Phosphorylation of the Synaptonemal Complex Protein Zip1 Regulates the Crossover/Noncrossover Decision during Yeast Meiosis

Interhomolog crossovers promote proper chromosome segregation during meiosis and are formed by the regulated repair of programmed double-strand breaks. This regulation requires components of the synaptonemal complex (SC), a proteinaceous structure formed between homologous chromosomes. In yeast, SC formation requires the “ZMM” genes, which encode a functionally diverse set of proteins, including the transverse filament protein, Zip1. In wild-type meiosis, Zmm proteins promote the biased resolution of recombination intermediates into crossovers that are distributed throughout the genome by interference. In contrast, noncrossovers are formed primarily through synthesis-dependent strand annealing mediated by the Sgs1 helicase. This work identifies a conserved region on the C terminus of Zip1 (called Zip1 4S), whose phosphorylation is required for the ZMM pathway of crossover formation. Zip1 4S phosphorylation is promoted both by double-strand breaks (DSBs) and the meiosis-specific kinase, MEK1/MRE4, demonstrating a role for MEK1 in the regulation of interhomolog crossover formation, as well as interhomolog bias. Failure to phosphorylate Zip1 4S results in meiotic prophase arrest, specifically in the absence of SGS1. This gain of function meiotic arrest phenotype is suppressed by spo11Δ, suggesting that it is due to unrepaired breaks triggering the meiotic recombination checkpoint. Epistasis experiments combining deletions of individual ZMM genes with sgs1-md zip1-4A indicate that Zip1 4S phosphorylation functions prior to the other ZMMs. These results suggest that phosphorylation of Zip1 at DSBs commits those breaks to repair via the ZMM pathway and provides a mechanism by which the crossover/noncrossover decision can be dynamically regulated during yeast meiosis.


Introduction
During meiosis, crossovers (COs), in combination with sister chromatid cohesion, physically connect homologous chromosomes, thereby promoting proper segregation at Meiosis I (MI). COs arise by the regulated repair of programmed double-strand breaks (DSBs), formed by the topoisomerase-like protein, Spo11 (SGD S000001014) [1]. This regulation involves components of the synaptonemal complex (SC), a meiosis-specific tripartite structure formed between homologous chromosomes [2]. SC formation begins by condensation of sister chromatids upon protein cores to form axial elements (AEs). AEs contain loops of chromatin tethered at their bases by meiosis-specific axis proteins, which in yeast include Hop1 (SGD S000001334), Red1 (SGD S000004253), and the cohesin kleisin subunit, Rec8 (SGD S000006211) [3][4][5]. Hotspot sequences are brought to the axes where DSBs result in the recruitment and activation of the meiosis-specific kinase, Mek1 (SGD S000005878), via Mec1/Tel1 (SGD S000000340/ S000000184) phosphorylation of Hop1 [1,6,7]. Mek1 promotes interhomolog (IH) recombination by antagonizing cohesion around DSBs and inhibiting the strand exchange activity of the mitotic recombinase, Rad51 (SGD S000000897), thereby facilitating strand invasion of homologous chromosomes by the meiosisspecific Dmc1 (SGD S000000981) recombinase [8][9][10][11]. For many organisms such as yeast and mammals, strand invasion is critical for bringing homologs together. Synapsis is achieved when homologous AEs are held together by the insertion of transverse filament (TF) proteins in the central region to form the SC [2]. In yeast, the TF protein is Zip1 (SGD S000002693) [12].
In yeast, whether a DSB is repaired as a CO or NCO has been proposed to be determined at, or just after, the time of DSB formation [13,24]. Furthermore, there is plasticity in the decision as evidenced by the existence of CO homeostasis [25]. When DSBs are reduced, CO levels are maintained at the expense of NCOs. This behavior suggests that mechanisms exist for regulating the CO/NCO decision after DSBs have been made, but the molecular bases for such regulation have been elusive.
Phosphorylation is a dynamic modification used to regulate different steps in recombination, in addition to IH bias. For example, the highly conserved cell cycle kinases Cdc28-Clb5/6 (Cdk1) (SGD S000000364-S000006324/S000003341) and Cdc7-Dbf4 (Dbf4-dependent kinase; DDK) (SGD S000002175-S000002459) are required for DSB formation, and dHJ resolution is triggered by the polo-like kinase Cdc5 (SGD S000004603) [26][27][28]. Whether Mek1 or DDK is directly involved in forming IH COs was unknown, because inactivation of these kinases prevents creation of IH recombination intermediates. In this work, phosphoproteomic analysis reveals conserved phosphorylation sites on the C terminus of Zip1 that are required for the ZMM pathway of DSB repair. Phosphorylation of these sites is dependent upon MEK1, revealing a previously unknown role for this kinase in the regulation of IH recombination and suggesting a link by which IH bias and the CO/NCO decision can be coordinated.

Zip1 Is Phosphorylated on Its C Terminus in a Region Required for Chromosome Synapsis
To identify phosphorylation sites regulated by Mek1, stable isotope labeling of amino acids in cell culture (SILAC) was performed using a mek1-as ndt80Δ diploid as described in [29]. mek1-as encodes a conditional version of Mek1 that can be inactivated by the purine analog 1-NA-PP1 [30]. Diploids lacking the meiosis-specific transcription factor NDT80 (SGD S000001166) arrest in meiotic prophase with fully synapsed chromosomes and unresolved dHJs [15,31]. A mek1-as ndt80Δ arg4 lys4 diploid was grown in synthetic medium containing  [18]). Red and blue lines represent duplexes of DNA from the nonsister chromatids of a pair of homologous chromosomes. The green box indicates "stabilization" of IH JMs leading to biased resolution as COs. (B) Schematic showing phosphosites on Zip1 detected from a stable isotope labeling of amino acids in cell culture (SILAC) experiment using the mek1-as ndt80Δ diploid, NH2221. A hyphen indicates both amino acids were phosphorylated on the same peptide. Gray boxes indicate predicted coiled-coil domains [12]. The pink box represents the 800-824 amino acid region required for synapsis [12]. (C) Average Light/Heavy (L/H) ratios of Zip1 phosphopeptides detected by SILAC. Ratios above the dashed line indicate possible regulation by Mek1. Error bars indicate the standard deviation for the L/H ratios in cases where more than one phosphopeptide was detected. Numbers above each bar indicate how many peptides containing the indicated phosphosite were observed. Pink indicates phosphosites in the 4S region. (D) Sequence of the region indicated by the pink box in B. Pink color indicates the serines mutated in the zip1-4A and zip1-4D alleles. Antibodies against phosphorylated S816 (indicated by an asterisk) were generated using the peptide underscored by the black line. (E) Clustal Omega alignment of TF proteins from different organisms in the Zip1 4S region. Numbers indicate amino acid position. Dashes indicate gaps. In the consensus, red = identical amino acids, green = amino acid is present in four species, blue = amino acids with conserved properties, ϕ = hydrophobic, [-] = acidic. Asterisks (*) represent residues predicted by NetPhos analysis to be phosphorylated in all five species.
doi:10.1371/journal.pbio.1002329.g001 Zip1 Phosphorylation Regulates Meiotic Recombination "heavy" or "light" versions of arginine and lysine and then transferred to Spo medium for 14 h. Mek1-as was inhibited in the heavy culture by addition of 1-NA-PP1 for 20 min, chromatin was isolated from both cultures, and equal amounts of protein were mixed and digested with trypsin. Phosphopeptides were enriched using immobilized metal affinity chromatography and analyzed by mass spectrometry (MS). Non-Mek1-dependent phosphopeptides should be present in equal amounts in both cultures giving light:heavy (L/H) peptide ratios of 1. Mek1-dependent phosphopeptides are predicted to be under-represented in the heavy culture and exhibit L/H ratios > 1. This approach revealed that one phosphoprotein potentially regulated by Mek1 is the TF protein Zip1.
MS analysis identified 18 phosphorylation sites on Zip1 (Fig 1B). Two of these phosphopeptides (S314 S318 and S712) exhibited L/H ratios around 1.0, indicating that equivalent amounts of Zip1 protein were present in the light and heavy cultures (Fig 1C). Of particular interest were three out of four adjacent serine residues in the C terminus (S815-S818, hereafter referred to as 4S) that exhibited L/H ratios >3, suggesting potential regulation by Mek1. These serines are located within a 25-amino acid region required for chromosome synapsis and WT levels of recombination, suggesting that phosphorylation of Zip1 at the C terminus plays a role in these processes (Fig 1B and 1D) [12].
Although TF proteins are structurally similar, they are poorly conserved at the primary sequence level [2]. It is therefore compelling that the 4S region of Zip1 is conserved among yeast, mammals and zebrafish TF proteins ( Fig 1E). Furthermore, NetPhos analysis [32] predicts that at least two of the residues in the conserved 4S region are phosphorylated in all five species. In contrast, no homology was observed in this region for the TF proteins from Drosophila melanogaster, Caenorhabditis elegans, or Arabidopsis thaliana.

Phosphorylation of Zip1 4S Promotes Negative Feedback Regulation of Spo11
The functional significance of Zip1 4S phosphorylation was tested using ZIP1 alleles in which all four serines were mutated either to alanine (zip1-4A) or aspartic acid (zip1-4D) to prevent or mimic phosphorylation, respectively. The zip1-4A diploid exhibited spore viability that is significantly reduced compared to WT (76.7% vs. 94.3%), but which is higher than the zip1Δ (45.9%) (p < 0.0001) (S1 Table). This decrease in spore viability is due the absence of negative charges as the zip1-4D diploid produced 96.2% viable spores. Meiotic time course analysis of steady state Zip1 protein levels revealed that 4S phosphorylation is required for Zip1 degradation (S1A Fig). The increased stability of Zip1-4A protein raised the possibility that the spore viability defect was indirectly due to the persistence of Zip1-4A. If this were true, then the zip1-4A mutant phenotype should be dominant, since the Zip1-4A protein should fail to get degraded even in the presence of the WT protein. In fact, the amount of Zip1 protein persists at higher levels in the ZIP1/zip1-4A heterozygote than the WT strain, yet spore viability in the heterozygote is indistinguishable from WT (94.5% ZIP1/zip1-4A vs 94.3% ZIP1; p = 0.149) (S1A Fig, S1 Table). Therefore, the failure to degrade Zip1-4A is not the reason for its mutant phenotype.
Meiotic progression was delayed in zip1-4A and zip1Δ diploids, but not zip1-4D, providing further evidence that negative charges at these positions are important for ZIP1 function (Fig  2A). Physical analysis using the HIS4::LEU2 hotspot on chromosome III was used to monitor DSB formation and repair [24]. This hotspot consists of a DSB site flanked by polymorphic XhoI restriction sites (S2A Fig). DSBs can be distinguished by Southern blot analysis using XhoI-digested DNA. DSBs in zip1-4A appeared at the same time as WT and zip1-4D, but were present in greater abundance than the zip1Δ at later time points and took longer to disappear (Fig 2B and 2C). As an alternative way of assessing DSB levels, accumulation of genome-wide Spo11 oligonucleotides was assayed. After DNA cleavage, Spo11 is removed by endonucleolytic digestion, resulting in oligonucleotide "tags" [33]. The zip1-4A mutant reproducibly accumulated significantly higher levels of Spo11 oligonucleotides than any of the other strains, including the zip1Δ (Fig 3). These results indicate that Spo11 down-regulation due to a failure of ZMM-mediated IH engagement is more impaired in the presence of the Zip1-4A protein than in the absence of Zip1 altogether.  Quantification of DSBs in various zip1 mutants using Spo11 oligonucleotides. Time courses of meiotic cultures of ZIP1 (NH2311), zip1::LEU2 NH2312, zip1-4A (NH2313), and zip1-4D (NH2314) containing Protein A-tagged Spo11 were performed in parallel as described in [14]. (A) Spo11-Protein A was immunoprecipitated using Protein A antibodies, and the covalently attached radioactively end-labeled oligonucleotides were detected by autoradiography (top panel). Free Spo11 was detected by probing immunoblots of the extracts with Protein A antibodies (middle panel). Extracts used for Spo11 immunoprecipitation were fractionated by SDS-PAGE and proteins stained with Coomassie Brilliant Blue to show that equivalent amounts of extract were used as input for the immunoprecipitation reactions (bottom panel). (B) Quantification of mean Spo11 oligonucleotide levels from four independent experiments. Error bars represent the standard deviation. Mutants are plotted with WT data collected in parallel.
zip1Δ. In contrast, NCOs were increased, with an even higher level of NCOs in zip1-4A compared to zip1Δ (Fig 2D and 2E). Joint molecules (JMs) indicative of single end invasion (SEI) and dHJ intermediates can be detected using two-dimensional gel electrophoresis [24]. Previous work has indicated that ZIP1, as well as the other ZMM genes, promotes the transition from DSBs into SEI intermediates as well as dHJs [13]. Like zip1Δ, SEIs and dHJs were delayed in zip1-4A (Fig 2F and 2G) (S2D Fig). The areas under the SEI and IH-dHJ curves appear to be greater in the zip1-4A diploid compared to WT and zip1Δ, as expected given the increased number of DSBs. It is also possible that intermediates are accumulating due to a slower rate of repair. The formation of intersister (IS) JMs is similarly delayed in both zip1-4A and zip1Δ. We conclude that phosphorylation of Zip1 4S is responsible for the requirement of ZIP1 in promoting stable strand invasion intermediates.

Phosphorylation of Zip1 4S Is Required for Making COs That Are Distributed by Interference
To complement the physical analysis, recombination was monitored genetically in a strain containing linked markers on chromosomes III, VII, and VIII [35]. In this strain, sporulation was slightly decreased by zip1-4A and zip1Δ compared to ZIP1 and zip1-4D ( Fig 4A). Consistent with previous results, zip1-4D spore viability was equal to WT, while the spore viability of zip1-4A was intermediate between WT and the zip1Δ (Fig 4B).
Map distances were reduced in the zip1-4A mutant on chromosome III, consistent with reduced crossing over measured at the HIS4::LEU2 hotspot ( Fig 4C). In contrast, map distances in zip1-4A were greater in three intervals on chromosome VII and one interval on chromosome VIII than either WT or zip1Δ. Notably, no CO defect was observed for zip1Δ on chromosome VII either. This differential effect on CO formation as a function of chromosome size was also observed with a mutant allele of the ZMM gene, ZIP3 [39]. Large chromosomes, such as chromosome VII, sustain disproportionately higher numbers of DSBs in zmm mutants [14]. Therefore, the zip1-4A increase in COs on chromosomes VII and VIII could be a consequence of continued Spo11 activity generating new breaks, resulting in more CO opportunities. Increasing the number of DSBs should also increase the number of opportunities for gene conversion and in fact, gene conversion events are elevated for loci on all chromosomes in zip1-4A ( Fig 4D).
A reduction in COs may lead to increased MI non-disjunction due to a failure to physically connect homologs. For both zip1Δ and zip1-4A, the distribution of viable spores in tetrads supports the idea that spore lethality is due to MI non-disjunction. Both mutants exhibited an increase in the number of 2 viable:2 inviable spores (predicted if one or more pair of homologs have non-disjoined to the same pole) and 0 viable:4 inviable spores (predicted if two or more pairs of homologs have disjoined to opposite poles) ( Fig 4E) [40].
Non-disjunction can be directly demonstrated for chromosome III where the codominant MATa (SGD S000124955) and MATα (SGD S000029699) alleles produce a distinctive nonmating (NM) phenotype when chromosome III is disomic. Chromosome III non-disjunction (in the absence of recombination) should therefore lead to tetrads with two viable spores, both of which are NMs. Out of 63 two viable spore tetrads obtained from zip1-4D, only one contained two NM spores (1.6%, Table 1). In contrast, the number of tetrads exhibiting chromosome III non-disjunction increased to 18.6% in zip1Δ and 14.9% in zip1-4A (Table 1). Another prediction of MI non-disjunction is that viable spores should be derived from segregation of sister chromatids for the other chromosomes. For chromosome VIII, sister versus nonsister spores can be determined using the centromere-linked marker URA3 (SGD S000000747) ( Table 1). For the zip1Δ and zip1-4A chromosome III disomic tetrads, equal numbers of either Ura + :Ura + or Ura − :Ura − sister spores were observed, as expected given that chromosomes III and VIII independently assort (Table 1).
Recombination was monitored in the intervals between MAT and HIS4 on the disomic III chromosomes. For both zip1Δ and zip1-4A, all but one of the disomic spores contained III homologs that were non-recombinant between MAT and HIS4 (Table 1). Therefore, failure to phosphorylate Zip1 4S results in reduced COs and increased non-disjunction of chromosome III, similar to zip1Δ.
To assess whether non-disjunction is occurring on other chromosomes, the segregation of chromosome III was followed using ADE2 (SGD S000005654), which was integrated adjacent to the centromere on chromosome III (Table 1). In the two viable spore tetrads that were not disomic for chromosome III (i.e., both spores were maters), random spore death should result in a 2:1 ratio of nonsister spores (Ade + : Ade − ) to sister spores (Ade + : Ade + or Ade − : Ade − ) for chromosome III. If, however, spore death is due to non-disjunction of a chromosome(s) other than III, then tetrads containing chromosome III sister spores should be over-represented in the two viable spore asci. The zip1-4D diploid exhibited a ratio of 1.3 nonsister/sister spore tetrads, which is not significantly different from the ratio of 2 predicted for random spore death (χ 2 , p = 0.357). In contrast, the nonsister/sister spore ratios for zip1Δ and zip1-4A were 0.2 and Genetic analysis of sporulation, spore viability, and recombination in various zip1 diploids. (A) ZIP1 (NHY957), zip1-4D (NH2241::p382), zip1Δ (NH2241::pRS304), and zip1-4A (NH2241::p382-4A) were sporulated for 2 d at 30°C. Average sporulation from five independent colonies was graphed; error bars indicate the standard deviation. (B) Average spore viability calculated from dissection of nine independent transformants from each diploid. Brackets indicate strain pairs whose spore viability is significantly different (χ 2 ; p < 0.0001). The number of tetrads dissected for each strain: ZIP1, 449; zip1-4D, 1,332; zip1-4A, 1,857; and zip1Δ, 4,012. The isogenic ZIP1 data shown in Panels B and E were previously published [35]. (C) Map distances for the indicated intervals on Chromosomes III, VII, and VIII were calculated using the formula from Perkins [36]. At least 940 tetrads were analyzed for each genotype. ZIP1 map distances were generated by pooling data from published NHY957 dissections [25,35,37]. p-Values indicating the significance between the map distances of different intervals for mutants and WT are shown in S1 Data. (D) The percent of four viable spore tetrads exhibiting non-2 + :2 − segregation (gene conversion) for the indicated loci. ZIP1 data were taken from [37]. (E) The distribution of viable spores in tetrads from the dissections shown in Panel B. (F) Interference values for adjacent intervals were calculated using the method of [38]. A solid line indicates interference was observed when either interval was used as the reference. A black broken line indicates interference was detected using one of the reference intervals. A red broken line indicates that no interference was observed using either reference interval. Numbers represent the average of the two ratios for that interval. Data for the ZIP1 strain in Panel F are from [37]. The aberrantly high value observed for the zip1Δ in the HIS4-LEU2/LEU2-CEN interval is likely due to the chance occurrence of an extra nonparental ditype (NPD).  Determined by both spores being non-maters and prototrophic for adenine, leucine and histidine. c Determined by both spores being non-maters but auxotrophic for at least one marker in at least one of the spores. d Significantly different from zip1-4D (p = 0.0007, χ 2 test). e Significantly different from zip1-4D (p = 0.007, χ 2 test). f Determined using two viable spore tetrads in which both spores were haploid (i.e., able to mate).  (Table 1), supporting the idea that non-disjunction of chromosomes other than III also contributes to the spore lethality.
COs from the ZMM pathway exhibit genetic interference, where a CO in one interval decreases the likelihood of COs in adjacent intervals [16]. One way of measuring interference is to divide tetrads into two classes based on crossing over in a reference interval: (1) those with no detectable CO in the interval (parental ditypes [PDs]) and (2) those with at least one detectable CO (nonparental ditypes [NPDs] and tetratypes [TTs]) [38]. Map distances for an adjacent test interval are then calculated for each class. If interference is acting, then the Class 2/ Class 1 map distance ratio should be <1. In the absence of interference, the ratio should be = 1. With one exception, zip1-4D exhibited similar levels of interference as WT for intervals on chromosome III and VII ( Fig 4F). In contrast, interference was absent or weakened in the zip1Δ and zip1-4A diploids on both chromosome III and VII. These data indicate that the bulk of the COs observed in zip1-4A occur via the noninterfering CO pathway.
MUS81 encodes a subunit of a structure-specific endonuclease, which in combination with MMS4, is responsible for generating COs along the noninterfering pathway [17,35]. A mus81Δ diploid produced 38.7% viable spores, similar to mus81Δ zip1-4D (40.0%), while zip1Δ exhibited 48.4% (S1 Table). The number of viable spores generated by mus81Δ zip1Δ (16.1%) is not significantly different from the number expected for compound defects in two independent CO pathways (17.4%, p = 0.35, χ 2 test). The residual spore viability most likely arises from COs created by alternative structure-specific endonucleases such as Yen1 [17,21]. In contrast, the mus81 zip1-4A diploid exhibited 20.2% viable spores, which is significantly lower than the 29.5% predicted based on the 75% spore viability observed for zip1-4A (p = 0.004, χ 2 test). The observation that the mus81Δ zip1-4A spore viability is similar to that of mus81Δ zip1Δ suggests that the higher spore viability in zip1-4A is due to MUS81-dependent COs.
Zip1 4S phosphorylation promotes interference on both chromosome III and chromosome VII, yet a CO deficit is only observed on the smaller chromosome. This apparent paradox can be explained if Zip1 4S phosphorylation is required for ZMM-mediated COs on all chromosomes, but the loss of these COs is obscured on larger chromosomes by an increased number of DSBs allowing the formation of more COs along the alternative MUS81-dependent pathway.

Phosphorylation of Zip1 4S Is Required for Efficient Chromosome Synapsis
The ZMM genes are required for chromosome synapsis [13]. ZIP1 and zip1-4D exhibited similar frequencies and timing of synapsis when monitored by staining spread chromosomes with antibodies against either Zip1 in the SK1 background ( [42]. PCs were abundant specifically in the zip1-4A mutant, consistent with a defect in SC formation (Fig 5C and 5D). The presence of Zip1-4A aggregates in PCs may make the protein less accessible to proteases, which could explain the stability of the Zip1-4A protein compared to the WT and Zip1-4D proteins.

Phosphorylation of Zip1 S816 Is Promoted by DSB Formation
Antibodies were generated using a peptide containing phosphorylated Zip1 S816 (Fig 1D). Extracts were generated from ndt80Δ, ndt80Δ zip1Δ, and ndt80Δ zip1-S816A strains 8 h after transfer to Spo medium. The NDT80 gene was deleted to prevent Zip1 degradation [27]. Antibodies specific to phosphorylated S816 (p-S816) exhibited a signal with extracts from ZIP1 but not ZIP1-S816A, and this signal was lost upon phosphatase treatment of the WT protein ( Fig  6A). In the course of purifying the phosphospecific antibodies, antibodies that specifically recognize the unphosphorylated form of Zip1 (unp-Zip1) were also obtained. As expected, Zip1 detected by these antibodies displayed a complementary signal to the p-S816 antibodies (e.g., low in the absence of phosphatase and increased after treatment with phosphatase) (Fig 6A). Equivalent amounts of Zip1 protein were present in the different samples, as shown by probing with an antibody that recognizes total Zip1. The Zip1 p-S816 antibodies are therefore highly specific for the phosphorylated protein and provide an excellent tool for monitoring the phosphorylation state of Zip1 at S816.
In WT cells, Zip1 S816 phosphorylation was coincident with DSB formation, indicated indirectly by phosphorylation of Hop1 ( Fig 6B) [43]. Zip1 protein was degraded at later time points, as noted previously by [27]. Although a low level of phosphorylation was observed in spo11Δ, this amount was reduced approximately 15-fold compared to WT, suggesting that DSBs promote Zip1 S816 phosphorylation (Fig 6B, 6D and 6E) [44]. We conclude that there is a low background level of Zip1 S816 phosphorylation in the absence of DSBs, and that phosphorylation of this amino acid is induced above this background by meiotic DSBs.
Whether DSB formation is sufficient for Zip1 S816 phosphorylation was determined using an sae2Δ/com1Δ (SGD S000003143) diploid where the ends of the breaks are covalently attached to Spo11 and are not resected [45,46]. Phosphorylation of Zip1 S816 was observed in this mutant, suggesting that DNA processing is not necessary (Fig 6B). In dmc1Δ and ndt80Δ diploids, which arrest with unrepaired DSBs and dHJs, respectively [15,47], Zip1 protein was not degraded and p-S816 also persisted.

Phosphorylation of Zip1 S816 Is Promoted by Mek1 Kinase Activity
The SILAC experiment showing that Zip1 S816-S818 phosphosites are potentially regulated by Mek1 is consistent with the induction of S816 phosphorylation by DSBs, since one consequence of making DSBs is activation of the kinase [6]. In fact, Zip1 S816 phosphorylation is reduced approximately 8-fold in a mek1-K199R diploid that encodes a catalytically inactive version of the kinase (Fig 6C, 6D and 6E) [48]. Mek1 activation occurs via the Mec1 and Tel1 checkpoint kinases, which phosphorylate the AE component Hop1 in response to DSBs, resulting in Mek1 recruitment to axes [7]. A 7-fold reduction in Zip1 S816 phosphorylation was observed in a mec1Δ strain, while a tel1Δ diploid appeared more like WT (Fig 6C, 6D and 6E). These results are similar to what was observed for phosphorylation of Hop1 by these kinases [6] and are consistent with Zip1 phosphorylation being promoted by activation of Mek1 via Hop1 phosphorylation. We conclude that Zip1 4S phosphorylation is induced upon DSB formation as a result of Mek1 activation and does not require the removal of Spo11 from the ends of the breaks or DSB resection.
The four adjacent serines containing S816 represent potential phosphorylation sites for the highly conserved, essential kinase, DDK [54]. DDK is an acid-directed kinase that prefers to phosphorylate amino acids immediately upstream of a negative charge [28,[55][56][57]. This charge can be provided either by priming phosphorylation or by negatively charged amino acids. To test whether DDK can directly phosphorylate Zip1, an in vitro kinase assay was developed using an analog-sensitive version of Cdc7, Cdc7-as [58]. CDC7 or cdc7-as was fused to sequences encoding FLAG epitopes and the proteins partially purified from ndt80Δ -arrested cells. When incubated with ATPγS, autophosphorylation of Cdc7, as well as trans phosphorylation of Dbf4, was detected with DDK using the semisynthetic epitope system ( Fig 7B). As expected, phosphorylation by DDK was unaffected by the addition of the DDK-as inhibitor, PP1 [58]. Cdc7-as-Dbf4 (DDK-as) was also able to use ATPγS, although not as efficiently as DDK (Fig 7B). Addition of PP1 to this reaction eliminated Cdc7-as autophosphorylation but not Dbf4 phosphorylation, indicating the presence of a contaminating kinase in the DDK-as preparation capable of modifying Dbf4. Specificity for DDK-as phosphorylation was achieved using the ATP analog, 6-benzyl-ATPγS (6-Bn-ATPγS) instead of ATPγS. In this situation, DDK-as, and not DDK, exhibited phosphorylation of Cdc7-as and Dbf4, and phosphorylation of both proteins was completely eliminated by the addition of PP1 (Fig 7B). DDK-as can therefore be used with 6-Bn-ATPγS to reveal direct targets of the kinase in vitro.
Unlike GST-Mek1-as, DDK-as directly phosphorylated MBP-Zip1 in vitro, and this phosphorylation was abolished by PP1 (Fig 7C, left). The semisynthetic epitope kinase assay detects phosphorylation anywhere on a protein and does not reveal specifically which amino acid(s) were phosphorylated. Kinase reactions were therefore performed with 6-Bn-ATP (which transfers phosphate instead of thio-phosphate), and the proteins were probed with p-S816 antibodies. Phosphorylated Zip1 S816 was observed, but not in the presence of inhibitor, confirming that Zip1 S816 is a direct target of DDK-as in vitro (Fig 7C, right).
To see whether DDK is capable of phosphorylating Zip1 S816 in vivo in mitotic cells, the Zip1 C terminus (amino acids 749-875) was fused to lexA (Unitprot P31080) (lexA-ZIP1C) and ectopically expressed in vegetative cells using the ADH1 (SGD S000005446) promoter in a cdc7-as mcm5-bob1 (SGD S000004264) strain. The mcm5-bob1 mutation bypasses the essential function of cdc7-as in DNA replication initiation and therefore allows cdc7-as strains to grow vegetatively in the presence of PP1 [59,60]. In the absence of inhibitor, Zip1 p-S816 was detected with the WT lexA-ZIP1C plasmid, but not the lexA-ZIP1C-4A mutant (Fig 7D). Phosphorylation of S816 was reduced when DDK-as was inactivated by PP1, confirming that Zip1 S816 is an in vivo target of DDK-as. The low level of residual phosphorylation is presumably equivalent to the DSB-independent phosphorylation observed in spo11Δ meiotic experiments. The fact that DDK can phosphorylate Zip1 S816 in vegetative cells suggests that constraints on DDK phosphorylation of Zip1 are imposed specifically during meiosis.
Inhibiting DDK during meiosis is predicted to reduce Zip1 S816 phosphorylation because DDK is required for making DSBs, which in turn promote Zip1 S816 phosphorylation [58] (Fig 6). This prediction was partially borne out in that sporulating cdc7-as cells in the presence of PP1 resulted in a slight reduction in phosphorylation, but the decrease was not as great as that observed for spo11Δ (Figs 6, 7E, 7F and 7G). One difference between spo11Δ and cdc7-as + PP1 strains is that spo11Δ cells are able to progress into the meiotic divisions and degrade Zip1, while cdc7-as + PP1 cells are arrested prior to NDT80 induction due to the requirement of DDK for NDT80 transcription [61]. Zip1 S816 phosphorylation was therefore assayed in an ndt80Δ spo11Δ diploid, which should phenocopy both the DSB and meiotic progression defects of cdc7-as + PP1. Zip1 S816 phosphorylation was increased over four fold in the ndt80Δ spo11Δ diploid compared to spo11Δ, suggesting that phosphorylation of Zip1 S816 is no longer constrained at the ndt80Δ arrest. The fact that S816 phosphorylation is slightly reduced in the absence of DDK activity suggests that some of this phosphorylation is occurring due to DDK. However, the fact that phosphorylation is still present indicates either that the DDK inhibition was incomplete, or more likely, that there is a second kinase capable of phosphorylating this site independently of DDK. Our preferred hypothesis is that during normal meiosis, Zip1 phosphorylation is regulated by Mek1 and is mediated by DDK, but we cannot rule out the possibility that a second, yet-identified kinase also plays a role in Zip1 regulation.
There is no negatively charged amino acid immediately downstream of the Zip1 4S sequence (Fig 1D), suggesting that a priming negative charge is not necessary for DDK phosphorylation of recombinant Zip1. This appears to be true in vivo as well, as alanine substitutions of any one of the four serines supported WT spore viability (Fig 7H, S1 Table). Various combinations of double alanine substitutions reduced spore viability significantly from WT, but only the zip1-S817A S818A double mutant was as low as the level observed for the zip1-4A diploid. Interestingly, in TF proteins from other species, the predicted phosphoserines in the region conserved with Zip1 4S flank a negatively charged amino acid equivalent to the position of Zip1 S817 (Fig 1E). This is consistent with the hypothesis that a negatively charged patch created by phosphorylation in this region is a conserved requirement for TF function. Kinase assays using GST-Mek1-as (GST: Unitprot P0A9D2) partially purified from meiotic yeast cells and MBP-Zip1 (MBP: Unitprot P19642) or Rad54 as indicated with the ATP analog, 6-Fu-ATPγS. Phosphorylated proteins were detected using the semisynthetic epitope system with α-hapten (α-hap) antibodies [53]. The presence of MBP-Zip1 was confirmed using antibodies against total Zip1 (α-Zip1). The 1-NA-PP1 inhibitor concentration was 1 μM. (B) Kinase assays comparing autophosphorylation of DDK-FLAG and DDK-as-FLAG partially purified from meiotic yeast cells in the presence of ATPγS or the ATP analog, 6-benzyl (Bn)-ATPγS. The PPI inhibitor was used at a concentration of 20 μM. (C) DDK-as-FLAG kinase assays with MBP-Zip1. Left panel: 6-Bn-ATPγS was used with α-hapten antibodies to detect phosphorylated proteins by the semisynthetic epitope system. Right panel: 6-Bn-ATP was used with α -phospho-S816 antibodies to specifically detect Zip1 p-S816. (D) Zip1-S816 phosphorylation in vegetative cells with and without DDK-as kinase activity. The haploid cdc7-as mcm5-bob1 strain, NH2309-6-3, was transformed with plasmids constitutively expressing lexA, lexA-ZIP1C 749-875 , or lexA-ZIP1C 749-875 -4A and grown in the presence of either dimethylsulfoxide (DMSO) or 15 μM PP1. Extracts were probed with α-pS816 or α-lexA antibodies to detect Zip1 p-S816 and LexA, respectively. (E) Zip1 S816 phosphorylation in meiotic cells with and without DDK-as kinase activity. The cdc7-as3-9myc (called cdc7-as for short) diploid, NH2322, was transferred to Spo medium and the culture split in half at the 0 hour time point. To one half, DMSO was added (-I) and to the other a final concentration of 20 μM PP1 was added. Time points were collected and protein extracts assayed for the total amount of Zip1 as well as Zip1 p-S816. (F) Zip1 S816 phosphorylation in an ndt80Δ spo11Δ (NH2337) diploid. (G) Composite gel using extracts from time points with the highest amount of Zip1 from the indicated strains were run on the same gel and probed for total Zip1, p-S816, and Arp7. Quantification was performed as in Fig 6E. (H) Spore viability (% s.v.) of different ZIP1 mutant alleles. Δ = zip1Δ; A = alanine, S and "." = serine, D = aspartic acid. * and § = significantly different from WT and zip1-4A, respectively. See also S1 Table. doi:10.1371/journal.pbio.1002329.g007

SGS1 and Zip1 4S Phosphorylation Are Both Required for Meiotic Progression
Zmm proteins are proposed to stabilize recombination intermediates by protecting them from the action of the Sgs1 helicase [16]. The sgs1-md allele has SGS1 under the control of the CLB2 promoter that is not expressed in meiotic cells, thereby allowing selective depletion of Sgs1 during meiosis [62]. Depleting SGS1 in the zip1-4A background produced an unexpected result: a large decrease in sporulation from 45.0% for sgs1-md zip1Δ to 8.5% for sgs1-md zip1-4A ( Fig  8A). The failure to sporulate is due to meiotic prophase arrest, as sgs1-md zip1-4A cells do not undergo the meiotic divisions, remaining mononucleate and failing to generate MI and Meiosis II (MII) spindles measured by staining of nuclei and spindles with 4', 6-diamidino-2-phenylindole (DAPI) and tubulin, respectively (Fig 8B and S4 Fig). The sgs1-md zip1-4A meiotic prophase arrest is due to the absence of Zip1 4S phosphorylation, as an sgs1-md zip1-4D strain went through the meiotic divisions more quickly and efficiently than sgs1-md zip1-4A (Fig 8B). The zip1-4A mutant alone reduced sporulation compared to WT in this background, but not to the low level observed in combination with sgs1-md. Therefore, SGS1 prevents the meiotic prophase arrest resulting from failure to phosphorylate Zip1 4S.
Analysis of recombination products in the sgs1-md background revealed that the excess of NCOs observed in zip1-4A is dependent upon SGS1. Previously it has been observed that depletion of SGS1 can partially suppress the CO defects of different zmm mutants to various extents [19,23,63]. While this phenotype was observed for the sgs1-md zip1Δ combination at the HIS4-LEU2 hotspot, COs were reduced in the sgs1-md zip1-4A diploid (Fig 8C and 8D). This result is consistent with the idea that many of the DSBs in sgs1-md zip1-4A cells are unavailable for repair.

Zip1 4S Phosphorylation Functions Upstream of Other ZMM Genes
The meiotic progression and sporulation defects observed for sgs1-md zip1-4A, but not sgs1-md zip1Δ, indicate that the presence of the Zip1-4A protein has a toxic gain of function phenotype in the absence of SGS1. One way to explain this phenotype would be if (1) Zip1 is recruited to a subset of DSBs where phosphorylation of Zip1 4S commits the breaks to repair by the ZMM pathway of interfering CO formation; (2) the absence of Zip1 4S phosphorylation blocks the break from being repaired, and (3) Sgs1 removes Zip1-4A from the breaks, thereby allowing them to be recycled and repaired via SDSA or the alternative Mus81-Mms4/Yen1 pathway. This model posits that the sgs1-md zip1-4A meiotic prophase arrest is due to unrepaired DSBs triggering the meiotic recombination checkpoint and is supported by the observation that preventing DSBs using spo11Δ restored sporulation to sgs1-md zip1-4A (Fig 8A).
Although the ZMM genes function in the pathway for producing interfering COs, the order in which they work has yet to be elucidated. The zip1-4A inhibition of sporulation in sgs1-md strains provides an assay by which to order ZMM genes relative to Zip1 4S phosphorylation. There are two ways that a ZMM gene could function prior to Zip1 4S phosphorylation. One possibility would be that a ZMM gene is required for Zip1-4S phosphorylation. In this case, the sgs1-md zmmΔ diploid would contain unphosphorylated Zip1 4S and should therefore phenocopy sgs1-md zip1-4A and fail to sporulate. This was not observed, however, as combining sgs1-md with a deletion of ZIP2, ZIP3, ZIP4, MSH5, MER3, or SPO16 resulted in high levels of sporulation (Fig 8A). Consistent with this observation, Zip1 S816 phosphorylation was unaffected in sgs1-md diploids containing mer3Δ, msh5Δ or zip3Δ (S1B Fig). Another possibility is that a ZMM gene is required for Zip1 4S function (for example if it was needed to recruit Zip1 to the DSB). In this scenario, the absence of the ZMM gene should phenocopy the spo11Δ zip1-4A sgs1-md diploid and allow sgs1-md zip1-4A zmmΔ diploids to sporulate. This was not observed, either, as all of the sgs1-md zip1-4A zmmΔ diploids exhibited decreased sporulation (Fig 8A). A simple interpretation of these data is that these ZMM genes function after Zip1 4S phosphorylation.

Discussion
It has recently become clear that whether a meiotic DSB is repaired as a CO or NCO is due primarily to the interplay between the "anti-CO" activities of helicases and the "pro-CO" activities of the functionally diverse set of Zmm proteins [16]. Yet, how this interplay might be modulated was unclear. Our discovery that phosphorylation of serines S815-818 in the Zip1 C terminus is required for the ZMM pathway of CO formation suggests a mechanism by which the CO/NCO decision can be regulated.

Phosphorylation of the Zip1 C Terminus Initiates the ZMM Pathway of Interfering COs
The Zip1 protein contains N and C terminal globular domains flanking coiled-coil regions [12]. In the SC, TF proteins such as Zip1 form homo-oligomers in which the carboxy (C) termini are in contact with AEs, while the amino (N) termini interact within the central region [2]. While the Zip1 N terminus is unnecessary for synapsis [12], phosphorylation of serines 815-818 within the C terminus is required for SC formation. Failure to phosphorylate Zip1 S815-818 produces a canonical zmm phenotype. There is a delay in the formation of SEI and dHJ recombination intermediates. COs are reduced on small chromosomes, resulting in MI non-disjunction and decreased spore viability. Although CO defects are less obvious on larger chromosomes, CO interference is weakened or absent on both small and large chromosomes, indicating that ZMM-mediated COs are affected on all chromosomes. NCOs are increased in an SGS1-dependent manner, as expected if DSBs that cannot be repaired via the ZMM pathway are channeled by SGS1 into pathways that generate NCOs either through SDSA or unbiased HJ resolution. In addition, the increased spore viability observed in zip1-4A relative to zip1Δ is dependent upon MUS81. Therefore, Zip1 4S phosphorylation is required to antagonize the anti-CO function of Sgs1.

Zip1 4S Phosphorylation Functions after DSB Formation But Prior to Other ZMM Genes
In yeast, it has previously been hypothesized that the CO/NCO decision is made soon after DSBs are generated [13,24]. In WT cells, Zip1 phosphorylation coincides with the appearance of phosphorylated Hop1, an indirect marker of DSBs. Furthermore, Zip1 phosphorylation is promoted by Mek1 kinase activity, which is activated in response to DSBs [6,51] and reduced when DSBs are prevented by deletion of SPO11. We propose that Zip1 associates with DSBs (either directly or through interaction with other proteins) and that 4S phosphorylation then commits those breaks to the ZMM pathway (Fig 9A). In fact, association of Zip1 and other Zmm proteins at DSBs has recently been shown to be promoted by the 9-1-1 checkpoint clamp complex that binds to the single strand DNA-double strand DNA junctions of resected ends [64]. Epistasis analysis indicates that Zip1 4S phosphorylation functions prior to all of the other ZMM genes. Although this finding is at odds with cytological studies showing that ZIP2 and ZIP3 are required for the assembly of Zip1 along the SC, it is consistent with the fact that Zip3 localization specifically to DSB sites (measured by chromatin immunoprecipitation) is dependent upon ZIP1 [2,39]. We propose that Zip1 has two temporally distinct functions with regard to SC formation. Its early function is to localize to DSBs where phosphorylation of the 4S region creates a patch of negative charges at the C terminus that promotes Zip1 interaction with downstream proteins, thereby committing the DSB to an interfering CO fate (Fig 9A). The resulting protein complex, which includes Zip2 and Zip3, is then necessary to nucleate Zip1 polymerization to allow synapsis.
Diploids lacking both SGS1 and ZIP1 are able to sporulate, whereas sgs1-md zip1-4A strains are not. Furthermore, the increased number of NCOs in zip1-4A is dependent upon SGS1.

Zip1 Phosphorylation Regulates Meiotic Recombination
These results suggest that Sgs1 is able to remove Zip1-4A from DSBs, thereby allowing breaks to be repaired by alternative pathways. In sgs1-md strains, the DSBs associated with Zip1-4A cannot be repaired and trigger the meiotic recombination checkpoint (Fig 9B). We further hypothesize that there is competition for DSBs between Sgs1 and Zip1 and this competition can be biased towards the ZMM pathway by phosphorylation of Zip1. This provides a mechanism for influencing which pathway is preferred to implement CO homeostasis. For example, when DSBs are reduced, CO levels could be kept constant by increasing the fraction of DSBs containing phosphorylated Zip1.
The requirement for Zip1 4S Phosphorylation for Interhomolog Engagement Obscures the CO Defect for zip1-4A on Large Chromosomes In addition to their effect on CO regulation, ZMMs are required for synapsis and to mediate a type of IH engagement that results in feedback inhibition of Spo11 [13,14]. This is also true for Zip1 4S phosphorylation as zip1-4A mutants exhibit higher levels of DSBs that persist longer than WT. Interestingly, DSBs in zip1-4A are present at higher levels and for longer than zip1Δ as well. This suggests that, while zip1Δ cells eventually adapt to the lack of ZMM-mediated IH engagement and down-regulate Spo11, the presence of the Zip1-4A protein actively prevents this adaptation.
When IH engagement is defective, additional DSBs are introduced disproportionately onto larger chromosomes [14]. We propose that some of these extra DSBs are converted into COs via the alternative MUS81-dependent pathway and that the higher number of DSBs formed in zip1-4A on larger chromosomes consequently increases the number of COs, thereby providing an explanation for the higher spore viability of zip1-4A compared to zip1Δ. These results underscore the complexity in analyzing zmm mutant phenotypes. While zip1Δ and other zmms are commonly considered to be CO-defective, both zip1Δ and zip1-4A may exhibit WT (or greater) levels of COs depending on the chromosome that is being assayed. The fact that interference is reduced on all chromosomes, large and small, indicates that the ZMM pathway is not functioning on large chromosomes either, but the ZMM CO defect is obscured by the creation of COs along the MUS81-independent pathway that do not exhibit interference [35].
In addition to being required for recombination and synapsis, Zip1 also functions at centromeres. Prior to DSB formation, Zip1 promotes homology-independent centromere coupling, which is then regulated by Mec1 phosphorylation of S75 in the Zip1 N terminus [65,66]. Both zip1-4A and zip1-4D are functional for nonhomologous centromere coupling, as might be expected given that the 4S sequence is in the C terminal globular domain (S5 Fig). After SC disassembly, Zip1 remains at centromeres and helps promote segregation of nonexchange chromosomes [67,68]. While it is possible that the higher spore viability observed for zip1-4A compared to zip1Δ is due to this post-pachytene chromosome segregation function, this idea would not explain why the increased spore viability is dependent upon MUS81 and therefore seems unlikely.

C Terminal Phosphorylation of TF Proteins May Be a Conserved Mechanism for Regulating Recombination
Phosphorylation of Zip1 S816 is promoted by two very different kinases: the meiosis-specific kinase, Mek1, and the highly conserved cell cycle kinase, DDK. A number of experiments indicate that DDK1 is directly phosphorylating Zip1, while the requirement for Mek1 is indirect. First, S816 is not contained within a Mek1 consensus sequence (RXXT), making it unlikely to be a direct substrate. In contrast, the sequence of consecutive serines in the Zip1 C terminus provides a good target for the addition of multiple phosphates by DDK, which normally has a preference for phosphorylating amino acids immediately upstream of a negative charge [28,[55][56][57]. Second, DDK-as, but not Mek1-as, phosphorylates recombinant Zip1 in vitro. The use of analog-sensitive kinases with the appropriate ATP analogs ensures that the phosphorylation observed in vitro is directly due to the analog-sensitive kinase. The specificity for DDK-as is further supported by the fact that Zip1 S816 phosphorylation is abolished in vitro by the DDK-as inhibitor, PP1. Third, the bulk of S816 phosphorylation observed in vegetative cells ectopically expressing the Zip1 C terminus is dependent upon DDK-as kinase activity, confirming that this reaction occurs in vivo, as well as in vitro.
Phosphorylation of Zip1 S816 is complex, however, as noted by the surprising observation that ndt80Δ spo11Δ-arrested cells exhibited high levels of Zip1 S816 phosphorylation in the absence of DSBs, while phosphorylation was reduced~15-fold in NDT80 spo11Δ diploids. Like spo11Δ ndt80Δ, inactivation of DDK-as in meiosis prevents DSBs and meiotic progression [57,60]. The fact that S816 phosphorylation was not abolished but only partially reduced in these cells, even though DDK-as was inactive, indicates a second kinase can phosphorylate S816 when the cells are arrested. One possibility is that Zip1 phosphorylation simply correlates with the time that cells spend in meiotic prophase-so that mutants that cause delays in meiotic progression such as sae2Δ, dmc1Δ, and ndt80Δ allow the phosphorylated form of the protein to accumulate. An alternative explanation is that the ndt80Δ arrest creates a situation in which the WT regulation of Zip1 phosphorylation is no longer working. It may be that, given enough time, DDK and/or the second kinase can phosphorylate Zip1 everywhere along the chromosome. The discovery of unregulated Zip1 S816 phosphorylation may explain the surprising observation that in mek1Δ ndt80Δ diploids, IH JMs that are suppressed early in meiosis (as expected due to the lack of IH bias) accumulate at the ndt80Δ arrest [69]. This result also raises the question as to why Mek1 regulation of S816 was observed in the SILAC experiment in which cells were arrested using ndt80Δ. One potential explanation is that phosphopeptides identified in the SILAC experiment were derived from chromatin-associated proteins, while the western blots probed with the α-pS816 antibodies were from whole cell extracts, which do not provide any information as to location of the Zip1 that is getting phosphorylated.
DDK is an essential protein kinase with an evolutionarily conserved role in the initiation of replication [70]. Given that DDK directly phosphorylates Zip1 S816 and is responsible for the bulk of phosphorylation observed when the Zip1 C terminus is ectopically expressed in vegetative cells, a reasonable hypothesis is that DDK has a role in regulating CO formation in yeast, raising the question as to whether this role is also evolutionarily conserved. In fact, the TF proteins from mammals and zebrafish contain homology to the yeast 4S region. This conservation is especially striking because overall homology between TF proteins is low [2]. In addition, multiple residues within the 4S region are predicted to be phosphorylated in mammals and zebrafish. This, along with the fact that many of the yeast Zmm proteins have orthologs in mammals, supports the idea that this mechanism for CO regulation is conserved [71]. This mechanism may not be universal, however, as no homology was found in the TF proteins from plants, although plants appear to contain at least some ZMM genes (e.g., [72]). In addition, TF proteins from both D. melanogaster and C. elegans do not contain the 4S region. These organisms do not use recombination to initiate synapsis and therefore may not need the homolog interaction function of TF proteins mediated by phosphorylation [2].

MEK1-Dependent Phosphorylation of Zip1 4S Provides a Link between IH Bias and the CO/NCO Decision
DSBs trigger the localized activation of Mek1 on the axis. A role for Mek1 in promoting IH bias is well-established as inactivation of Mek1 results in IS repair [9,43]. This work shows that Mek1 regulates IH CO formation as well, by promoting phosphorylation of the C terminus of Zip1. Mek1 can therefore be considered a "master regulator" of meiotic recombination in that it phosphorylates substrates that both suppress IS recombination while at the same time promoting interfering CO formation.
The observation that Zip1 S816 is phosphorylated in vegetative cells but is induced by Mek1 and DSBs during WT meiosis suggests that Zip1 phosphorylation is constrained in meiotic cells and that Mek1 kinase activity overcomes this constraint. How Mek1 does this remains to be established. DSB formation is timed to occur after DNA replication [1]. DDK phosphorylation of Mer2 (SGD S000003782) promotes the recruitment of Spo11 to axes where DSBs occur. Recent work has shown that DDK travels with the replication fork, phosphorylating Mer2 and then moving on [73]. Since Zip1 is not phosphorylated until after DSBs have been made, it seems unlikely that replication fork-associated DDK is responsible. Therefore, one possibility is that Mek1 phosphorylation of itself or some other protein results in the recruitment of DDK (or potentially the alternative kinase) to DSBs, either by creating a "landing pad" for the kinase or by changing the chromosomal architecture to make breaks more accessible. For example, Mek1 has been proposed to antagonize Rec8-mediated cohesion locally around DSBs to release an "homology searching tentacle", which could then attract the kinase [9]. We propose that Mek1 kinase activity is the link by which IH bias and IH CO formation are coordinated to ensure proper chromosome segregation during meiosis.

Yeast Strains
The genotypes of all strains used in this work can be found in S2 Table. With the exception of experiments shown in S3 Fig and S5 Fig, all strains were derived from the SK1 background. Genes were deleted by polymerase chain reaction (PCR)-based methods using the kanMX6, natMX4, or hphMX4 markers that confer resistance to G418, nourseothricin, and Hygromycin B, respectively [74][75][76]. PCR fragments using Kluyveromyces lactis URA3 gene as a selectable marker were also used to make precise gene deletions (pKlU, A. Neiman, Stony Brook University). Unless stated otherwise, all deletions were confirmed using colony PCR.
To enable integration of the ZIP1 TRP1 (SGD S000002414) plasmid, p382 (generously provided by A. Hochwagen, New York University), the first 222 bp of the TRP1 open reading frame (ORF) were deleted from NHY1210 and NHY1215. The presence of the deletion was confirmed by failure to grow on SD-Trp medium. ZIP1 was then deleted with various markers, and the two haploids mated to make diploids. The p382 plasmid was targeted to integrate into the 3' end of the TRP1 gene by digestion with BsgI. All transformants were checked by PCR using internal ZIP1 primers to confirm the presence of the plasmid. The sgs1-md allele was introduced into the NHY1210 zip1 trp1 and NHY1215 zip1 trp1 haploids by replacing the SGS1 promoter with the GAL1 promoter using kanMX6 [75]. SPO11, SAE2 and various ZMM genes were then replaced with the K. lactis URA3 gene. Various zip1 alleles were introduced by integration of plasmids as described above into each haploid and mated to generate homozygous diploids.
CDC7 and cdc7-as were tagged with three copies of the FLAG epitope using pFA6a-3FLAG-kanMX6 [80] in S2683 and RKY1145 containing ndt80Δ::hphMX4. The haploids were then mated to make NH2082 and NNH2072.

Plasmids
Mutations were introduced into p382 and pBTM116-ZIP1C 750-875 (below) by site-directed mutagenesis (QuikChange kit, Stratagene) and confirmed by DNA sequencing by the Stony Brook University DNA Sequencing Facility. All alleles were sequenced in their entirety to ensure that no additional mutations were introduced. To ectopically express the C terminus of Zip1 in vegetative cells, a 383 bp fragment containing codons 749-875 was amplified by PCR. A BamHI site was engineered immediately upstream of ZIP1 codon 750 in frame with lexA in pBTM116 [82] with a PstI site located immediately after the ZIP1 stop codon. Ligation of the BamHI/PstI ZIP1C fragment into BamHI/PstI digested pBTM116 created pBTM116-ZIP1C-WT. The S815A S816A S817A S818 mutations were then introduced by site-directed mutagenesis. The pXC6 plasmid containing ADE2 mek1-K199R was constructed by subcloning a 2.9 kb NotI/SalI fragment from pLP34 [48] into NotI/Sal-digested pRS302. Digestion of pXC6 with StuI targets integration into ade2. pLT11 and pLW20 contain HOP1 URA3 and MEK1 ADE2, respectively [43]. For MBP-ZIP1 expression and purification, a 2.6 kb fragment of ZIP1 was amplified by PCR in which XbaI and HindIII restriction sites were engineered immediately upstream of the ATG or downstream of the stop codon, respectively. This fragment was then cloned into pMAL-TEV (provided by Patrick Sung, Yale University) to fuse ZIP1 to the gene encoding Maltose-Binding Protein (MBP) with a TEV cleavage site engineered between MBP and Zip1 to make pMAL-TEV-ZIP1-WT. pLW3 is a high copy number 2μ plasmid containing URA3 and GST-mek1-as [30].

Meiotic Time Courses
Cells were sporulated at 30°C as described in [53] with the exception of the time courses used for the physical analyses in Fig 2, which used 0.5% potassium acetate instead of 2% potassium as the Spo medium. Physical analysis of the HIS4/LEU2 hotspot was performed as described in [24]. Chromosome spreads were performed as described in [83]. All primary antibodies were diluted 1:100. The Zip1 antibodies used for cytology were provided by A. MacQueen (Wesleyan) (BR strains) or S. Keeney (Sloan-Kettering Cancer Research Center) (SK1 strains). Meiotic progression was analyzed by staining cells with DAPI and using fluorescent microscopy to count the number of mononucleate, binucleate (MI), and tetranucleate (MII) cells. Two hundred cells were counted for each time point of each strain in each replicate.

Identification of Phosphorylation Sites on Zip1
SILAC experiments were performed as described in [29] by growing the mek1-as ndt80Δ diploid, NH2221, in synthetic medium containing either arginine/lysine or arginine/lysine with stable heavy isotopes of 13 C and 15 N and then transferring the cells to 2% potassium acetate for 14 h to allow cells to arrest at pachytene [29]. 1-NA-PP1 was added to a final concentration of 1 μM to the heavy culture for 20 min to inactivate Mek1-as. Phosphopeptides were enriched and analyzed as described in [29].

Western Blots and Phosphatase Experiments
Protein extracts were prepared using the tri-chloroacetic acid (TCA) method as described by [65]. For phosphatase experiments, a modification of a protocol from Pedro San-Segundo was used [86]. 60 μl of TCA extracts in 1 X Laemmli sample buffer were diluted~1:10 with 525 μl 1 X PMP buffer (50 mM HEPES, pH 7.5, 100 mM NaCl, 2 mM dithiolthreitol) from which 195 μl was aliquoted into two tubes. Alkaline phosphatase (5 μl) (Roche, #11097075001) was added to one tube, and 5 μl 1 X PMP buffer was added to the other tube. After incubation at 30°C for 30 min, the samples were put on ice and reprecipitated with TCA prior to fractionation by SDS-polyacrylamide gel electrophoresis and immunoblot analysis. After incubation with the appropriate secondary antibodies coupled to horse radish peroxidase (HRP-see below), the blots were developed using the Pierce ECL2 western blotting substrate (Thermo Scientific) and exposed to film.
To compare the fraction of S816 phosphorylated Zip1 in different strains, two sets of identical samples were run on the same gel and blotted to a membrane. The membranes were then cut vertically to separate the two sets and then horizontally to separate the Zip1 proteins from Arp7. The top half of each blot was probed with either total Zip1 or α-pS816 antibodies and the bottom halves with α-Arp7 antibodies. Quantification was performed by scanning the films using a Canon LiDE 110 scanner and analyzing the bands using ImageQuant TL 7.0 software. Both total Zip1 and pS816 signals were normalized by the amount of Arp7 loaded in the same lane, and the percent of phosphorylated Zip1 was calculated by pS816/Arp7 pS816 /Total Zip1/Arp7 Total Zip1 Ã 100.

Antibodies
Antibodies specific for phosphorylated Zip1 S816 were generated by injecting a rabbit with the peptide, Ac-CDEFDLS(pS)SSND-amide (Covance Research Products). To isolate p-S816 antibodies, the serum was first passed over a column containing the unphosphorylated peptide. Antibodies that bound to the column (called 1907) were eluted and used to detect Zip1 lacking S816 phosphorylation. The flow through was loaded onto a column containing the phosphorylated peptide. These Zip1 phospho-S816-specific antibodies (called 1907-p) were eluted and stored at −80°C. The 1907 and 1907-p antibodies were used at 1:250 and 1:2,000 dilution, respectively.

Protein Purification
GST-Mek1-as was partially purified from NH520/pLW3 as described in [53]. Cdc7-3FLAG and Cdc7-as-3-FLAG were partially purified out of the ndt80Δ diploids, NH2082 and NH2072, respectively, as described in [88]. Eight liters of cells incubated for 8 h in Spo medium at 30ºC were used to make protein extracts, which were then frozen in 10 ml aliquots. One 10 ml aliquot was used for the FLAG antibody purification. The Rad54 protein was purified as described in [89] and was generously supplied by Patrick Sung. pH 7.4, 10% glycerol, 0.5 mM EDTA, 0.01% Igepal and 1 mM DTT) containing a cocktail of protease inhibitors (aprotinin, chymostatin, leupeptin, and pepstatin A at 5 μg/ml each and phenylmethylsulfonyl fluoride at 1 mM) and 500 mM KCl. The lysate was cleared by ultracentrifugation (100,000 X g for 45 min) and then incubated with 1 mL of Amylose resin (New England Biolabs) for 2 h. The matrix was washed three times with 20 mL of K buffer containing 500 mM KCl and 0.1% Igepal before MBP-Zip1 protein was eluted with 5 mL of K buffer containing 500 mM KCl and 10 mM Maltose. The eluate was further concentrated and filter dialyzed against K buffer plus 500 mM KCl. The purified MBP-Zip1 protein was frozen in liquid nitrogen and stored at −80°C as small aliquots.

Spo11 Oligonucleotide Analysis
Diploids were made fresh by mating haploids of the appropriate genotypes immediately prior to each experiment. Extracts for Spo11 oligo time courses were prepared as in [90] with the following modification: immunoprecipitations utilized 10 μl IgG Sepharose Fast Flow beads (GE) per 0.2 ml of extract. Immunoprecipitations were then performed exactly as in [14]. A. Time courses were performed at 30°C. Isogenic diploids containing either zip1Δ (NH2297::pRS304), ZIP1 (NH2297::p382), zip1-4A (NH2297::p382-4A), or ZIP1/zip1-4A (NH2297::p382::p382-4A) were sporulated, and protein extracts were generated for the indicated time points. Equal amounts of extract were loaded for each time point for each strain. The gels were cut in half and the top half was probed with antibodies that recognize total Zip1, while the bottom half was probed with Arp7 as a loading control. B. Zip1 S816 phosphorylation in sgs1-md diploids containing different zmm mutants. Protein extracts from time courses from the sgs1-md diploids containing either mer3Δ (NH2267::p382)), msh5Δ (NH2265::p382), or zip3Δ (NH2243::p382) were probed with either the pS816 phospho-specific antibody or the total Zip1 antibody. The Arp7 loading control was performed as described (p382-4A), or pRS304 (vector). Twenty hours after transfer to Spo medium at 30°C, chromosomes were spread and stained with anti-myc (9E10) and anti-Red1 antibodies [30]. Spreads with substantial Red1 staining, indicating mid-prophase nuclei, were scored for the number of Ctf19 foci. Because haploids contain 16 nonhomologous chromosomes, pairing of nonhomologous centromeres should result in eight foci, as is observed for ZIP1 [65,66]. In contrast, zip1Δ exhibits closer to the 16 foci expected if centromeres are unpaired. Both zip1-4D and zip1-4A appear to be fully functional for nonhomologous centromere association. B. Homologous centromere associations: Chromosomes from CTF19-myc zip Δ (NH2247) carrying various ZIP1 alleles (ZIP1, zip1D, zip1-4D and zip1-4A) were spread 20 h after transfer to Spo medium and stained with anti-myc and anti-Red1 antibodies. Nuclear spreads with linear Red1 staining, indicating mid-prophase, were scored for the number of Ctf19 foci. Like zip4Δ, zip2Δ diploids exhibit~16 foci since the chromosomes are homologously aligned and their centromeres are associated via Zip1 [66]. The zip1Δ zip2Δ mutant forms~24 foci, consistent with the requirement for ZIP1 to hold homologous centromeres together. This number is less than the 32 foci expected if all chromosomes are unassociated, presumably due to centromere proximal axial associations. zip1-4D zip2Δ, and zip1-4A zip2Δ exhibit Ctf19 focus numbers similar to zip2Δ, indicating these alleles are functional for homologous centromere association during prophase. (TIF) S1