SWITCH 1/DYAD is a novel WINGS APART-LIKE antagonist that maintains sister chromatid cohesion in meiosis

Mitosis and meiosis both rely on cohesin, which embraces the sister chromatids and plays a crucial role for the faithful distribution of chromosomes to daughter cells. Prior to the cleavage by Separase at anaphase onset, cohesin is largely removed from chromosomes by the non-proteolytic action of WINGS APART-LIKE (WAPL), a mechanism referred to as the prophase pathway. To prevent the premature loss of sister chromatid cohesion, WAPL is inhibited in early mitosis by Sororin. However, Sororin homologs have only been found to function as WAPL inhibitors during mitosis in vertebrates and Drosophila. Here we show that SWITCH 1/DYAD defines a novel WAPL antagonist that acts in meiosis of Arabidopsis. Crucially, SWI1 becomes dispensable for sister chromatid cohesion in the absence of WAPL. Despite the lack of any sequence similarities, we found that SWI1 is regulated and functions in a similar manner as Sororin hence likely representing a case of convergent molecular evolution across the eukaryotic kingdom.

manner as Sororin hence likely representing a case of convergent molecular evolution across the eukaryotic kingdom.

Introduction
The tight regulation of sister chromatid cohesion is essential for accurate chromosome segregation during mitosis and meiosis. During S-phase, the genomic DNA is duplicated resulting in the formation of two sister chromatids per chromosomes. The newly formed sister chromatids are held together by the cohesin complex, which builds a ring-like structure embracing the chromatids.
Besides sister chromatid cohesion, the cohesin complex is crucial for genome stability, DNA repair, chromatin structure organization, and gene expression 1-4 .
The cohesin complex is highly conserved in the eukaryotic kingdom with homologs present from animals to plants comprising four core subunits: SMC1 and SMC3, two ATPases that belong to the family of structural maintenance of chromosomes (SMC) proteins, the heat-repeat domain protein SCC3/SA and one α-kleisin component RAD21/SCC1, which is replaced in meiosis by REC8/SYN1. The presence of cohesin on chromosomes is very dynamic. Cohesin is already loaded onto chromosomes by the SCC2-SCC4 loader complex during the G1 phase of the cell cycle. Sister chromatid cohesion is established in the subsequent S-phase and regulated by several cohesin accessory proteins, including the PRECOCIOUS DISSOCIATION OF SISTER 5 (PDS5) and WINGS APART-LIKE (WAPL) [5][6][7] . PDS5 assists the acetylation of the SMC3 subunit by Establishment of cohesion 1 (Eco1)/Chromosome Transmission Fidelity 7 (CTF7), needed to close the cohesin ring [8][9][10] . Cohesin is then maintained on chromosomes until late G2 in the mitotic cell cycle and early prophase I in meiosis, respectively. As cell division is approaching metaphase, cohesin, especially on chromosome arms, undergoes tremendous removal mediated by the cohesin dissociation factor WAPL, a process known as prophase pathway of cohesin removal [11][12][13][14] . At the centromeric regions, cohesin is largely protected by the Shugoshin-PP2A complex 15,16 . This centromeric cohesin is released by a Separase-dependent proteolytic cleavage of the kleisin subunit RAD21/REC8, thereby allowing the separation of sister chromatids at anaphase onset (anaphase II in meiosis).
To prevent a premature release of sister chromatid cohesion in mitosis, especially on chromosome arms, Sororin counteracts the releasing force of WAPL by binding to PDS5 and displacing WAPL from PDS5 11,[17][18][19] . However, Sororin has so far only been identified in vertebrates. More recently, an ortholog of Sororin, named Dalmatian, was found in Drosophila, which exert both Sororin's cohesin stabilizing and Shugoshin's cohesin protecting functions in mitosis 20 .
In late prophase, Sororin is recognized by the APC/C Cdh1 (Anaphasepromoting complex/cyclosome) and degraded by the ubiquitin-proteasome pathway, thereby releasing its repression of WAPL and activating the prophase removal of cohesin 18,21 . Phosphorylation through Cdk1 (cyclin-dependent kinase 1) and Aurora B kinase serves thereby as a signal for the degradation of Sororin 22,23 .
In contrast to mitosis, it is not clear how sister chromatid cohesion is protected during early meiotic prophase I. Notably, Sororin does not seem to play a role for the regulation of meiotic cohesion. Although Sororin is present in male meiosis in mouse, it is exclusively localized on the central regions of the synaptonemal complex (SC) and not on the axial/lateral elements of SC where the cohesin complex is found 24 . This localization pattern makes Sororin unlikely, at least in mouse, be the protector of cohesin. This conclusion is substantiated by the finding that the localization of Sororin in the central region of the SC is not dependent on the meiosis-specific subunits REC8 and SMC1β 24 .
In contrast, WAPL has been found to remove meiotic cohesin at late prophase in most if not all organisms studied including Arabidopsis and other plants 11,14,[25][26][27] . Thus, it remains a puzzle how the activity of WAPL is inhibited in early meiotic prophase I especially since no obvious sequence homolog of Sororin or Dalmatian has been identified in the plant lineage and other major branches of the eukaryotic kingdom 28 .
Here, we report that the previously identified SWI1 gene in Arabidopsis encodes a novel WAPL inhibitor. Despite any sequence similarities between SWI1 and Sororin, we further reveal that SWI1 antagonizes WAPL in prophase I of meiosis through a similar strategy as Sororin in mitosis. Moreover, SWI1 turned out to be amazingly similarly regulated in Arabidopsis as Sororin in vertebrates.

Removal of meiotic cohesin is mediated to large extent by WAPL
To get an understanding of cohesin dynamics during meiosis, we followed the expression and localization of a previously generated functional REC8-GFP reporter in male meiocytes by live cell imaging 29 . We observed that the majority of cohesin (~90%) in the wildtype, but not in the previously described wapl1 wapl2 double mutant 11 , is already largely released from chromatin prior to anaphase I indicating that the impact of the WAPL-dependent prophase pathway on cohesin removal is very strong in male meiosis of Arabidopsis ( Fig. 1a- Supplementary Video S1, S2).
To follow WAPL1, we generated a WAPL1-GFP reporter, which fully complemented the wapl1 wapl2 defects (Supplementary Fig. S1) and accumulated in somatic cells of the anther and in male meiocytes. In meiocytes, the WAPL1-GFP signal showed a homogeneous distribution in the nucleoplasm from pre-meiosis until leptotene, suggesting no or only a very weak interaction of WAPL1 with chromatin ( Fig. 1d i,ii). Subsequently, foci and/or short stretches of WAPL1-GFP appeared in the nucleus at late leptotene/early zygotene, coinciding with the eviction of cohesin from chromatin (Fig. 1d iii). The accumulation of WAPL1-GFP signal on chromatin became more prominent in zygotene and pachytene, which is consistent with the progressive release of cohesin (Fig. 1c, Fig. 1d iv,v). In metaphase I, WAPL1-GFP was found at condensed chromosomes (Fig.1d vi). While WAPL1-GFP signal is still present in the nucleus after the first meiotic division until tetrad stage, it was not localized to chromatin any longer ( Fig. 1d vii,viii). This localization pattern was confirmed by immuno-localization of WAPL1-GFP using an antibody against GFP ( Supplementary Fig. S1c).

SWI1 is expressed in early meiosis
The observation that WAPL1 is already present in early prophase at a time point when REC8 removal from chromatin has not started, suggested the existence of a WAPL repressor that might prevent WAPL from localizing to chromatin and unloading cohesin prematurely. However, no obvious sequence homolog of Sororin, the only known WAPL repressor in mitosis, exists in Arabidopsis 28 . We reasoned that a potential repressor of WAPL during meiosis should have all or least some of the following characteristics: First, mutants of this repressor should experience premature loss of sister chromatid cohesion and hence probably have a strong mutant phenotype in meiosis. In turn, this makes it likely that such a mutant has already been identified due to the extensive search for meiotic mutants in Arabidopsis. Second, this repressor would probably be a protein of unknown molecular function. Third, as a regulator of sister chromatid cohesion, this factor should interact with the cohesin complex and hence, its correct localization to chromatin may also depend on a functional cohesin complex.
The gene SWITCH1 (SWI1), also known as DYAD, was previously identified based on its requirement for sister chromatid cohesion in meiosis [30][31][32] . SWI1 encodes for a protein of unknown biochemical function and its mechanism of action has been unresolved up to now. However, SWI1 was previously reported to be exclusively expressed in interphase prior to meiosis and could neither be detected in leptotene nor in any subsequent meiotic stage 30,31 . This expression pattern is difficult to reconcile with the swi1 mutant phenotype, e.g., a failure to assemble the chromosome axis and to establish sister chromatid cohesion. Therefore, we revisited the expression pattern of SWI1 in both male and female meiocytes by generating a genomic reporter in which the coding region of GFP was inserted directly before the STOP codon of SWI1. Expression of this reporter in swi1 mutants could fully restore a wild-type meiotic program ( Supplementary Fig. S2). To stage the expression of SWI1, we also generated a functional reporter line for the chromosome axis protein ASYNAPTIC 3 (ASY3), where RFP was used as a fluorescent protein ( Supplementary Fig. S3).
Consistent with previous reports, SWI1 was first detected as numerous foci/short stretches in interphase nuclei of both male and female meiocytes ( Fig.   2a; Supplementary Fig. S4). In addition, the SWI1-GFP signal was present in leptotene and became even stronger as cells progressed through leptotene as staged by the migration of the nucleolus to one side of the nucleus [33][34][35] and the appearance of an ASY3 signal on condensing chromosomes (Fig. 2a, Supplementary Fig. S4). This analysis also showed that SWI1 is chromatin associated. In zygotene, when chromosomes further condensed, highlighted by ASY3-RFP, the SWI1 signal strongly declined until it was not detectable any longer in late pachytene (Fig. 2a, Supplementary Fig. S4).
To confirm that SWI1 reaches its expression peak in late leptotene and decreases by zygotene, we constructed a reporter line for ZYP1b, a component of the central element of the synaptonemal complex. Since a fusion of ZYP1b to RFP resulted in only a very weak fluorescent signal, we generated a ZYP1b-GFP fusion along with a fusion of SWI1 to RFP, which could also restore full fertility and meiotic progression of swi1 mutants ( Supplementary Fig. S5, S6). In late leptotene, the SWI1-RFP signal is strongly present on chromosomes while no signal for ZYP1b was detected (Fig. 2b). From zygotene onwards, when short stretches of ZYP1b indicate partially synapsed chromosomes, the SWI1 signal was hardly detectable, corroborating that SWI1 is largely absent from chromosomes after zygotene corresponding to the removal of REC8 (Fig. 1a).

Chromatin association of SWI1 and REC8 is mutually dependent on each other
Establishment of sister chromatid cohesion has been shown to be compromised during meiosis in swi1 and cohesin components, e.g., REC8 and SMC3, were found to be not properly bound to chromosomes in this mutant 30 . Using live cell imaging and immuno detection assays, we confirmed these cohesion defects by studying REC8-GFP in three different mutant alleles, swi1-2, swi1-3, and swi1-4, that showed identical REC-GFP localization defects (Fig. 5a, Supplementary Fig.   S7).
To address whether SWI1 localization also depends on cohesin, we introgressed the SWI1-GFP reporter into rec8 mutants. While no obvious differences were found in interphase in comparison to swi1 mutants complemented by the expression of SWI1-GFP, we found that SWI1 did not properly localize to chromatin in rec8 mutants in prophase (Fig. 2c). This failure was not due to chromatin fragmentation present in rec8 since we observed the same pattern when the SWI1 reporter was introgressed into rec8 spo11 double mutants in which the endonuclease SPORULATION DEFECTIVE 11 (SPO11) is not functional and hence no double strand breaks are formed.
However, immuno-localization experiments using an antibody against GFP corroborated that residual levels of SWI1 remain on chromatin in rec8 mutants that expressed the SWI1-GFP reporter construct. This suggested that chromatin association of SWI1 also relies on other factors in addition to the REC8-containing cohesin ( Supplementary Fig. S8a).

SWI1 interacts with PDS5 family proteins
A direct interaction of SWI1 with one of the cohesin components is a likely explanation for the observation that proper SWI1 localization is dependent on cohesin. To explore this possibility, we tested the interaction of SWI1 with all core cohesin subunits including SMC1, SMC3, REC8 and SCC3 by yeast twohybrid assays. However, SWI1 did not interact with any of these proteins ( Supplementary Fig. S9a). We further investigated the interaction of SWI1 with the cohesin accessory proteins PDS5A, one of the five PDS5 genes in Arabidopsis, and WAPL1, one of the two WAPL homologs. While we did not find an interaction of SWI1 with WAPL1, SWI1 strongly interacted with the N-terminus but not the C-terminus of PDS5A (Fig. 3a, Supplementary Fig. S9b). The interaction domain of SWI1 was then determined to reside in the N-terminal 300 amino acids as the C-terminal domain from amino acid 301-639 failed to bind to N-terminus of PDS5A (Fig. 3a). This interaction was confirmed by GST pull down assay with recombinant proteins purified from E. coli, and by bimolecular fluorescence complementation (BiFC) assay in tobacco leaves (Fig. 3b,c).
Whether SWI1 also interacts with the other four PDS5 paralogs present in Arabidopsis, was next addressed in BiFC assay. While PDS5B and PDS5D only weakly bound to SWI1, an even stronger interaction of SWI1 with PDS5C and PD55E than with PDS5A was found, indicating that SWI1 has the potential to regulate all PDS5 proteins in Arabidopsis.

SWI1 antagonizes WAPL
PDS5 has been shown to form a complex with WAPL in several vertebrates and yeast 12,13,36,37 . Correspondingly, we found that Arabidopsis WAPL1 bound to the N-but not the C-terminus of PDS5A by yeast two-hybrid and BiFC assays (Supplementary Fig. S9b and c). Thus, WAPL1 and SWI1 interact, at least broadly, with the same region of PDS5. Sororin is known to bind to PDS5 and displace WAPL from the cohesin complex 18 . To assess whether SWI1 may act similarly as Sororin by dislodging WAPL from PDS5, we first compared the binding affinity of PDS5A with SWI1 and WAPL1 by using a ratiometric BiFC (rBiFC) system 38 that allows quantification of the interaction strength. The rBiFC assay revealed that the interaction between SWI1 and PDS5A is stronger than the interaction of WAPL1 with PDS5A ( Fig. 4a,b). To further explore the relationship of these three proteins, we perform an in vitro competition experiment. To this end, we loaded recombinant WAPL1-PDS5A heterodimers co-purified from E. coli onto PDS5Abound beads and incubated them with increasing concentrations of SWI1. With increasing concentrations of SWI1, more WAPL1 protein could be released from PDS5A into the supernatant (Fig. 4c). Conversely, more SWI1 was bound to PDS5 with increasing concentrations of SWI1.
The displacement of WAPL from PDS5 by SWI1 was further confirmed by a competitive binding assay in tobacco leaf cells (Fig. 4d). While the simultaneous presence of WAPL1 tagged with mTuquiose did not affect the interaction of SWI1 with PDS5A, the co-expression of SWI1-mTurquiose resulted in a strong reduction of the BiFC signal from WAPL1-PDS5A interaction (Fig. 4d).
Thus, despite any sequence similarities, SWI1 appears to act in a similar fashion as Sororin in animals.
Therefore, we speculated that the absence of WAPL should restore the presence of REC8 on chromatin in swi1 mutants. To this end, we generated the triple mutant swi1 wapl1 wapl2 containing in addition the REC8-GFP reporter.
REC8 localization was then analyzed in male meiocytes at different meiotic stages of this triple mutant in comparison to the wildtype, swi1 and wapl1 wapl2 double mutants. In contrast to swi1 mutants ( Fig. 5a,b, Supplementary Fig. S7, Video S3), REC8 localization in swi1 wapl1 wapl2 mutants was nearly identical to the pattern found in wapl1 wapl2 double mutants, i.e., residing on chromosomes till metaphase I (Fig. 1a,b, 5c,d, Supplementary Video S4). Note that due to the failure of chromosome axis formation and of the aberrant migration of nucleolus in swi1 mutants, the meiotic stages in swi1 mutants were determined by the morphology of meiocytes in combination with the number of nuclei in tapetal cells 29,34 . The restoration of cohesion in the swi1 wapl1 wapl2 and the resemblance to the wapl1 wapl2 mutant phenotype was further confirmed by chromosome spread analysis (Fig. 5c). Since swi1 mutants do not have an obvious growth defect and since we also could not detect SWI1 outside of meiocytes, we conclude that SWI1 specifically maintains cohesion in meiosis by antagonizing WAPL. We also found that the putative SWI1 homologs from maize and rice, AMEIOTIC 1 (AM1), which likewise are required for meiotic progression and cohesion establishment 39,40 , both interacted with Arabidopsis PDS5A in a yeast two-hybrid interaction assay (Fig. 3d). Thus, it is likely that the SWI1 function as a WAPL antagonist in meiosis is conserved in flowering plants and, given the presence of SWI1 homologs in moss, possibly in all land plants.

SWI1 presence on chromatin is controlled by Cdk-cyclin activity
A crucial question is how WAPL is liberated from the inhibition by SWI1 in late prophase to mediate the release of cohesin (Fig. 1a-c). In vertebrate mitosis, this problem is solved by the phosphorylation dependent release of Sororin from chromatin. Two kinases have been observed to participate in this regulation, Cyclin-dependent kinase 1 (Cdk1) and Aurora B 22,23 . We observed that SWI1

contains 13 consensus Cdk phosphorylation sites, 12 [S/T]P and 1 [S/T]Px[R/K]
sites. We found that at least 7 of these sites can be phosphorylated in an in vitro kinase assay by CDKA;1, the Arabidopsis Cdk1/Cdk2 homolog, together with the meiotic cyclin SOLO DANCERS (SDS) (Fig. 6a, Supplementary Table S1).
To address whether the analogies between SWI1 and Sororin would extend to phospho-regulation, we introgressed the SWI1-GFP reporter, together with the ASY3-RFP reporter for staging, into weak loss-of-function alleles of cdka;1 (CDKA;1 T161D ) 41 . Similar to the wildtype, SWI1 is present on chromatin in CDKA;1 T161D plants until leptotene (Fig. 6b). However, the SWI1 signal does not decline as strongly in CDKA;1 T161D plants as in the complemented swi1 mutants.
To test whether the phosphorylation of SWI1 is essential for its release from chromosomes at late prophase I, we generated de-phospho mutant constructs. The localization pattern of SWI1 with four mutated CDK phosphorylation sites in the N-terminus of SWI1 (SWI1 4A -GFP), was indistinguishable from wildtype SWI1 protein (Fig. 6c). However, mutating all 13 or only the C-terminal nine phosphorylation sites in SWI1 (SWI1 13A -GFP and SWI1 9A -GFP), resulted in extended occupancy of SWI1 on chromosomes, reminiscent of the pattern found in CDKA;1 T161D and sds mutants (Fig. 6c,   Supplementary Fig. S8b). Note that SWI1 13A -GFP and SWI1 9A -GFP seems to be functional since the cohesion defects in early prophase I were completely rescued in swi1 mutants harboring either version ( Supplementary Fig. S10d,e; for effects in later stages of meiosis, see below).
To complement this analysis, we also generated a phosphomimic version of SWI1 in which the Serine or Threonine of all 13 CDK phosphorylation sites were mutated to the negatively charged amino acid Aspartate (SWI1 13D -GFP) and introduced this construct into wild-type plants. SWI1 13D -GFP showed the same localization pattern as the wild-type version, indicating that the phosphomimic SWI1 version is recognized by its releasing factors (Fig. 7). Moreover, we did not find any reduction in fertility of these plants ( Supplementary Fig. S11).
Taken together, these findings corroborated that mitosis in vertebrates and meiosis in plants (Arabidopsis) utilize a similar mechanism to control the presence of the WAPL inhibitors on chromatin through phosphorylation by CDKcyclin complexes. However, the observation that SWI1 was not prematurely removed from chromatin by mimicking its phosphorylation indicates that phosphorylation is necessary but not sufficient for SWI1 removal hinting at a higher order coordination of SWI1 phosphorylation and the machinery involved in controlling its stability.

Chromatin release of SWI1 is important for WAPL action
Our above presented cytological and biochemical data suggested that the timely release of SWI1 is needed for WAPL to remove cohesin. To test this in vivo, we made use of the dephospho-mutant version of SWI1 13A -GFP that complemented the early defects of swi1 mutants ( Supplementary Fig. S10d, see above). Notably, swi1 mutant harboring SWI1 13A -GFP were to a large degree infertile as seen by their short siliques and strongly reduced pollen viability (Supplementary Fig.  S10a-c,i). Since these defects precluded discerning between a dominant effect as expected from interfering with WAPL versus a partial functionality of SWI1 13A -GFP, we switched to wild-type plants harboring the SWI1 13A -GFP construct (SWI1 13A -GFP/WT) for the following analysis. While the vegetative growth of these plants was not affected, they also suffered from a drastic fertility reduction in 51 out of 55 T1 transformants similar to swi1 mutants expressing the SWI1 13A -GFP mutant version ( Supplementary Fig. S10a-c,i), indicating that it is not the lack of a functional version that causes this phenotype.
Chromosome spread analysis showed that chromosome pairing and synapsis was not altered in SWI1 13A -GFP/WT (n=88) consistent with the restoration of these defects in swi1 mutants by the same construct (Fig. 8a i). The first obvious defects were found at diakinesis. Whereas 5 clearly discernable bivilents are then present in the wildtype, chromosomes were entangled and clustered in SWI1 13A -GFP/WT (51 out of 101 meiocytes analyzed) (Fig. 8a ii,xvi).
We therefore speculated that the prolonged retention of SWI1 might result in an extended abundance of cohesin on chromatin. To address this question, plants expressing a SWI1 13A version without a fluorescent tag were generated and combined with plants harboring the REC8-GFP reporter. Based on the time-resolved quantification of REC8-GFP signal in male meiocytes, we found that in comparison to wildtype, REC8-GFP signal showed a decreased speed of removal in SWI1 13A plants (1/2 removal time, 14.66 ± 0.58h, n=3 in SWI1 13A versus 11.33 ± 1.15h, n=3 in wildtype) (Fig. 8b, Supplementary Video S5, S6). At metaphase I, instead of ~10% (n=3) REC8-GFP signal retained in the wildtype, twice the signal, i.e., ~20% (n=3) was observed in SWI1 13A plants (student's ttest P<0.0001) partially resembling the retention of REC8-GFP in wapl1 wapl2 mutants. However, it has to be noted that the level of REC8-GFP withholding in wapl1 wapl2 is higher than in SWI1 13A plants (~55% versus ~10%) (Fig. 1a,  Consistent with such a scenario is the observation that the eviction of REC8 starts apparently earlier in SWI1 13A versus the wildtype (Fig. 8b). In any case, our data strongly suggest that a vast (more than 90%) removal of REC8 is crucial for meiosis and even an increase from 10 to 20% is sufficient to cause an over cohesive effect underlining the importance of the WAPL-PDS5-SWI1 regulatory triangle.

SWI1 abundance is controlled by the APC/C
Our results show that the release of SWI1 from chromosomes is regulated by CDKA;1-mediated phosphorylation. However, the degradation pathway for SWI1 is still obscure. An analysis of SWI1 by the GPS-ARM algorithm 42 S13b), we mutated all 5 D-boxes (SWI1-Δ5D-box-GFP). Plant expressing SWI1-Δ5D-box-GFP showed an extremely prolonged abundance of SWI1 that only disappeared in tetrads, suggesting that SWI1 is targeted by the APC/C for degradation from zygotene onwards (Fig. 9a, Supplementary Fig. S10 f-i). We also observed reduced fertility of SWI1-Δ5D-box-GFP expressing plants consistent with the prolonged presence of SWI1 on chromatin. However, the reduction in fertility was less severe in plants expressing SWI1-Δ5D-box-GFP than in wapl1 wapl2 mutants or in SWI1 13A -GFP expressing plants ( Supplementary Fig. S10 f-i). Again, we cannot exclude a compromised function of SWI1-Δ5D-box-GFP due to the many point mutations introduced and, consistent with an affected functionality, we also observed that SWI1-Δ5D-box-GFP had a slightly less pronounced chromosome association than the nonmutated SWI1-GFP version (compare Fig. 9a with Fig. 2a).
To hence seek further evidence for a possible proteolytic control of SWI1, we performed a cell free degradation assay by incubating protein extracts from flower buds with the purified C-terminal half of the SWI1 protein (His-GST-SWI1 301-639 ). We found that SWI1 301-639 degradation started in mock-treated samples after 15 min of incubation time and the majority of the protein (80%) was not detectable any longer by 90 min. In contrast, SWI1 301-639 disappeared at a much slower rate in samples treated with the proteasome inhibitor MG132 and after 90 min, more than 50% of the protein was still present (Fig. 9b i,ii, 9c).
Since we found that phosphorylation is required for the release of SWI1 from chromatin, we next compared the degradation kinetics of wild-type SWI1 301-639 with the mutated SWI1 301-639/9A version. Indeed, the nonphosphorylatable version SWI1 301-639/9A was stabilized in comparison to the phosphorylatable version and showed similar turnover kinetics as MG132treated extracts (Fig. 9b iii, 9c). To further assess whether the degradation of SWI1 is mediated through the phosphorylation of SWI1 by CDKs, we treated the protein extracts with Roscovitine, a potent CDK inhibitor 43 . In comparison to the mock-treated sample, SWI1 301-639 was also stabilized under Roscovitine treatment, substantiating that CDK-dependent phosphorylation marks SWI1 for 26S proteasome-dependent degradation which relies on D-boxes and thus, is likely mediated by the APC/C.

Discussion
The precise establishment, maintenance, and removal of sister chromatid cohesion is essential for faithful chromosome segregation in both mitosis and meiosis. In contrast to the well-described mechanisms of cohesion regulation in mitosis 17,18,20 , much less is known about the control of cohesion in meiosis. Our study in Arabidopsis provides evidence that the prophase pathway of cohesion regulation exists in meiosis including the inhibition of the cohesin remodeler WAPL by a new type of inhibitor represented by the previously identified protein SWI1 that functions and is regulated in a amazingly similar fashion as Sororin in animals. Given that both animals and plants have WAPL homologs and that the lineage that led to plants and to animals were very early separated in eukaryotic evolution, much earlier than the separation of the predecessors of animals and yeast, it is likely that a basic prophase pathway of cohesin removal is very ancient and was probably present in the last common ancestor of animals and plants.
However, the repression of WAPL appears to have evolved independently in animals and plants and hence is likely younger in evolutionary terms.
Remarkably, the two independent WAPL regulators, Sororin and SWI1, target the same cohesin subunit, i.e. PDS5, and are themselves controlled by a similar mechanisms, i.e. Cdk phosphorylation. Our finding that SWI1 from Arabidopsis can also bind to PDS5 from maize and rice indicates that the function of SWI1 as a WAPL antagonist is conserved in flowering plants. Moreover, the presence of SWI1 homologs in moss gives rise to the hypothesis that SWI1 appeared very early in the plant lineage.

A model of the regulation of the prophase pathway in plant meiosis
Based on our results, we propose the following model of how SWI1 prevents the premature removal of sister chromatid cohesion in Arabidopsis (Fig. 10): SWI1 starts to be expressed and is recruited onto chromosomes by interacting with PDS5 proteins during very early meiosis, likely already during the premeiotic S phase. The recruitment of SWI1 is dependent, at least partially, on the formation of the cohesin complex (Fig. 2c). After entry into meiosis, cohesin needs to be maintained until late prophase to facilitate multiple processes of meiosis, e.g., double-strand break (DSB) repair, chromosome pairing, and homologous recombination 44,45 . The maintenance of cohesin is achieved by SWI1 that has a higher affinity towards PDS5 than WAPL, thereby displacing WAPL from PDS5, consistent with the dispersed localization of WAPL at early stages in prophase I (Fig. 1d). Given the stronger interaction strength between SWI1 and PDS5 versus WAPL and PDS5, it seems likely that both complexes do not co-exist or that at least the SWI1-PDS5 complex is much more prominent than a WAPL-PDS5 complex if all three proteins are equally present. SWI1 is released from chromatin by CDKA;1-dependent phosphorylation and subjected to degradation, likely in an APC/C-mediated manner (Fig. 10).
However, CDKA;1 phosphorylation of SWI1 does not appears to be sufficient for degradation and possibly, the degradation machinery needs to be activated as well, perhaps depending on CDKA;1 phosphorylation as well. The removal of SWI1 allows the interaction of WAPL with PDS5 as indicated by the tight chromosome association of WAPL at late prophase (Fig. 1d, Supplementary Fig.   S1c), thereby activating the prophase pathway of cohesin removal.

The complex nature of the cohesin complex in plants
The retained signal of SWI1-GFP in rec8 mutants suggests that SWI1 might also localize to other cohesin complexes that do not contain REC8. Unlike most other organisms that have only one mitotic RAD21 gene, three paralogs of the kleisin subunit, RAD21.1/SYN2, RAD21.2/SYN3, and RAD21.3/SYN4, have been identified in Arabidopsis next to the meiotic paralog REC8/SYN1. Similarly rice and other plants also have several SCC1/RAD21 genes 1 . Among the Arabidopsis RAD21 genes, especially RAD21.2 has been found to play an important role in reproduction, i.e., meiosis and subsequent gametogenesis, next to a function in vegetative growth since knockdown of RAD21.2 in meiocytes impaired chromosome synapsis and SC formation 46,47 . This together with the observation that sister chromatids are still bound at centromeres in the absence of REC8 until metaphase I indicates that at least two different kleisins contribute to sister chromatid cohesion. However, RAD21.2 was unexpectedly detected to be predominantly present in the nucleolus of meiocytes and not along chromatin 46 .
Thus, the function of this putative kleisin is still obscure and it is also not clear whether SWI1 can regulate different types cohesin complexes in meiosis.
Conversely, wapl and swi1 mutants do not show any mitotic defects raising the question whether there is a prophase pathway in mitosis in Arabidopsis.

SWI1 function beyond the control of cohesion
Although the premature removal of REC8 and with that the REC8-dependent cohesion loss are suppressed by the absence of WAPL1 and WAPL2, swi1 wapl1 wapl2 plants are still completely sterile similarly to the swi1 single mutants and much more affected than wapl1 wapl2 double mutants ( Supplementary Fig. S14).
This implies that SWI1 might have further roles in meiosis and/or might be involved in other biological processes after meiosis. Indeed, in addition to the cohesion defects, swi1 mutants are also compromised in the specification of meiocytes resulting in the formation of multiple rather than a single female meiocytes 48,49 . These defects are specific to swi1 in Arabidopsis and have not been reported for am1 mutants in maize and rice 39,40 . However, AM1 also appears to regulate the entry into meiosis and very early meiotic events.
Interestingly, we found that the number of ovules with a single female meiocyte is significantly higher in the swi1 wapl1 wapl2 mutants (14.3%, n=126) than in swi1 mutants (3.9%, n=128) (Chi-squared test, p=0.00395, Supplementary Fig. S15). Whether the loss of cohesin contributes to the formation of multiple meiocytes is not clear as yet. The germline in plants has to be established post-embryonically and there is accumulating evidence that the specification of meiocytes also involves a substantial reprogramming of chromatin possibly contributing to the repression of mitotic regulators and stem cell genes [50][51][52] . In this context it is interesting to note that the pattern of histone modifications is altered in swi1 mutants 53 . Our finding that SWI1 binds to PDS5 opens a new perspective here given that PDS5 plays a broad role in regulating plant growth and development 54 . Thus, it is tempting to speculate that PDS5 is also involved in meiocyte specification in Arabidopsis. Further work is required to explore the role of SWI1 as a regulator of PDS5 containing complexes.

Plasmid construction and plant transformation
To create the SWI1 reporters, the genomic sequence of SWI1 was amplified by PCR and inserted into pDONR221 vector by gateway BP reactions (Supplemental Table 2). The SmaI restriction site was then introduced in front of the stop codon by PCR. All constructs were then linearized by SmaI restriction and ligated to GFP or RFP fragments, followed by gateway LR reactions with the destination vector pGWB501 59 . WAPL1-GFP and ASY3-RFP reporters were created by using same strategy as described above. For the ZYP1b-GFP reporter, the AscI restriction site was inserted into pDONR221-ZYP1b between the 464-465aa of ZYP1b by PCR. Following the linearization by AscI, the mEGFP fragment was inserted into pDONR221-ZYP1b. The resulting ZYP1b-GFP expressing cassette was integrated into the destination vector pGWB501 by a gateway LR reaction.
Primers used for the creation of these constructs are shown in Supplementary

Yeast two-hybrid assay
The coding sequences of the respective genes were amplified by PCR from cDNA with primers flanking the attB recombination sites and subcloned into pDONR223 vector by gateway BP reactions (Supplementary Table S2). The resulting constructs were then integrated into the pGADT7-GW or pGBKT7-GW vectors by gateway LR reactions. Yeast two-hybrid assays were performed according to the Matchmaker Gold Yeast two-hybrid system manual (Clontech).
Different combinations of constructs were co-transformed into yeast strain AH109 using the polyethylene glycol/lithium acetate method as described in the manual of Clontech. Yeast cells harboring the relevant constructs were grown on the SD/-Leu -Trp and SD/-Leu -Trp -His plates to test for protein-protein interactions.
The resulting constructs were integrated by gateway LR reactions into pHMGWA or pHGGWA vectors for the His MBP-or HisGST-tagged fusions, respectively. For the heterologous expression, the constructs were transformed into the E. coli

Chromosome spread analysis
Chromosome spreads were performed as described previously 60 . In brief, fresh flower buds were fixed in 75% ethanol and 25% acetic acid for 48 h at 4°C, followed by two washing steps with 75% ethanol and stored in 75% ethanol at 4°C. For chromosome spreading, flower buds were digested in an enzyme solution (10mM citrate buffer containing 1.5% cellulose, 1.5% pectolyase, and 1.5% cytohelicase) for 3 h at 37°C and then transferred onto a glass slide, followed by dispersing with a bended needle. Spreading was performed on a 46°C hotplate by adding an approximately 10 μl drop of 45% acetic acid. The slide was then rinsed by ice-cold ethanol/acetic acid (3:1) and mounted in Vectashield with DAPI (Vector Laboratories) to observe the DNA.

In vitro kinase assay
CDKA;1-SDS complexes were expressed and purified as described in Harashima and Schnittger 61 with slight modification using Strep-Tactin agarose (iba) instead of Ni-NTA agarose for the purification. Briefly, the CDKA;1-SDS complexes were first purified by Strep-Tactin agarose (iba), followed by desalting with PD MiniTrap G-25 (GE Healthcare). The quality of purified kinase complexes was checked by CBB staining and immunoblotting ( Supplementary   Fig. S15). Kinase assays were performed by incubating the kinase complexes with HisMBP-SWI1 1-300 or HisMBP-SWI1 301-639 in the reaction buffer containing 50 mM Tris-HCl, pH 7.5, 10mM MgCl2, 0.5 mM ATP and 5mM DTT for 90 min. The CBB stained gel after kinase reaction is shown in Fig. S9.

Cell free degradation assay
Wild-type Arabidopsis flower buds were freshly harvested and immediately grounded into powder in liquid nitrogen. Subsequently, total soluble proteins were extracted in the degradation buffer containing containing 25 mM Tris-HCl (pH 7.5), 10 mM NaCl, 10 mM MgCl2, 4 mM PMSF, 5 mM DTT, and 10 mM ATP as previously described 62 . The supernatant was harvested after two 10-min centrifugations at 16,000 g at 4°C and protein concentration was measured using the Bio-Rad protein assay. Two hundred nanograms of recombinant proteins (HisGST-SWI1 301-639 or HisGST-SWI1 301-639/9A ) were mixed with 150 µL protein extracts (600 µg in total) containing either DMSO, 50 μm MG132, or 5 μm Roscovitine. The reactions were incubated at 22°C and protein samples were collected at the indicated intervals followed by western blot analysis.

Bimolecular Fluorescence Complementation Assay
The coding sequences of SWI1, PDS5 paralogs and WAPl1 were amplified from cDNA and subcloned into pDONR221-P3P2 or pDONR221-P1P4. The resulting constructs were subsequently integrated into the pBiFC-2in1-NN destination vector using a gateway LR reaction 38  Laboratories) were added onto the slides (1:500 dilution) followed by 2 h incubation at 37°C in a moisture chamber. After three times washing in PBST, the chromosomes were counterstained with anti-fade DAPI solution (Vector Laboratories). The images were captured using the Leica SP8 laser scanning microscopy.

Sample preparation and LC-MS/MS data acquisition
The protein mixtures were reduced with dithiothreitol, alkylated with chloroacetamide, and digested with trypsin. These digested samples were desalted using StageTips with C18 Empore disk membranes (3 M) 63  Peptides with a charge of 1, greater than 6, or with unassigned charge state were excluded from fragmentation for MS 2 ; dynamic exclusion for 20s prevented repeated selection of precursors.

Confocal microscopy and sample preparation
For protein localization analyses, young anthers or ovules were dissected and imaged using an Leica TCS SP8 inverted confocal microscope. For tracing the dynamics of cohesin in swi1 wapl1 wapl2 mutants, live cell imaging was performed as described by Prusicki et al. 29 . Briefly, one single fresh flower bud was detached from the flower and dissected with two anthers exposed.
Subsequently, the isolated bud including the pedicel and a short part of the floral stem was embedded into the Arabidopsis Apex Culture Medium (ACM) and then covered by one drop of 2% agarose. The sample was then subjected to constant image capture with 15 min intervals by using an upright Zeiss LSM880 confocal microscope with Airyscan.
For analyzing the dynamics of cohesin during meiosis, live cell imaging was performed with the anthers of the wildtype and wapl1 wapl2 mutant plants harboring a functional REC8-GFP reporter. To quantify the dynamics of the REC8-GFP signal, the metaphase I was denoted as 0 h and the REC8-GFP signal from at least 20 meiocytes was calculated for every one hour prior to metaphase I by using the image processing software Fiji 65 . Representative videos for the wildtype and wapl1 wapl2 mutants are shown in the supplementary Video S1 and S2, respectively.

Pollen viability assay
The Peterson staining method was used to analyze the pollen viability 66 .

Data availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 67 partner repository with the dataset identifier PXD009959.