A ZIP1 Separation-of-Function Allele Reveals that Meiotic Centromere Pairing Drives Meiotic Segregation of Achiasmate Chromosomes in Budding Yeast

In meiosis I, homologous chromosomes segregate away from each other - the first of two rounds of chromosome segregation that allow the formation of haploid gametes. In prophase I, homologous partners become joined along their length by the synaptonemal complex (SC) and crossovers form between the homologs to generate links called chiasmata. The chiasmata allow the homologs to act as a single unit, called a bivalent, as the chromosomes attach to the microtubules that will ultimately pull them away from each other at anaphase I. Recent studies, in several organisms, have shown that when the SC disassembles at the end of prophase, residual SC proteins remain at the homologous centromeres providing an additional link between the homologs. In budding yeast, this centromere pairing is correlated with improved segregation of the paired partners in anaphase. However, the causal relationship of prophase centromere pairing and subsequent disjunction in anaphase has been difficult to demonstrate as has been the relationship between SC assembly and the assembly of the centromere pairing apparatus. Here, a series of in-frame deletion mutants of the SC component Zip1 were used to address these questions. The identification of separation-of-function alleles that disrupt centromere pairing, but not SC assembly, have made it possible to demonstrate that centromere pairing and SC assembly have mechanistically distinct features and that prophase centromere pairing function of Zip1 drives disjunction of the paired partners in anaphase I. AUTHOR SUMMARY The generation of gametes requires the completion of a specialized cell división called meiosis. This division is unique in that it produces cells (gametes) with half the normal number of chromosomes (such that when two gametes fuse the normal chromosome number is restored). Chromosome number is reduced in meiosis by following a single round of chromosome duplication with two rounds of segregation. In the first round, meiosis I, homologous chromosomes first pair with each other, then attach to cellular cables, called microtubules, that pull them to opposite sides of the cell. It has long been known that the homologous partners become linked to each other by genetic recombination in a way that helps them behave as a single unit when they attach to the microtubules that will ultimately pull them apart. Recently, it was shown, in budding yeast and other organisms, that homologous partners can also pair at their centromeres. Here we show that this centromere pairing also contributes to proper segregation of the partners away from each other at meiosis I, and demonstrate that one protein involved in this process is able to participate in multiple mechanisms that help homologous chromosomes to pair with each other before being segregated in meiosis I.

correlated with proper segregation, as zip1 deletion mutants have no centromere pairing and also 86 segregate achiasmate partners randomly (Fig. 1A)  other studies (15) have suggested that in the SC, Zip1 is in the form of head-to-head dimers ( Fig.  90 1 B). These dimers, in turn are thought to assemble in a ladder-like structure with the N-termini 91 in the center of the SC and the C-termini associated with the axes of the homologous partners 92 (Fig. 1 B). This model has been extrapolated to other organisms because the basic structure of 93 transverse filament components, like Zip1, are believed to be conserved even though their amino 94 acid sequences have diverged (reviewed in (16)). 95 Tung and Roeder (1998) used an ordered series of in-frame deletions of ZIP1 to identify 96 ways in which different regions of the protein contributed to SC structure and function ( Fig. 1  97 C). This was before the discovery that Zip1 is also involved in promoting centromere coupling 98 and centromere pairing. We have re-constructed this deletion series to evaluate the ways in 99 which different regions of Zip1 contribute to these centromere-associated functions. This 100 information could be used to reveal relationships in the underlying mechanisms of centromere 101 coupling, centromere pairing and SC assembly, and identify to separation-of-function alleles that 102 would reveal more specifically contributions made to these processes by Zip1. These approaches 103 make clear that centromere coupling, centromere pairing, and SC assembly all require certain 104 parts of the Zip1 protein that are not required by the others -suggesting mechanistic differences 105 in these phenomena. Second, they provide a clear demonstration that centromere pairing in 106 prophase, distinct from other SC-related functions of Zip1, drives disjunction of achiasmate 107 partner chromosomes in anaphase I. 108 109 RESULTS 110 The N and C terminal globular domains of Zip1 are essential for centromere coupling. 111 A series of nine in-frame deletion mutants ( Fig. 1 C) were tested to determine which 112 regions of the ZIP1 coding sequence are essential for the homology independent centromere 113 coupling that occurs in early meiotic prophase. Centromere coupling was assayed by monitoring 114 the numbers of kinetochore foci (Mtw1-MYC) in chromosome spreads from prophase meiotic 115 cells (10, 12) ( Fig. 2 A). Diploid yeast have sixteen pairs of homologous chromosomes. When 116 the centromeres of the thirty-two chromosomes are coupled they form on average sixteen Mtw1-117 MYC foci (Fig. 2 B, ZIP1, blue line). Mutants that are defective in coupling exhibit higher 118 numbers of Mtw1-MYC foci (Fig. 2 B, zip1∆, red line). The experiment was done in strains 119 lacking SPO11, which encodes the endonuclease responsible for creating programmed double 120 strand DNA (17)). This blocks meiotic progression beyond the coupling stage and prevents the 121 homologous alignment of chromosomes (12, 18). The strains also featured GFP-tagged copies of 122 the centromeres of chromosome I. Briefly, 256 repeats of the lac operon sequence was inserted 123 adjacent to the centromere of chromosome I (CEN1) and the cells were engineered to express 124 lacI-GFP, which localizes to the lacO array (19). In the centromere coupling stage, the two 125 CEN1-GFP foci are nearly always separate because coupling is usually between non-126 homologous partner chromosomes ( Fig. 2 A) (10). 127 The mutants could be assigned to one of three groups based on their coupling phenotypes 128 (Fig. 2  III, allowing the plasmids to behave as single copy mini-chromosomes in yeast. One plasmid is 147 tagged with tdTomato-tetR hybrid proteins at a tet operon operator array (21), the other is tagged 148 with GFP, as described above for chromosome I. Previous work has shown that such achiasmate 149 model chromosomes disjoin properly in most meioses (22-24) and this segregation at anaphase I 150 is correlated with the ability of their centromeres to pair late in prophase (5). To increase the 151 synchrony of meiotic progression in this experiment NDT80, which promotes the transition out 152 of prophase and into pro-metaphase, was placed under the control of an estradiol-inducible 153 promotor (25-27). Meiotic cells were allowed to accumulate in pachytene of prophase, then 154 induced to synchronously exit pachytene and enter pro-metaphase. We scored segregation of the 155 plasmids in the first meiotic division by monitoring the location of their GFP and tdTomato-156 tagged centromeres in anaphase I cells, identified by their two separated chromatin masses (Fig 3  157 A). 158 Wild-type cells, under these conditions, exhibited 28% non-disjunction of the CEN 159 plasmid pair (Fig. 3 B). The loss of Zip1 function can result in a pachytene arrest in some strain 160 backgrounds (28) including the strain used in these experiments. Reducing the sporulation 161 temperature to 23°C, as was done here, can permit a partial bypass of the arrest (28). Still several 162 of the mutations (zip1Δ, zip1-C2, zip1-C1, and zip1-NM2) yielded very few anaphase cells, and 163 failed to sporulate, presumably due to the pachytene arrest. These observations are consistent 164 with previously published work (14). Of the remaining mutants, the zip1-N1 mutant showed 165 significantly elevated non-disjunction of the centromere plasmids (Fig. 3 B). The zip1-N1 mutant 166 exhibits only mild defects in progression through meiosis, SC formation, sporulation efficiency, 167 and the segregation of chiasmate chromosomes (14) and Figure S1), suggesting that amino acids 168 23-163 are more critical for mediating the segregation of achiasmate partners than for SC 169 assembly and function. 170 Because achiasmate segregation is correlated with prior centromere pairing (7, 8), we 171 tested whether the zip1-N1 mutants were proficient in centromere pairing. Wild-type and zip1-N1 172 cells containing the GFP and tdTomato tagged centromere plasmids were induced to sporulate 173 and harvested five -seven hours later when pachytene cells are prevalent. Chromosome spreads 174 were then prepared and the distance between the tdTomato and GFP foci were measured in 175 spreads exhibiting the condensed chromatin typical of pachytene cells (Fig. 4 A). The average 176 centromere-centromere distance was significantly greater in zip1-N1 mutants (Fig. 4 B) 177 consistent with a loss of pairing. When spreads with an inter-centromere distance of less than 0.6 178 μm were scored as "paired" (see example in Fig. 4 A), the zip1-N1 mutation was found to exhibit 179 a significant reduction in the frequency centromere pairing between the achiasmate plasmids 180 (Fig. 4 C). 181 182 The N-terminus of Zip1 is necessary for efficient localization to kinetochores 183 Failure of centromere pairing in the zip1-N1 mutant could be due to a failure of Zip1 to 184 associate with centromeres. To test this, we analyzed the co-localization of the Zip protein with 185 kinetochores in ZIP1 and zip1-N1 strains. The experiments were done in a zip4 strain 186 background to allow visualization of Zip1 localization independently of an SC structure. Images 187 were collected using structured illumination microscopy and the level of co-localization was 188 determined using ImageJ software (see Materials and Methods). Every ZIP1 spread analyzed 189 showed significantly more co-localization of Zip1 and Mtw1 than was found in a randomized 190 sample ( Fig. 5 A), consistent with earlier work (9, 10, 12), while many of the zip1-N1 spreads 191 showed no significant co-localization above the randomized control ( Fig. 5 B). Consistent with 192 these results, zip1-N1 strains showed significantly lower levels of co-localization with Zip1 than 193 was seen in ZIP1 strains ( Our analysis of a set of in-frame Zip1 deletions has added to our understanding of the 211 functional domains of the Zip1 protein, helping to ascribe particular Zip1 functions to specific 212 regions of the protein. Zip1 is critical for SC assembly and processes that depend on SC 213 assembly, including crossover formation and progression through pachytene (28). More recently 214 it has become clear that Zip1 acts at centromeres both early in prophase, where centromeres 215 become associated in a homology-independent fashion (centromere coupling), and later when 216 homologous centromeres, or the centromeres of achiasmate chromosomes, become associated by 217 remnants of the SC that remain at the centromeres after SC disassembly (reviewed in (30)). The 218 experiments here were intended to clarify whether SC assembly, centromere coupling, and 219 centromere pairing incorporate Zip1 in the same or different mechanisms, and if there are 220 differences in the regions of Zip1 that are critical to each function. 221 222 Centromere coupling and SC assembly 223 Prior work has shown convincingly that the structure that mediates centromere coupling 224 is distinct from mature SC (9, 10, 20, 31). Several proteins (Zip2, Zip3, Zip4, Ecm11, Gmc2, and 225 Red1) known to be essential for SC assembly are not required for centromere coupling. But the 226 domains of Zip1 that are required for centromere coupling have not been defined. The 227 experiments here reinforce that the requirements for Zip1 for centromere coupling and SC 228 assembly are quite different. First, centromere coupling was proficient in zip1-C2 mutants, which 229 have severe defects in SC assembly. But these mutants exhibit little Zip1 expression, which may 230 be due to the lack of a nuclear localization signal (32). Thus, this result is difficult to interpret 231 other than to suggest that centromere coupling may require far less Zip1 than does SC assembly. 232 Notably, the zip1-M1 mutation, which also blocks SC assembly, is proficient in centromere 233 coupling. The zip1-M1 mutation, which eliminates amino acids 244-511, has a unique SC defect. 234 The Zip1-M1 protein efficiently localizes to the axes of the homologous partners, but does not 235 efficiently cross-bridge the axes (Fig. 1 C; (14)). This defect may reflect an inability of Zip1 236 molecules from opposite axes to associate with one another (as in Fig.1 B) or may reflect an 237 inability of Zip1 to associate with central element proteins that promote or stabilize the cross-238 bridging of axes by Zip1. In either case, such cross-bridging must not be important for 239 centromere coupling, and is consistent with the finding that the central element proteins Ecm11 240 and Gmc2 are also not required for centromere coupling (31). Together these findings suggest 241 that centromere coupling is probably not mediated by a structure that includes SC-like cross-242 bridging. The only protein, beyond Zip1, that is known to be required for centromere coupling is 243 the cohesin component Rec8 (9) (the requirements for the other cohesin subunits have yet to be 244 reported). It may be that centromere coupling is mediated by the cohesin-dependent 245 accumulation of Zip1 at early prophase centromeres (9, 29), followed by interactions between 246 Zip1 molecules that promote the association of centromere pairs. 247

Centromere pairing and SC assembly 248
Experiments performed mainly in a mouse spermatocyte model (3, 6) suggest that the SYCP1 249 (the functional homolog of Zip1) that persists at paired centromeres, after SC disassembly, is 250 accompanied by other SC proteins. This suggests that centromere pairing could be mediated by a 251 conventional SC structure. But the identity of regions of Zip1 that are critical for centromere 252 pairing, and whether they are distinct from the regions necessary for SC assembly, have not been 253 addressed. Our work suggests that there are significant differences in the requirements for Zip1 254 function in centromere pairing and SC assembly. We arrive at this conclusion following an 255 evaluation of the centromere pairing phenotypes of the zip1-N1 in-frame deletion. Prior work had 256 shown this allele had no measurable differences from the wild-type ZIP1 allele in spore viability, 257 crossover frequency, and genetic interference, and a slight defect in the continuity of mature 258 linear SC structures (14). In our strain background the zip1-N1 mutation also exhibited wild-type 259 levels of spore viability, and structured illumination microscopy confirmed the slight 260 discontinuity in some SC structures in the zip1-N1 background (Fig. S1). However, in 261 centromere pairing assays the zip1-N1 mutants showed major defects. In the zip1-N1 mutant the 262 centromeres of natural chromosome bivalents were more likely to become disengaged in 263 chromosome spreads than was seen with wild-type controls, but the defect was not as severe as is 264 seen in zip1 strainssuggesting that there are regions outside of the N1 region that also 265 promote association of the bivalent centromeres. It could be that these other regions are 266 influencing things like cross-over frequency or distribution, that along with centromere-pairing 267 help keep bivalent centromeres associated in the natural chromosome pairing assays. When we 268 used achiasmate centromere plasmids, in which such functions cannot contribute to centromere 269 association, then the zip1-N1 phenotype becomes severe. The zip1-N1 mutant showed a dramatic 270 reduction in the pairing of plasmid centromeres. The fact that the Zip1-N1 protein is proficient 271 for SC assembly but highly defective in centromere pairing suggests that the N-terminus imbues 272 functions on the protein that are specifically required for centromere pairing. The mechanism of 273 centromere pairing remains unclear as does the role of the Zip1 N-terminus, but kinetochore co-274 localization experiments suggest that this region of Zip1 promotes localization to, or 275 maintenance of, Zip1 at the centromeres in late prophase. The fact that early prophase 276 centromere coupling is normal in zip1-N1 mutants reinforces that coupling and pairing are 277 fundamentally distinct processes and that the N1 region is not necessary for localization of Zip1 278 to centromeres in early prophase when coupling occurs. impacts other processes such as synapsis, crossover formation, genetic interference, and the 288 pachytene checkpoint, making it impossible to formally name centromere pairing, and not some 289 other SC-related function as the driver of achiasmate segregation. The zip1-N1 separation-of-290 function allele, because it is largely wild-type for these other functions of Zip1, has made it 291 possible to demonstrate in a compelling way that centromere-pairing in prophase is a requisite 292 step in a process that mediates the segregation of achiasmate partners in anaphase. 293 The mechanistic question of how prophase centromere pairing drives disjunction remains 294 to be answered. The fact that in yeast, mice and Drosophila, the majority of the centromeric SC 295 components have been lost from the centromeres well before the partners begin to attach to 296 microtubules makes this even more mysterious. The zip1-N1 allele, which specifically targets 297 centromere associations of Zip1, and the centromere pairing process, will be an important tool 298 for addressing these questions. 299 300

301
Strains 302 We created the same nine deletion mutants of ZIP1 that Tung and Roeder had studied for their 303 work in SC formation (14) by using standard PCR and two-step-gene-replacement methods (33, 304 34). All mutant versions of ZIP1 were confirmed by PCR and sequencing. The native ZIP1 305 promoter was unaltered in these strains allowing each mutant protein to be expressed at the 306 appropriate level and time. Culturing of strains was as described previously (20). Strain 307 genotypes are listed in Table S1. 308

309
Centromere coupling assay 310 Centromere coupling was monitored largely as described previously (12 Centromere pairing in pachytene was assessed using published methods (7) but with the 359 centromere plasmids described above. Sporulation was done at 30°C. Chromosome spreads were 360 prepared as described in (37), with the following modifications: Cells were harvested 5-7 hours 361 after induction of sporulation at 30°C. After chromosome spreads were created and dried 362 overnight, the slides were rinsed gently with 0.4% Photoflo (Kodak). Each slide was then 363 incubated with PBS/4% milk at room temperature for 30 minutes in a wet chamber. Milk was 364 drained off of the slide, and primary antibody diluted in PBS/4% milk was incubated on the slide 365 overnight at 4°C. A control slide with PBS/4% milk was used for each experiment. The 366 following day, the slides were washed in PBS, and incubated with secondary antibody diluted in 367 PBS/4% milk for 2 hours in a wet chamber at room temperature. The slides were gently washed 368 in PBS. DAPI (4',6-diamidino-2-phenylindole, used at 1µg/ml) was added to each slide and 369 allowed to incubate at room temperature for 10 minutes. Slides were then washed gently in PBS 370 and 0.4% Photoflo, then allowed to dry completely before a coverslip was mounted. Antibodies 371 are described in the previous section. Only cells that exhibited "ropey" DAPI staining were 372 scored in this assay, and were disqualified for assessment if there was more than one GFP focus 373 or more than one tdTomato focus. In these cells, the distance between the center of the green 374 focus and the center of the red focus was measured using AxioVision software. The distributions 375 of distances in the ZIP1 and zip1-N1 strains were determined to be significantly different with 376 the Kolmogorov-Smirnov test (Kolmogorov-Smirnov D=0.4032; P=0.0002) using the Prism 6.0 377 software package. As in previous work (7), foci with center-to-center distances less than or 378 equal to 0.6 µm were scored as paired (these foci are typically touching or overlapping). The 379 frequency of pairing (distance less than 0.6 µm) in the ZIP1 (32 of 50) and zip1-N1 (14 of 63) 380 chromosome spreads was found to be significantly different (p<0.0001) using Fisher's Exact test 381 performed with the Prism 6.0 software package. 382 Synaptonemal complex evaluation by structured illumination microscopy. 383 Chromosome spreads were prepared according to the protocol of Grubb and colleagues 384 (37) as described above, and harvested from sporulation cultures five hours after placing cells in 385 sporulation medium at 30°C. To visualize the axial elements (Red1) and transverse elements 386 (Zip1) of the SC by indirect fluorescence microscopy, chromosome spreads were stained with 387 following primary and secondary antibodies: guinea pig anti-Red1 antibody (1:1000), goat anti-388 Guinea pig Alexa 488 antibody (Invitrogen) (1:1000), and rabbit anti-Zip1 antibody (1:800), 389 donkey anti-rabbit Alexa 568 antibody (Invitrogen) (1:1000). Chromosome spreads were imaged 390 with a Deltavision OMX-SR structured illumination microscope (SIM). 391 Mtw1-Zip1 co-localization assay 392 Chromosome spreads were prepared according to the protocol of (37) as described above. 393 All strains carried the zip4 to prevent SC assembly. Chromosomes were stained with primary 394 antibodies: mouse anti-MYC (Mtw1-13xMYC) (Developmental Studies Hybridoma Bank) at 395 1:20 dilution and rabbit anti-Zip1 antibody at 1:1000 dilution and secondary antibodies Alexa 396 488 donkey anti-mouse (Invitrogen) at 1:1000 dilution and Alexa 568 goat anti-rabbit 397 (Invitrogen) at 1:1000 dilution. Chromosome spreads were imaged with a Deltavision OMX-SR 398 structured illumination microscope (SIM). Acquired images were converted to binary images 399 using ImageJ software and the number of overlapping Mtw1-13xMYC and Zip1 foci were 400 scored using the imageJ plugin, JACoP. To determine whether co-localization occurred at 401 frequencies that were significantly higher than expected for random overlaps given the number 402 of Mtw1 and Zip1 foci in each image, the foci in each image were randomized in one thousand 403 simulations, then the frequency of random overlaps was determined and compared to the 404 observed overlap. Costes' P-value was then calculated to evaluate the statistical significance of 405 the difference between the frequency of observed versus random overlap (38). In addition, the 406 average co-localization observed for all of the ZIP1 spreads (26 spreads, 238 Mtw1 foci, 33 co-407 localized with Zip1) and all of the zip1-N1 spreads (18 spreads, 279 Mtw1 foci, 12 co-localized 408 with Zip1) was determined and the statistical significance of the difference determined using 409 Fisher's two-tailed exact test (p=0.0001). The experiment presented is one of two performed, 410 both with the same outcome (significantly reduced Mtw1-Zip1 co-localization in the zip1-N1 411 mutant). 412 Centromere pairing of natural chromosomes 413 The chromosome spreads used in the experiment above were used to assay the number of 414 distinct Mtw1-13xMYC foci in ZIP1, zip1-N1 and zip1 chromosome spreads. With complete 415 pairing of the homologous chromosomes, the thirty-two kinetochores should appear as sixteen 416 Mtw1-13xMYC foci. In the absence of pairing, kinetochores from the paired homologs can 417 sometimes separate far enough to be resolved as individual foci (the homologs remain tethered 418 by crossovers and probably other constraints), thus giving higher numbers of Mtw1-13xMYC 419 fociin theory up to thirty-two foci. The SIM images described in the preceding section were 420 converted to binary images using ImageJ software and the number of Mtw1-13xMYC foci tallied 421 for each spread using the Analyze Particles function in ImageJ. The average number of Mtw1-422 13xMYC foci per spread was determined for each genotype (ZIP1, zip1-N1, and zip1) and the 423 statistical significance of the observed differences between the genotypes was calculated with 424 one-way ANOVA and multiplicity adjusted P values were obtained with Sidak's multiple 425 comparisons testing using Prism 7.0. 426 427 Acknowledgements 428 We thank Marta Kasperzyk for the guinea pig Red1 antibody, Rebecca Boumil for the mouse 429 anti-Zip1 antibody, and Jingrong Chen and Susannah Rankin for help with antibody preparation. 430 We thank Lori Garman for help with statistical analysis. We are grateful to colleagues in the 431 Program in Cell Cycle and Cancer Biology for helpful discussions of this work. The work was 432 supported by NIGMS grant R01GM087377 to DD.

535
The axial element protein is shown in green and Zip1 is shown in Red. Each panel 536 presents representative spreads from A. ZIP1, B. zip1D and C. zip1-N1 strains.

537
Panels to the right are larger images of individual chromosomes. The results in our 538 strains are in keeping with the more comprehensive previous study of SC assembly in 539 zip1-N1 mutants (Tung and Roeder, 1998) in that the zip1-N1 strain exhibited slightly 540 less continuous Zip1 staining in pachytene-like spreads than was observed with the 541 wild-type control strain. It is not clear if this reflects a slight reduction in assembly 542 kinetics, or reduced continuity of the Zip1 in the mature SC of the zip1-N1 strain. D.

543
Tetrads were dissected to assess spore viability in ZIP1 and zip1-N1 strains. Though in 544 this sample set the zip1-N1 exhibited slightly lower spore viability than the wild-type 545 control, as in prior studies (Tung and Roeder, 1998)   Zip1 (orange) mediates centromere coupling (green arrowheads) between non-homologous 556 partner chromosomes (light blue and purple). As the cell proceeds through later stages of 557 meiosis, homologs pair and the mature synaptonemal complex (SC) structure zips the 558 chromosomes together. After pachytene, the SC disassembles, except at the centromeres (blue 559 arrowhead). B. The Zip1 protein is predicted to have globular domains at its ends spanning a 560 longer coiled-coil and forms parallel dimers with N-termini in the center of the SC (denoted by 561 N) and the C-termini along the axial elements (denoted by C). C. We evaluated the same nine 562 ZIP1 deletion mutants previously described by Tung and colleagues (Tung & Roeder, 1998 Table S2. 579 580 Figure 3. Centromere plasmid disjunction requires the N-terminus of Zip1. A. 581 Representative binucleate cells with disjoined (a ZIP1 cell) and non-disjoined (a zip1-N1 cell) 582 centromere plasmids. The segregation of CEN plasmids in anaphase I was assessed by 583 monitoring the tetR-tdTomato and lacI-GFP foci localized to tet and lac operator repeats, 584 respectively, inserted into a plasmid that contains 5.