The conserved Cdc14 phosphatase ensures effective resolution of meiotic recombination intermediates

Meiotic defects derived from incorrect DNA repair during gametogenesis can lead to mutations, aneuploidies and infertility. Effective and coordinated resolution of meiotic recombination intermediates is necessary to accomplish both rounds of successful chromosome segregation. Cdc14 is an evolutionarily conserved dual-specificity phosphatase required for mitotic exit and meiotic progression. Mutations that inactivate the phosphatase lead to meiotic failure. Here, we have identified previously undescribed roles of Cdc14 in ensuring correct meiotic recombination. We found that recombination intermediates accumulate during prophase I when Cdc14 is depleted. Furthermore, Cdc14 plays a role in correct homolog disjunction at the end of anaphase I, both by modulating the timely removal of arm-cohesion between sister chromatids and by promoting elimination of SPO11dependent entanglements. We also demonstrate that Cdc14 is required for correct sister chromatid segregation during the second meiotic division, independent of centromeric cohesion but dependent on the correct reduplication of SPBs during meiosis II, and on the activity of the Holliday Junction resolvase Yen1/GEN1. Timely activation of Yen1/GEN1 in anaphase I and II is impaired in the meiosis defective allele, cdc143HA. Based on these new findings, we propose previously undescribed functions of Cdc14 in the regulation of meiotic recombination; roles that are independent of sister chromatid cohesion, spindle dynamics and the metabolism of gamete morphogenesis. 2 Author Summary Meiotic recombination is fundamental for sexual reproduction, with efficient and orchestrated resolution of recombination intermediates critical for correct chromosome segregation. Homologous recombination is initiated by the introduction of programmed DNA Double-Strand Breakds (DSBs) followed by the formation of complex branched DNA intermediates, including double Holliday Junctions (dHJs). These recombination intermediates are eventually repaired into crossover or noncrossover products. In some cases, unresolved recombination intermediates, or toxic repair products, might persist until the metaphase to anaphase transition, requiring a set of late-acting repair enzymes to process them. Unrestrained activity of these enzymes, however, is equally detrimental for genome integrity, thus several layers of regulation tightly control them. For example, in budding yeast meiosis, Yen1/GEN1 is mainly activated during the second meiotic division, although how it is activated is unknown. Here, we have identified that the phosphatase Cdc14 is required during meiotic divisions for timely nuclear localization and activation of Yen1 in budding yeast meiosis. Additionally, we have been able to identify previously undescribed roles of Cdc14 in controlling meiotic recombination. Strikingly, we found that levels of recombination intermediates increase during prophase I in cdc14 meiotic deficient cells, indicating that Cdc14 plays a direct role in monitoring meiotic DSB repair, possibly in Yen1-independent manner. Resolution of recombination intermediates in the absence of Cdc14 is dependent on SGS1 and MUS81/MMS4, otherwise accumulating different types of aberrant recombination intermediates and a highly reduced efficiency in CO formation. Deficient resolution of JMs in cdc14 meiotic cells, together with difficulties in SPB reduplication, likely contribute to the missegregation problems observed during the second meiotic division. Introduction Meiotic recombination is initiated by the conserved Spo11 transesterase, which introduces numerous DNA Double-Strand Breaks (DSBs) into the genome [1]. Association of single strand DNA binding proteins, including RPA, Rad51 and the meiosis specific recombinase, Dmc1, with the resected DSB ends promotes strand invasion into the intact homologous non-sister chromatid template, which culminates in the formation of a displacement loop (D-loop). Nascent D-loops can be processed through the Synthesis Dependent Strand Annealing (SDSA) repair pathway to generate non-crossover (NCO) repair products [2]. Alternatively, stabilization of the D-loop followed by second-end capture gives rise to double Holliday Junction (dHJ) intermediates. In budding yeast meiosis, dHJs are most frequently resolved into crossovers (COs) through the combined action of Mlh1-Mlh3 (MutLγ) and Exo1 [3, 4]. A second class of COs arises from the resolution of recombination intermediates via the Structure 3 Selective Endonuclease (SSE) Mus81/Mms4 [5]. In addition, multiple rounds of strand exchange over a stabilized D-loop, can give rise to three and four interconnected duplexes, also known as multichromatid joint molecules (mc-JMs) [6]. Further, both 3’-ends might also participate in multiinvasion events on two or three chromatids, creating heavily branched DNA structures [7]. These complex DNA species must be efficiently processed to prevent the risk of becoming potential hazards for the genome integrity due to their capacity to form aberrant crossovers, or other forms of toxic repair products. Resolution of all types of JMs requires the orchestrated action of a set of helicases and endonucleases, capable of displacing multi-branched DNA assemblies. The Bloom’s syndrome helicase (BLM/Sgs1), in addition to be involved in SDSA repair, plays a prominent role in resolving aberrant “off-pathway” JMs, including mc-JMs [8-12]. It is less clear which enzymatic complexes take over the function of removing any unresolved branched DNA intermediates once meiotic cells leave the pachytene stage. Upon activation of NDT80-dependent transcription, and activation of the budding yeast polo-like kinase Cdc5, cells abandon the pachytene stage concomitantly with a surge of enzymatic activity that culminates in the resolution of ZMMdependent, and independent, dHJs to give rise to COs and some NCO products [13, 14]. Cdc5 phosphorylation is required to stimulate the function of a set of SSEs, most notoriously Mus81-Mms4 [15, 16]. The Mus81-Mms4 complex acts on branched DNA substrates that have not been resolved by the canonical ZMM-dependent Class-I CO pathway [4, 10]. On the other hand, Yen1GEN1, a conserved member of the Rad2/XPG family of SSEs, appears to act much later in meiosis, despite the fact that Yen1 is considered a prototypical HJ resolvase [17]. In meiotic cells, this discrepancy is easily explained since the activity of Yen1 is negatively regulated by CDK-dependent phosphorylation, preventing the activation of the nuclease during prophase I [15]. Abolition of CDK-phosphorylation in Yen1, turns on the protein’s enzymatic activity prematurely and promotes its nuclear localization. Therefore, counteracting the phosphorylation of Yen1 generates the fully active form of the nuclease [15, 18]. In mitotic cells, dephosphorylation of Yen1 is carried out by the Cdc14 phosphatase [19, 20]. Activation of Yen1 during anaphase by Cdc14 allows the resolution of persistent repair intermediates that otherwise impose a physical impediment to chromosome segregation. It is currently unknown whether Cdc14 also dephosphorylates Yen1 in meiosis. Furthermore, it is unclear whether Yen1 might be briefly activated following anaphase I onset, or whether it is limited to anaphase II exclusively [4, 15]. Cdc14 is a well-conserved dual-specificity phosphatase, and it has been defined as a key component in the reversal of CDK phosphorylation during the completion of mitosis [21]. In budding yeast, Cdc14 activity is essential and cells lacking this phosphatase will remain arrested at the end of mitosis [22, 23]. Additionally, Cdc14 regulates transcriptional silencing at the rDNA, and other repeated sequences [24-26]. Several pathways regulate Cdc14 localization and activity. The RENT complex, required for transcription inhibition at the rDNA intergenic spacers (IGS), retains Cdc14 at the nucleolus until anaphase through its interaction with the Cif1/Net1 protein [27, 28]. Cdc14 is released from the nucleolus by the sequential action of two additional pathways. The FEAR (Cdc Fourteen Early 4 Anaphase Release) pathway promotes the early release of Cdc14 through the phosphorylation of Cif1/Net1 and Cdc14 [29-34]. Late in anaphase, the Mitotic Exit Network (MEN) keeps Cdc14 in its released state, allowing the de-phosphorylation of additional substrates, important for the full termination of mitosis and cytokinesis [35, 36]. In addition to curtailing CDK activity at the end of mitosis, Cdc14 plays critical roles promoting chromosome segregation [37]. In S. cerevisiae, it is required for the correct segregation of the ribosomal gene array (rDNA) and telomeres [38-40]. Lack of transcriptional silencing at these regions prevents the loading of condensins at the rDNA, and subtelomeric regions [26, 38, 41], which leads to unresolved linkages, intertwined sister chromatids, and non-disjunction during mid-anaphase. Notably, a number of functions have been allocated to Cdc14 before its anaphase-triggered release. In particular, Cdc14 has been involved in the completion of late-replicating regions in the rDNA locus, and other parts of the genome [42]. Additionally, anaphase-independent release of Cdc14 was observed upon induction of DNA damage, where generation of DSBs promoted the transitory release of Cdc14 from the nucleolus, targeting the SPB component Spc110 for dephosphorylation [43]. Cdc14 is also critically required for the completion of meiosis [44]. In budding yeast, Cdc14 is released from the nucleolus in two waves; the first during anaphase I and the second in anaphase II [45-47]. Contrary to mitotic cells, Cdc14 release from the nucleolus requires the essential function of the FEAR pathway whereas it appears that the MEN pathway is dispensable [48-50]. Additionally, some components of the MEN pathway, such as Cdc15, have functionally differentiated to fulfil a requirement in spore morphogenesis [48, 49]. Meiotic depletion of the MEN component Cdc15 leads to progression through both meiotic divisions but impairs spore formation. This meiotic function of Cdc15 appears to be independent of Cdc14 [49]. Curiously, inactivation of the FEAR pathway in meiosis allows exclusively one round of nuclear division to take place, culminating with the formation of asci carrying two diploid spores (dyads) [45, 46, 48, 51, 52]. Premature activation of FEAR blocks spindle assembly during meiosis, a process normally averted by PP2ACdc55 [53-55]. On the other hand, inactivation of Cdc14 function by means of FEAR mutants, or by employing conditional temperature sensitive alleles of the phosphatase, impairs the second round of chromosome segregation [45, 46]. Furthermore, absence of the phosphatase creates chromosome entanglements, a problem that can be reverted by deleting SPO11 [46]. Occasionally, the second meiotic division takes place over a single meiosis II spindle in the absence of Cdc14 and subsequently, these cells carry out aberrant chromosome segregation [54]. These interesting observations prompted the identification of another critical function of Cdc14 at the meiosis I to meiosis II transition, which is the licensing of SPB re-duplication/half-bridge separation. Asymmetrical enrichment of Cdc14 on a single SPB during anaphase I promotes SPB re-duplication, otherwise preventing the initial step of morphogenesis of new SPBs [56]. Cdc14 is highly conserved and orthologs have been identified in other organisms, including fission yeast, nematodes, insects, and vertebrates [57]. The human genome contains three Cdc14 paralogs, 5 hCDC14A, hCDC14B and hCDC14C [58, 59]. Depletion of Cdc14A, required for mitotic CDK inactivation, leads to defects in centrosome duplication, mitosis and cytokinesis [60]. In vertebrates, CDC14B exits the nucleolus in response to DNA damage [61, 62], a process conserved in fission yeast [63], and more recently also observed in budding yeast [43]. Thus, in addition to playing important roles in the DNA Damage Response (DDR) [61] it appears that both Cdc14A and Cdc14B might be required for efficient repair of damaged DNA [62]. Here, we describe various previously unknown roles of Cdc14 in budding yeast. First, we reveal its direct involvement in ensuring correct meiotic recombination by suppressing the accumulation of recombination intermediates during prophase I. We also show that Cdc14 functions in resolving a subset of aberrant JMs, presumably by regulating other substrates during, or concomitantly with, the completion of prophase I. Finally, we demonstrate that Cdc14 is required for the timely activation of the conserved HJ resolvase Yen1/GEN1 during both meiotic divisions. Results cdc143HA is a novel meiosis-specific separation-of-function allele of CDC14 During the course of unrelated studies, we generated the cdc143HA allele carrying an endogenous Cterminal tagged version of the CDC14 gene with three copies of the hemagglutinin epitope (3HA). Unlike other temperature sensitive alleles of cdc14 used in many studies (see introduction), mitotic growth was not affected in diploid cells carrying the cdc143HA allele when grown at different temperatures on solid or liquid media (Fig 1A). Thus, we conclude that cdc143HA cells behave like CDC14 at all temperatures tested under standard unchallenged growth conditions. Unexpectedly, when homozygous SK1 background diploid cdc143HA cells were induced to sporulate in liquid media, very few asci that contained spores were observed. In contrast, CDC14 control strains analyzed in parallel showed extensive tetrad formation (asci containing four spores) (Fig 1B). Following an extended incubation time in sporulation media, cells with untagged and tagged Cdc14 were sampled at 24, 48, 72, or 96 hours, fixed and analyzed under the microscope. Over 90% of cells in the CDC14 diploid completed sporulation after 24 hours (Fig 1B). Conversely, cdc143HA strains displayed as few as 10% of asci containing one or two spores at all times examined (Fig 1B). To confirm the observed defect in liquid media during sporogenesis of cdc143HA cells, we further induced sporulation in solid media and assessed the presence of N,N-bisformyldityrosine in the sporulated cultures. Dityrosine is a component that comprises approximately 50% of the mass of the outer spore wall layer [64]. Dityrosine is an autofluorescent molecule that possesses an excitation and emission wavelength of 325nm and 400nm respectively. Thus, when sporulated cultures are exposed under the UV325nm light, spores emit visible light that can be detected using standard photographic methods (Fig 1C). CDC14 and cdc143HA cells were plated on supplemented sporulation media and they were incubated for at least three days. After the incubation period, plates were directly illuminated with


Introduction
Meiotic recombination is initiated by the conserved Spo11 transesterase, which introduces numerous DNA Double-Strand Breaks (DSBs) into the genome [1]. Association of single strand DNA binding proteins, including RPA, Rad51 and the meiosis specific recombinase, Dmc1, with the resected DSB ends promotes strand invasion into the intact homologous non-sister chromatid template, which culminates in the formation of a displacement loop (D-loop). Nascent D-loops can be processed through the Synthesis Dependent Strand Annealing (SDSA) repair pathway to generate non-crossover (NCO) repair products [2]. Alternatively, stabilization of the D-loop followed by second-end capture gives rise to double Holliday Junction (dHJ) intermediates. In budding yeast meiosis, dHJs are most frequently resolved into crossovers (COs) through the combined action of Mlh1-Mlh3 (MutLγ) and Exo1 [3,4].
A second class of COs arises from the resolution of recombination intermediates via the Structure Selective Endonuclease (SSE) Mus81/Mms4 [5]. In addition, multiple rounds of strand exchange over a stabilized D-loop, can give rise to three and four interconnected duplexes, also known as multichromatid joint molecules (mc-JMs) [6]. Further, both 3'-ends might also participate in multiinvasion events on two or three chromatids, creating heavily branched DNA structures [7]. These complex DNA species must be efficiently processed to prevent the risk of becoming potential hazards for the genome integrity due to their capacity to form aberrant crossovers, or other forms of toxic repair products. Resolution of all types of JMs requires the orchestrated action of a set of helicases and endonucleases, capable of displacing multi-branched DNA assemblies. The Bloom's syndrome helicase (BLM/Sgs1), in addition to be involved in SDSA repair, plays a prominent role in resolving aberrant "off-pathway" JMs, including mc-JMs [8][9][10][11][12].
It is less clear which enzymatic complexes take over the function of removing any unresolved branched DNA intermediates once meiotic cells leave the pachytene stage. Upon activation of NDT80-dependent transcription, and activation of the budding yeast polo-like kinase Cdc5, cells abandon the pachytene stage concomitantly with a surge of enzymatic activity that culminates in the resolution of ZMMdependent, and independent, dHJs to give rise to COs and some NCO products [13,14]. Cdc5 phosphorylation is required to stimulate the function of a set of SSEs, most notoriously Mus81-Mms4 [15,16]. The Mus81-Mms4 complex acts on branched DNA substrates that have not been resolved by the canonical ZMM-dependent Class-I CO pathway [4,10]. On the other hand, Yen1 GEN1 , a conserved member of the Rad2/XPG family of SSEs, appears to act much later in meiosis, despite the fact that Yen1 is considered a prototypical HJ resolvase [17]. In meiotic cells, this discrepancy is easily explained since the activity of Yen1 is negatively regulated by CDK-dependent phosphorylation, preventing the activation of the nuclease during prophase I [15]. Abolition of CDK-phosphorylation in Yen1, turns on the protein's enzymatic activity prematurely and promotes its nuclear localization. Therefore, counteracting the phosphorylation of Yen1 generates the fully active form of the nuclease [15,18]. In mitotic cells, dephosphorylation of Yen1 is carried out by the Cdc14 phosphatase [19,20]. Activation of Yen1 during anaphase by Cdc14 allows the resolution of persistent repair intermediates that otherwise impose a physical impediment to chromosome segregation. It is currently unknown whether Cdc14 also dephosphorylates Yen1 in meiosis. Furthermore, it is unclear whether Yen1 might be briefly activated following anaphase I onset, or whether it is limited to anaphase II exclusively [4,15].
Cdc14 is a well-conserved dual-specificity phosphatase, and it has been defined as a key component in the reversal of CDK phosphorylation during the completion of mitosis [21]. In budding yeast, Cdc14 activity is essential and cells lacking this phosphatase will remain arrested at the end of mitosis [22,23]. Additionally, Cdc14 regulates transcriptional silencing at the rDNA, and other repeated sequences [24][25][26]. Several pathways regulate Cdc14 localization and activity. The RENT complex, required for transcription inhibition at the rDNA intergenic spacers (IGS), retains Cdc14 at the nucleolus until anaphase through its interaction with the Cif1/Net1 protein [27,28]. Cdc14 is released from the nucleolus by the sequential action of two additional pathways. The FEAR (Cdc Fourteen Early hCDC14A, hCDC14B and hCDC14C [58,59]. Depletion of Cdc14A, required for mitotic CDK inactivation, leads to defects in centrosome duplication, mitosis and cytokinesis [60]. In vertebrates, CDC14B exits the nucleolus in response to DNA damage [61,62], a process conserved in fission yeast [63], and more recently also observed in budding yeast [43]. Thus, in addition to playing important roles in the DNA Damage Response (DDR) [61] it appears that both Cdc14A and Cdc14B might be required for efficient repair of damaged DNA [62].
Here, we describe various previously unknown roles of Cdc14 in budding yeast. First, we reveal its direct involvement in ensuring correct meiotic recombination by suppressing the accumulation of recombination intermediates during prophase I. We also show that Cdc14 functions in resolving a subset of aberrant JMs, presumably by regulating other substrates during, or concomitantly with, the completion of prophase I. Finally, we demonstrate that Cdc14 is required for the timely activation of the conserved HJ resolvase Yen1/GEN1 during both meiotic divisions.

cdc14 3HA is a novel meiosis-specific separation-of-function allele of CDC14
During the course of unrelated studies, we generated the cdc14 3HA allele carrying an endogenous Cterminal tagged version of the CDC14 gene with three copies of the hemagglutinin epitope (3HA).
Unlike other temperature sensitive alleles of cdc14 used in many studies (see introduction), mitotic growth was not affected in diploid cells carrying the cdc14 3HA allele when grown at different temperatures on solid or liquid media ( Fig 1A). Thus, we conclude that cdc14 3HA cells behave like CDC14 at all temperatures tested under standard unchallenged growth conditions. Unexpectedly, when homozygous SK1 background diploid cdc14 3HA cells were induced to sporulate in liquid media, very few asci that contained spores were observed. In contrast, CDC14 control strains analyzed in parallel showed extensive tetrad formation (asci containing four spores) ( Fig 1B). Following an extended incubation time in sporulation media, cells with untagged and tagged Cdc14 were sampled at 24, 48, 72, or 96 hours, fixed and analyzed under the microscope. Over 90% of cells in the CDC14 diploid completed sporulation after 24 hours ( Fig 1B). Conversely, cdc14 3HA strains displayed as few as 10% of asci containing one or two spores at all times examined (Fig 1B).
To confirm the observed defect in liquid media during sporogenesis of cdc14 3HA cells, we further induced sporulation in solid media and assessed the presence of N,N-bisformyldityrosine in the sporulated cultures. Dityrosine is a component that comprises approximately 50% of the mass of the outer spore wall layer [64]. Dityrosine is an autofluorescent molecule that possesses an excitation and emission wavelength of 325nm and 400nm respectively. Thus, when sporulated cultures are exposed under the UV 325nm light, spores emit visible light that can be detected using standard photographic methods (Fig 1C). CDC14 and cdc14 3HA cells were plated on supplemented sporulation media and they were incubated for at least three days. After the incubation period, plates were directly illuminated with both white light, and UV light and they were photographed (Fig 1C, left and right panels respectively).
As anticipated, wild-type diploid strains emitted autofluorescence when exposed to UV light. On the other hand, we could not observe any UV-induced autofluorescence from cdc14 3HA diploid colonies ( Fig 1C and S2B Fig). All together, we confirmed that cdc14 3HA diploid strains cannot efficiently complete sporulation despite their apparently normal mitotic growth.

Bulk DNA synthesis is not altered in cdc14 3HA mutant cells
To gain greater insight into why cdc14 3HA mutants fail to form spores, we followed different stagespecific markers to characterize the mutant phenotype in detail. First, we sought to find whether the initiation of the meiotic program was altered by studying meiotic DNA replication kinetics. During the mitotic cell cycle, released Cdc14 has been shown to de-phosphorylate replication factors, such as Sld2 and Dpb2 [65]. In addition, the lack of Cdc14 activity has been linked to problems during DNA replication termination [42]. Thus, we considered the possibility that affecting the regulatory circuit involving the Cdc14 phosphatase could cause bulk DNA replication problems during meiotic S-phase.
FACS analysis of DNA content revealed that S-phase progression occurred with nearly identical kinetics in both control and cdc14 3HA strains ( Fig 1D). This result suggests that initiation of meiotic Sphase is not altered in cdc14 3HA cells.

cdc14 3HA cells initiate both meiotic divisions
In most organisms, meiosis is characterized by a single round of DNA replication followed by two rounds of nuclear divisions [66], thus we examined whether cdc14 3HA cells were capable of separating the nuclear content during both meiotic divisions. To achieve this, synchronous cultures of CDC14 and cdc14 3HA cells were cultured in parallel and samples were collected and fixed every two hours for at least 10-12 hours and up to 24 hours after induction of meiosis. Fixed samples were stained with 4',6diamidino-2-phenylindole (DAPI) and counted under the fluorescence microscope for nuclear division.
Results showed that CDC14 cells initiated nuclear divisions around five hours after transfer to sporulation medium and most cells displayed more than one DAPI-stained nucleus by ten hours ( Fig   1E). On the other hand, cdc14 3HA cells displayed subtly slower kinetics of nuclear divisions as well as an overall lower amount of cells exiting the one-nucleus stage by ten hours. Prolonged time-courses showed that, eventually, most cdc14 3HA cells exited the one nuclei stage, although with a ~60 min delay compared to CDC14 control cells (Fig 1E; S1 Fig).
Previous work has shown that most cells in temperature sensitive mutants of cdc14 complete meiosis with two nuclei, due to the inability to assemble two fully functional meiosis II spindles [44,46,54,67]. Furthermore, in the very few cells that continue to the second round of chromosome segregation in cdc14 ts mutants, MII spindles do not assemble normally because of problems in SPB reduplication/half-bridge separation at the meiosis I-to-meiosis II transition [56]. SPB half-bridge separation likely requires Cdc14-mediated de-phosphorylation of the SPB component Sfi1 during anaphase I [68]. However, we noticed that meiotic nuclear divisions went past the two-nuclei stage in a high proportion of cdc14 3HA cells (S1A Fig), distinguishing this cdc14 3HA allele from those temperature sensitive alleles described in earlier studies.
Next, we wanted to test whether the phenotype described for cdc14 3HA could be reproduced under other forms of meiosis-specific inactivation/depletion. To achieve this, we replaced the endogenous promoter of CDC14 with the CLB2 promoter, which is actively repressed in meiotic cells [69]. Cells expressing CDC14 under the control of the CLB2 promoter (cdc14-md) recapitulated the sporulation defect of cdc14 3HA (Fig 4F). Meiotic nuclear divisions took place in the cdc14-md but most cells remained binucleated following extended periods in meiosis (S1A Fig). Thus, the terminal meiotic phenotype of cdc14-md appears to be slightly different from that of cdc14 3HA .

Cdc14 protein levels are depleted during meiosis in cdc14 3HA cells
To discern the cause of the cdc14 3HA phenotype, we analyzed Cdc14 3HA protein levels during the entire length of meiotic time-courses. CDC14 and cdc14 3HA cells were synchronized and induced to sporulate in parallel. Samples were collected every 2 hours and protein extracts obtained from each sample.
Western blot analysis showed that cdc14 3HA cells accumulate reduced levels of the tagged protein, even at time zero, prior to meiotic induction. Levels of Cdc14 3HA protein become untraceable after 6-8 hours into meiosis (Fig 2A-B). To directly ascertain if Cdc14 protein levels varied from mitotically dividing cells to meiotic cells, in both the control and cdc14 3HA mutant strains, we prepared protein extracts from exponentially growing cultures. We also included the cdc14-md strain in our assay, to compare protein levels for all meiosis-specific mutants tested. The overall levels of Cdc14 3HA were lower in mitotic and meiotic dividing cells when compared to the wild type. On the other hand, Cdc14 levels in the mitotically dividing cdc14-md allele were similar to the wild type, and were higher than cdc14 3HA protein levels when growing exponentially ( Fig 2C). Strikingly, although levels of Cdc14 were almost invariant for the wild-type strain in mitotic or meiotic cells (Exp. vs Mei0), both cdc14 3HA and cdc14md showed a drastic depletion with meiosis onset (Fig 2C). Finally, we compared protein levels of cdc14 3HA with those from the cdc14-md strains (Fig 2D). We found that depletion of Cdc14 was more effective in the cdc14-md allele, while in cdc14 3HA it occurred more gradually. This is not related to the ability to progress into meiosis, since all strains underwent meiotic S-phase with similar kinetics ( Fig   2D; bottom panels). Currently, it is unknown to us what triggers the selective protein degradation process of the Cdc14 3HA phosphatase. We exclude the nature of the HA tag itself, as Smk1-3HA protein was included as internal control in our western blots, and proteins levels were stable, and even increased during mid to late meiosis (Fig 2A). Increasing the copy number of different cdc14 alleles, as well as overproduction of the Cdc14 3HA protein restored sporulation (Fig 1F and S2 Fig). These results confirm that reduction of phosphatase protein levels, but not its full depletion, is causing the meiotic phenotype in the cdc14 3HA cells.

cdc14 3HA cells undergo two rounds of SPB duplication and separation before degenerating
Following the identification of the meiotic cdc14 3HA phenotype, we proceeded to study at the cellular and molecular level what might be preventing cdc14 3HA cells from completing sporulation when the protein levels of the phosphatase are reduced. As mentioned above, most cdc14 3HA cells eventually passed the two-nuclei stage observed as terminal meiotic phenotype in cdc14-md (this study) and other cdc14 temperature sensitive mutants [44,46,67]. Thus, we monitored the kinetics and morphology of the meiotic SPBs and spindles during the first and the second divisions. Both CDC14 and cdc14 3HA cells, carrying GFP-tubulin and Spc29-CFP to visualize spindles and SPBs respectively, were induced to undergo synchronous meiosis. Samples were then collected and studied under the microscope at times where MI, MII, and later stages in the sporogenic process take place [70][71][72]. In addition to the kinetics of nuclear divisions, the number of SPBs per individual cell were counted. Cells with two or more Spc29-CFP foci were considered to have initiated one or both meiotic divisions ( Fig 2E). Furthermore, nuclear segregation was monitored simultaneously by looking at the separation of SPB-associated DAPI masses ( Fig 2F). We observed that by eight hours into meiosis, CDC14 and cdc14 3HA cultures contained around 75% and 60% of cells with more than one SPB, respectively. Among the population of cells that showed multiple SPBs, 56% of wild-type cells presented more than two Spc29 foci; therefore, they had initiated meiosis II. However, only 40% of cdc14 3HA cells displayed more than two Spc29 foci, confirming a slight reduction of cells forming meiosis II spindles carrying four SPBs. This observation correlates with the slightly lower frequency of cdc14 3HA cells containing more than two DAPI masses ( Fig 2F; S1A Fig). As expected, after 24 hours of meiotic induction, wild-type CDC14 cells showed over 90% of cells with three or four SPBs. In marked contrast, the cdc14 3HA meiotic mutant possessed more than two thirds of cells lacking a single Spc29-CFP focus, suggesting that in the absence of Cdc14 the structural integrity of SPBs becomes compromised once meiotic divisions take place (

Meiotic spindle dynamics are not altered in the cdc14 3HA mutant
Since we found that SPB re-duplication and separation was not fully impeded in the cdc14 3HA mutant, along with the fact that we could detect cells past the second nuclear division, we investigated whether the observed SPBs assembled functional spindles at both MI and MII. In addition, we assessed if the kinetics of both chromosomal divisions were altered in the mutant. To test that, wild-type and cdc14 3HA strains carrying Tub1-GFP and Spc29-CFP were induced to undergo meiosis synchronously and at appropriate times, they were transferred to an incubator chamber for image acquisition using live-cell microscopy ( Fig 3A-B; S1 and S2 Video). To analyze the first meiotic division, the time between metaphase I and the end of telophase I was recorded. Thus, for CDC14 cells, the MI spindle extensionto-breakage period lasted around 53 min under our experimental conditions. Cells from the cdc14 3HA culture completed MI in 58 min, on average, which was not statistically different from control cells ( Fig   3C). Next, we estimated the length of MII by timing the period between metaphase II spindles and the lengthening of the microtubule bundles until the tubulin signal became diffuse (Fig 3B). The average length of MII divisions was 80 min and 86 min for CDC14 and cdc14 3HA cells, respectively ( Fig 3C).
Taken together, we conclude that the detected reduction in Cdc14 3HA abundance throughout meiosis affected neither the lifespan, nor dynamic, of meiosis I and meiosis II spindles. Rec8-GFP signal (Fig 3D-F). 30 minutes after anaphase I onset, 50% of wild-type cells still displayed pericentromeric signal and the remaining 50% were lacking Rec8-GFP signal completely. By 60 minutes past anaphase I, >85% of cells displayed no Rec8-GFP signal and less than 15% still displayed pericentromeric Rec8 (Fig 3F). On the other hand, cdc14 3HA cells displayed analogous stepwise loss of cohesins as the wild type did, however, ~40% of mutant cells displayed pericentromeric Rec8-GFP signal, which was ~35% less than in wild type, whereas ~60% of cdc14 3HA cells displayed diffused Rec8-GFP signal still associated with the chromatin. After 30 minutes into anaphase I, ~50% of mutant cells displayed pericentromeric Rec8 signal, a comparable frequency to the observed in the wild type.
The remaining 50% of cells showed no detectable Rec8-GFP signal, matching the frequencies displayed by the wild type at the same time point. After 60 minutes following anaphase I, ~25% of mutant cells displayed pericentromeric Rec8, which was moderately higher than for the wild type ( Fig 3F).
Altogether, cdc14 3HA cells are able of undergoing stepwise cohesin loss at key transitions during meiosis, although with slightly reduced efficiency in comparison to the wild type.

cdc14 3HA mutants missegregate sister chromatids during meiosis II
While monitoring late meiotic events in cdc14 3HA mutants, we observed a high frequency of cells with more than four nuclei (S1B and S3C Figs). This might be due to impediments during chromosome segregation in the first and/or the second division. To test this possibility, we analyzed the nuclear morphology, along with the meiotic spindle, in individual cells undergoing the first and second divisions ( Fig 4A-B). Cells presenting a meiosis I (MI) spindles with a length of ≥5μm were considered to be in late anaphase I, and therefore, two DAPI masses should be distinguishable. On the other hand, cells presenting DAPI-stained bridges at the end of anaphase I might reflect those cells with some form of chromosome entanglement [46].
Wild-type cells showed clear separation of nuclear content in 74% of the anaphases analyzed, and only ~25% of anaphase I cells presented DAPI-stained bridges (Fig 4A). By contrast, cdc14 3HA cells had only ~20% fully resolved two nuclei after anaphase I. ~80% of cdc14 3HA mutant anaphase cells presented some kind of chromosomal bridge, suggesting that despite having undergone MI spindle elongation correctly, some form of DNA entwining was not properly resolved. Similar results were observed for the cdc14-md allele, where ~80% of anaphase I cells showed entanglements ( Fig 4A).
The observed anaphase I bridges could be due to transitory entanglement of lagging, under-condensed chromosomal domains, that could be eventually resolved before the initiation of the second meiotic division. Furthermore, a subtle delay in cohesin removal was observed in cdc14 3HA anaphase I cells ( Fig   3F), which might be responsible, for at least a fraction of these bridged DAPI masses. To test if the anaphase I bridges were eventually eliminated or persisted until the second meiotic division, we looked at metaphase II cells and scored the configuration of their nuclear masses ( Fig 4B). The wild type presented over 76% of metaphase II cells with single individualized nuclei located within each MII spindle. By contrast, 75% of metaphase II cells in the cdc14 3HA mutant showed DNA threads connecting both nuclear masses. Strikingly, we were unable to detect metaphase II cells with 4-SPBs for the cdc14md mutant, thus we could not perform such analysis in this background. The severity of the SPB reduplication/half-bridge separation phenotype in this mutant seems to be more drastic than it is observed for the cdc14 3HA mutant allele.
We conclude that diminished levels of Cdc14 in meiosis results in the abnormal formation of nuclear bridges after the first meiotic division and that these bridges persist at least until metaphase II.
To determine whether the DNA bridges present at metaphase II in cdc14 mutant cells are the consequence or the cause of other problems arising during chromosome segregation, we constructed strains carrying fluorescent markers on specific chromosomes and followed their segregation [73]. We made use of tetO repeats integrated at the interstitial lys2 locus of chromosome II, together with the presence of the tetR repressor gene fused with GFP, which associates to these repeats [74,75]. A second strain with a different GFP reporter was made, but in this case, it carried lacO repeats in a centromereproximal region at the trp1 locus of chromosome IV, with a lacI repressor gene fused to GFP integrated at the ura3 locus [76]. First, we analyzed strains carrying LYS2-GFP on both homologs to determine chromosome segregation fidelity in cells that had completed both meiotic divisions. In this way, we were able to visualize four individual dots segregating to each of the four nuclei within a single tetranucleated cell. The results showed that chromosome II segregated correctly in 95% of wild-type cells. Notably, only 78% of cdc14 3HA tetranucleated cells presented the correct GFP distribution, whereas the remaining 22% of mutant cells displayed, at least, one of the four nuclei lacking a GFP focus ( Fig 4C; S4 Fig). To understand the origin of chromosome missegregation in the cdc14 3HA cells we further analyzed the first and the second division separately. In order to determine correct homolog disjunction, we first analyzed binucleated cells using strains carrying the CENIV-GFP tag in homozygosity (S5A Fig). We found that the cdc14 3HA mutation did not enhance missegregation of homologs during the first meiotic division (Fig 4D). Very similar results were previously reported for homozygous CENV-GFP in cdc14-1 mutants [54]. Next, we analyzed sister chromatid segregation in the second meiotic division. Thus, heterozygous strains for the CENIV-GFP marker were used; 79% of wild-type cells showed correct sister separation after meiosis II. In contrast, 50% of tetranucleated cdc14 3HA cells displayed problems segregating sister centromeres ( Fig 4E). Taken all together, we conclude that accurate sister chromatid segregation in meiosis requires unperturbed Cdc14 activity.

cdc14 3HA
We considered that the increase in sister chromatid missegregation after the second division in cdc14 3HA cells could be in part originated from aberrant recombination causing entanglements between sister chromatids. If this is the case, we should be able to alleviate some of the sporulation problems associated to the cdc14 3HA mutation by eliminating meiotic recombination. To test this hypothesis we combined the cdc14 3HA allele with the spo11 mutation, which prevents DSB formation [1]. The absence of DSBs improved the separation of DAPI mases in anaphase I in the cdc14 3HA mutant ( Fig 4A). Therefore, recombination leads to the formation of anaphase I bridges in cdc14 3HA strains. Intriguingly, sister missegregation after meiosis II could not be tested in the spo11 cdc14 3HA double mutant because the SPB re-duplication defect, previously described, was highly aggravated [56].
On the other hand, the sporulation defect was partially rescued in all combinations of spo11 with cdc14 meiotic-defective mutants tested ( Fig 4F).

Meiotic recombination is impaired in cdc14 3HA mutants
Elimination of Spo11-dependent DSBs robustly alleviates problems with homologous chromosome segregation in FEAR mutants or in a cdc14-1 background [46]. We have confirmed this rescue by introducing spo11 deficiency in cdc14 3HA and cdc14-md mutants (Fig 4A and 4F). Thus, it seems that Cdc14 could play a role in promoting accurate repair of Spo11-dependent DSBs. To gain greater insight into the function of Cdc14 during meiotic recombination, we analyzed the well-studied HIS4::LEU2 (HIS4-LEU2) meiotic recombination hotspot (Fig 5A; [77][78][79][80]) in the presence of the cdc14 3HA and cdc14-md mutations. First, we analyzed total levels of DSBs, CO and NCOs products (Fig 5B-D).
Levels of DSBs and COs were very similar in the three strains. In contrast, cdc14 3HA and cdc14-md mutants showed a small reduction in NCO levels in comparison to the wild-type strain (Fig 5C-D).
Furthermore, we noticed a slower turnover of DNA species of higher molecular weight, corresponding with branched recombination intermediates [6] (Fig 5A and 5E). To distinguish whether the accumulation of these intermediates was due to slower turnover or increased formation of those branched species, we studied their abundance in the ndt80Δ mutant background, where resolution of JMs is largely prevented due to the lack of CDC5-dependent JM-resolution promoting activity [14].
Surprisingly, levels of high molecular weight species accumulated twice as much in the cdc14 3HA mutant over the wild type (Fig 5F and 5H). This unbalanced JM accumulation was accompanied by moderately higher levels of DSBs in the cdc14 3HA mutant. In contrast, no differences in NCO levels was observed in the ndt80Δ background between the two strains ( Fig 5G-H). Thus, CDC14 appears to counteract the accumulation of a population of high molecular-weight intermediates of recombination during, or before, prophase I.
Next, we wondered how the higher levels of JMs originated in the mutant. For that, we examined earlier recombination intermediates by 2D-gel analysis using the same hotspot ( Fig 6A and S6 Fig). Different DNA species were quantified at two different time points from synchronous meiotic ndt80Δ cultures.
Results confirmed that the total levels of JMs were increased in the cdc14 3HA and the cdc14-md mutants, with respect to the wild-type strain ( We wanted to monitor if the resolution stage of DSB repair could also be affected in cdc14 mutants, and not just the initial accumulation of recombination intermediates. Since cdc14 mutants already (ES) addition to the media [14], allowed us to follow the efficiency of recombination intermediates resolution into CO and NCO products, in both mus81Δ and mus81Δ cdc14 3HA mutants (Fig 6D-E).
Induced expression of CDC5 in the mus81Δ mutant efficiently led to the formation of COs, only one hour after ES addition to the medium. CO levels did not increase further after the 8-hour time-point.
NCOs, on the other hand, steadily increased reaching their maximum at 17 hours in SPM+ES (24h). In contrast, fewer COs were detected one hour after CDC5 induction in the presence of cdc14 3HA mutation, and final levels did not increase at the end of the 24-hour time-course. NCOs levels were also lower in comparison to the mus81Δ mutant background alone, and total levels did not rise over the period when CDC5 was expressed ( Fig 6E). Thus, it appears that when comparable levels of JMs accumulate, Cdc14 could be required for efficient formation of COs and NCOs, at least when meiotic DSB-repair takes place in a mus81Δ mutant background. To confirm that Cdc14 plays a relevant role in the late resolution of repair intermediates, we decided to study the efficiency of repair in a NDT80 background. To this end, we took advantage of the well-known negative effect on DSB-repair when depleting Sgs1 in combination with mutations in one or more of the SSEs like mms4/mus81 and yen1/slx1/slx4 [4,10].
Strikingly, the cdc14-md sgs1-md mms4 triple mutant displayed very high levels of unrepaired JMs after 24 hours in meiosis in comparison to the sgs1-md mms4 double mutant alone (Fig 6F). Total JMs in cdc14-md sgs1-md mms4 accumulated to analogous levels as in the sgs1-md mms4 yen1 triple mutant ( Fig 6G). Furthermore, the reduced formation of COs in the sgs1-md mms4 cdc14 strain was also nearly identical to the one observed for the triple mutant sgs1-md mms4 yen1 (Fig 6H-I).
Thus, Cdc14 is required for proper processing of recombination intermediates. These results also suggest that the chromosome non-disjunction events observed in FEAR, cdc14-1, and cdc14-md mutants ([45, 46]; this study), as well as the sister chromatid missegregation in cdc14 3HA mutants, may be a consequence, at least in part, of the abnormal accumulation and processing of Spo11-dependent JMs in the absence of Cdc14 activity.

Yen1 nuclear localization and activity requires full levels of Cdc14 activity
The Cdc14 requirement for the correct processing of meiotic JMs led us to search for suitable candidates that could be regulated by the phosphatase in meiosis. We have determined that the absence of Cdc14 leads to similar defects in JM processing and repair that the absence of Yen1 (Fig. 6F-I). Additionally, Yen1 has been described as a target of Cdc14 in mitotic cells [19,20,81]. Thus, to determine whether Yen1 could be a target of Cdc14 during meiosis, we first tested the subcellular localization of Yen1, which is known to change in a phosphorylation-dependent and cell-cycle stage-specific manner, both in mitotic and meiotic cells [15]. In mitotic cells, Yen1 localization is cytoplasmic when CDK activity is high, but it enters the nucleus upon anaphase onset. Cdc14 release in anaphase reverts CDKdependent phosphorylation of the endonuclease, allowing the un-phosphorylated Yen1 to enter the nucleus at this stage [19,20,81]. We tested Yen1 localization once cells initiated anaphase I and found that nuclear Yen1 was more infrequent in cdc14 3HA mutants than it was in wild type ( Fig 7A). This result indicates that Yen1 nuclear localization is reduced in the presence of cdc14 3HA mutation. We also tested if the activity of the nuclease changed during a meiotic time-course in the cdc14 3HA mutant ( Fig   7B-C). Samples were collected every two hours from synchronized meiotic cultures of wild-type and cdc14 3HA cells, each of them carrying a functional tagged version of the YEN1 gene fused to 9 copies of the MYC epitope [18]. From each time-point, Yen1-9Myc was immunoprecipitated from the meiotic extracts and beads carrying the protein were incubated with a synthetic Holliday Junction substrate [82].
Active Yen1 should be able to cleave the HJ substrate whereas the phosphorylated inactive form of the nuclease would not. In wild-type diploid cells, Yen1-9myc did not display significant nuclease activity during early stages of the meiotic time course, prior to the accumulation of Cdc5, except for the residual endonuclease activity observed at t=0 hours ( Fig 7B). Approximately 2 hr after the accumulation of Cdc5, we detected robust HJ processing, consistent with activation of Yen1 at late stages of meiosis [15]). Notably, the resolvase activity displayed by Yen1 in cdc14 3HA cultures was extremely low for the duration of the time course (Fig 7B-C). Meiotic S-phase was completed with similar timing in wild-type and cdc14 cells suggesting that both strains were transiting at similar kinetics through the time course (Fig 7D). Moreover, the Cdc5 protein (marking the exit from prophase I) was detected at the same time and at the same levels in both wild-type and cdc14 cultures (Fig. 7B). Thus, these results strongly suggest that Cdc14 promotes nuclear accumulation of Yen1 during anaphase I and is required for its enzymatic activity at late stages of meiosis.

Constitutively active Yen1 supresses the accumulation of aberrant recombination intermediates in cdc14 3H A mutants
Next, we sought to eliminate the requirement of Cdc14 phosphatase for Yen1 activation by directly using a phosphorylation-resistant version of the nuclease (YEN1 ON ). Mutation of nine of the CDK consensus sites of phosphorylation in Yen1 renders the protein constitutively active [18,20]. We combined the cdc14 3HA allele with YEN1 ON and tested for sporulation efficiency. Similar to the elimination of Spo11-dependent DSBs, unrestrained activity of Yen1 partly rescued the sporulation defect of cdc14 3HA (Fig 7E). Since enhanced sporulation efficiency might not necessarily reflect improved chromosome segregation (Fig. 4), we directly tested the effect of YEN1 ON in chromosome segregation during both meiotic divisions ( Fig 7F). First, we checked if the presence of Yen1 ON would help resolving DAPI-stained DNA bridges observed at late anaphase I in cdc14 3HA mutants. Indeed, a two-fold improvement over cdc14 3HA single mutants was observed. Furthermore, missegregation of sister chromatids was fully rescued to the levels of YEN1 ON single mutant in the cdc14 3HA YEN1 ON double mutant in comparison to the cdc14 3HA single mutant ( Fig 7F). Moreover, to confirm if the improvement in chromosome segregation by YEN1 ON was due to the restoration of Yen1's nuclease activity, we tested this possibility using a resolution assay in a cdc14 3HA mutant background (Fig 7G).
Unlike Yen1 WT , Yen1 ON displayed potent resolvase activity in the absence of Cdc14 activity.
Altogether, by eliminating the requirement for Cdc14 in de-phosphorylating Yen1, we partly rescue the chromosome missegregation defect and the lack of resolvase activity observed in cdc14 3HA mutants, confirming that Yen1 activation by Cdc14 is important for late resolution of toxic recombination intermediates.

Discussion
Meiotic recombination is fundamental for sexual reproduction, ensuring that homologous chromosomes segregate correctly at the first meiotic division as well as to facilitate distinct allele combinations that sustain evolution. Homologous recombination is initiated by the introduction of programmed DSBs followed by homolog pairing and DNA strand exchange. Those early steps in the recombination process lead to the formation of stable JMs, which are ultimately resolved into two main classes of HR repair products, known as COs and NCOs (Fig 8). Recombining chromosomes may also contain intermediates consisting of three-and four-armed DNA structures such as mc-JMs, where three and four DNA duplexes stay connected (Fig 8; [8,9,79,[83][84][85][86][87]). In some cases, unresolved recombination intermediates might persist until the metaphase to anaphase transition where a subset of late-acting nucleases process them in order to safeguard genome integrity [88]. Unrestrained activity of such nucleases can interfere with the correct allotment of COs and NCOs in meiosis [18], thus, they are tightly controlled by several layers of regulation. Posttranslational modifications carried out principally by key cell cycle kinases, like CDK, DDK, and PLK1, are the most extended form of regulation [15,16]. Additionally, at least one of the late acting nucleases, Yen1, has been described to be modulated through the highly conserved Cdc14 phosphatase in mitotic cells [19,20,81]. Given the essentiality of CDC14 in S. cerevisiae, most meiotic studies performed to date have required the use of conditional temperature sensitive alleles of cdc14 that cannot complete meiosis following the first round of chromosome segregation [22,46,56,67]. Despite the large amount of valuable data that has been collected over the past years using these alleles, other less conspicuous functionalities of Cdc14, particularly those affecting meiosis II, might have been precluded from being discovered when studying meiosis in cdc14 ts mutants. Here, using two different meiotic-deficient alleles of CDC14 we have been able to identify previously undetected functions of the phosphatase during meiosis, particularly those affecting the second meiotic division (Fig 8). Furthermore, we found the unprecedented requirement of

Novel insights into Cdc14 functionalities by using an epitope-tagged version of the phosphatase
In the present work we describe an allele of CDC14 (cdc14 3HA ) that displays no obvious defects during unchallenged mitotic divisions; nonetheless, it is strongly deficient in sporulation during meiosis.
cdc14 3HA is a C-terminal tagged version of CDC14 with three copies of hemagglutinin, which has been used in other studies to monitor its behaviour during mitosis [27, 33,36] and meiosis [50]. Protein levels of the phosphatase are severely reduced in homozygous cdc14 3HA meiotic cells (Fig 2A-D), a problem also perceived during the mitotic cell cycle, although not as dramatic. No such problems have been reported when 3xHA is fused at the N-terminal end of Cdc14 ([46]; this study), or when cdc14 3HA is present in heterozygosity together with the wild-type CDC14 allele [89], making this phenotype exclusive for the C-terminally tagged version of the gene. cdc14 3HA mutant cells complete meiotic DNA replication and both nuclear divisions with kinetics similar to those of the wild type, suggesting that both rounds of chromosome segregation take place in cdc14 3HA meiotic cells. Nonetheless, the cdc14 3HA mutant fails to form spores once cells completed both nuclear divisions. We exclude that the process which initiates spore formation was affected since we visualized pro-spore membrane components and found them to appear with similar kinetics in the mutant strain as they did in the wild type (S7 Fig). Other groups have also described similar defects in sporulation in the absence of Cdc14 activity, or components of the MEN pathway [49,90]. Critical for our findings was to detect correct SPB reduplication and separation in a high proportion of cdc14 3HA cells, but not in the more restrictive cdc14md strain (Fig 2E-F; S3 Fig). Furthermore, many cdc14 3HA cells were able to assemble functional tetrapolar spindles at metaphase/anaphase II confirming that problems arising in the cdc14 3HA mutant somewhat differ from those in the cdc14-md as well as in the widely employed, thermo-sensitive alleles, cdc14-1, -3 or other FEAR mutants [46, 54,56]. One possibility is that differences in the mechanism by which Cdc14 activity is depleted during meiosis might be responsible for the observed variations in the meiotic behaviour. Even so, most of the described phenotypical features regarding the recombination-dependent missegregation are reproduced for cdc14 3HA (Fig 4), so our allele recapitulates those impaired functionalities linked to the recombination defect previously observed. Notwithstanding, meiotic divisions occur in most cdc14 3HA cells with no discernible delay in spindle dynamics when compared to wild-type cells (Fig 3A-C). A very subtle delay, observed during both divisions, could in part arise from difficulties in SPB re-duplication and half-bridge separation, as described in cdc14-1 cells [56]. However, it is worth mentioning that the frequencies of cdc14 3HA cells presenting anaphase I bridges is consistently higher than those cells displaying persistent Rec8 signal during anaphase I (Figs 3F and   4A), thus it is unlikely that inefficient Rec8 removal is the only source of most DNA bridges, but instead a contributing factor. This possibility is still compatible with a SPO11-dependent formation of entanglements. Further experiments will need to be performed to unveil the complex nature of these DAPI structures appearing in late anaphase I in certain mutants. Strikingly, levels of sister chromatids missegregation, but not of homologs, are substantial once cdc14 3HA cells entered the second meiotic division, suggesting that residual unresolved recombination intermediates could affect sister chromatid disjunction more acutely than that of homologs. These results again highlight the importance of safeguarding correct genome repair during the entire length of the meiotic cell cycle. Non-disjunction at either first or second division could likewise bring severe consequences to the correct ploidy of the developing gametes.

CDC14-dependent resolution of recombination intermediates via multiple mechanisms
The appearance of most types of JMs should occur from zygotene until the end of pachytene stage [80], far earlier than the first release of Cdc14 at anaphase I that has been reported during meiosis.
Nevertheless, this is not the first time that the requirement of Cdc14 has been linked to cell cycle stages that precedes its bulk release from the nucleolus. Cdc14 has been involved in completion of latereplicating regions in the rDNA, and other parts of the genome. Such problems arise due to insufficient dosage of replication factors, although those deficiencies in principle seemed to be contingent on the reduced Cdc14 activity in preceding mitoses in the cdc14-1 allele used for such study [42].
Notwithstanding, collision amidst transcription and replication due to absence of Cdc14-mediated transcriptional silencing during S-phase is expected also to hamper the completion of DNA replication.
Irrespectively, the insufficient dosage of replication factors in cdc14 cells stems from two main reasons: i) nuclear import is constitutively impaired in cdc14 cells, preventing essential replication factors to enter into the nucleus properly, such is the case for certain subunits of RPA; and ii) reduced transcription of genes controlled by MBF and SBF transcription complexes in G1 which translates into reduced protein levels. Furthermore, Swi6, a transcription factor member of MBF and SCF complexes, requires the de-phosphorylation of its phospho-Serine 160 by Cdc14, in order to be imported to the nucleus upon mitotic exit [91]. Swi6 has also been shown to be important in meiotic cells [92] and the expression of genes like RAD54 and RAD51 is downregulated in a swi6 mutant. Furthermore, intragenic recombination frequencies between his4-4 and his4-290 is considerably decreased in the swi6 mutant, suggesting that the earlier accumulation of JMs might be the consequence of a reduced dosage of recombination proteins. By extrapolation, the defective processing of recombination intermediates observed at the HIS4-LEU2 recombination hotspot in cdc14 3HA mutant might be the consequence of the reduced activity of the Cdc14 3HA phosphatase during previous mitotic divisions before entering into meiosis, given the reduced levels of phosphatase displayed in this mutant also in cycling cells (Fig 2C).
Nonetheless, a cdc14-md mutant did not display reduced protein levels in mitotic cultures while still manifested high levels of JMs in prophase I. Thus, it appears unlikely that problems arising outside meiosis would cause this abnormal accumulation of JMs. In contrast, we favour a role of Cdc14 in the direct regulation of substrates required for correct processing of JMs in prophase I [4,10,15].
Interestingly, the conserved SSE, Yen1 GEN1 is a critical substrate of Cdc14 during budding yeast mitosis, and it exhibits a phosphorylation-regulated nucleocytoplasmic shuttling behaviour similar to that of Swi6 [93]. Namely, CDK-dependent phosphorylation restricts Yen1 from entering the nucleus and becoming active, whereas reversal of that phosphorylation by Cdc14 allows Yen1 to enter the nucleus and resolve entangled DNA-structures [19,20,81]. Nonetheless, Yen1 nuclear import and activation appears to be concomitant with bulk Cdc14 release from the nucleolus during anaphase. This is why a role for Yen1 in safeguarding chromosome segregation has been proposed, especially, during the second meiotic division. Our results support such a conclusion, since higher frequencies of aneuploidies and missegregation are detected during meiosis II for a cdc14 3HA meiotic mutant (Fig 4E). This is consistent with the demonstrated ability of Cdc14 to activate Yen1 (Fig 7A-C). Puzzlingly, the residual resolvase activity of Yen1 observed at time=0 in wild-type meiotic cultures might reflect certain level of early functionality (Fig 7B-C), perhaps important for the repair of any residual DNA damage carried over from the preceding mitosis. Such early resolvase activity is not observed in cdc14 3HA mutants ( Fig   7B-C). Taking into account that the levels of JMs in mus81Δ ndt80Δ-arrested cells are similar to those observed in cdc14 ndt80Δ mutants (S6 Fig), it is possible that SSEs play important functions in processing "off pathway" JMs generated as consequence of secondary DNA damage, which might be unrelated to SPO11-induced DNA repair. Such mechanisms could safeguard genome integrity at both, beginning and end, of meiosis putting Yen1 and Cdc14 as part of a specialized surveillance repair pathway that ensures DSB-repair fidelity, while preventing that other sources of unrepaired secondary damage to interfere with the canonical inter-homolog recombination repair. Whether or not Yen1 acts during anaphase I is less clear. Nuclear localization of Yen1 is moderately high at anaphase I ( Fig 7A) and maximal during anaphase II [18], but its endonucleolytic activity is only extensively turned on once MII spindles appear ( [18]; Fig 7C). Perhaps, minimal nuclear localization does not necessarily boost Yen1 activity, which might require additional regulatory modifications only present at the second stage of meiotic divisions [94]. Interestingly, it has been shown that Cdc14 only gets fully released to the cytoplasm during the second meiotic division, whereas the first division confines Cdc14 diffusion within just the nucleoplasm [47].
Absence of Cdc14 activity in cells depleted for SGS1 and MMS4 displayed the same recombination problems as the triple mutant sgs1 mms4 yen1 (Fig 6F-I). Additionally, unrestrained action of Yen1 (Yen1 ON ) restores faithful chromosome segregation and resolvase activity in cdc14 3HA diploid cells.
Furthermore, Yen1 ON partly restores sporulation in both cdc14 mutants. Considering all the observations mentioned above, we conclude that the lack of Yen1 activity is somewhat responsible, but not entirely, for the severity of the cdc14 3HA meiotic phenotype.

Additional meiosis-specific substrates of Cdc14 during gametogenesis.
It has been known for some time that cdc14 and FEAR mutants display problems segregating chromosomes in a SPO11-dependent way [45,46]. Consistently, the cdc14 3HA sporulation and homolog nondisjunction defects are somewhat solved when eliminating Spo11 activity (Fig 4). In recent years, human orthologs of Cdc14 phosphatase have received increased attention due to their involvement in key processes like DDR, DNA repair and cell cycle control. Furthermore, recent findings point to recessive variants of the phosphatase to be directly responsible of human diseases, like deafness and male infertility [101]. Thus, in order to comprehend the underlying factors that trigger those conditions, a deeper understanding of the genetic and molecular mechanisms that are responsible for the countless functionalities of the Cdc14 phosphatase during gametogenesis and HR repair will be required.

Yeast strains and plasmids
All strains were SK1, as detailed in S1 Table. Standard yeast manipulation procedures and growth media were utilized. To introduce the 3HA tag at the C-terminal end of Cdc14, the CDC14ha3-pKan R plasmid containing the last ~500bps of the CDC14 gene, was cloned in frame with 3xHA using the NotI restriction site, and containing the CLB1 terminator was used. The plasmid was linearized using a unique restriction site located within the CDC14 sequence and transformed into a SK1 haploid strain.
Alternatively, CDC14 was tagged using the PCR-based method described in [102] using the plasmid pFA6a-3HA-kanMX6. The phenotype of CDC14-3HA strains obtained from both tagging methods was checked and the sporulation defect was identical. Transformants containing correct tag integration were identified and tested by western blot for the presence of the tag. Southern blot analysis and/or PCR was performed to confirm the integration at the endogenous locus. Multi-copy plasmids used in S2 Fig. were originally described in [103].

Synchronous meiotic time courses
Induction of synchronous meiosis was carried out according to the established protocols for standard assays [104]. All pre-growth and meiotic time courses were carried out at 30 o C unless specified otherwise. For cdc14-1 meiosis, the culture was kept at 23 o C and shifted to 30 o C 2 hours after transferring into sporulation medium (SPM). Aliquots were removed at the specified times, and subjected to various analyses.

DNA manipulation, extraction and southern blot detection
Standard DNA extraction was performed as in [105]. For studies at the HIS4-LEU2 recombination hotspot, the protocol described in [106] was followed. For 2D gel agarose electrophoresis, cell cultures were photo-crosslinked with trioxalen (Merck) using long-wave UV light before DNA extraction, in order to visualize recombination intermediates by standard southern blotting techniques at the HIS4-LEU2 hotspot [106].

Time-lapse imaging, immunofluorescence, microscopy, and image analysis
Time-lapse experiments were performed as in [107], with small variations. In brief, 1 ml aliquots from synchronous meiotic cultures were taken at specific times, diluted 1:9 in fresh SPM (kept at 30 o C).
300μl of diluted cells were placed in suitable multi-well slides (80821 uncoated, ibidi). Slides were placed in a temperature-controlled incubation chamber from a Multidimensional Microscopy System Leica AF6000 LX. Images were taken at multiple positions and channels every 5, 10 or 15 minutes, depending on the experiment. Image acquisition was carried out using a CCD Hamamatsu 9100-02 camera. The system was controlled by the proprietary software, Leica Application Suite X, version 3.30. For preparations of fixed cells and immunofluorescence, aliquots were fixed and prepared as described in [105] and [18], respectively. Chromosomal DNA was stained with 1μg/ml 4,6-diamino-2phenylimide (DAPI). Images were recorded and analyzed using a Deltavision (DV3) workstation from Applied Precision Inc. with a Photometrics CoolSnap HQ (10-20MHz) air cooled CCD camera and controlled by SoftWorx image acquisition and deconvolution software.

Nuclease assays
For nuclease assays, myc9-tagged Yen1 and Yen1 ON were immuno-affinity purified from yeast using anti-Myc agarose beads (9E10) and washed extensively. The beads (approx. volume 10 μl) were then mixed with 10 μl cleavage buffer (50 mM Tris-HCl pH 7.5, 3 mM MgCl2) and 15 ng of 5'-Cy3-endlabeled synthetic Holliday junction X0 DNA. After 1 h incubation at 37 o C with gentle rotation, reactions were stopped by addition of 2.5 μl of 10 mg/ml proteinase K and 2% SDS, followed by incubation for 30 min at 37 o C. Loading buffer (3 μl) was then added and fluorescently-labelled products were separated by 10% native PAGE and analyzed using a Typhoon scanner and ImageQuant software.
Resolution activity was calculated by determining the fraction of nicked duplex DNA product relative to the sum of the intact substrate and resolution product. The protein input was estimated by western blot.

Data analysis and Biostatistics
Data was compiled and analyzed using Excel, LibreOffice Calc, and SPSS Statistical Data Editor. For multiple comparisons, analysis of variance (one-way ANOVA) was performed. For pairwise comparisons, two-tailed unpaired t-tests were used using IBM SPSS Statistics and SigmaPlot.

23.
Hartwell           (Fig 6D). Asterisk indicates meiosis-specific non-characterized recombination products.  Table. S. cerevisiae strains used in this study. SK1 background strains were used throughout the study, unless specified otherwise.