Identification of Rec8 Phosphorylation Sites
A tandem affinity purification (TAP) tag was used to isolate Rec8 from extracts of diploid yeast cells arrested in metaphase of meiosis I. Purified proteins were analyzed by gel electrophoresis (see Figure S1
A available online) or digested in solution with different proteases for mass spectrometric peptide identification. In addition to Rec8, we detected the cohesin subunits Smc1, Smc3, Scc3, and Pds5 (Table S1
). Consistent with previous work (Matos et al., 2008; Petronczki et al., 2006
), Rec8 was associated with the protein kinases Cdc5/PLK and Hrr25/CK1δ/
. Interestingly, Rec8 also copurified with the meiosis-specific recombination proteins Dmc1 and Hop1.
Analysis of Rec8 peptides covering 95% of the sequence revealed phosphorylation of eight serine or threonine residues. Two Ser-Ser sequences carried a phosphate group on either one of the two residues (A, blue residues; Figure S1
A). If Rec8 phosphorylation were important for its cleavage by separase, mutation of these to alanine, which cannot be phosphorylated, should block the meiosis I division. However, substitution of all 12 residues has little effect on the meiotic progression of homozygous rec8-12A
cells (data not shown). There are two possible explanations for this finding: either the phosphorylation of Rec8 is unimportant, or additional residues are phosphorylated when primary sites are mutated. To investigate the latter explanation, we mapped phosphorylation sites within Rec8-12A purified from meiotic cells. This revealed nine phosphorylated residues and one phosphate group in each of three Ser-Ser or Thr-Thr sequences (A, green residues; Figure S1
B). These additional residues were also mutated to alanine, and the resulting Rec8-24A protein was subjected to a third round of phosphosite mapping, which uncovered two more phosphorylated residues (A, orange residues; Figure S1
C). Interestingly, all 26 phosphorylation sites are located within the central region of Rec8, which is poorly, if at all, conserved among kleisins. Of these phosphorylation sites, nine were not identified in a previous analysis of Rec8 (Brar et al., 2006
Rec8 Phosphorylation Is Required for Chiasmata Resolution upon APC/C Activation
Nonphosphorylatable Rec8 Persists on Chromatin and Blocks Chiasmata Resolution
To analyze the consequences of preventing Rec8's phosphorylation, we used live imaging to observe GFP-tagged versions of Rec8 or Rec8-24A together with Pds1-RFP and the spindle pole body (SPB) component Cnm67-RFP (B). In addition, we measured protein levels by immunoblot analysis of protein extracts (Figure S2
A). Most of the Rec8-GFP disappears from chromatin at the same time as Pds1-RFP destruction and SPB segregation, after which Rec8-GFP persists exclusively as faint “centromeric” clusters associated with each spindle pole until disappearing from view as centromeres disperse around the time spindle poles separate during meiosis II. Strikingly, Rec8-24A-GFP persists throughout the nucleus long after Pds1 destruction and remains at high levels even after SPB reduplication and separation in meiosis II (B). Homozygous rec8-24A-GFP
cells separate SPBs, express proteins required to enter metaphase I, and degrade Pds1-RFP with kinetics comparable to that of REC8-GFP
cells (C; Figure S2
A). The Rec8-24A protein does not, therefore, cause a significant delay in entry into and progression through meiosis I. Rec8-24A remains at high levels beyond meiosis I also in cells lacking Sgo1, an inhibitor of cohesin removal from chromatin (Figure S2
B). We conclude that the nonphosphorylatable Rec8-24A protein resists removal from chromatin and degradation at the metaphase I-to-anaphase I transition.
To investigate the role of Rec8 phosphorylation in meiotic chromosome segregation, we imaged homozygous REC8-ha and rec8-24A-ha strains containing Pds1-RFP and a tet repressor-GFP fusion (TetR-GFP), which binds to tet operators integrated at LYS2 on the arms of both chromosome II homologs (D). After S phase, TetR-GFP bound to tetO marks all four LYS2 sister sequences, and the free fraction labels the nucleus. Recombination causes the marked LYS2 loci to coalesce into a single GFP dot during prophase I. In wild-type cells, the degradation of Pds1-RFP triggers loss of sister chromatid cohesion on chromosome arms: the LYS2-GFP dot splits into two pairs of GFP foci, which segregate into the two daughter nuclei resulting from the meiosis I division. Neither the splitting of the LYS2-GFP dot nor nuclear division occurs upon Pds1-RFP destruction in rec8-24A cells. To address whether rec8-24A hinders the resolution of chiasmata, we eliminated Spo11, the endonuclease that initiates recombination. Crucially, deletion of SPO11 restores the meiosis I, but not the meiosis II, division in rec8-24A cells (E). This also implies that sister centromeres are properly mono-orientated at meiosis I in these cells. We conclude that separase activation fails to trigger the conversion of bivalent chromosomes to dyads in rec8-24A/rec8-24A diploids.
Nonphosphorylatable Rec8 Is Resistant to Separase
If Rec8-24A blocked the meiosis I division due to its persistence on chromatin, it should prevent nuclear division in a dominant manner. To test this, we imaged Pds1-RFP and homozygous GFP-marked LYS2
loci in REC8-ha/REC8
heterozygotes (A). Despite the frequent splitting of sister LYS2
sequences upon Pds1-RFP destruction, the first meiotic division fails to take place in most (73%) rec8-24A-ha/REC8
cells. Importantly, the deletion of SPO11
restores this division (Figure S2
C), from which we conclude that Rec8-24A is a dominant inhibitor of chiasmata resolution. Surprisingly, nuclear division in meiosis II occurs with only a small delay, suggesting that a critical amount of sister chromatid cohesion may be required to resist spindle forces effectively.
Rec8 Phosphorylation Is Required for Its Cleavage
To investigate whether phosphorylation is necessary for Rec8's cleavage by separase, we used the C-terminal Myc and Ha tags to compare the abundance of Rec8 cleavage products in heterozygous REC8-myc/REC8-ha and REC8-myc/rec8-24A-ha cells (B). To facilitate detection of the short-lived cleavage products, we synchronized our meiotic cultures by using a pachytene arrest/release protocol. After transfer to sporulation medium, cells arrest in pachytene due to a deletion of the NDT80 gene. Cells are then released to synchronously progress through meiosis I by expressing NDT80 from an estradiol-inducible promoter. In REC8-myc/REC8-ha cells, the full-length proteins of both Rec8 versions start to decline, and their cleavage products accumulate (transiently) 45 min after spindle formation (B, left). Cleavage of wild-type Rec8-myc proceeds with similar kinetics in REC8-myc/rec8-24A-ha cells, but this is neither accompanied by a major decline in full-length Rec8-24A-ha protein nor by the appearance of Ha-tagged cleavage products, and 70% of the cells fail to undergo the first nuclear division (B, right). Although Rec8-24A is not cleaved by separase, it does not hinder the activation of the protease. Next, we measured the association of Ha- and Myc-tagged proteins with chromatin from anaphase I cells (C). Wild-type Rec8-ha and Rec8-myc colocalize and accumulate exclusively within pericentric chromatin surrounding each SPB. In contrast, in cells coexpressing Rec8-24A-ha and Rec8-myc, Ha-tagged protein colocalizes with the bulk of chromatin, and only Myc-tagged protein surrounds the SPBs. These data imply that Rec8-24A is neither cleaved nor removed from chromatin upon activation of separase in meiosis I. As a consequence, it inhibits meiosis I nuclear division in a dominant manner.
Finally, to demonstrate that nonphosphorylatable Rec8 is a poor substrate for separase in vitro, we incubated chromatin isolated from meiotic REC8-myc/REC8-ha and REC8-myc/rec8-24A-ha cells with extracts from mitotic cells that overproduce separase (D). Separation of these reactions into an insoluble chromatin fraction and supernatant revealed that both Rec8-myc and Rec8-ha are cleaved by wild-type separase, but not by a “catalytic-dead” version. In the presence of active separase, full-length Rec8-myc and Rec8-ha disappear from the chromatin fraction while a cleavage product appears in the supernatant. Rec8-24A, in contrast, is poorly cleaved and remains in the chromatin pellet, even when wild-type Rec8-myc is readily cleaved by separase in the same extract. These data suggest that the cleavability of Rec8 depends on its phosphorylation status rather than on any meiosis-specific regulation of separase.
Phosphomimetic Rec8 Mutants Cause Precocious Separation of Sister Centromeres
The finding that Rec8 phosphorylation promotes its cleavage suggests that cohesin's persistence at centromeres until meiosis II might be conferred by Rec8's selective dephosphorylation by PP2A at this location. If so, replacement of serines or threonines whose phosphorylation promotes cleavage by a phosphomimetic residue such as aspartate might confer phosphorylation-independent cleavage. PP2A should not protect the phosphomimetic form, which would be cleaved at centromeres at the same time as cleavage along chromosome arms, leading to precocious sister centromere separation and nondisjunction at meiosis II. To test this, we replaced with aspartate the 12 serines and threonines from our first round of phosphosite mapping plus 2 residues close to the separase cleavage sites, creating the rec8-14D allele. In addition, we created rec8-D mutants with different subsets of these substitutions (A). To analyze sister chromatid cohesion, one copy of chromosome V was marked with GFP at the URA3 locus, 35 kb from the centromere.
Phosphomimetic Rec8 Mutants Are Not Protected at Centromeres
Due to monopolin activity, sister centromeres segregate to the same pole at anaphase I in 90% of wild-type cells, a phenomenon unaltered by any of the rec8-D mutations (B). In contrast, rec8-14D and, to a lesser extent, rec8-7D-I and rec8-4D cause a noticeable increase in the frequency of anaphase I cells with separated sister URA3 sequences (B). rec8-14D, rec8-7D-I, and rec8-4D also cause large increases in the frequency of sister centromere nondisjunction at anaphase II (40%, 33%, and 24%, respectively; C). We conclude that phosphomimetic substitutions within Rec8's N-terminal half cause the precocious separation of sister centromeres. To test whether this phenotype is due to the cleavage of centromeric cohesin at meiosis I, we used immunofluorescence microscopy to detect Rec8 in metaphase II cells. Rec8 is observed in the vicinity of SPBs in most wild-type cells (98%), but only rarely in rec8-D mutants with phosphomimicking substitutions in the N terminus (rec8-14D, 2%; rec8-7D-I, 14%; rec8-4D, 28%; D).
We also analyzed the rec8-14D
allele by using live imaging. Homozygous rec8-14D-GFP
cells progress through meiosis with normal kinetics, as judged by the separation of RFP-marked SPBs and the degradation of Pds1-RFP (Figure S3
). However, the Rec8-14D-GFP protein fails to persist at centromeres after the degradation of Pds1 in meiosis I (E). Importantly, the disappearance of Rec8-14D-GFP at anaphase I is abolished by the esp1-2
mutation, which inactivates separase at 34°C (F). Rec8-14D is not, therefore, removed from chromosomes by a separase-independent mechanism (Yu and Koshland, 2005
). We conclude that phosphomimetic substitutions cause Rec8 to be cleaved by separase at centromeres as well as along chromosome arms during meiosis I.
Phosphomimetic Rec8 Restores the First Nuclear Division in Monopolin Mutants
To address whether Rec8-14D is cleaved at centromeres at the same time as along chromosome arms at meiosis I, we tested whether the aspartate substitutions suppress the inability of mam1Δ cells, which lack monopolin, to undergo the first meiotic division. Sister kinetochores are pulled to opposite poles at meiosis I in mam1Δ cells but cannot disjoin due to the resistance of centromeric cohesin to separase activity. This results in an accumulation of Pds1-negative, mononucleate cells with a single bipolar spindle (A). A failure to protect centromeric cohesin from separase, as occurs in sgo1Δ or rts1Δ mutants, or in cells that produce Scc1 in meiosis instead of Rec8, enables mam1Δ cells to divide their nuclei at meiosis I. rec8-14D has a similar effect (A). Due to chiasmata, sister centromeres segregate to opposite poles in only 76% of cases in rec8-14D mam1Δ cells (B). However, the elimination of recombination by deleting SPO11 enables almost all rec8-14D mam1Δ spo11Δ cells to disjoin sister centromeres at meiosis I (B). A corollary is that Rec8-14D is not simply defective in conferring sister centromere cohesion because if it were, efficient biorientation of sister centromeres would not be possible in rec8-14D mam1Δ spo11Δ triple mutants. Rec8-14D creates cohesion at centromeres, but it cannot persist after meiosis I separase activation.
Phosphomimetic Rec8 Supports Sister Centromere Cohesion in Metaphase I and Does Not Affect the Localization of Sgo1-PP2A
Phosphomimetic Rec8 Does Not Alter the Association of Sgo1 and PP2A with Centromeres
Substitution of serines and threonines by aspartate causes precocious cleavage of centromeric Rec8 either because it mimics the effect of phosphorylation, which is both necessary and sufficient to confer cleavability by separase, or because it somehow prevents the association of Sgo1 or PP2A with centromeres. If the latter were the case, PP2A's crucial substrate could be a protein other than Rec8. However, live imaging of Sgo1-GFP and the kinetochore protein Mtw1-RFP reveal similar levels of Sgo1 during metaphase I in REC8 and rec8-14D cells (C). Likewise, the rec8-14D allele has no detectable effect on the localization of Rts1-GFP at kinetochores. The levels of Sgo1 and Rts1 at kinetochores drop markedly as cells enter anaphase I, only to increase again at metaphase II. On chromosome spreads, however, both proteins can still be detected in the vicinity of SPBs during anaphase I (D). We conclude that the precocious cleavage of centromeric Rec8-14D at meiosis I cannot be caused by the loss of Sgo1 or PP2A from centromeres. Instead, it must be due to PP2A's inability to prevent cleavage of Rec8-14D.
The Protein Kinases Hrr25/CK1δ/
, DDK, and Cdc5/PLK Bind to Rec8
What kinases are responsible for the Rec8 phosphorylation necessary for its cleavage? In mitotic cells, Cdc5/PLK promotes cohesin cleavage by phosphorylating Rec8's mitotic counterpart Scc1 (Alexandru et al., 2001; Hornig and Uhlmann, 2004
). Because Cdc5 also regulates Rec8 phosphorylation, it has been assumed, but never demonstrated, that Cdc5 also promotes cohesin cleavage in meiosis (Brar et al., 2006; Clyne et al., 2003; Lee and Amon, 2003
). However, certain mutations allow for the efficient cleavage of Rec8 prior to Cdc5's appearance. For example, mnd2Δ ndt80Δ
cells activate the meiosis-specific APC/C-Ama1 prematurely due to the absence of the APC/C inhibitor Mnd2. This causes separase activation and Rec8 cleavage in the absence of Cdc5, whose accumulation depends on Ndt80 (Oelschlaegel et al., 2005; Penkner et al., 2005
). Phosphorylation of Rec8 by Cdc5 is not, therefore, obligatory for cleavage, and other protein kinases must be involved. Good candidates are Hrr25/CK1δ/
and DDK because these kinases are required for the normal phosphorylation of Rec8 not only in metaphase I (Matos et al., 2008; Petronczki et al., 2006
) but also during prophase I (A).
Interaction of Rec8 with the Protein Kinases Hrr25, DDK, and Cdc5
Rec8 might be expected to copurify with its kinases. Thus, we measured the abundance of the three kinases in immunoprecipitates of Rec8-ha prepared at different times after the induction of meiosis. Hrr25 and Cdc7 coprecipitate with Rec8 from early and mid-prophase I onward, respectively. Cdc5 copurifies with Rec8 only from metaphase I forward (B). Interestingly, neither Hrr25 nor Cdc7 associate with Scc1 when expressed during meiosis instead of Rec8. In contrast, Cdc5 associates with both kleisin subunits and, if anything, preferentially with Scc1 (C). Importantly, association of each kinase is independent of the activity of the others. Thus, neither inhibition of the analog-sensitive Hrr25-as kinase with 1NM-PP1 (D) nor deletion of CDC7 (only possible in bob1 mutants) (E) nor depletion of Cdc5 (F) has any effect on Rec8's association with the remaining two kinases. There is, therefore, no evidence that the activity of earlier kinases “primes” Rec8 to associate with later ones.
DDK and Hrr25 Are Required for Rec8 Cleavage in mnd2Δ Mutants
To investigate whether Hrr25 and/or DDK promotes cohesin cleavage during prophase I, we filmed Rec8-GFP and URA3 sister sequences marked with RFP in ndt80Δ cells in the presence or absence of Mnd2. In ndt80Δ cells containing Mnd2, Rec8-GFP persists on chromosomes and sisters remain tightly associated (A, left). In the absence of Mnd2, Rec8-GFP disappears from chromosomes ~2 hr after its accumulation, and this is accompanied by sister separation (A, right). Crucially, both Rec8's disappearance and sister separation are greatly delayed in ndt80Δ mnd2Δ cells homozygous for rec8-24A-GFP (B, right), confirming that Rec8 cleavage is promoted by phosphorylation under these conditions. The levels of Rec8-24A-GFP do eventually decline, possibly due to residual phosphorylation and the persistence of separase activity.
Hrr25 and DDK Promote Cohesin Destruction upon APC/C Activation in Prophase I
To reduce DDK activity, ndt80Δ mnd2Δ
cells containing the temperature-sensitive cdc7-4
allele were shifted to 31°C. Under these conditions, DNA replication proceeds normally, whereas recombination and monopolar attachment are defective (Matos et al., 2008
). Reduced DDK activity causes a modest delay in the removal of Rec8 (C, second panel). Rec8 removal is similarly delayed in bob1
mutants lacking the entire CDC7
gene (Figure S4
A). To inhibit Hrr25, we treated ndt80Δ mnd2Δ
cells containing hrr25-as
with 1NM-PP1. This also causes a moderate delay in Rec8's disappearance (C, third panel). Remarkably, simultaneous inhibition of both kinases has a far greater effect, blocking Rec8's disappearance for several hours (C, fourth panel). We also detected Rec8 by immunoblotting in protein extracts from ndt80Δ mnd2Δ
cells, whose UBR1
gene had been deleted to stabilize Rec8's C-terminal cleavage product (Buonomo et al., 2000
) (Figure S4
B). Consistent with our imaging results, full-length Rec8 remains at high levels, and production of the cleavage product is diminished only when both kinases are inhibited or phosphosites are mutated to alanine. These data suggest that DDK and Hrr25 both promote the cleavage-dependent removal of Rec8 from chromosomes, at least when separase is activated prematurely during prophase I in ndt80Δ mnd2Δ
mutants. Accordingly, Rec8 on chromatin isolated from prophase I-arrested cdc7-4 hrr25-as
double mutants is a very poor substrate for separase in vitro (E).
If the kinases exerted their effect by phosphorylating Rec8, then the delayed removal of Rec8 should be abrogated by replacing REC8
with the rec8-14D
allele. This was partly the case. The Rec8-14D-GFP protein disappears more rapidly than Rec8-GFP when Cdc7 and Hrr25 are inhibited, either separately or together, although not as rapidly as in cells with active Cdc7 and Hrr25 kinases (D). The aspartate substitutions also increase the cleavage of Rec8 on chromatin from cdc7-4 hrr25-as
cells by separase in vitro (E). Because Cdc5 is not expressed in ndt80Δ
mutants (Clyne et al., 2003
), it cannot have any role in facilitating Rec8 cleavage in these cells. Even when expressed ectopically during prophase I from the DMC1
promoter, Cdc5 fails to accelerate Rec8's removal in ndt80Δ mnd2Δ
cells with or without Hrr25 activity (Figures S4
C and S4D).
DDK and Hrr25, but Not Cdc5, Are Required for Rec8 Cleavage at Anaphase I
Are DDK and Hrr25 also important for Rec8's removal from chromosome arms at the onset of anaphase I? To address this, we filmed cells containing Rec8-GFP and Pds1-RFP. Inactivation of either Cdc7 or Hrr25 alone has little effect on the kinetics of Rec8's disappearance (A, top right and bottom left). In contrast, simultaneous inhibition of both kinases causes Rec8 to persist on chromosome arms for several hours after Pds1 destruction (A, bottom right), even in cells lacking Sgo1 (Figure S5
A). Due to bioriented sister centromeres, cdc7-4
single mutants fail to undergo the first meiotic division unless protection of centromeric cohesin from separase is abrogated (Matos et al., 2008; Petronczki et al., 2006
). In contrast, most (74%) cdc7-4 hrr25-as
double mutant cells fail to divide their nuclei at meiosis I even in the absence of Sgo1, indicating a strong delay in Rec8 cleavage (Figure S5
B). Importantly, the Rec8-14D protein disappears upon Pds1 degradation in the double kinase mutant cells (B). This implies that the persistence of wild-type Rec8 is due to a reduction in its phosphorylation. Our data suggest that phosphorylation of Rec8 by either DDK or Hrr25 promotes cohesin cleavage on chromosome arms. In the absence of both kinases, Rec8 is much more resistant to separase. Cells that produce Scc1 during meiosis instead of Rec8 undergo nuclear division after inhibition of both DDK and Hrr25, demonstrating that Rec8's dependence on these kinases for its cleavage is not a general property of α-kleisins (Figure S5
C). Hrr25's activity in monopolar attachment depends on its binding to the Mam1 subunit of monopolin, which is dispensable for Rec8 cleavage. To confirm that Hrr25 promotes Rec8 cleavage independently of monopolin, we analyzed hrr25-zo
strains, in which Hrr25 cannot bind to Mam1 (Petronczki et al., 2006
). As expected, Rec8 disappears with normal kinetics in both hrr25-zo
single and hrr25-zo cdc7-4
double mutants (Figure S5
Rec8 Phosphorylation by Hrr25 and DDK, but Not Cdc5, Is Required for Cleavage at Anaphase I
Our results imply that Cdc5 cannot alone promote Rec8 cleavage at the onset of anaphase I. This does not exclude an auxiliary role. The fact that Cdc5 is required for Pds1 degradation in meiosis I (Clyne et al., 2003; Lee and Amon, 2003
) has so far precluded any rigorous analysis of its function in cohesin cleavage. However, we have discovered that elimination of the meiosis-specific APC/C activator Ama1 renders Pds1 degradation independent of Cdc5 (Figure S6
A). This enabled us to analyze the effects of Cdc5 depletion on processes that normally depend on Pds1 degradation. In ama1Δ
mutants, Rec8 is cleaved with similar kinetics as in wild-type cells (C, top left). Inactivation of either Hrr25 or Cdc7 has little if any effect, and simultaneous inhibition of both kinases blocks Rec8 cleavage, as observed in Ama1-containing cells (C, top right, middle left, and middle right, respectively). Interestingly, Cdc5 depletion does not retard Rec8's removal from chromosomes upon Pds1 degradation in ama1Δ
cells (C, bottom left), even when Hrr25 is also inactivated (C, bottom right). This experiment also revealed that little or no Rec8 persists at centromeres after Pds1 destruction in ama1Δ
cells lacking Cdc5 (C, bottom left). We confirmed this surprising observation by detecting Rec8 on chromosome spreads (Figure S6
B). These data suggest that Cdc5 has no direct role in promoting Rec8 cleavage. Indeed, it seems to play a quite different role, namely, helping to protect centromeric Rec8 from separase. In conclusion, we propose that cohesin cleavage at meiosis I requires the phosphorylation of Rec8 by DDK and Hrr25. By removing these phosphate groups, PP2A bound to Sgo1 protects centromeric cohesin from separase (). Because phosphomimicking mutations in Rec8's N terminus stimulate cleavage several hundred residues away, separase might recognize phosphorylated Rec8 with its large noncatalytic domain rather than its C-terminal protease domain (Figures S7
A and S7B).
Model for the Control of Rec8 Cleavage by Rec8 Kinases, Sgo1-PP2A, and Separase