Cell-specific modulation of nuclear pore complexes controls cell cycle entry during asymmetric division

The acquisition of cellular identity is coupled to changes in the nuclear periphery and nuclear pore complexes (NPCs). Whether and how these changes determine cell fate remains unclear. We have uncovered a mechanism regulating NPC acetylation to direct cell fate after asymmetric division in budding yeast. The lysine deacetylase Hos3 associates specifically with daughter cell NPCs during mitosis to delay cell cycle entry (Start). Hos3-dependent deacetylation of nuclear basket and central channel nucleoporins establishes daughter cell-specific nuclear accumulation of the transcriptional repressor Whi5 during anaphase and perinuclear silencing of the CLN2 gene in the following G1 phase. Hos3-dependent coordination of both events restrains Start in daughter but not in mother cells. We propose that deacetylation modulates transport-dependent and - independent functions of NPCs, leading to differential cell cycle progression in mother and daughter cells. Similar mechanisms might regulate NPC functions in specific cell types and/or cell cycle stages in multicellular organisms.


Introduction
Asymmetric cell division is a conserved mechanism that generates diversity in cell populations. Asymmetric divisions are present not only in unicellular organisms such as bacteria, budding yeasts, and flagellates, but also play a major role in stem cell selfrenewal and tissue homeostasis during metazoan development (Knoblich, 2010;Li, 2013).
During asymmetric division, unequal partitioning of cell fate determinants ultimately leads to distinct identities in the newly generated cells. This usually requires the establishment of a cell polarity axis that directs the distribution of fate determinants.
Subsequent positioning of the cell division plane perpendicular to the fate determinant axis leads to their asymmetric partitioning between daughter cells after cytokinesis.
The nuclear envelope (NE) limits free diffusion of molecules between the nucleus and the cytoplasm. Transport between these two compartments is mediated by nuclear pore complexes (NPCs), macromolecular assemblies composed of approximately 30 nucleoporins that form a channel across the NE (Ibarra and Hetzer, 2015;Knockenhauer and Schwartz, 2016;Terry et al., 2007). Nucleo-cytoplasmic transport of proteins and RNA determines chromatin accessibility to transcription factors and mRNA export, and therefore is intimately tied to the regulation of gene expression and cell fate determination (Capelson et al., 2010a;Yang et al., 2013). In addition, the NE and nucleoporins associated with the nuclear basket of NPCs can directly interact with the nuclear genome to regulate gene expression and thus affect cell differentiation [reviewed in (Akhtar and Gasser, 2007;Capelson et al., 2010a;Mattout et al., 2015;Zuleger et al., 2014)]. In particular, the nuclear periphery is a transcriptionally repressive environment its homolog the Rb tumour suppressor in mammals. In yeast, a key event driving the G1/S transition is the dilution of Whi5 activity by cell growth, whereby the volume increase in daughter cells during G1 lowers the concentration of Whi5 below a critical threshold (Schmoller et al., 2015). This allows Cyclin-dependent kinase (Cdk) complexes to inactivate Whi5, which is then evicted from the nucleus (Costanzo et al., 2004;de Bruin et al., 2004). Interestingly, the G1 concentration of Whi5 is higher in daughter cells than in mother cells (Liu et al., 2015;Schmoller et al., 2015). The mechanism establishing this Whi5 concentration asymmetry is not known.
Here, we discovered that the distinct cell cycle program in budding yeast daughters compared to mothers is due to daughter-specific association of the class II histone deacetylase (HDAC) Hos3 with NPCs. We identify the mechanism that recruits Hos3 to the NPCs in the daughter cell during mitosis. Further, we demonstrate that Hos3mediated lysine deacetylation of distinct nucleoporins within NPCs establishes asymmetric segregation of the Whi5 transcriptional repressor and perinuclear tethering of the G1/S cyclin gene CLN2 in daughters. We find that both of these Hos3-coordinated functions contribute to inhibit Start in daughter cells. Thus, cell-specific regulation of NPCs by lysine deacetylation can direct differences in cell identity during asymmetric division.

Hos3 inhibits cell cycle entry exclusively in daughter cells
Commitment to a new division cycle in budding yeast occurs first in mother cells, and only later in daughter cells. Although the lysine deacetylase Hos3 has been implicated in the control of G1 length and gene expression, it is not known to control the distinct cell cycle programs in mother and daughter cells (Huang et al., 2009;Wang et al., 2009). To investigate this hypothesis, we determined G1 duration in wild type and hos3∆ mother and daughter cells by evaluating the interval between cytokinesis and bud emergence.
Specifically, we measured the disappearance of Myo1-GFP from the bud neck to its reappearance to the new bud site, normalized by the rate of growth a during this period (aTG1) (Figure 1A). This revealed a shortened G1 in hos3∆ daughter cells, but not mother cells, relative to wild type ( Figure 1B-C, S1A and Supplementary Table S1).
G1 is divided in two periods, T1 and T2, separated by nuclear export of the transcriptional repressor Whi5 ( Figure 1A). Whi5 nuclear export determines commitment to the next cell cycle (Start) by allowing expression of MBF/SBF target genes required for budding and DNA replication (Turner et al., 2012). Analysis of Whi5-GFP nuclear export kinetics (T1) showed that HOS3 deletion reduces aT1 in daughter cells relative to wild-type daughters (mean and standard deviation: WT: 12.6 ± 8.3, N = 55; hos3∆: 6.3 ± 3.5, N = 64, p < 0.0001, Mann-Whitney test) (Figure 1, D-E). Shortening of G1 in hos3∆ daughter cells relative to wild type daughters was observed even for cells born at similar sizes ( Figure 1G). In contrast, the cellular growth rate and the scaled time between Whi5 export and budding (aT 2 ), ascribed to molecular noise (Di Talia et al., 2007), were similar in hos3∆ and wild type cells (Figure S1B-C). In addition, a catalytically inactive mutant of Hos3 (Hos3 H196E, D231N , or Hos3 EN ) (Wang and Collins, 2014) showed advanced cell cycle entry exclusively in daughter cells in a similar manner to hos3∆ (Figure 1F-H).
Therefore, Hos3 activity inhibits Start in daughter cells.

Hos3 is recruited to the nuclear periphery of daughter cells during anaphase
Hos3 is present in the cytoplasm during interphase but is recruited to the daughter side of the septin-based ring at the mother-bud neck during mitosis (Wang and Collins, 2014).
To further investigate the subcellular localization of Hos3-GFP during mitosis, we performed time-lapse fluorescence microscopy. Unexpectedly, we discovered that Hos3-GFP associates with the nuclear periphery of the daughter cell, but not the mother cell, during anaphase. The perinuclear localization of Hos3 in daughter cells is coincident with migration of the anaphase nucleus from mother to daughter through the bud neck ( Figure   2A, S2A, movies S1-2). Hos3 disappeared from both the bud neck and the nuclear periphery 1-2 minutes before cytokinesis, marked by contraction of the acto-myosin ring and septin ring splitting, visualized by mCherry-tagged Myo1 and Cdc3, respectively (Figure S2B-C and movie S3). Hos3 protein levels did not change dramatically during mitosis, suggesting that protein degradation is not the major driver of its localization dynamics ( Figure 1B and S2D). The catalytically inactive protein Hos3 EN -GFP showed enrichment at the bud neck and nuclear periphery identical to the wild type protein ( Figure S2E), indicating that the deacetylase activity of Hos3 is not required for its localization dynamics.
We next asked whether Hos3 enrichment at the nuclear periphery is dependent on passage of the nucleus across the bud neck and into the daughter cell. To test this hypothesis, we first examined Hos3-GFP localization in cells arrested in metaphase by depletion of the APC activator Cdc20; these cells develop nuclear protrusions extending across the bud neck and into the daughter cell compartment (Kirchenbauer and Liakopoulos, 2013). Hos3-GFP decorated nuclear protrusions and remained confined in the daughter cell ( Figure S3A). Next, we examined cells that completed anaphase within the mother cell due to loss of the dynein microtubule motor Dyn1. In dyn1∆ mutants, Hos3-GFP was recruited to the perinuclear region only after migration of the anaphase nucleus into the daughter compartment (20/20 cells with delayed nuclear migration) ( Figure 2C, S3B-C, and movie S4). Thus, nuclear migration through the bud neck is required for Hos3 targeting to the perinuclear region.
We next investigated the relationship between the bud neck localization of Hos3 in early mitosis and its subsequent perinuclear enrichment. Hos3 localization to the bud neck depends on septins and the septin-interacting protein, Hsl7 (Wang and Collins, 2014). We found that mild perturbation of septin ring structure in temperature sensitive (ts) cdc12-1 mutant cells led to symmetric localization of Hos3 at both the bud neck and the nuclear periphery during anaphase ( Figure S4). Moreover, deletion of Hsl7 prevented perinuclear localization of Hos3-GFP, which was instead detected at the anaphase daughter spindle pole body (dSPB) (71/74 cells) (Figure 2D and S5A). Thus, proper recruitment of Hos3 to the septin ring is required for its enrichment at the nuclear periphery of daughter cells.
We note that enrichment of Hos3 at the dSPB, but not at the nuclear periphery, was reported previously using ectopically expressed Hos3-GFP (Wang and Collins, 2014). Indeed, we observed a similar localization pattern when Hos3-GFP was expressed from the strong ADH1 promoter ( Figure S5B, movie S5), suggesting that Hos3 is preferentially detected at the dSPB when over-expressed. Hos3 was also enriched at the dSPB after inhibition of mitotic exit in cdc15-1 ts mutants at 37 ºC, which in addition retained Hos3-GFP in the bud neck ( Figure S5C). We conclude that Hos3 recruitment to the bud neck is dispensable for its localization to the dSPB but is essential for its enrichment at the daughter cell nuclear periphery.
Finally, we asked whether the perinuclear localization of Hos3 reflects an association with the cytoplasmic or the nucleoplasmic side of the nuclear envelope. Inactivation of nuclear import factors Kap95, Kap60 and Kap122 did not affect the perinuclear localization of Hos3-GFP (data not shown). In contrast, inactivation of the importin Mtr10 impaired Hos3 recruitment to the nuclear periphery, which was instead found at the bud neck and the dSPB during anaphase (42/52 cells; Figure 2E). Notably, changes in Hos3-GFP localization in dyn1, hsl7 and mtr10 cells were not due to differences in Hos3-GFP levels, as these were comparable across the mentioned strains ( Figure S5D).
Together, the above data suggest that Hos3 is imported into the daughter nucleus during its migration across the bud neck, in a manner dependent on Mtr10. The highly regulated asymmetric enrichment of Hos3 at the nuclear periphery in daughters suggests that it performs an important function in daughter cell nuclei, and might underlie the Hos3mediated inhibition of Start in daughters.

Hos3 associates with nuclear pores in daughter cells to delay Start
To examine whether perinuclear Hos3 is sufficient to delay Start, we asked if targeting Hos3 to the mother cell nucleus prolongs G1 in these cells. We generated Hos3-GFP fused with the nuclear localization signal of SV40 TAg (Kalderon et al., 1984) (Hos3-NLS-GFP) and found that it localized to the periphery of both mother and daughter cell nuclei throughout the cell cycle ( Figure 3A and S6A-B). Importantly, we found that Whi5 nuclear export (T1) was delayed in both mother and daughter cells expressing symmetrically distributed Hos3-NLS by time-lapse imaging (Figure 3, B-C).
Furthermore, over-expression of Hos3-NLS impaired growth on solid media ( Figure 3D and S6C) and led to a dramatic increase in cell volume, consistent with a severe cell cycle delay ( Figure 3E). These phenotypes arising from Hos3-NLS over-expression required both the constitutive nuclear localization and the catalytic activity of Hos3, as cells grew normally when over-expressing Hos3-GFP (without the NLS) or Hos3 EN -NLS-GFP ( Figure 3D and S6C). We conclude that the perinuclear localization of active Hos3 is sufficient to inhibit Start.
Notably, Hos3-NLS did not accumulate in the nucleoplasm but was instead enriched in the nuclear periphery like its wild-type counterpart (see Figure 3A), and was detected by immuno-electron microscopy on the nucleoplasmic side of nuclear pore complexes (NPCs) ( Figure S6D). Moreover, artificial NPC clustering caused by deletion of NUP133 (Pemberton et al., 1995) led to clustering of Hos3-GFP and Hos3-GFP-NLS, which partially co-localized with Nup60-mCherry ( Figure 4A, S7A-C, and movie S6). These findings suggested that Hos3 associates with NPCs. To test this hypothesis, we investigated the role of distinct NPC components in the perinuclear localization of Hos3.
We found that the perinuclear association of Hos3-NLS was independent of the nuclear pore basket components Nup2 and Mlp1/2, but was dependent on Nup60 (Rout et al., 2000) (Figure S7D-E). By time-lapse microscopy, we found that although Hos3-GFP localized to the anaphase daughter nuclei in nup60∆ cells, it was no longer retained at the nuclear periphery. Instead, Hos3-GFP was homogeneously distributed throughout the nucleoplasm in nup60∆ anaphase cells ( Figure 4B and movie S7). Furthermore, we detected a direct interaction between epitope-tagged Hos3 and Nup60 in logarithmically growing cultures by co-immunoprecipitation. Interestingly, this interaction was much stronger in extracts prepared from Cdc20-depleted cells that are arrested in metaphase, when Hos3 localizes to the nuclear periphery ( Figure 4C and S7F), supporting the notion that Hos3 associates with nuclear pore basket components during nuclear migration into the bud.
To test if Hos3 interactions with the NPC could play a role in the inhibition of Start, we determined if artificial tethering of Hos3 to NPCs with an FKBP-FRB system (Gallego et al., 2013) affected the time from cytokinesis to Whi5 nuclear export. Indeed, we found that addition of rapamycin to HOS3-FKBP NUP49-FRB cells caused immediate recruitment of Hos3 to the nuclear periphery, and strongly and specifically delayed Start ( Figure 4D). We conclude that retention of Hos3 at the nuclear periphery through interactions with the NPC during anaphase is sufficient to delay Whi5 export and cell cycle entry.

Hos3 is required for the enrichment of Whi5 in daughter cell nuclei
To gain insight into the mechanism by which Hos3 inhibits Start in daughter nuclei, we next asked whether the presence of Hos3 alters the protein composition at the nuclear periphery. To facilitate the identification of proteins whose subnuclear distribution is regulated by Hos3, we examined the localization of 217 nuclear proteins fused to GFP, in wild type and HOS3-NLS cells (Figure 5A-B and Supplementary Table S2). We did not observe Hos3-dependent changes for 207 of these candidates, including more than 30 nucleoporins.
In contrast, 10 proteins showed altered localization in HOS3-NLS relative to wild type control strains, and belonged to two major categories. Group I includes proteins that showed reduced perinuclear localization in Hos3-NLS cells. These were all involved in nucleo-cytoplasmic transport and comprised the karyopherins Kap95 and Kap123, the mRNA export regulators Mtr2 and Mex67, and the TREX-2 component Sac3 ( Figure 5B and S8A). Group II proteins showed the inverse pattern: these proteins localized to the nucleoplasm in wild type cells and became enriched at the nuclear periphery in the presence of Hos3-NLS. This group was more heterogeneous and included the tRNA export factor Los1, the telomeric Ku complex components Yku70 and Yku80, and the UBX domain-containing Ubx7 (Figure 5B and S8A). In addition, the karyopherin Kap114 was weakly enriched in the nucleoplasm of wild type cells, and this enrichment was increased in Hos3-NLS cells.
The identification of group I proteins as potential Hos3 effectors raised the possibility that Hos3 regulates nucleo-cytoplasmic transport dynamics to delay Start in daughter cells. The nuclear concentration of the S-phase inhibitor Whi5 is intimately linked to the regulation of the G1/S transition. Indeed, the probability of entering the cell cycle is inversely proportional to the Whi5 nuclear concentration, and Start is triggered when cell growth dilutes Whi5 below a critical level (Liu et al., 2015;Schmoller et al., 2015).
Notably, the Whi5 nuclear concentration is higher in daughter than mother cells, although the reason for this asymmetry has not been elucidated. We therefore determined the nuclear concentration of Whi5-GFP in wild type and hos3Δ cells right after cytokinesis.
We found that the Whi5 nuclear concentration was higher in daughter than in mother cells. This asymmetry was slightly lower than previously described, possibly owing to differences in strain background or imaging conditions. Strikingly however, the asymmetric nuclear concentration of Whi5 was dependent on Hos3 ( Figure 5C and S8B).
We next focused on the regulation of karyopherins by Hos3. Specifically, we asked whether Hos3 regulates the nuclear levels of the Kap95 importin and of the Msn5 exportin, which together control Whi5 nuclear levels (Kosugi et al., 2009;Taberner et al., 2009). We found that after cytokinesis, the total amount and the concentration of Kap95-GFP and Msn5-GFP are higher in mother than in daughter nuclei. Furthermore, these asymmetries were dependent on Hos3 (Figure 5D-F and S8C-D). Thus, nuclear Kap95 and Msn5 are distributed asymmetrically and with the opposite polarity of their cargo Whi5, which is more concentrated in daughter than in mother nuclei. Notably, deletion of Hos3 did not affect the asymmetric distribution of NPCs marked with Nup49-mCherry ( Figure S8E), which are partially retained in mother cells due to a septin-based diffusion barrier (Shcheprova et al., 2008). These results suggest that Hos3 inhibits cell cycle entry in daughter cells through modulation of Whi5 nucleo-cytoplasmic transport in late anaphase. We propose that this leads to a higher Whi5 concentration in daughter nuclei, and consequently prolongs G1 duration of these cells.
To test whether regulation of Whi5 levels accounts entirely for the role of Hos3 in Start, we examined G1 length in cells lacking both Hos3 and Whi5. Deletion of WHI5 leads to advanced Start due to derepression of SBF/MBF target genes (Costanzo et al., 2004;de Bruin et al., 2004). Intriguingly, hos3∆ whi5∆ cells showed shorter G1 duration than either single mutant (Figure 5G and S9). This suggests that in addition to its role controlling Whi5 nuclear levels, Hos3 delays Start through additional, Whi5-independent mechanism(s).

Hos3 inhibits CLN2 expression in parallel to the CLN3 repressor Ace2
To further investigate how the Hos3 interaction with the NPC during anaphase delays Start in daughter cells, we examined the relationship between Hos3 and known regulators of cell cycle entry. Control of Start is composed of two distinct modules, each controlled by a different G1 cyclin (Di Talia et al., 2007). Cln3-Cdk complexes initiate Start through inactivation of Whi5 in early G1, followed by Cln1/2-Cdk complexes, which complete Whi5 inactivation in a positive feedback loop (Cross and Tinkelenberg, 1991;Dirick and Nasmyth, 1991;Skotheim et al., 2008) (Figure S10A). The daughter-specific factor Ace2 delays G1 in daughters by repressing CLN3 (Di Talia et al., 2009;Laabs et al., 2003). As Ace2 is retained in daughter nuclei during G1, we asked whether its localization dynamics depended on Hos3. However, the nuclear residence time of Ace2-GFP was not affected in hos3∆ cells ( Figure S10B). Moreover, advanced Start in hos3∆ cells was independent of the presence of CLN3, but required the presence of CLN2 (Figure S10C-D). Finally, deletion of Hos3 and Ace2 advanced Start to similar extents, and hos3∆ ace2∆ double mutants showed additive effects, suggesting that Ace2 and Hos3 function in distinct pathways (Figure S10D-E). Together, these results show that Hos3 delays Start in daughter cells in parallel to the Ace2 -Cln3 module, and possibly by inhibiting the Cln2-associated module.

Hos3 controls CLN2 expression and subnuclear localization of the CLN2 gene locus
We next analyzed CLN2 expression by time-lapse microscopy of wild type and hos3∆ cells expressing fast-folding, unstable GFP expressed from the CLN2 locus under the control of the CLN2 promoter (Bean et al., 2006). This analysis showed that the time of CLN2 expression relative to cytokinesis, determined using Myo1-mCherry, was advanced in hos3∆ daughter cells (Figure 6A-B and S11). The deacetylase Rpd3 binds to the CLN2 promoter to repress it in G1 phase (Takahata et al., 2009;Wang et al., 2009). We could detect Rpd3 but not Hos3 at the CLN2 promoter (Figure S12), suggesting that Rpd3 and Hos3 repress CLN2 through different mechanisms.
The yeast nuclear periphery and NPCs play an important role in three-dimensional genome organization and gene expression. For instance, yeast telomeres are anchored at the nuclear periphery, which is associated with gene silencing (Taddei and Gasser, 2012;Zimmer and Fabre, 2011). As the CLN2 gene locus is positioned near a telomere, just 65 kb away from TEL16R (Figure S13A), we hypothesized that CLN2 might be positioned in the perinuclear region, where it would be poised for Hos3-dependent regulation.
Confocal 3D imaging showed that CLN2 labeled with the LacO/LacI-GFP system was adjacent to the NE (Nup49-mCherry) in 88% of unbudded cells and in 20% of budded cells (Figure 6C-D, and S13B-C). Thus, CLN2 is perinuclear in G1 and is displaced towards their nuclear interior in S phase. This cell-cycle dependent localization was specific to CLN2, because subtelomeric loci (TEL4R and TEL12R) localized to the NE region in a similar fraction (»88%) of G1 and S-phase cells ( Figure 6E). Strikingly, the G1-specific perinuclear association of CLN2 was independent of WHI5 but required Hos3, as the fraction of G1 cells with CLN2 at the NE was reduced to less than 40% in hos3∆ cells (Figure 6D and S13C). Hos3 deletion also led to displacement of TEL12R but not of TEL4R from the NE, suggesting that Hos3 affects the perinuclear localization of multiple gene loci (Figure 6E and S13D).
To determine if Hos3-dependent perinuclear association of CLN2 might contribute to the role of Hos3 in Start, we evaluated ectopic recruitment of CLN2 to the NE. CLN2 containing LexA recognition sites within the 3'UTR was recruited to the NE via a LexA-Yif1 fusion protein (Schober et al., 2009). Expression of LexA-Yif1, but not of LexA alone, increased G1 duration in both mother and daughter cells, indicating that targeting of CLN2 to the NE is sufficient to delay Start. This delay was not bypassed in hos3∆ cells, suggesting that Hos3 targets CLN2 to the nuclear envelope but is dispensable for its repression once it is localized at the NE ( Figure 6F).

Daughter cell cycle control requires deacetylation of NPC components
The association of Hos3 with Nup60 raised the possibility that at least some of its functions in cell cycle entry might involve its association and/or deacetylation of NPC components. Consistent with the first possibility, time-lapse imaging showed that deletion of NUP60 impaired CLN2 perinuclear recruitment, but did not affect Whi5 asymmetric segregation and only moderately reduced G1 length in daughters (Fig. S14).
Interestingly, multiple nucleoporins are acetylated in yeast (Henriksen et al., 2012), and we found that the acetylation levels of the nuclear basket protein Nup60, and of the central channel components Nup49, Nup53 and Nup57 were reduced in HOS3-NLS relative to hos3∆ strains (Fig. 7A), suggesting that these proteins are Hos3 substrates in daughter cells. We next generated point mutants of these nucleoporins in which previously identified acetylated lysines (Henriksen et al., 2012) were changed to asparagine, whose biophysical properties resemble those of acetylated lysine and thus may mimic constitutive acetylation (Ben-Shahar et al., 2008). Consistent with this, lysine-to-asparagine (KN) alleles of Nup60, Nup49, Nup53 and Nup57 did not perturb cell growth or Nup60 perinuclear localization, and their combination conferred resistance to the toxic effects of Hos3-NLS over-expression ( Fig. S15 and Table S5).
We next used time-lapse microscopy to determine whether nucleoporin KN alleles affect asymmetric Whi5 nuclear concentration and perinuclear tethering of CLN2 in daughter cells. We found that asymmetric partitioning of Whi5-GFP was mildly affected in nup60 KN cells, and was severely impaired in cells expressing KN alleles of central channel nucleoporins, to levels similar to cells lacking Hos3 activity ( Fig. 7B and S16A).
Finally, we examined progression through Start in single and double KN mutant strains. This analysis showed a moderate reduction of T1 in single KN mutant daughters relative to wild type daughters of similar size (Fig. 7D, E, G and S16B). Remarkably, T1 duration was strongly reduced in nup49 KN nup60 KN double mutant daughters to levels comparable to hos3 mutants ( Fig. 7F-G). Thus, a combination of these two KN alleles mimics the effect of Hos3 inactivation in daughter cell cycle control, suggesting that constitutive acetylation of Nup60 and Nup49 is sufficient to bypass Hos3-mediated inhibition of cell cycle Start in daughter cells.

Discussion
Here, we have identified a mechanism that establishes differences in NPC acetylation and cell fate between mother and daughter cells in Saccharomyces cerevisiae. The Hos3 deacetylase, which localizes to the daughter side of the bud neck during mitosis, associates with NPCs in the daughter cell to direct its cell cycle program. This stands in stark contrast to other known fate-determinant partition mechanisms, which rely on compartmentalization and polarization of the cell cortex. The enrichment of Hos3 at the nuclear periphery in daughter cells was dependent on septin ring integrity, on the septinassociated protein Hsl7, and on passage of the nucleus through the bud neck. Although it is formally possible that local protein degradation contributes to Hos3 localization dynamics, we speculate that NPC association is triggered by the close proximity between the daughter side of the septin ring and NPCs traversing the bud neck.
Daughter-specific inheritance of Hos3 delayed cell cycle commitment of daughter cells.
Indeed, deletion of Hos3 shortened G1 duration in daughters but did not affect cell cycle progression of mother cells. Conversely, artificial recruitment of Hos3 to the NPCs of mother cells was sufficient to delay their G1/S transition (Start), indicating that Hos3 acts at the NPCs to inhibit cell cycle entry. Hos3 promotes activity of the spindle orientation checkpoint, a mechanism that prevents exit from mitosis in cells with misoriented spindles (Wang and Collins, 2014). This function is likely associated with localization of Hos3 to the daughter spindle pole. Our findings reveal that an additional, separate pool of Hos3 localizes to the NPC to control cell cycle entry. Furthermore, our data suggest that Hos3 delays Start through mechanisms independent from Ace2, which represses Cln3 transcription in daughter cells (Di Talia et al., 2009;Laabs et al., 2003), and from Rpd3, a type I HDAC that associates with the CLN2 promoter and favors its repression (Huang et al., 2009).
Our results define two key processes regulated by Hos3 to inhibit the G1/S transition.
Firstly, Hos3 is required for enrichment of the Start inhibitor Whi5 in daughter nuclei (Liu et al., 2015;Schmoller et al., 2015). Whereas Hos3 promotes higher concentration of Whi5 in daughter nuclei, it counters the accumulation of karyopherins Kap95 and Msn5 in daughters, which control nuclear levels of Whi5 (Kosugi et al., 2009;Taberner et al., 2009). Since the concentration of Whi5 and the probability to enter the cell cycle are inversely correlated, this provides a mechanism by which Hos3 can control Start in daughter cells. Secondly, we find that Hos3 confines the CLN2 gene locus to the nuclear periphery in G1 daughter cells, and that perinuclear CLN2 anchoring delays Start. Hos3 might target CLN2 to repressive perinuclear domains or to NPCs in daughter cells.
Nucleoporins can directly activate gene expression at NPCs in yeast (Casolari et al., 2004;Taddei et al., 2006) or at the nucleoplasm, in metazoans (Capelson et al., 2010b;Kalverda et al., 2010;Liang et al., 2013); however, silencing of subtelomeric genes after association with NPCs has also been reported (Feuerbach et al., 2002;Van de Vosse et al., 2013). In any case, perinuclear silencing of CLN2 would dampen the positive feedback on Cdk1, further contributing to delaying Start in daughter cells. Interestingly, acetylation by the SAGA complex plays a major role in the recruitment of chromatin to NPCs and the nuclear periphery (Denoth-Lippuner et al., 2014;Dultz et al., 2016). Whether Hos3 counteracts SAGA to control Start in daughter cells remains to be determined.
Our data indicates that Hos3 acts to repress Start by modulating the functions of nuclear pore complexes. Notably, Hos3 physically interacts with the nuclear basket component Nup60, and perinuclear Hos3 deacetylates basket and central channel nucleoporins.
Mutation of individual acetylated lysines in either Nup60 or the central channel component Nup49 perturbs Whi5 and CLN2 asymmetries, and a combination of these mutations entirely disrupts Start inhibition in daughter cells. Based on these observations, we propose that Hos3 inhibits cell cycle entry in daughter cells through deacetylation of the nuclear basket (which predominantly modulates CLN2 tethering) and of the central channel, which affects both CLN2 positioning and Whi5 partitioning (Fig. 7H). Although both Whi5 nuclear levels and CLN2 gene positioning are Hos3-dependent, these two processes are likely regulated independently of each other. Supporting this possibility, Whi5 is not required for CLN2 perinuclear anchoring in G1, and nup60 mutations perturb CLN2 anchoring but not Whi5 asymmetry. It is important to note that Hos3 may have additional substrates at the NPC that may also play a role in cell cycle regulation.
Furthermore, Hos3 may control the nucleo-cytoplasmic shuttling of other factors besides Whi5, and sub-nuclear positioning of other genes besides CLN2, which may contribute to the daughter-specific Start delay.
In summary, we reveal that asymmetric partitioning of a lysine deacetylase during mitosis modulates NPC properties in daughter cells to promote two distinct mechanisms underlying their cellular identity. These data reveal that modulation of NPC functions plays a key role in the control of cell cycle progression, and establish deacetylation as a mechanism to regulate NPCs in different cell cycle stages and cell types. The segregation mechanism of Hos3 may be specific to budding yeasts, which have a specialized type of nuclear division. However, differential expression or localization of Hos3-like HDACs might modulate NPC acetylation in other organisms in a cell cycle-or cell type-specific manner. Interestingly, the human homologues of yeast Nup60 and Nup49 (Nup153 and Nup58, respectively) are also acetylated (Choudhary et al., 2009). Moreover, an interplay between NPCs and tissue-specific class-II HDACs has been linked to the regulation of gene expression and positioning in cultured animal cells (Brown et al., 2008;Kehat et al., 2011;Zullo et al., 2012) and Nup153 is involved in ES cell pluripotency through gene silencing (Jacinto et al., 2015). Whether any of these processes is controlled by acetylation of NPCs is not known. Thus, our findings open the possibility that nucleoporin deacetylation might be a conserved mechanism to regulate NPC functions in specific cell types. It will be of interest to determine whether as in budding yeast, NPC deacetylation can affect the establishment of cell identity in multicellular organisms.

Experimental Procedures
Strains, cell growth: S. cerevisiae strains are derivatives of S288c except when noted (strains ; Table S3). Gene deletions and insertions were generated by standard one-step PCR-based methods (Janke et al., 2004) (oligos; Table S4). Site-directed mutagenesis (QuickChange, Agilent Technologies) as used to generate the HOS3pr-HOS3 EN -GFP. KN mutants were generated using CRISPR/Cas9 to replace acetylated lysines [identified in (Henriksen et al., 2012)] with asparagine, and are listed in Table S5. Positive clones were generated and verified by sequencing as described (Laughery et al., 2015). At least two independent clones per strain were analyzed, with identical results. LacO arrays were inserted using a two-step PCR based method (Rohner et al., 2008). LexA or LexA-Yif1 was expressed from pRS425-based plasmids under the control of the ADH1 promoter.
For G1 arrest, cells were grown in YPDA (yeast extract, peptone, dextrose, and adenine) to log phase, synchronized with 15 µg/ml alpha-factor (Sigma-Aldrich) for 2 h and released in fresh YPDA at 30°C. For induction of the GAL1,10 promoter, cells were grown overnight in YP+Raffinose medium, washed, shifted to YP+Galactose medium for 4 hours, and then plated on YP+Gal plates.
Fluorescence microscopy: Time-lapse imaging was performed using a spinning-disk confocal microscope (Revolution XD; Andor Technology) with a Plan Apochromat 100x, 1.45 NA objective equipped with a dual-mode electron-modifying charge-coupled device camera (iXon 897 E Dual Mode EM-CCD camera; Andor Technology), temperaturecontrolled microscopy chamber, a z-stepper and an automated stage. Images were analyzed on 2D maximum projections. For epifluorescence microscopy, a Leica AF 6000 wide-field microscope with an Andor DU-885K-CSO-#VP camera was used. Maximum projections are shown throughout, except in Fig. 1A, where one confocal section is shown for clarity. For tethering of Hos3 to nuclear pores, inducible dimerization of FK506-binding protein (FKBP) and FKBP-rapamycin binding (FRB) domain was used with C-terminally tagged proteins, as described (Gallego et al., 2013). The final strain harbors the tor1-1 mutation and lacks the endogenous FPR1 gene rendering growth insensitive to rapamycin. Cells were incubated with 25 µM rapamycin throughout the imaging experiment. Quantification of GFP fusion protein abundance was determined in background-subtracted 2D sum projections of whole-cell Z-stacks, with the nuclear area defined by Nup49-mCherry. The position of GFP-LacI spots was mapped in confocal sections by assigning the GFP spot to one of three nuclear zones of equal area, as described (Hediger et al., 2002). Analysis of G1 duration was performed essentially as described (Di Talia et al., 2007) except that cell size was measured on DIC stacks using ImageJ and a customized version of the BudJ3D plug-in (Ferrezuelo et al., 2012) to calculate cell volume using bright-field images. Growth rates were determined by linear regression of volumetric data over time. Time-lapse series of 4 µm stacks spaced 0.2-0.3 µm were acquired every 2-4 minutes.
Immunoprecipitation: Cells were grown to log phase or synchronized by Cdc20  Hos3-myc was immunoprecipitated in extracts prepared from metaphase-arrested cells (Cdc20 depletion) and from asynchronously growing cultures (log phase). Proteins were detected by western blot using the indicated antibodies. (D) Hos3-FRB-GFP recruitment to NPCs (co-localization with Nup49-FKBP-RFP) upon rapamycin addition prolongs T1 duration in mother cells (right; N=25 cells). Rapamycin (25 µM) was added immediately before imaging. Association of Hos3-GFP-FRB with nuclear pores was observed within 5 minutes of rapamycin addition. T1 in daughter cells was inhibited for the duration of the movie (3 hours) and is not plotted. Lines represent the mean. In (A-B), time is indicated in minutes; t=0 marks the last frame before nuclear entry into the bud.
Scale bars, 2 µm. Fluorescence intensity of CLN2pr-GFP in mother and daughter cells relative to cytokinesis (Myo1-mCherry disappearance; t=0) in wild type and hos3∆. Images were acquired at 3 min intervals. The GFP signal increases shortly after cytokinesis in WT and hos3∆ mother cells. Arrows indicate the interval between cytokinesis and the halfmaximum of CLN2pr-GFP intensity in daughter cells. GFP accumulates in the wild type