahimi et al., 2018). Unexpectedly, here we have found that nup132⁺ deletion causes a SUMO-dependent rescue in centromere-clustering defects in csi1Δ cells, and that this is at least partially mediated through SUMOylation of Lem2, given that rescue is suppressed upon expression of Lem2^K7R. Interestingly, the rescue is associated with SUMOylation-dependent enhancement of Lem2 localisation at the SPB. However, deletion of Bqt4 has also been shown to increase Lem2 localisation to the SPB, even in the absence of Csi1 (Ebrahimi et al., 2018), yet we find that bqt4Δ csi1Δ cells show no alleviation in centromere-clustering defects. Taken together, these observations suggest that deletion of nup132⁺ causes SUMOylation-dependent enhancement of Lem2 localisation at SPB in csi1Δ cells; however, this localisation is not sufficient to rescue centromere-clustering defects, and SUMOylation plays an additional role in enhancing centromere clustering adjacent to the SPB.

How might SUMOylation influence centromere clustering pathways? SUMOylated proteins can be targeted by STUbL complexes resulting in ubiquitin-dependent extraction and/or proteasomal degradation, as seen for Pli1 in nup132Δ cells (Uzunova et al., 2007; Nie and Boddy, 2015, 2016). However, analyses in slx8Δ and ufd1ΔCt^213-342 backgrounds revealed no evidence that enhancement of centromere clustering in nup132Δ cells involves targeting of SUMOylated substrates via STUbL and Ufd1 pathways. We therefore speculate that a SUMO ‘molecular glue’ mechanism might be enhancing centromere clustering, similar to what has been described for promyelocytic leukaemia nuclear bodies (PML-NBs) (Shen et al., 2006; Corpet et al., 2020). PML-NBs are membrane-free compartments that regulate a number of processes including transcriptional control and DNA repair, and importantly both SUMOylation of PML and internal SIMs are key in NB formation through non-covalent SUMO–SIM interactions. We hypothesise that SUMOylated Lem2 interacts with as-yet-undiscovered SIM domains within Lem2-associated proteins, enhancing both localisation of Lem2 to the SPB and recruitment of centromere-clustering factors. Of note, although PML-NB formation is mediated by SUMOylation, the process is also driven by liquid–liquid phase separation in intrinsically disordered proteins (Banani et al., 2017; Shin and Brangwynne, 2017; Zhang et al., 2020). Strikingly, LEM2 (also known as LEMD2) has been observed to form liquid–liquid phase droplets in human cells (von Appen et al., 2020); an unexplored question is whether S. pombe Lem2 also has the ability to phase separate, and whether SUMOylation of Lem2 could mediate such processes, similar to the SUMO-mediated formation of PML-NBs.

We observe that expression of Lem2^K7R is sufficient to partially but not fully suppress the nup132Δ-mediated rescue of centromere clustering in csi1Δ cells. Western blot analysis confirmed that the K7R mutation is sufficient to abolish Lem2 SUMOylation, arguing against the possibility that this partial effect is explained by residual Lem2 SUMOylation at remaining lysine residues. We therefore suspect that the reason why the effects of nup132⁺ deletion are not fully suppressed by expression of Lem2^K7R is that SUMOylation of one or more additional proteins also contributes to the effects. Factors related to centromeres and the nuclear periphery are highly enriched amongst SUMOylated proteins (Kohler et al., 2015); Lem2 might therefore be one of several proteins whose SUMOylation can contribute to enhanced centromere clustering in circumstances where this process is impaired. Perhaps related to this, it has been shown previously that deletion of nup132⁺ disrupts the normal dynamic disassembly and reassembly of the outer kinetochore during meiotic prophase in S. pombe (Yang et al., 2015); it is tempting to speculate that this might be caused by constitutive hyper-SUMOylation of component proteins due to the loss of localised Ulp1-mediated deSUMOylation.

We note that although suppressing Lem2 SUMOylation (Lem2^K7R) rescues centromere-silencing defects in nup132Δ cells, it conversely exacerbates TBZ sensitivity in this background. This is despite the fact that reducing global polySUMOylation through overexpression of Ulp1 or Pmt3^KallR alleviates TBZ sensitivity. This suggests that although SUMOylation of Lem2 plays at least a partial role in both reducing pericentromeric silencing and simultaneously enhancing centromere clustering, hyper-SUMOylation of further unidentified protein(s) is likely responsible for TBZ sensitivity. Identifying these proteins might be challenging, for example if multiple factors contribute redundantly. However, TBZ sensitivity can be caused by reduced accumulation of cohesins at centromeres, and interestingly, studies in human cell culture have revealed that hyper-SUMOylation of RAD21 and other cohesin subunits reduces their chromatin association (Wagner et al., 2019). S. pombe Rad21 has also been found to be SUMOylated (Kohler et al., 2015); hence this could be a promising candidate for a protein whose hyper-SUMOylation could contribute to TBZ sensitivity in nup132Δ cells. Levels of human RAD21 SUMOylation are regulated through the specific localisation of the deSUMOylating enzyme SENP6 at centromeric and telomeric domains (Wagner et al., 2019). Given the localisation of centromeres to the nuclear envelope in fission yeast, an attractive model is that only centromere-localised Rad21 that is within proximity of nuclear envelope-tethered Ulp1 is deSUMOylated, allowing centromere-specific deSUMOylation and enhanced chromatin binding at specific loci.

Through analysis of nup132Δ cells, we have found that hyper-SUMOylation at the nuclear periphery both impairs centromeric silencing and enhances clustering of centromeres at the SPB. Both physical clustering (Muller et al., 2019) and enrichment for SUMOylated proteins (Abrieu and Liakopoulos, 2019), appear to be common features of centromeres across diverse species, and although changes in centromere clustering have been linked to human carcinoma progression (Verrelle et al., 2021), mechanisms of clustering remain poorly defined. It will be interesting to see whether SUMO plays a conserved role in promoting centromere clustering, possibly acting as a ‘molecular glue’ to facilitate protein–protein interactions. In addition, we show that Lem2 is a key SUMO substrate in the context of both centromere silencing and clustering, and present a novel model whereby SUMOylation plays an important role in modulating the balance of Lem2 interactions with partner proteins to coordinate its diverse functions. Mammalian Lem2 is critical for embryonic development (Tapia et al., 2015), and variants in LEM2 have been linked to several human diseases including juvenile cataracts, arrhythmic cardiomyopathy and a novel nuclear envelopathy with progeria-like symptoms (Boone et al., 2016; Abdelfatah et al., 2019; Marbach et al., 2019), highlighting the potential clinical importance of understanding the regulation of Lem2-mediated pathways. In S. pombe, it has previously been suggested that post-translational modification could account for observed differences in Lem2 function in different nutritional conditions (Martin Caballero et al., 2022); it is an intriguing possibility that changes in Lem2 SUMOylation status might mediate the response to environmental cues.

Strains and plasmids used in this study are listed in Tables S2 and S3, respectively, and are available upon request. For deletion and epitope tagging (with Flag or GFP), a PCR-based method was used to amplify resistance cassettes flanked by 80 bp target site homology for integration at endogenous loci by homologous recombination (Bahler et al., 1998). For Pli1^K3R and Lem2^K7R strains, long homology-containing fragments incorporating the relevant lysine to arginine mutations were generated by fusion PCR, and integrated at endogenous loci by transformation into pli1Δ::ura4⁺ or lem2Δ::ura4⁺ strains, respectively, followed by selection on 5-FOA. For expression of C-terminally GFP-tagged Lem2 or Lem2^K7R, constructs were assembled in the pREP41 plasmid using Gibson assembly methods. For overexpression of a mature Pmt3 construct that does not require processing by Ulp1, a pREP41-myc-his-pmt3⁺ plasmid (Jongjitwimol et al., 2014) was modified by site directed mutagenesis to insert a stop codon immediately after residue G111 (resulting in a terminal di-glycine). The Pmt3^KallR overexpression plasmid was subsequently generated from the pREP41-myc-his-mature-pmt3⁺ plasmid by further rounds of site-directed mutagenesis. To generate the Ulp1 overexpression plasmid, genomic ulp1⁺ was amplified by PCR and integrated into pREP41-myc-his plasmid using Gibson assembly methods. To tether Ulp1 to the nuclear envelope, Gibson assembly methods were used to generate a GFP-nup107-his-myc-Ulp1 plasmid, and the construct was inserted at the endogenous arg3⁺ locus by CRISPR/Cas9 editing as previously described (Torres-Garcia et al., 2020). Genomic modifications and plasmids were verified by sequencing.

Cells were grown in YES medium, with the exception of strains expressing pREP41 plasmids which were maintained in PMG –Leu (Sabatinos and Forsburg, 2010). For spotting assays, 10-fold serial dilutions were plated onto non-selective medium, or medium supplemented with 1 g/l 5-FOA (Melford Laboratories) or 20 µg/ml TBZ (Sigma-Aldrich).

Enrichment for ubiquitylated peptides using PTMScan^® beads broadly followed the detailed protocol published previously (Udeshi et al., 2013), with the following exceptions to apply the method in S. pombe (a broad outline is shown in Fig. S1A). Approximately 3×10⁹ wild-type S. pombe cells were grown to mid-log phase in YES liquid, washed twice with PBS and harvested. Cells were lysed by bead-beating in a denaturing lysis buffer [8 M Urea, 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 6 mM MgCl₂, 1 mM PMSF, 4 mM NEM, 50 µM PR-619, 20 µM MG-132, 4 mM 1, 10-phenanthroline, 1× complete EDTA-free protease inhibitor cocktail (Roche) and 1 mM pefabloc] using acid washed beads and a VXR basic Vibrax^® at 4°C. The lysate was clarified by centrifugation at 17,000 g for 30 mins at 4°C, and a Bradford assay was performed to assess protein concentration. Reduction, alkylation, Lys-C and trypsin digestion steps were performed as described previously (Udeshi et al., 2013). Peptide reverse-phase offline fractionation was performed on a Dionex Ultimate 5000 HPLC (Thermo Fisher Scientific) as described previously (Udeshi et al., 2013) with minimal variations. In brief, peptides were loaded on a Zorbax 300-Extend-C-18 (5 µm, 4.6×250 mm) column (Agilent) at a constant flow rate of 1 ml/min. Peptides were separated at constant flow of 1ml/min, according to the following gradient. The percentage of buffer B was increased linearly from zero to: 3% at 5 mins; 5% at 7 mins; 15% at 35 mins; 20% at 45 mins; 35% at 53 mins; 50% at 60 mins; and finally, 70% at 62 mins. The percentage of buffer B was held at 70% for 3 mins and then returned gradually to zero at 70 mins. The total time of the gradient was 82 mins. Fractions were collected at 1 min time slices until the 64th minute, and vacuum centrifuged to dryness. Dried peptides were resuspended in 187.5 µl IAP buffer (50 mM MOPS pH 7.2, 10 mM sodium phosphate and 50 mM NaCl), and fractions were pooled in a serpentine, non-contiguous manner such that every eighth fraction was combined to generate eight final fractions containing 1.5 ml of resuspended peptides. DiGly remnant enrichment using PTMScan^® beads was performed as described previously (Udeshi et al., 2013).

LC-MS/MS analyses were performed on Q Exactive Plus and Q Exactive mass spectrometers (both Thermo Fisher Scientific). For the Q Exactive Plus analysis, liquid chromatography for the LC-MS/MS runs was performed on an EASY-nLC 1000 liquid chromatography system (Thermo Fisher Scientific) coupled to spectrometers via modified NanoFlex sources (Thermo Scientific). Peptides were loaded onto 250 mm×75 µm PicoFrit (C18, 2 µm medium) analytical columns (New Objective) at a maximum pressure of 80 MPa. Solutions A and B for ultra-performance LC were 0.1% formic acid (FA) in water and acetonitrile (ACN), respectively. Peptides were eluted into the mass spectrometer at a flow rate of 200 nl/min using a gradient that incorporated a linear phase from 6% B to 30% B in 80 min, followed by a steeper phase and wash. The gradient run time was ∼120 min. The Q Exactive Plus mass spectrometer was operated in the data-dependent mode acquiring HCD MS/MS scans in the 300–1800 m/z scan range with a resolution of 17,500 after each MS1 scan (resolution 70,000) on the 12 most-abundant ions using an MS1 ion target of 3×10⁶ ions and an MS2 target of 5×10⁵ ions. The maximum ion time utilized for MS/MS scans was 120 ms; the HCD-normalized collision energy was set to 25 and the dynamic exclusion time was set to 20 s. The peptide match and isotope exclusion functions were enabled. The Q Exactive mass spectrometer (Thermo Fisher Scientific) was coupled on-line to a 50 cm Easy-Spray column (Thermo Fisher Scientific), which was assembled on an Easy-Spray source and operated constantly at 50°C. Mobile phase A consisted of 0.1% FA, whereas mobile phase B consisted of 80% ACN and 0.1% FA. Peptides were loaded onto the column at a flow rate of 0.3 μl/min and eluted at a flow rate of 0.25 μl/min according to the following gradient: 2 to 40% buffer B in 120 min, then to 95% in 11 min (total run time of 160 min). Survey scans were performed at 70,000 resolution (scan range 350–1400 m/z) with an ion target of 10⁶ and injection time of 20 ms. MS2 was performed with an ion target of 5×10⁴, injection time of 60 ms and HCD fragmentation with normalized collision energy of 27. The isolation window in the quadrupole was set at 2.0 Thomson. Only ions with charge between 2 and 7 were selected for MS2.

The MaxQuant software platform (Cox and Mann, 2008) version 1.6.1.0 was used to process the raw files and search was conducted against the Schizosaccharomyces pombe (July, 2016) protein database, using the Andromeda search engine (Cox et al., 2011). For the first search, peptide tolerance was set to 20 ppm whereas for the main search peptide tolerance was set to 4.5 ppm. Isotope mass tolerance was 2 ppm and maximum charge set to 7. Digestion mode was set to specific with trypsin allowing maximum of two missed cleavages. Carbamidomethylation of cysteine was set as a fixed modification. Oxidation of methionine, and the diGly residue on lysine were set as variable modifications. Peptide and protein identifications were filtered to a 1% false discovery rate (FDR).

Total RNA was extracted from 1×10⁷ mid-log phase cells using the Masterpure Yeast RNA Purification Kit (Epicentre), according to the manufacturer's instructions. 1 µg of extracted RNA was treated with TURBO DNase (Ambion) for 1 h at 37°C, and reverse transcription was performed using random hexamers (Roche) and Superscript III reverse transcriptase (Invitrogen). Lightcycler 480 SYBR Green (Roche) and primers (_qact1_F: 5′-GTTTCGCTGGAGATGATG-3′; _qact1_R: 5′-ATACCACGCTTGCTTTGAG-3′; _qura4_F: 5′-CGTGGTCTCTTGCTTTGG-3′; _qura4_R: 5′-GTAGTCGCTTTGAAGGTTAGG-3′) were used for qPCR quantification of imr:ura4⁺ transcript levels relative to act1⁺. Data presented represent three biological replicates and error bars represent one standard deviation. P-values were calculated using a two-tailed unpaired Student's t-test.

Immunoaffinity purifications were performed essentially as previously described (Oeffinger et al., 2007). For Pli1-Flag IP, cultures were grown to mid-log phase in YES medium, and 3×10⁸ cells were harvested in 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, 1× complete EDTA-free protease inhibitor cocktail (Roche). Cells were lysed by bead beading, and supernatant was clarified by two 10-min centrifugations at 17,000 g and 4°C. Extracts were incubated with pre-equilibrated protein G–agarose and anti-Flag M2 (Merck) for 1 h at 4°C. For Lem2–GFP IP, cultures were grown to mid-log phase in supplemented minimal medium and 3×10⁸ cells were harvested in 50 mM Hepes pH 7.5, 150 mM NaCl, 10% glycerol (v/v), 1% NP40 (v/v), 0.5% sodium deoxycholate (w/v), 1 mM EDTA, 2 mM PMSF, 50 mM N-ethylmaleimide 1× complete EDTA-free protease inhibitor cocktail (Roche). Cells were lysed by bead beading, and supernatant was clarified by 10 mins centrifugation at 400 g. Extracts were incubated with anti-GFP antibody (11814460001, Sigma-Aldrich; 1:100 dilution), for 90 mins at 4°C, and then pre-equilibrated protein G–agarose added for a further 1 h incubation. For both Pli1–Flag and Lem2–GFP IPs, beads were washed three times with lysis buffer, and proteins eluted in gel loading buffer [150 mM Tris-HCl pH 6.8, 8 M urea, 2.5% (w/v) SDS, 20% (v/v) glycerol, 10% (w/v) 2-mercaptoethanol, 3% (w/v) DTT, 0.1% (w/v) Bromophenol Blue] and analysed by immunoblotting using anti-Flag M2 (F3165, Merck, 1:1000 dilution), anti-GFP (A-11122, Thermo Fisher Scientific; 1:1000 dilution), anti-SUMO (rabbit-polyclonal anti-SUMO from Sarah Lambert, Institut Curie, Orsay, France; 1:5000 dilution) and anti-Tat1 (mouse monoclonal anti-tubulin from Keith Gull, University of Oxford, UK; 1:200 dilution).

Cells expressing Lem2–GFP, GFP–Cnp1 and Sid4–RFP were grown to mid-log phase in YES medium (except for those strains expressing plasmids which were grown in PMG −Leu) and embedded in low-melting point agarose. Imaging was performed at 25°C using a Nikon Ti2 inverted microscope, equipped with a 100×1.49 NA Apo TIRF objective and a Teledyne Photometrics Prime 95B camera. Images were acquired with NIS-elements (version 5.1), with Z-stacks taken at 0.25 µm intervals. Maximum intensity Z-projections were made in ImageJ. For centromere clustering analyses, chi-squared (χ²) tests were performed to calculate P-values for differences in the proportions of cells displaying centromeres ‘clustered’ versus ‘unclustered’ (one GFP–Cnp1 focus versus two or three GFP–Cnp1 foci). Similarly, for analysis of Lem2 localisation, χ² tests were used to calculate P-values for differences in the proportion of cells in which Lem2 was ‘present’ versus ‘absent’ at the SPB (denoted by Sid4–RFP).

We thank Yasushi Hiraoka, Robin Allshire, Genevieve Thon, Matthew Whitby and Julia Cooper for strains, Felicity Watts for the Pmt3 expression plasmid, Sarah Lambert for anti-SUMO antibody, and Keith Gull for the anti-Tat1 antibody. We are also grateful to Sigurd Braun for advice on detection of SUMOylated Lem2, and Ivan Matic for assistance with proteomic analyses. Imaging was performed in Centre Optical Instrumentation Laboratory (COIL), which is supported by a core grant (203149) to the Wellcome Centre for Cell Biology at the University of Edinburgh.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260868.reviewer-comments.pdf.