Hog1 activation delays mitotic exit via phosphorylation of Net1

To study whether osmostress resulted in a delay after G2, we synchronized cells at early mitosis by means of expressing CDC20 under the control of the inducible galactose promoter (GAL1p-CDC20), which arrested cells in metaphase. The GAL1p-CDC20 expression system consists of the replacement of the CDC20 promoter by the GAL1 promoter, and thus cells arrest in metaphase in the absence of galactose and reenter into the cell cycle again in the presence of galactose. This is a well-established and accepted tool for the synchronization of cells when analyzing specific mitotic events (e.g., refs. 51 and 52). The ability of cells to progress into a new cell cycle was analyzed by flow cytometry after release in control conditions or upon osmostress (SI Appendix, Fig. S1A). Control cells progressed normally through mitosis and entered a new cell cycle 50 to 60 min after release. Remarkably, when released in the presence of 0.4 M NaCl, cells showed progression to a new G1 phase only after 90 min. Additionally, when mitosis progression was monitored by means of degradation of the mitotic cyclin Clb2, a delay in Clb2 degradation was observed in osmostressed compared with untreated cells (SI Appendix, Fig. S1B). These data indicate that osmostress induces a striking delay after release from early mitosis. However, after an initial delay, cells were able to restore cell-cycle progression (SI Appendix, Fig. S1A), suggesting that their viability is not compromised. GAL1p-CDC20 cells promoted osmoadaptation upon stress similar to wild-type (W303) cells (SI Appendix, Figs. S1C and S2C). Similar results were obtained when cells were synchronized using α-factor and released in control conditions or stressed with 0.4 M NaCl when entering mitosis (SI Appendix, Fig. S2 A and B). This is consistent with the idea that cells are able to adapt to the environmental change and survive despite being exposed to stressful conditions. Of note, cells expressing CDC20 from the MET3 promoter [where Cdc20 expression was regulated by the presence of methionine in the culture media (53)] also showed delayed progression into a new cell cycle upon stress when compared with control conditions (SI Appendix, Fig. S2D), thereby indicating that the phenotype observed upon osmostress does not depend on the synchronization method.

To verify the involvement of the MAPK pathway in the observed phenomena upon osmostress after release from early mitosis synchronized cells, we compared a GAL1p-CDC20 strain (from here on, WT) with a strain additionally bearing a temperature-sensitive SLN1 allele (sln1^ts), which leads to constitutive genetic activation of Hog1 independent of external stimuli when exposed to the restrictive temperature (54, 55). It is worth mentioning that activation of the HOG pathway by genetic manipulation has served to unravel the direct role of Hog1 in the cell cycle (4, 7, 8, 10, 15). Cell-cycle progression and Clb2 protein dynamics were monitored under these conditions. Consistent with what was observed upon salt addition, genetic activation of Hog1 promoted a delay in the progression into a new G1 phase for cells released from metaphase arrest at 37 °C (SI Appendix, Fig. S1D). This delay was also reflected in a defect in full timely degradation of Clb2 in the sln1^ts strain (SI Appendix, Fig. S1E). Remarkably, for both cell-cycle progression and Clb2 degradation and reaccumulation, HOG1 deletion in the sln1^ts background fully suppressed the delay caused by Hog1 activation (SI Appendix, Fig. S1 D and E). Taken together, these data indicate that Hog1 mediates a delay when cells are released from early mitosis.

Then, we attempted to determine which mitotic event(s) was impaired upon stress. Initially, we monitored the dynamics of nuclear division by DAPI staining of GAL1p-CDC20 cells released from metaphase arrest at the restrictive temperature (Fig. 1A). In the WT strain, most of the cells at time 0 showed a clear metaphase phenotype, with one nuclear signal localized on the mother cell. Over time, the percentage of binucleated anaphase cells progressively increased, then augmenting the percentage of late anaphase/telophase cells (separated nuclei with mother and daughter cells still attached to one another), to finally almost fully reach the typical G1 phenotype of separated cells with a concentrated nuclear signal at 40 min (Fig. 1 A, Top). sln1^ts nuclear distribution over time differed greatly from the WT. Cells progressed normally through anaphase but, although showing seemingly separated nuclei, failed to reach physical separation of mother and daughter (Fig. 1 A, Middle). Once again, sln1^ts hog1 cells mimicked the behavior of the control strain (Fig. 1 A, Bottom). These data indicate that the activation of Hog1 results in a delay in the late stages of mitosis.

A crucial step of successful mitosis is the establishment of a functional spindle for chromosome segregation. To test whether mitotic spindle stability or elongation was affected, cells carrying GAL1p-CDC20 were synchronized and released at the restrictive temperature, and tubulin was immunostained to measure spindle length in WT, sln1^ts, and sln1^ts hog1 cells (Fig. 1B). All strains were observed to show a timely enrichment in the number of cells with anaphase spindles (>2 µm), thereby indicating that Hog1 activation does not impair mitotic spindle elongation. On the other hand, the persistence of longer spindles observed in sln1^ts cells suggests a delay in spindle disassembly.

The phosphatase Cdc14 is a master regulator of progression through mitosis (reviewed in refs. 56 and 57), and its localization is modulated by interaction with the nucleolar protein Net1 (20, 21). We thus assessed whether Cdc14 localization is impaired upon Hog1 activation. GAL1p-CDC20 strains additionally bearing a enhanced GFP (yeGFP)-tagged version of endogenous Cdc14 were analyzed by fluorescence microscopy upon release from metaphase arrest at the restrictive temperature in WT, sln1^ts, and sln1^ts hog1 strains on fixed cells (Fig. 1C). The control strain displayed a rapid release of Cdc14, which reached its maximum 15 min after release from metaphase arrest. In contrast, the sln1^ts strain failed to initiate Cdc14 release as promptly. Additionally, successful relocalization of Cdc14 in this strain occurred in only ∼71% of the cells compared with ∼92% of control cells at the same time point (15 min after release). sln1^ts hog1 cells mimicked the pattern of the control strain. These observations indicate that Hog1 activation impairs the timely release of Cdc14. Given the crucial function of this phosphatase in mitotic exit, this finding could explain the delay observed in cell-cycle progression.

Upon stress-induced activation, the MAPK cascade, through its effector Hog1, is responsible for a rapid alteration in the phosphoproteome of the cell to mediate an effective response to environmental changes (58). We found that Cdc14 release is impaired upon genetic activation of Hog1 and thus tested whether this phosphatase is a direct target of Hog1. Cdc14 phosphorylation was assessed in an in vitro kinase assay using glutathione S-transferase (GST)-tagged versions of Cdc14, Hog1, and its activator Pbs2 (in its constitutively active form, Pbs2^EE) purified from Escherichia coli (Fig. 2A). Under the experimental conditions tested, Hog1 did not phosphorylate Cdc14. However, Cdc14 release has been reported to depend on its binding to the nucleolar protein Net1 and the initial release is in fact promoted by Clb2–Cdc28–dependent phosphorylation on Net1 (22, 28). We then hypothesized that Hog1 acts directly on Net1 to negatively regulate the release of Cdc14. To test for Net1 phosphorylation, we performed an in vitro kinase assay with GST-tagged Net1 (Fig. 2B). Net1 was phosphorylated by Hog1, thereby indicating that Net1 is a direct substrate of the kinase. To further explore Net1 as a substrate of Hog1, we screened the functional domains of Net1 (SI Appendix, Fig. S3A) and generated a series of GST-tagged truncations of the protein, which were tested for in vitro phosphorylation (SI Appendix, Fig. S3B). These analyses showed that the Hog1-dependent phosphorylation sites were present in the first 511 amino acids of the N terminus of the protein (a fragment we called F2). Strikingly, this region contains the Cdc14-binding domain (amino acids 1 to 341) (59) and Clb2–Cdc28 phosphorylation sites responsible for promoting Cdc14 release (22, 28). The Net1 F2 sequence contains several S/T-P sites that can be phosphorylated by Hog1. All these serine and threonine residues were mutated to alanine, alone or in combination, expressed as recombinant GST–Net1 F2, and tested for Hog1 phosphorylation in vitro as done previously (SI Appendix, Fig. S4). The combination of alanine mutations on T62 and S385 almost fully abolished the phosphorylation signal in comparison with the WT (Fig. 2C).

To validate Net1 as a target of the osmostress-induced response, we tested its phosphorylation status in vivo. GAL1p-CDC20 strains bearing tandem affinity purification (TAP)-tagged endogenous Net1 were synchronized in metaphase, cultures were divided into two groups, and one was stressed with 0.4 M NaCl for 5 min (SI Appendix, Fig. S3C). Net1 mobility shift was analyzed on Phos-tag gels in WT cells and hog1 and Net1^T62A,S385A mutants (Fig. 2D). In response to osmotic shock, the control strain showed a reduced mobility shift of Net1 signal when compared with the untreated sample. Importantly, no clear differences between the treated and untreated hog1 and Net1^T62A,S385A mutant strains were observed. This result suggests that, upon osmotic stress, Net1 is phosphorylated in vivo by Hog1 on residues T62 and S385.

Since Hog1 phosphorylates Net1 and activation of this kinase alters Cdc14 release, we envisioned that these phosphorylations could affect the stability of the Net1–Cdc14 complex. To address this question, we tested whether Hog1-dependent phosphorylation of Net1 interferes with the capacity of Clb2–Cdc28 to modify Net1 itself, thus affecting Cdc14 release. To this end, we performed a two-step in vitro kinase assay in which GST–Net1 F2 was incubated sequentially with activated Hog1 and Clb2–Cdc28 (SI Appendix, Fig. S5A). The Clb2–Cdc28 complex phosphorylated Net1 in vitro, regardless of Hog1, indicating that Hog1 does not prevent Clb2–Cdc28 phosphorylation on Net1. To rule out that the residues required for phosphorylation of Hog1 and Clb2–Cdc28 could be the same, Net1 F2 wild type and T62A, S385A were tested in an in vitro kinase assay with Clb2–Cdc28 (SI Appendix, Fig. S5B). Clb2–Cdc28 could phosphorylate the alanine mutant, suggesting that Clb2–Cdc28 and Hog1 phosphorylations on Net1 occur on different residues.

Next, the interaction between Net1 and Cdc14 was reconstituted in vitro in a multistep binding assay in order to recapitulate the contribution of Clb2–Cdc28 and Hog1 to the stability of the complex (Fig. 3 C and D and schematic representation in Fig. 3 A and B). Briefly, purified Cdc14 and Net1 F2 proteins from E. coli were allowed to interact, and the complex was retained on an affinity resin. To promote the disassembly of the complex, samples were then incubated with Clb2–Cdc28 immunoprecipitated from metaphase-arrested yeast cells. To test whether Hog1 induced stabilization of the Net1–Cdc14 complex, purified Hog1 and Pbs2^EE from E. coli were added to the reaction prior to Clb2–Cdc28. Samples were extensively washed to release all unspecifically bound proteins and analyzed by Western blot (Fig. 3C). Finally, the relative % of Net1 binding was quantified (Fig. 3D). The unspecific binding of Net1 to the resin was tested in the absence of Cdc14 (Fig. 3C, sample 1). The assay performed with unmodified Net1 F2 showed the interaction between Net1 and Cdc14 (Fig. 3 A, C, and D, sample 2). Additionally, as expected, the addition of Clb2–Cdc28 caused a marked destabilization of the complex (decreased GST–Net1 binding; Fig. 3 A, C, and D, sample 3), an observation consistent with the findings of a previous study (22). However, of note, when Hog1 was added to the reaction, the interaction between Net1 and Cdc14 appeared to be stabilized, even in the presence of Clb2–Cdc28 (Fig. 3 A, C, and D, sample 4). These results thus indicate that Hog1 is indeed involved in modulating Cdc14 release. To further confirm the critical role of Hog1 phosphorylation on Net1 residues T62 and S385, the assay was then performed using the Net1 F2^T62A,S385A mutant (Fig. 3 C and D and schematic representation in Fig. 3B). The reconstitution of the complex with Cdc14 was similar to that achieved with the wild-type Net1 (Fig. 3 B–D, sample 6 compared with sample 2), indicating that the two sites are not required to promote or sustain binding. Furthermore, the introduction of Clb2–Cdc28 resulted in the destabilization of the Net1–Cdc14 complex to a similar extent as that of wild-type Net1 (Fig. 3 B–D, sample 7 compared with sample 3). These data show that the modification of residues T62 and S385 per se does not interfere with the role of Clb2–Cdc28 in facilitating Cdc14 release from Net1. Importantly, prior addition of Hog1 to the reaction involving Net1 F2^T62A,S385A did not promote stabilization of the Net1–Cdc14 complex (Fig. 3 B–D, sample 8), thereby demonstrating that Hog1 does phosphorylate the two residues of Net1 to impair Cdc14 release. Of note, the presence of Hog1/Pbs2 alone did not alter Net1–Cdc14 complex stability (neither for the wild type nor for the alanine mutant) (SI Appendix, Fig. S5 C and D). To study whether Hog1-dependent phosphorylation of the residues T62 and S385 of Net1 is sufficient to cause the observed impairment of Cdc14 release (Fig. 1C), we monitored Cdc14 localization in sln1^ts, sln1^ts Net1^T62,S385A, and sln1^ts hog1 strains upon release from metaphase arrest at the restrictive temperature, as previously described (Fig. 3E). The sln1^ts strain showed a clear delay in Cdc14 release when compared with sln1^ts hog1. The sln1^ts Net1^T62A,S385A strain showed overall higher levels of nucleolar release compared with sln1^ts and promoted a prompt increase in Cdc14 release already at 5 min after exit from metaphase arrest, as occurred in the sln1^ts hog1 strain.

Altogether, these data indicate that Hog1-dependent phosphorylation of Net1 is required for the stabilization of Cdc14 binding and thus the prevention of Cdc14 release.

The early functions of Cdc14 include the promotion of rDNA and telomere segregation, so-called late segregating regions, as they divide after all other genomic regions have segregated (33–35, 37, 38). To test whether the segregation of these regions is affected upon activation of Hog1, we performed fluorescence time-lapse microscopy on GAL1p-CDC20 strains bearing a TetO/TetR system, where a yellow fluorescent protein (YFP)-tagged reporter could be recruited to a specific genomic locus (60, 61). rDNA segregation was monitored over time after release from metaphase arrest at the restrictive temperature on WT, sln1^ts, sln1^ts Net1^T62A,S385A, and sln1^ts hog1 strains bearing TetR-YFP tetO:450Kb-ChrXII (Fig. 4 A and B). The control strain started segregating the rDNA region at around 20 min after release, reaching ∼80% segregation at 30 min. As expected, the sln1^ts Net1^T62A,S385A and sln1^ts hog1 strains showed similar behavior. Strikingly, the sln1^ts strain started to segregate the rDNA only at 30 min after release, and full release was not reached in ∼30% of the cells. These observations thus indicate that Hog1 activation has an impact on rDNA segregation.

Telomere segregation was monitored on a similar set of strains bearing TetR-YFP tetO:1061Kb-ChrXII (Fig. 4 C and D). In this case, the dynamics of segregation were slower and less homogeneous among cells of the same population, with the control strain reaching ∼80% segregation only at 60 min. Of note, sln1^ts Net1^T62A,S385A and sln1^ts hog1 reached a similar extent of segregation. In contrast, the sln1^ts strain was unable to fully segregate the telomeric region of chromosome XII during the time course. The segregation of the telomeric region upon genetic activation of Hog1 was also assessed in cells reaching mitosis from α-factor block/release (SI Appendix, Fig. S6 A and B). Consistently, the sln1^ts strain displayed a clear defect in telomere segregation compared with sln1^ts hog1 and this defect was almost fully reversed in the sln1^ts Net1^T62A,S385A mutant strain. This observation therefore indicates that Hog1 phosphorylation of Net1 affects the segregation of this genomic region. Additionally, the mutations T62A and S385A on Net1 did not have any effect on telomere segregation in a SLN1 wild-type background (SI Appendix, Fig. S6C). Taken together, these data indicate that genetic Hog1 activation impairs the segregation of rDNA and telomeres. To measure whether these segregation defects compromise genomic integrity, we measured Rad52-YFP foci to assess the presence of recombination foci. Net1^T62A,S385A cells displayed significantly higher levels of Rad52 foci compared with Net1 wild-type cells (SI Appendix, Fig. S6D). Finally, to assess whether segregation defects could ultimately affect bud-neck structures, we monitored by time-lapse microscopy the localization of Myo1, a protein involved in bud-neck morphogenesis and cytokinesis (SI Appendix, Fig. S6 E and F). The sln1^ts strain released at the restrictive temperature displayed a clear defect in Myo1 localization compared with the WT. This delay in the actomyosin ring disassembly could be accountable for the pronounced defect, in contrast with the delay observed in Cdc14 release, in cell-cycle completion observed upon Hog1 activation.

Hog1 activation in metaphase-arrested cells induces a delay in cell-cycle progression that appears to be linked to Net1 phosphorylation on residues T62 and S385. We thus tested whether the phosphorylation of Net1 is required to maximize cell survival upon osmostress. To this end, WT, Net1^T62A,S385A, and hog1 strains were presynchronized in metaphase and released either in yeast extract peptone galactose (YPGal) or in YPGal supplemented with 1 M NaCl. Cell growth was monitored for 66 h after release in the two conditions (Fig. 5A). All strains were equally able to grow in YPGal (control), thereby indicating that neither HOG1 deletion nor T62 and S385 mutations on Net1 affected cell fitness. When grown in media containing NaCl, the WT strain adapted over time and reached normal growth. As expected, the hog1 strain did not proliferate under these conditions as it lacked the capacity to adapt to high osmolarity. Strikingly, the Net1^T62A,S385A strain showed osmosensitivity compared with the WT strain, thereby indicating that the T62 and S385 residues of Net1 are involved in osmoadaptation during mitosis. To corroborate this finding, we also tested the same set of strains for growth on solid YPGal media in the presence or absence of 0.8 M NaCl upon release from metaphase arrest (Fig. 5B). Once again, the WT strain grew in NaCl, whereas the condition was lethal for the hog1 strain. Similar to what was observed in liquid media, the Net1^{T62A, S385A} strain showed less growth than the WT cells on the NaCl plate. The reduced viability of the unphosphorylatable mutant of Net1 was also observed in asynchronous cultures (SI Appendix, Figs. S7 and S8 A and B), confirming its disadvantage for growth upon osmostress. Of note, Net1^T62A,S385A cells also displayed defective growth upon osmostress compared with a W303 wild-type background (SI Appendix, Fig. S8C). Altogether, these data show that Hog1 phosphorylation of Net1 on T62 and S385 is required for proper adaptation to high osmolarity.

All data discussed in this study are included in the main text and SI Appendix.