Centromere clustering stabilizes meiotic homolog pairing

During meiosis, each chromosome must selectively pair and synapse with its own unique homolog to enable crossover formation and subsequent segregation. How homolog pairing is maintained in early meiosis to ensure synapsis occurs exclusively between homologs is unknown. We aimed to further understand this process by utilizing a unique Drosophila meiotic mutant, Mcm5A7. We found that Mcm5A7 mutants are proficient in homolog pairing at meiotic onset yet fail to maintain pairing as meiotic synapsis ensues, causing seemingly-normal synapsis between non-homologous loci. This pairing defect corresponds with a reduction of SMC1-dependent centromere clustering at meiotic onset. Overexpressing SMC1 in this mutant significantly restores centromere clustering, homolog pairing, and crossover formation. These data indicate that the initial meiotic pairing of homologs is not sufficient to yield synapsis between exclusively between homologs and provide a model in which meiotic homolog pairing must be stabilized by SMC1-dependent centromere clustering to ensure proper synapsis.

formation between homologs (reviewed in Page and Hawley 2004). 52 Perhaps the most enigmatic event within early meiosis is the mechanism by which 53 a meiotic chromosome selectively pairs and synapses with its unique homologous 54 partner. Initial homolog pairing is believed to be facilitated through early meiotic 55 chromosome movement and telomere or the centromere clustering (for reviews, see In Drosophila, eight centromeres aggregate into one or two diffraction-limited clusters at 182 the onset of meiosis, which is defined cytologically as zygotene. Centromeres remain 183 clustered through pachytene (Takeo et al. 2011). To determine whether centromere 184 clustering at the onset of meiosis is associated with initial homolog pairing, we quantified 185 the foci number of CID, the CENP-A homolog (Henikoff et al. 2002), in zygotene nuclei in 186 wild-type and Mcm5 A7 mutants at zygotene (Figure 4a). We observed a mean of 2 CID 187 foci in wild-type, demonstrating centromere clustering. In Mcm5 A7 mutants, we see a 188 significant increase in CID foci, with a mean of 4.8 per nucleus (p < 0.001, unpaired T-189 test). These results show that in Mcm5 A7 mutants, centromeres are not heterologously 190 clustered entering meiosis, even though chromosome arms are paired. 191 Next, we determined whether centromeres cluster in pachytene in Mcm5 A7 192 mutants. As shown in Figure 4b, we observe a mean of 1.7 CID foci in early pachytene 193 nuclei of wild-type, compared to 5.6 foci in Mcm5 A7 mutants (p < 0.001, unpaired T-test). 194 In mid-pachytene, Mcm5 A7 mutants exhibit a mean of 5.2 CID foci, significantly higher 195 than wild-type (1.6 CID foci; p < 0.001, unpaired T-test) (Figure 4c). We conclude that 196 centromere clustering is perturbed throughout early and mid-pachytene in Mcm5 A7 197 mutants.
In the regions assessed, we observed more than four CID foci in most Mcm5 A7 199 nuclei (Figure 4a,  Using Mcm5 A7 mutants, we observed that a decrease in centromeric SMC1 at meiotic 238 onset is associated with a reduction in meiotic centromere clustering and homologous 239 chromosome pairing in pachytene, but not chromosome pairing in zygotene. Thus, we 240 hypothesized that the centromeric-SMC1 defect at meiotic onset causes the reduction in 241 centromere clustering, and that centromere clustering defects cause the defect in pairing 242

maintenance. 243
To test this hypothesis, we attempted to restore SMC1 localization at the meiotic 244 centromere in Mcm5 A7 mutants by exogenously expressing SMC1 (Gyuricza et al. 2016) 245 in the background of Mcm5 A7 (nos>Smc1; Mcm5 A7 ) (Supplemental Figure 5). Using 246 quantitative microscopy, we found that centromeric-SMC1 is significantly higher in 247 nos>Smc1; Mcm5 A7 than in Mcm5 A7 mutants at meiotic onset (***p < 0.0001, unpaired T-248 test) (Figure 6a). We next assayed centromere clustering at early pachytene, when we 249 first observe pairing defects in Mcm5 A7 mutants ( Figure 1b); as shown in Figure 6b, 250 centromere clustering was significantly increased in nos>Smc1; Mcm5 A7 as compared to 251 Mcm5 A7 (***p < 0.0001, unpaired T-test), indicating that the increase in centromeric-SMC1 252 localization at meiotic onset partially rescues the early pachytene centromere clustering 253 We reasoned that if SMC1-dependent centromere clustering is partially rescued at 255 early pachytene in nos>Smc1; Mcm5 A7 , then the pairing defect at this stage will be 256 attenuated. To test this, we examined pairing frequency of X and 3R at early pachytene 257 in nos>Smc1; Mcm5 A7 flies ( Figure 6c). We see a significant pairing increase in 258 nos>Smc1; Mcm5 A7 mutants compared to Mcm5 A7 mutants (pairing frequency of 71% and 259 59%, respectively, **p = 0.0066, chi-square). From these data, we propose that SMC1-260 dependent centromere clustering in early meiosis promotes the stabilization of meiotic 261 homolog pairing, giving rise to homosynapsis. 262 We initially hypothesized that a lack of homolog pairing results in the loss of meiotic 263    According to this model, the enrichment of SMC1 at the centromere and 306 chromosome movements in pre-meiotic stages yield centromere clustering at meiotic 307 onset. While chromosome arms and centromeres enter meiosis paired, heterologous 308 centromere clustering serves as a mechanism to stabilize the pairing, resisting forces 309 generated by synapsis nucleation and/or diffusion that may otherwise push paired 310 chromosomes apart.
As the SC extends between the arms of homologs, DSBs are formed and subsequently repaired via HR to yield crossovers, which promote accurate 312 disjunction at the end of meiosis. 313 In Mcm5 A7 mutants, coordinated pre-meiotic centromere-directed movements 314 occur, yet there is not sufficient SMC1 enriched at the centromere to yield centromere 315 clustering. Thus, at meiotic onset, arms are paired, but centromeres are not clustered. 316 As SC nucleation occurs, the stabilization provided by centromere clustering is absent 317 and chromosome arms move freely in response to SC nucleation and/or diffusion. As 318 synapsis extends, the SC is formed between nearby chromosomes, regardless 319 homology, yielding heterologous synapsis. During instances of heterosynapsis, DSBs 320 are made but cannot be repaired via HR without a homologous template. Therefore, 321 overall crossover levels are reduced, and nondisjunction occurs at high frequency in 322 The centromere clustering-dependent pairing model highlights that initial meiotic 324 pairing is not sufficient to yield homosynapsis, indicating that pairing may be a two-step 325 process. Initial homolog pairing must occur, but a stabilization step must be enforced for 326 proper synapsis. In Drosophila, this stabilization is provided by SMC1-dependent 327 centromere clustering. We propose that, to ensure stabilization of the initial pairing event, 328 centromere clusters act as anchors at the nuclear envelope, maintaining the rigid AE 329 Though we cannot rule out SC aberrations in Mcm5 A7 mutants, our data reveal no 349 structural defects, supporting the notion that "normal" synapsis is largely homology-350 independent(Rog and Dernburg 2013). However, results from this study suggest that 351 synapsis initiation may require homology.  Crossovers on chromosome 2L were measured by crossing virgin net dpp ho dp b 391 pr cn / + females of desired genotype to net dpp ho dp b pr cn males. Vials of flies were 392 flipped after three days of mating. Resulting progeny were scored for all phenotypic 393 markers. Similarly, crossovers on chromosome X were measured by crossing virgin y sc 394 cv v g f y + / + females to y sc cv v g f males. Progeny were assessed for all phenotypic 395

markers. 396
To calculate intersister recombination, R(1)2, y 1 w hd80k17 f 1 / y 1 females with desired 397 genotype were crossed to y 1 w 1118 and progeny was scored for phenotypic markers. whole mount germaria were taken Zeiss LSM880 confocal laser scanning microscope 498 using 63x/0.65 NA oil immersion objective with a 2x zoom using ZEN software. Figure  499 2b: Images were obtain using AIRY-Scan on Zeiss LSM880 confocal laser scanning 500 microscope using40oil immersion objective. Images were saved as .czi files and 501 processed using FIJI (Schindelin et al. 2012). 502 503

Live cell imaging 504
Ovaries were dissected in 10S Voltalef oil. The muscular sheath around each ovariole 505 was removed and ovarioles were manually separated. Individual ovarioles were 506 transferred to a drop of oil on coverslip. Videos were collected with an on an inverted 507 Zeiss Axioobserver Z1 with motorized XYZ spinning-disc confocal microscope operated 508 by Metamorph coupled to a sCMOS (Hamamatsuorca) camera and a temperature control 509 chamber. All images were acquired with the Plan-Apochromat 100x/1.4 oil objective lens. projection. Three-dimensional tracking of spinning-disc data was performed using Imaris 556 software (Bitplane). The CID::RFP signal was tracked using the 'spots' function with an 557 expected diameter of 0.3 μm. Automatically generated tracks were then edited manually to eliminate inappropriate connections, including connections between foci in different 559 nuclei or between foci of different sizes or intensity when more likely assignments were 560 apparent or multiple spots assigned to the same focus. 561 To remove global movements of the germarium, each nucleus containing a 562 CID::RFP focus was assigned to the nearest fusome foci. Then, the position of the 563 reference fusome was subtracted from each CID::RFP focus for each time point of the 564 tracking to get the relative tracks. These relative tracks were then compiled using a 565 custom MATLAB (MathWorks) routine that computes the minimum volume of the ellipsoid 566 that encloses all of the three-dimensional points of the trajectory. 567 To analyze centromere trajectories: Positions of individual centromeres were 568 tracked every 10 seconds during 8 minutes to quantify the volume covered by each 569 centromere. This raw volume was then corrected both for overall movements of the tissue 570 and for variations in total nuclear volume. First, we subtracted the motion of the 571 germarium using the position of the fusome as a reference within each cyst. Second, to 572 take into account the nuclear volume at 8cc, we computed the relative volume, which is 573 the raw volume divided by the mean value of the nuclear volume at 8cc stage. Finally, we 574 normalized durations of each track by calculating the relative covered volume per second 575 (as shown in Figure 3e). 576 577

Transparent Reporting 578
Each microscopy experiment performed in this study was repeated independently at least 579 two times. We did not use explicit power analysis; rather, in each experiment, at least 8 580 independent germaria were imaged, and meiotic cells within the germaria were quantified,