Multiple pools of PP2A regulate spindle assembly, kinetochore attachments and cohesion in Drosophila oocytes

ABSTRACT

Meiosis in female oocytes lacks centrosomes, the microtubule-organizing centers. In Drosophila oocytes, meiotic spindle assembly depends on the chromosomal passenger complex (CPC). To investigate the mechanisms that regulate Aurora B activity, we examined the role of protein phosphatase 2A (PP2A) in Drosophila oocyte meiosis. We found that both forms of PP2A, B55 and B56, antagonize the Aurora B spindle assembly function, suggesting that a balance between Aurora B and PP2A activity maintains the oocyte spindle during meiosis I. PP2A-B56, which has a B subunit encoded by two partially redundant paralogs, wdb and wrd, is also required for maintenance of sister chromatid cohesion, establishment of end-on microtubule attachments, and metaphase I arrest in oocytes. WDB recruitment to the centromeres depends on BUBR1, MEI-S332 and kinetochore protein SPC105R. Although BUBR1 stabilizes microtubule attachments in Drosophila oocytes, it is not required for cohesion maintenance during meiosis I. We propose at least three populations of PP2A-B56 regulate meiosis, two of which depend on SPC105R and a third that is associated with the spindle.

Summary: Oocyte chromosomes direct meiotic spindle assembly. PP2A is important for this process, regulating microtubule stability, sister chromatid cohesion, microtubule attachments and metaphase I arrest.

INTRODUCTION

Chromosome segregation errors during meiosis I or II result in zygotic aneuploidy that is often lethal. In humans, aneuploidy is a leading cause of infertility, spontaneous abortion and birth defects, such as Down syndrome, Klinefelter syndrome and Turner syndrome (Kitajima, 2018; Mihajlović and FitzHarris, 2018; Webster and Schuh, 2017). In females of many species, the meiotic spindle assembles in the absence of centrosomes and is directed largely by the chromosomes (Dumont and Desai, 2012; Radford et al., 2017). Two pathways have been implicated in chromosome-directed spindle assembly: the chromosomal passenger complex (CPC) and the Ran pathway (Carazo-Salas et al., 2001, 1999; Cesario and McKim, 2011; Colombié et al., 2008; Radford et al., 2012b). In Drosophila, the CPC is required for spindle and kinetochore assembly (Colombié et al., 2008; Radford et al., 2015, 2012b). Aurora B is the catalytic component, while three other subunits – INCENP, Survivin (also known as Deterin) and Borealin – control the localization of the CPC (Carmena et al., 2012; Trivedi and Stukenberg, 2020; van der Horst and Lens, 2014). Possible targets of the CPC include kinetochore proteins (Emanuele et al., 2008; Radford et al., 2015) and kinesin family proteins such as the kinesin-6 Subito (the Drosophila MKLP2 homolog) and the kinesin-14 NCD (Beaven et al., 2017; Jang et al., 2005).

Phosphorylation is reversible, and the balance between kinases and phosphatases is an important mechanism for the regulation of cell division, including the initiation of prometaphase and the metaphase to anaphase transitions. In meiosis, there are two divisions with different properties and requirements. Therefore, the unique properties of each division and the transition between division types probably depends on mechanisms that regulate the antagonism between Aurora B and phosphatase activities (Keating et al., 2020). Oocytes in many organisms, including insects and mammals, undergo a long prophase arrest, which may be a cause of errors in chromosome segregation (Kitajima, 2018; Mihajlović and FitzHarris, 2018; Webster and Schuh, 2017). The first meiotic division involves several unique features that enable the segregation of homologous chromosomes rather than sister chromatids. During metaphase I, for example, sister centromeres must remain fused for them to co-orient and segregate to the same spindle pole (Wang et al., 2019; Watanabe, 2012). Cohesion along the chromosome arms must be released for anaphase I, while cohesion around the centromeres must be maintained for meiosis II (Keating et al., 2020). The nuclear envelope is not re-established between divisions; therefore, the meiotic spindle undergoes a transformation and reorganization between meiosis I and II (Riparbelli and Callaini, 2005; Riparbelli et al., 2002). These features of meiosis in oocytes may be regulated by antagonism between kinases and phosphatases.

In many of its functions, Aurora B is antagonized by protein phosphatase 1 (PP1) or protein phosphatase 2A (PP2A) (Funabiki and Wynne, 2013; Nilsson, 2019; Saurin, 2018). We previously have shown that one of the Drosophila PP1 paralogs, PP1-87B, regulates three meiosis I activities: kinetochore–microtubule (KT–MT) attachment stability, co-orientation of sister centromeres and chromosome alignment on the spindle (Wang et al., 2019). In this article, we report on studies to understand the functions of PP2A in Drosophila female meiosis.

PP2A is a highly conserved heterotrimeric serine/threonine phosphatase composed of a scaffolding A subunit, a catalytic C subunit, and a variable B subunit. The Drosophila genome, like the genomes of other organisms, encodes two B-type subunits, B55 (encoded by twins, tws) and B56 (encoded by widerborst, wdb; and well-rounded, wrd). Drosophila PP2A-B55 has several targets, many of which are phosphorylated by CDK1 and implicated in regulating progression through G2 and mitosis (Kim et al., 2012; Rangone et al., 2011; Von Stetina et al., 2008; Wang et al., 2011b). Depletion of B56 in Drosophila cell lines has a more severe defect in mitosis than depletion of B55 (Chen et al., 2007). The functions attributed to PP2A-B56 in mitotic cells include stabilization of KT–MT attachments and protection of cohesion (Foley and Kapoor, 2013; Funabiki and Wynne, 2013; Saurin, 2018; Trivedi and Stukenberg, 2016). Loss of PP2A in mouse oocytes also leads to defects in spindle assembly, chromosome alignment and stabilization of KT–MT attachments (Hu et al., 2014; Tang et al., 2016; Yoshida et al., 2015). In Caenorhabditis elegans, PP2A is required in meiosis for spindle assembly and chromosome congression (Bel Borja et al., 2020).

Our studies on PP2A were initiated because we found that both PP2A complexes antagonize Aurora B in the process of spindle assembly and maintenance. When the two Drosophila B56 paralogs, WRD and WDB, were depleted, additional defects in meiosis I were observed. PP2A-B56 WRD is required for sister chromatid cohesion and the stabilization of microtubule attachments. WRD localizes to meiotic kinetochores, and this depends on kinetochore protein SPC105R (the Drosophila KNL1 homolog), which was previously shown to be required for sister centromere cohesion in meiosis (Radford et al., 2015; Wang et al., 2019). WRD localization also depends on BUBR1 and MEI-S332 (the Drosophila Shugoshin homolog), which is surprising because neither of these proteins are required for meiosis I cohesion. BUBR1–MEI-S332–PP2A-B56 could be only one module that regulates meiotic cohesion. We propose that Dalmatian, the Drosophila ortholog of Sororin (Yamada et al., 2017), is part of a second module that has a cohesion protection function during meiosis I.

RESULTS

Sustained Aurora B activity is required to maintain the oocyte meiotic spindle

In the Drosophila ovary, the nuclear envelope breaks down and spindle assembly begins in stage-13 oocytes (Gilliland et al., 2009a). By stage 14, a bipolar spindle forms. In previous work, we found that the CPC, through the activity of Aurora B kinase, is required for assembly of a bipolar spindle in Drosophila oocytes (Radford et al., 2012b). In these experiments, germline-specific RNAi was used to deplete Aurora B or the targeting subunit INCENP prior to nuclear envelope breakdown. To determine the effect of depleting Aurora B activity during prometaphase, we treated stage-14 oocytes with the drug Binucleine 2 (BN2), which inhibits the kinase activity of Aurora B (Eggert et al., 2004; Smurnyy et al., 2010). Spindle assembly was monitored using immunofluorescence for tubulin, the kinetochore protein SPC105R and the central spindle protein Subito. In control experiments, when wild-type females were treated with only DMSO, all oocytes assembled a bipolar spindle with central anti-parallel microtubules containing Subito or INCENP (Fig. 1A,B). In wild-type oocytes treated with 50 µM BN2 for 1 h, the spindle was absent in ∼50% of oocytes and faint in most of the rest (Fig. 1A, n=84; Fig. S1). These results show that maintaining the meiotic spindle depends on Aurora B activity.

Fig. 1.

Constant Aurora B activity is required to maintain spindle microtubules and SPC105R localization in oocytes. (A) Wild-type (WT) oocytes treated with either 0.001% DMSO or 50 µM BN2. Subito was present in 18/18 DMSO-treated control oocytes and absent in 42/56 BN2-treated oocytes. Subito localization correlated with the presence of microtubules. In the merged images (left), DNA is shown in blue, Subito in red, tubulin in green, and CENP-C in white. (B) WT oocytes treated with either 0.001% DMSO or 50 µM BN2 immunostained for INCENP (red), tubulin (green), and Aurora B (white). DNA is shown in blue. See Fig. 3F for number of oocytes in each category. (C) Localization of kinetochore protein SPC105R (white) was reduced in WT oocytes treated with BN2, compared to levels in DMSO-treated controls. SPC105R localization was retained in mts^HMJ22483-RNAi oocytes (mts) treated with BN2. Tubulin is shown in green and Subito is shown in red. (D) SPC105R intensity in control and mts-RNAi oocytes treated with DMSO or BN2. SPC105R intensity was measured in WT oocytes treated with DMSO (n=91) or BN2 (n=158), and mts^HMJ22483-RNAi oocytes treated with DMSO (n=35) or BN2 (n=138). Mean±s.d. are shown. ****P<0.0001 (unpaired two-tailed t-test). All images are maximum intensity projections of z-stacks. Scale bars: 5 µm.

To determine the effect of inhibiting Aurora B on CPC localization, we treated stage-14 oocytes with BN2 and observed the effect on INCENP localization. When Aurora B activity was inhibited using BN2, INCENP also localized to the chromosomes, showing that upon BN2 treatment, the CPC moves from the microtubules to the chromosomes (Fig. 1B). This is similar to our previous observation that INCENP localizes to the meiotic chromosomes when Aurora B is depleted using RNAi (Radford et al., 2012b; Wang et al., 2021). A significant difference is that the BN2 treatment was applied after the spindle assembled, whereas RNAi depletes Aurora B before the spindle assembles. Thus, the results with BN2 show that the CPC is required to maintain the meiotic spindle, and when Aurora B is inhibited, INCENP moves from the microtubules to the chromosomes. In addition, Aurora B moved to the chromosomes, demonstrating that Aurora B associates with INCENP and localizes to the chromosomes in the absence of phosphorylation (Fig. 1B).

The Drosophila oocyte meiotic spindle has two types of microtubules based on the nature of their plus ends. The kinetochore microtubules have their plus ends at a kinetochore. The anti-parallel microtubules have their plus ends within the central spindle (Jang et al., 2005; Radford et al., 2015). Subito, a kinesin-6 required for interpolar microtubule assembly of the central spindle (Jang et al., 2005), was lost upon BN2 treatment (Fig. 1A), demonstrating that Subito localization depends on Aurora B activity and/or microtubules. SPC105R, which is required for kinetochore microtubules (Radford et al., 2015), was also reduced, although not eliminated (Fig. 1C,D). The localization of centromere proteins was not reduced, showing the effect of BN2 was specific to the loading of kinetochore proteins (Fig. S1E). These results show that maintenance of the two major organizers of oocyte spindle microtubules depends on sustained Aurora B activity.

Kinesin-13 KLP10A is regulated by the CPC

The kinesin-13 homolog KLP10A depolymerizes spindle microtubules in oocytes (Radford et al., 2012a). In Xenopus extracts, the spindle depolymerizing activity of kinesin-13 MCAK (also known as KIF2C) is downregulated by Aurora B (Sampath et al., 2004). To test the hypothesis that Aurora B negatively regulates KLP10A, we examined meiotic spindles in Klp10A-RNAi oocytes treated with BN2. In Klp10A-RNAi control oocytes, the spindle is very long and disorganized (Fig. 2B). Unlike wild-type oocytes, however, the spindle was retained in 87% of BN2-treated Klp10A-RNAi oocytes (Fig. 2A,C), suggesting that KLP10A is a possible CPC target and that the maintenance of the spindle could depend on the inhibition of depolymerizing factors. The spindle phenotype of BN2-treated Klp10A-RNAi oocytes was variable. In most (77%) oocytes, central spindle and kinetochore microtubules appeared to be intact. In a minority (23%) of oocytes, the spindle was retained, but Subito or INCENP localization was abnormal, indicating that the central spindle and/or kinetochore microtubules were absent (Fig. 2D). These results suggest KLP10A is inactivated by CPC phosphorylation and is activated when dephosphorylated by PP2A. Although we lack direct evidence of phosphorylation of KLP10A by Aurora B, Rui et al. (2020) have found that KLP10A activity is increased in mts-RNAi neurons, which is similar to our results.

Fig. 2.

Aurora B antagonizes KLP10A. (A) Wild-type (WT) oocytes treated with BN2. Most wild-type oocytes treated with BN2 failed to form a spindle (16/20). (B) Untreated klp10A-RNAi oocyte showing characteristic long spindle phenotype (Radford et al., 2012a). (C) Most Klp10A-RNAi oocytes treated with BN2 developed a long spindle (39/45; P<0.0001, Fisher's exact test). A representative long spindle is shown. (D) In some Klp10A-RNAi oocytes treated with BN2 (9/45), the spindle was detached from the chromosomes, as if the kinetochore attachments were destabilized. A representative detached spindle is shown. In A–D, centromeres are shown in white (CID), tubulin in green, Subito in red and DNA in blue. The tubulin channel is shown in the bottom panels. All images are maximum intensity projections of z-stacks. Scale bars: 5 µm.

PP2A-B55 and PP2A-B56 antagonize the CPC in spindle maintenance

To explain the loss of spindle in BN2-treated oocytes, we tested the hypothesis that a phosphatase (PP1 or PP2A) dephosphorylates Aurora B targets. To identify the phosphatase responsible for antagonizing Aurora B, shRNA expression under the control of the UAS/Gal4 system was used for germline-specific RNAi of each phosphatase. Each shRNA was expressed using matα4-GAL-VP16 (referred to hereafter as matα-GAL4), which induces expression of UAS-controlled transgenes after premeiotic DNA replication but throughout most stages of meiotic prophase during oocyte development (Radford et al., 2012b). During this time, the oocyte grows and matures, but does not divide, while the target gene expression is depleted. We refer to oocytes expressing an shRNA using matα-GAL4 as ‘RNAi oocytes’.

PP1-87B is the most important PP1 paralog in oocytes (Wang et al., 2019). Pp1-87B-RNAi oocytes were treated with BN2, and spindle loss was measured. If PP1-87B opposes the Aurora B spindle assembly function in oocytes, then addition of BN2 after depletion of PP1-87B would not cause loss of spindle microtubules. However, 76% of Pp1-87B-RNAi oocytes treated with BN2 had reduced spindle microtubules relative to Pp1-87B-RNAi solvent-treated controls (Fig. S1A,C). Consistent with loss of the spindle, Subito localization was also absent in Pp1-87B-RNAi oocytes treated with BN2 (Fig. S1B). In contrast, SPC105R localization and some KT–MT fibers were retained in Pp1-87B-RNAi oocytes treated with BN2 (Fig. S1B,D). These observations show that Pp1-87B-RNAi oocytes are partially resistant to BN2 treatment, suggesting that PP1-87B antagonizes Aurora B at the kinetochores but not the spindle, and could be related to our previous observation that loss of PP1-87B stabilizes KT–MT attachments (Wang et al., 2019).

All PP2A complexes contain the same A (PP2A-29B) and catalytic C (MTS, Microtubule star) subunits. Depletion of the A subunit (Pp2A-29B, GLC01651) or C subunit (mts, HMS04478) of PP2A using matα-GAL4 resulted in no mature stage-14 oocytes, indicating that PP2A activity is required for oocyte development. However, mts^HMJ22483/matα-GAL4 females were sterile and produced stage-14 oocytes, probably due to partial depletion of mts RNA by expression of HMJ22483 (Table S1). When mts^HMJ22483-RNAi oocytes were treated with BN2, a bipolar spindle and Subito localization were retained in 100% of oocytes (Fig. 1C, Fig. 3A–C,F). SPC105R localization was also retained in most oocytes (Fig. 1C,D). These results suggest that PP2A antagonizes the role of the CPC in maintaining the kinetochores and the microtubules of the bipolar spindle.

Fig. 3.

PP2A antagonizes Aurora B activity. (A–E) Representative images of wild-type (WT) oocytes treated with either (A) DMSO or (B) 50 µM BN2, as well as (C) mts^HMJ22483-RNAi (mts), (D) tws^GL00670-RNAi (tws) and (E) wdb^HMS01864-RNAi (wdb^HMS) ooctyes treated with 50 µM BN2, showing INCENP (red), centromeres (CID, white), tubulin (green) and DNA (blue). Tubulin is shown in the right-hand panels. (F) Qualitative (left) and quantitative (right) assessment of spindle assembly in wild-type DMSO-treated oocytes (WT+DMSO; n=11) and wild-type BN2-treated oocytes (n=84), as well as mts (n=39), tws (n=25), wdb^A5 (n=30) and wdb^HMS01864 (n=50) RNAi oocytes treated with BN2. For the qualitative phenotype comparison, ****P<0.0001 (Fisher's exact test). For the quantitative assessment of spindle intensity, all comparisons were ****P<0.0001 (unpaired two-tailed t-test) except for tws, which was **P=0.005. Relative spindle intensity was calculated as described in Materials and Methods and is presented as a violin plot with dashed lines indicating the median and dotted lines indicating the quartiles. (G) WT oocytes treated with 0.001% DMSO had pINCENP (19/20). (H) WT oocytes treated with 50 µM BN2 had pINCENP at a reduced frequency (6/15; P=0.007, Fisher's exact test). The BN2-treated oocytes with pINCENP tended to have some residual spindle assembly. (I) Retention of pINCENP was observed when wdb-RNAi oocytes were treated with 50 µM BN2 (n=15/15; P<0.001, Chi-square). In G–I, the merged images show pINCENP (red), INCENP (white), tubulin (green) and DNA (blue). All images are maximum intensity projections of z-stacks. Scale bars: 5 µm.

PP2A exists in at least two complexes, with the A and C subunits associating with either a B55 (TWS) or one of two B56 (WDB or WRD) subunits (Chen et al., 2007). In order to determine which PP2A complex antagonizes Aurora B, BN2 treatment was performed in the presence of B subunit RNAi. An shRNA line for tws (GL00670) caused complete sterility when expressed using matα-GAL4 and reduced the mRNA to 0% of wild-type levels (Table S1). In tws^GL00670-RNAi oocytes treated with BN2, oocyte spindle assembly was restored in ∼80% of oocytes (Fig. 3D,F). For the B56 subunit, we focused on WDB because it is essential and WRD is not (see below). Two shRNA lines were used for wdb: one generated by TRiP (HMS01864; Ni et al., 2011) and one generated in our laboratory (A5). Both shRNAs efficiently knocked down wdb RNA (Table S1), and the oocytes had similar phenotypes. In both wdb^HMS01864- and wdb^A5-RNAi oocytes treated with BN2, the spindle was present in ∼80% of oocytes (Fig. 3E,F). These results demonstrate that both B55 and B56 PP2A complexes antagonize Aurora B activity in the process of spindle assembly and maintenance.

To directly test antagonism between PP2A-B56 and Aurora B, we used an antibody for an Aurora B target, phosphorylation of INCENP (pINCENP) at a conserved serine in the C-terminal domain (Salimian et al., 2011; Wang et al., 2011a). Whereas solvent-treated controls had robust pINCENP in the central spindle, the chromosome-localized INCENP in BN2-treated oocytes had lost its phosphorylation (Fig. 3G,H). Depletion of WDB restored localization of INCENP and pINCENP to the central spindle (Fig. 3I). These results show that Aurora B phosphorylation of INCENP is antagonized by PP2A-B56 and is not required for chromosome localization of the CPC.

PP2A-B56 is required for establishing end-on kinetochore–microtubule attachments

To examine the function of PP2A during meiosis, we examined the phenotypes of oocytes lacking B55 or B56 subunits in the absence of BN2 treatment. As detailed below, the single RNAi oocytes (mts, tws, wdb, wrd) had relatively mild spindle phenotypes, suggesting that loss of Aurora B antagonism was not associated with gross spindle organization defects. For example, meiotic spindles in tws^GL00670-RNAi oocytes were similar to those of wild-type oocytes. The spindles were usually bipolar, bundles of microtubules terminated at the kinetochores and the centromeres were oriented towards a pole, suggesting that MT–KT attachments formed normally (Fig. 4A). These data suggest that TWS (B55) is not essential for bipolar spindle assembly. However, we observed more severe defects in kinetochore attachments (Fig. 4) and sister chromatid cohesion (Fig. 5) in oocytes depleted of wrd and wdb.

Fig. 4.

Spindle assembly in PP2A-RNAi oocytes. (A) Wild-type (WT) and mts^HMJ22483 (mts)-, tws^GL00670 (tws)- or wdb^A5-RNAi oocytes with INCENP in red, centromeres (CID) in white, tubulin in green and DNA in blue. Two examples of mts show some of the variation observed. For example, a low frequency of separated chromosomes were observed in mts- and wdb-RNAi oocytes (see Fig. 5). (B) Oocytes stained as in A. Arrows in the magnified views on the right show examples of end-on attachments in wild-type oocytes and lateral attachments in wrd^Δ; wdb^HMS01864 oocytes. (C) Assessment of end-on and lateral KT–MT attachments in wild-type (n=83), wrd^GL00671; wdb^HMS01864 (wrd^GL; wdb^HMS; n=41) and wrd^Δ; wdb^HMS01864 (wrd^Δ; wdb^HMS; n=45). Mean and standard deviation are shown. See Fig. S2 for single RNAi oocyte data. All images are maximum intensity projections of z-stacks. Scale bars: 5 µm.

Fig. 5.

Loss of cohesion in PP2A-RNAi oocytes. (A) Stage-14 oocytes from single (wrd^GL00671 and wdb^HMS01864), double (wrd^GL00671; wdb^HMS01864 and wrd^Δ; wdb^HMS01864) and triple (Ndc80; wrd^Δ; wdb^HMS01864) knockdown females, with INCENP in red, centromeres (CID) in white, tubulin in green and DNA in blue. Centromeres are shown below the merged images, and the number of centromere foci is indicated. All images are maximum intensity projections of z-stacks. Scale bars: 5 μm. (B) Percent chromosome phenotype for each genotype. Precocious anaphase is indicated by the separation of chromosomes towards the poles. (C) Number of centromere foci per nucleus in each genotype, with mean and s.d. For B and C, wild-type n=51, mts^HMJ22483 (mts) n=50, tws^GL00670 (tws) n=32, wdb^A5 n=39, wdb^HMS01864 (wdb^HMS) n=44, wrd^GL00671 (wrd^GL) n=13, wrd^GL00671; wdb^HMS01864 n=33, wrd^Δ; wdb^HMS01864 n=28 and Ndc80; wrd^Δ; wdb^HMS01864 n=20 oocytes. *P<0.05; ***P<0.001; ns, not significant (unpaired one-tailed t-test).

The two Drosophila B56-type paralogs, WDB and WRD, share 68% amino acid identity. Null mutants of wrd are viable and fertile (Hahn et al., 2010; Moazzen et al., 2009). An shRNA targeting wrd (GL00671) substantially reduced mRNA levels (Table S1). Consistent with the phenotype of wrd mutants, females expressing GL00671 in the germline were fertile, and nondisjunction was rare (Table S2). In contrast, ubiquitous expression of wdb shRNA caused lethality (see Materials and Methods), the same as homozygous null mutations of wdb. Thus, WDB appears to be the more important B56 subunit. However, the germline wdb-RNAi phenotypes were milder than expected. Maternal expression of either wdb shRNA caused reduced fertility, but the females were not sterile and meiotic nondisjunction was low (Table S2). The spindle morphology of wdb-RNAi oocytes was similar to that of wild-type oocytes; they were bipolar, assembled a robust central spindle, and most kinetochores made end-on attachments to microtubules (Fig. 4A; Fig. S2D). These data suggest that WDB is not essential for bipolar spindle assembly.

WDB and WRD are partially redundant in mitosis (Chen et al., 2007). To test for redundancy in meiosis, two genotypes were generated to reduce expression of both WDB and WRD: females expressing shRNA to target both wdb and wrd (wrd^GL00671; wdb^HMS01864), and females expressing wdb shRNA and hemizygous for a wrd mutation (wrd^Δ; wdb^HMS01864) (see Materials and Methods). When using matα-GAL4 for expression of each shRNA, the double knockdown females were completely sterile, unlike the single RNAi females. In addition, when using nos-GAL4-VP16 (referred to hereafter as nos-GAL4), which promotes expression of the shRNA in premeiotic germ cells, no oocytes were produced, also unlike the single RNAi females. These results demonstrate that WRD and WDB are partially redundant in the mitotic and meiotic germline and that expression of WRD is sufficient for fertility. In contrast, expression of WDB, but not WRD, is sufficient for viability.

In wild-type oocytes, most of the kinetochores in stage-14 oocytes have end-on attachments, defined as when bundles of microtubules end at the centromeres (Fig. 1A,C, Fig. 3A, Fig. 4A,B; Fig. S2A). Similarly, most attachments were scored as end-on in mts^HMJ22483, tws^GL00670, wdb^HMS01864 and wdb^A5 single RNAi oocytes (Fig. 4A; Fig. S2D). In contrast, wrd^GL00671; wdb^HMS01864 and wrd^Δ; wdb^HMS01864 oocytes had a high frequency of lateral microtubule attachments, defined as when the centromeres are positioned along the side of microtubule bundles (Fig. 4B,C; Fig. S2B,C). Assuming that lateral KT–MT attachments form first (Itoh et al., 2018; Shrestha et al., 2017), these data show that PP2A is required for the conversion from lateral to end-on kinetochore attachments in oocytes. Because these PP2A-B56-depleted oocytes enter precocious anaphase, these results also suggest that lateral attachments are sufficient to move the chromosomes towards the poles, as also shown in Drosophila cell lines (Feijão et al., 2013).

PP2A-B56 protects meiotic sister chromatid cohesion

Drosophila oocytes arrest in metaphase I at developmental stage 14 (>90%; Fig. 5B), and do not proceed into anaphase I and meiosis II until activated by passage through the oviduct (Heifetz et al., 2001). Similarly, we found that most (∼75%) tws-, wdb- and wrd-RNAi oocytes were arrested in metaphase I (Figs 4A, 5B). However, the chromosomes in wrd^GL00671; wdb^HMS01864 and wrd^Δ; wdb^HMS01864 oocytes were frequently observed to be separating towards the poles, indicative of entering anaphase (Figs 4B, 5B). These results suggest that the metaphase I arrest does not occur in oocytes lacking both B56 subunits.

Precocious entry into anaphase could be caused by loss of sister chromatid cohesion. To determine whether cohesion of the centromeres was affected, we counted the number of centromere (CID or CENP-C) foci in the oocytes of each genotype. Each pair of sister centromeres during wild-type meiosis I appear to be fused in a process that requires sister chromatid cohesion (Wang et al., 2019). Consistent with this, we observed approximately eight centromere foci in wild-type oocytes, as expected from four bi-oriented bivalents at metaphase I. In wdb-RNAi oocytes (HMS01864 and A5), there was an increase in the frequency of oocytes (45%) with greater than eight centromere foci compared to 14% in wild-type oocytes, indicating that the sister chromatids were separating prematurely (Fig. 5C). However, the sister centromere separation and precocious anaphase phenotypes of wdb-RNAi oocytes were relatively mild compared to those of other cohesion-defective mutants (Gyuricza et al., 2016; Wang et al., 2019). In contrast, the severity of the precocious anaphase and sister centromere separation phenotypes was dramatically increased in both wrd^GL00671; wdb^HMS01864 and wrd^Δ; wdb^HMS01864 oocytes (Fig. 5A,C). These results can be explained if PP2A-B56 is required for the protection of sister chromatid cohesion during meiosis I.

We previously found that two mechanisms can cause precocious sister centromere separation: loss of cohesion or inappropriate stabilization of KT–MT attachments (Wang et al., 2019). The inappropriate stabilization of KT–MT attachments depends on NDC80 and end-on KT–MT attachments. To test the role of KT–MT attachments in the wrd^Δ; wdb^HMS01864 phenotype, we examined wrd^Δ; wdb^HMS01864; Ndc80-RNAi oocytes. In Ndc80-RNAi oocytes, end-on microtubule attachments are absent. The wrd^Δ; wdb^HMS01864; Ndc80-RNAi oocytes exhibited both precocious anaphase and centromere separation (Fig. 5A), suggesting that stabilization of end-on KT–MT attachments was not required for the wrd^Δ; wdb^HMS01864 phenotype. Because end-on attachments are necessary for separating sister centromeres in the presence of intact cohesins (Wang et al., 2019), these results are consistent with loss of cohesion in wrd^Δ; wdb^HMS01864 oocytes. Ndc80 RNAi had little effect on the precocious anaphase phenotype of wrd^Δ; wdb^HMS01864 RNAi oocytes (Fig. 5B), which confirms that lateral attachments are sufficient for movement of the chromosomes towards the poles.

Stage-14 oocytes can be induced to proceed past the metaphase I arrest by incubation in certain buffers (Endow and Komma, 1997; Horner and Wolfner, 2008; Page and Orr-Weaver, 1997). Our methods use a modified Robb's buffer to prevent premature activation (Theurkauf and Hawley, 1992). Even after control oocytes were incubated for 1 h in modified Robb's buffer, metaphase I arrest was usually maintained (Fig. S3). We were surprised, therefore, to find that the 1 h incubation in modified Robb's buffer with BN2 treatment induced precocious anaphase in wdb-RNAi oocytes (Fig. 3E). To investigate if this was an effect of the BN2 treatment, wdb-RNAi oocytes were incubated in modified Robb's buffer for 1 h without BN2. In these conditions, precocious separation of homologous chromosomes was observed in ∼50% of oocytes (Fig. S3). These phenotypes were present in wdb-RNAi oocytes but not in tws-RNAi oocytes. These data suggest that 1 h incubation in modified Robb's buffer induces wdb-RNAi oocytes to lose their arm cohesion and precociously enter anaphase I. This result can be explained if PP2A-B56 protects cohesion on the chromosome arms.

PP2A-B56 is required for bi-orientation of homologous chromosomes

To examine the effects of PP2A depletion on chromosome alignment and bi-orientation, we used fluorescence in situ hybridization (FISH). FISH probes were used that detected the pericentromeric regions of all three major chromosomes: the AACAC repeat (second chromosome), the dodeca repeat (third chromosome) and the 359-basepair (bp) repeat (X chromosome). In wild-type oocytes, correct bi-orientation is indicated when each pair of homologous centromeres separates towards opposite poles (Fig. 6A). A low frequency of bi-orientation defects was observed in tws-RNAi oocytes (Fig. 6B,G), showing that the TWS (B55) subunit is not required for making the correct microtubule attachments during prometaphase I.

Fig. 6.

Analysis of bi-orientation in PP2A-RNAi oocytes. (A–F) FISH was performed using probes for the three major chromosomes (Ch) in (A) wild-type (WT) oocytes and RNAi oocytes of (B) tws^GL00670 (tws), (C) wdb^HMS01864 (wdb^HMS), (D) wrd^Δ; wdb^HMS01864 and (E) wrd^GL00671; wdb^HMS01864 (wrd^GL; wdb^HMS), as well as (F) wdb^HMS01864 incubated for 1 h in modified Robb's buffer. The paracentric FISH probes were for the X chromosome (359-bp repeat; Alexa Fluor 594, purple), the second chromosome (AACAC; Cy3, red) and the third chromosome (dodeca; Cy5, white). An example of one FISH probe is shown in the lower panels (chromosome X, 2 or 3). Examples of mono-orientation are shown in D–F. Panels D and F also show examples of precocious anaphase associated with separation of homologous chromosomes (arrows indicate Ch X and Ch 2). Tubulin is shown in green, and DNA is shown in blue. All images are maximum intensity projections of z-stacks. Scale bars: 5 μm. (G) Relative frequency of mono-oriented and bi-oriented centromeres in wild-type (n=117) and RNAi oocytes of mts^HMJ22483 (mts; n=60), tws (n=72), wdb^A5 (n=61), wdb^HMS01864 (n=66), wrd^GL00671; wdb^HMS01864 (n=51), wrd^Δ; wdb^HMS01864 (n=73), wild-type incubated for 1 h (n=45), wdb^A5 incubated for 1 h (n=98) and wdb^HMS01864 incubated for 1 h (n=87).

The frequency of bi-orientation defects in wdb^A5- and wdb^HMS01864-RNAi oocytes was also low (Fig. 6C,G). This was most likely due to B56 redundancy, because oocytes depleted of both wrd and wdb had a much higher frequency of bi-orientation defects (Fig. 6D,E,G). In some oocytes depleted of both wrd and wdb, however, pairs of homologous centromeres were in different chromatin masses and had separated towards opposite poles (Fig. 6D, arrows), indicating precocious anaphase due to loss of cohesion on the chromosome arms. In contrast, pairs of homologous centromeres in wild-type metaphase I oocytes were separated but remained connected by chiasmata and were, therefore, usually within the same chromatin mass. Similarly, in wdb-RNAi oocytes incubated for 1 h in Robb's buffer, the two FISH signals for each homolog were usually in different chromatin masses, showing that arm cohesion had been released, allowing the chromosomes to move towards the poles (Fig. 6F). Precocious anaphase in wdb-RNAi oocytes after 1 h incubation was not associated with a high frequency of bi-orientation defects (Fig. 6G).

Evidence for redundant mechanisms to recruit PP2A-B56

Because the function of kinases and phosphatases often depends on their localization, we examined the localization of WDB using either antibodies (Pinto and Orr-Weaver, 2017; Sathyanarayanan et al., 2004) or HA-tagged transgenes (Bischof et al., 2013; Hannus et al., 2002). In wild-type oocytes, we found that WDB localizes prominently to the centromere regions (Fig. 7A; Fig. S4A). In many oocytes, it was also possible to observe weaker localization to the chromosome arms (Fig. 7B; Fig. S4B). Surprisingly, centromeric WDB was still detected using either the antibody or the HA-tagged transgene in wdb-RNAi oocytes (Fig. 7C; Fig. S4C). WDB was also observed in wrd^Δ; wdb^HMS01864 oocytes (Fig. S2B), showing that the localization is not due to cross-reactivity with WRD. Thus, shRNA expression was effective enough to yield a mutant phenotype but did not eliminate WDB expression. As described below, there appear to be multiple pools of PP2A-B56, and the centromere-localized WDB may not be sufficient for some PP2A-B56 functions.

Fig. 7.

Localization of WDB in metaphase I of meiosis. Stage-14 oocytes showing WDB (red), centromeres (CID; white), tubulin (green) and DNA (blue). The single channel images beneath show WDB. Images show (A,B) wild-type (WT) oocytes; (C) wdb^HMS01864-RNAi (wdb^HMS) and (D) Spc105R-RNAi oocytes; and homozygous (E) sunn and (F) mei-S332 mutant oocytes. WDB was detected using a polyclonal antibody (Pinto and Orr-Weaver, 2017; Sathyanarayanan et al., 2004). The WDB channel in B is shown with higher levels and magnification to show threads of WDB signal (arrow), corresponding to PP2A on the chromosome arms. All images are maximum intensity projections of z-stacks. Scale bars: 5 μm. (G) WDB intensity measured in wild type (n=197), wdb^HMS01864 RNAi (wdb; n=234), wrd^Δ; wdb^HMS01864 RNAi (n=120), mei-S332/+ (n=127), mei-S332 (n=173), Spc105R RNAi (n=83) and sunn (n=103). The mei-S332-mutant and Spc105R-RNAi genotypes, but not the sunn mutant, have significantly lower WDB intensity than the wild type (unpaired, two-tailed t-test; ****P<0.0001). Mean±s.d. are shown.

The strongest WDB accumulation colocalized with centromere markers, including CID and CENP-C, suggesting it is enriched in the centromere regions. Therefore, we tested whether WDB localization depends on SPC105R, which is a kinetochore protein previously shown to be required for sister centromere cohesion in meiosis I (Radford et al., 2015; Wang et al., 2019). Spc105R-RNAi oocytes lacked WDB localization (Fig. 7D,G), showing that SPC105R is required for the recruitment of WDB. To determine whether cohesion is required for WDB localization, we used sunn-mutant oocytes. SUNN is a stromalin-related protein required for sister chromatid cohesion in meiosis (Krishnan et al., 2014). WDB was present at the centromeres in sunn-mutant oocytes (Fig. 7E,G), showing that cohesion is not required for WDB localization.

In human cell lines, PP2A is recruited to the kinetochores by BUBR1 (also known as BUB1B; Kruse et al., 2013; Xu et al., 2013) and to the centromeres by shugoshin 1 (Kitajima et al., 2006; Riedel et al., 2006; Tang et al., 2006). Consistent with these studies, WDB localization was reduced in mei-S332/+ heterozygotes and mei-S332-mutant oocytes (Fig. 7F,G; Fig. S4D,E,G). WDB was also absent from the centromeres in BubR1-RNAi oocytes (Fig. S4F,G). There was no effect of reduced Spc105R, BubR1 or mei-S332 expression on centromere proteins CID or CENP-C (Fig. S4H), suggesting that SPC105R recruits PP2A-B56 via BUBR1 and MEI-S332. These results were surprising, because BubR1-RNAi oocytes do not have a cohesion defect or a precocious anaphase phenotype (Wang et al., 2019). In addition, Drosophila mei-S332 mutants are viable, suggesting that MEI-S332 is not required for cohesion in mitosis. Chromosome segregation errors in mei-S332 mutants primarily involve sister chromatids during meiosis II (Kerrebrock et al., 1995; Tang et al., 1998) and, while MEI-S332 localizes to the centromere regions during meiosis I, the sister centromeres remain fused in mei-S332-mutant oocytes (see below) (Kerrebrock et al., 1995; Moore et al., 1998). Therefore, although both MEI-S332 and BUBR1 are required to recruit WDB to the meiotic kinetochores, unlike oocytes lacking PP2A, the absence of BUBR1 or MEI-S332 does not cause meiosis I cohesion defects. In addition, BubR1-RNAi oocytes were sensitive to BN2, suggesting that a kinetochore-independent pool of WDB is sufficient for PP2A to destabilize the meiotic spindle (Fig. S5).

Dalmatian may protect cohesion during meiosis I

The absence of a strong meiosis I phenotype in mei-S332 mutants could be explained if another protein recruits PP2A-B56 to the centromeres. Because we only examined WDB localization, it is possible this hypothetical pathway recruits WRD and not WDB. However, a mei-S332; wrd double mutant is viable (Pinto and Orr-Weaver, 2017), suggesting that in mitotic cells, WDB functions in the absence of MEI-S332. A candidate for a second protein that recruits PP2A is Dalmatian (DMT), which is a Sororin ortholog that has been proposed to recruit PP2A in Drosophila mitotic cells (Yamada et al., 2017). Consistent with this hypothesis, DMT colocalized with MEI-S332 at the centromeres in metaphase I oocytes (Fig. 8A). Mutations of dmt also cause lethality (Salzberg et al., 1994). Therefore, to test the function of DMT in meiosis, we created shRNA lines to target dmt for tissue-specific RNAi. The strongest of the two shRNA lines reduced mRNA levels to 5% of wild-type levels, caused sterility when expressed in oocytes with either nos-GAL4 or matα-GAL4, and caused lethality when expressed ubiquitously (Tables S1, S2), showing that the RNAi was effective. However, in dmt/matα-GAL4-RNAi oocytes, there were no centromere separation defects to indicate a loss of cohesion (Fig. 8B,C,F). One explanation for the absence of a defect in dmt-RNAi oocytes could be redundancy with MEI-S332. Therefore, we constructed mei-S332-mutant females expressing dmt-RNAi using matα-GAL4. These females, however, also had no centromere separation defects during meiosis I (Fig. 8D–F).

Fig. 8.

Localization of cohesion protection proteins in meiosis. (A) Wild-type (WT) and (B) dmt-RNAi oocytes showing DMT in red, MEI-S332 in white, tubulin in green and DNA in blue. Also shown are grayscale images of the DMT and MEI-SS332 channels. (C–E) Representative images showing (C) a dmt-RNAi oocyte, (D) a mei-S332 mutant oocyte, and (E) a mei-S332 mutant dmt-RNAi oocyte. Images show Subito (red), centromeres (CENP-C; white), tubulin (green), and DNA (blue). (F) Summary of mean±s.d. number of centromere foci in wild-type (n=51), dmt-RNAi (dmt³⁰⁵; n=15), mei-S332 mutant (n=23), and dmt-RNAi mei-S332 mutant (n=21) oocytes. There are no significant differences between the three data sets by one-way ANOVA; however, the number of centromere foci was significantly different in dmt-RNAi mei-S332 mutant ooctyes and mei-S332 mutant oocytes by an unpaired one-tailed t-test (P=0.013). (G) DMT localization (red) in early prophase/pachytene oocytes. Oocytes are marked by C(3)G (green) along with DNA (blue). (H,I) DMT localization when the dmt-targeting shRNA is expressed using nos-GAL4. DMT localization is visible in these late-stage pachytene (H) and prophase oocytes (I). All images are maximum intensity projections of z-stacks. Scale bars: 5 μm.

Another explanation for the absence of a meiotic defect in dmt-RNAi oocytes is that the DMT protein is stable and loaded prior to expression of the shRNA by matα-GAL4, such as in early meiotic S phase or prophase. Consistent with this hypothesis, DMT localization to the centromeres was observed in dmt-RNAi/matα-GAL4 oocytes (Fig. 8B). Furthermore, we found that DMT localized to the meiotic chromosomes in the germarium, which contains oocytes early in prophase undergoing pachytene (Fig. 8G–I). In contrast, we did not observe localization of WDB at these stages (Fig. S6). This result shows that DMT is loaded onto the chromosomes prior to the domain of shRNA expression by matα-GAL4. DMT may only be loaded at a specific time in meiosis, such as premeiotic S phase, and could be stable enough to persist on the chromosomes without replenishment until metaphase I.

If DMT protein is loaded prior to the domain of matα-GAL4 expression and stably maintained, shRNA expressed during S phase and early pachytene by nos-GAL4 should have a defect in meiosis. To test this possibility, we examined RNAi of cohesin components, which are an example of proteins stably loaded onto the meiotic chromosomes in premeiotic S phase (Gyuricza et al., 2016). Nondisjunction was observed when nos-GAL4 was used to induce expression of shRNAs for meiotic cohesins ord (32%, n=1270) and sunn (15.6%, n=486). In contrast, nondisjunction was not observed with matα-GAL4 and the same shRNAs for ord (0%, n=1622) or sunn (0%, n=547). These results can be explained if cohesins are stably maintained without new transcription throughout most of prophase. In contrast, expressing an shRNA to mei-S332 with either nos-GAL4 (40%, n=331) or matα-GAL4 (27.8%, n=208) caused high levels of nondisjunction. These results show that mei-S332 transcription is required late in prophase or during prometaphase I. If DMT behaves like the cohesins, nos-GAL4 should be necessary for dmt RNAi to affect meiosis. However, a loss of centromere cohesion was not observed in dmt/nos-GAL4 RNAi oocytes (average of 7.9 foci, n=10). Finally, to test the hypothesis that MEI-S332 and DMT both recruit PP2A during meiosis I, we expressed dmt RNAi using nos-GAL4 in mei-S332-mutant females. These females were sterile, but unlike dmt RNAi/nos-GAL4 females, very few oocytes were produced, which can be explained if DMT and MEI-S332 are redundant in the mitotic germline divisions and may separately recruit PP2A-B56. In summary, these data are consistent with the conclusion that Drosophila oocytes load cohesins and a protector of cohesin, DMT, during premeiotic S phase.

DISCUSSION

Initiation and maintenance of the meiotic spindle

When the Aurora B inhibitor BN2 was added to prometaphase I oocytes, they lost their meiotic spindles. The implication is that constant Aurora B activity is required to maintain the spindle. Furthermore, knockdown of PP2A components suppressed this BN2-induced loss of the meiotic spindle. Therefore, PP2A probably removes activating phosphorylation events on meiotic spindle proteins. This was shown directly using phosphorylated INCENP, which is a target of Aurora B. Based on indirect evidence, kinetochore protein SPC105R and kinesins KLP10A and Subito also depend on this balance of phosphorylation. Depletions of either PP2A subtype, B55 (by knockdown of TWS) or B56 (by knockdown of WDB), suppressed BN2-induced spindle loss, even though these two complexes are usually targeted to different substrates. These results can be explained if PP2A-B55 depletion reduces the activity of PP2A-B56. This is plausible, because B56 activity may depend on Polo kinase, which in turn may be regulated by PP2A-B55 (Rangone et al., 2011; Wang et al., 2011b). Alternatively, PP2A-B55 may directly regulate Aurora B activity via CDK1 (Hümmer and Mayer, 2009; Kitagawa et al., 2014).

In CPC-depleted Xenopus egg extracts, microtubule stabilization around chromatin fails, whereas when both the CPC and kinesin-13 MCAK are depleted, microtubules are stabilized (Sampath et al., 2004). Similarly, we found that depletion of kinesin-13 KLP10A partially alleviated the spindle loss in BN2-treated oocytes. These results indicate that the CPC negatively regulates spindle-depolymerizing factors. In contrast, the absence of spindle assembly in CPC RNAi oocytes is not suppressed by simultaneous knockdown of kinesin-13 KLP10A (Radford et al., 2012a,b). The different effects of KLP10A knockdown on oocytes with aurora B inhibited by either RNAi or BN2 treatment is most likely due to the timing of CPC depletion. In aurora B-RNAi oocytes, loss of a kinesin-13 does not restore spindle assembly because factors required for the initiation of spindle assembly are not activated. In contrast, spindle disassembly in oocytes treated with BN2 depends on depolymerizing activities, and therefore requires kinesin-13.

PP2A-B56 probably regulates factors that maintain spindle integrity. Localization and activity of motor proteins such as the kinesin-6 Subito and the kinesin-14 NCD may depend on high levels of Aurora B near the chromosomes (Beaven et al., 2017; Das et al., 2018). Interestingly, Subito contains one predicted PP2A-B56 binding motif (FDNIQESEE; Hertz et al., 2016), which is in a region we proposed negatively regulates Subito activity (Das et al., 2018). Regulation of several spindle assembly factors by antagonism between Aurora B and PP2A is probably a conserved activity, as PP2A has been shown to oppose Aurora B (AURKB) activity in the central spindle of HeLa cells (Bastos et al., 2014) and specifically in the regulation of the Subito homolog MKLP2 (Fung et al., 2017).

PP2A-B56 is required to maintain meiotic cohesion

In addition to the role of antagonizing Aurora B in spindle maintenance, PP2A-B56 is required for two additional meiosis I processes: sister chromatid cohesion and stabilization of end-on KT–MT attachments. The strongest sister chromatid cohesion phenotypes were observed when both B56 subunits, WDB and WRD, were depleted. The loss of cohesion resulted in the separation of sister centromeres and, as expected from the loss of arm cohesion, precocious anaphase (McKim et al., 1993). PP2A-B56 maintains cohesion by dephosphorylation of cohesin subunits, preventing Separase from cleaving the Kleisin subunit (Gutiérrez-Caballero et al., 2012). In many cell types, PP2A-B56 is recruited by shugoshin. Indeed, we observed a reduction in WDB localization in a mei-S332 (the Drosophila shugoshin homolog) mutant. However, there is a striking difference between the phenotype from loss of PP2A-B56 and MEI-S332, even though the latter is required for its localization. The lack of a meiosis I cohesion defect in mei-S332 mutants could be explained if a low level of PP2A-B56 depends on a MEI-S332-independent mechanism for recruitment to the chromosomes. Similar evidence for a second pathway to recruit PP2A has been found in Drosophila male meiosis (Pinto and Orr-Weaver, 2017). Although WDB localization is observed in both prophase I and metaphase I spermatocytes, it is eliminated only from the metaphase I spermatocytes in mei-S332 mutants. Thus, the prophase I localization of WDB does not depend on MEI-S332.

The Drosophila Sororin homolog Dalmatian, and not MEI-S332, is required for cohesion in Drosophila mitotic cells (Yamada et al., 2017) and has sequence features, including the LxxIxE motif, that suggest it recruits PP2A-B56 (Hertz and Nilsson, 2017). However, a cohesion-loss phenotype was not observed in dmt-RNAi oocytes. There are two possible explanations for these results. First, DMT, like the meiotic cohesins required for cohesion, is a stable protein that is only loaded onto chromosomes during premeiotic S phase (Gyuricza et al., 2016). Indeed, we observed DMT localization early in meiotic prophase. Second, DMT and MEI-S332 may redundantly protect cohesion in oocytes. We propose that PP2A-B56 recruitment by DMT occurs during S phase and ensures protection during the long oocyte prophase and into metaphase I. PP2A-B56 recruitment by MEI-S332 occurs later, once the nuclear envelope breaks down, and also protects cohesion during the meiotic divisions. Testing this hypothesis requires a precise knockout of DMT during premeiotic S phase but after the mitotic divisions of the germline.

PP2A-B56 regulates kinetochore–microtubule attachments

When both B56 subunits, WDB and WRD, were knocked out, we observed loss of end-on KT–MT attachments and bi-orientation defects. As in other systems, it is likely that lateral interactions are the first contacts between kinetochores and microtubules (Kalantzaki et al., 2015; Shrestha et al., 2017; Tanaka, 2010), and conversion to end-on attachments depends on factors like CENP-E and kinesin-13 MCAK (Shrestha and Draviam, 2013; Wandke et al., 2012). The bi-orientation errors in PP2A-B56-depleted oocytes could be a consequence of the attachment defects. PP2A, possibly via recruitment by BUBR1, can stabilize KT–MT attachments in mitotic cells (Foley et al., 2011; Kruse et al., 2013; Suijkerbuijk et al., 2012; Xu et al., 2013) and meiotic cells (Tang et al., 2016). Similarly, in Drosophila oocytes, we have shown that BUBR1 stabilizes KT–MT attachments (Wang et al., 2019). In the absence of end-on attachments, lateral attachments persist (Feijão et al., 2013; Radford et al., 2015). We propose at least two possible targets of PP2A-B56 in regulating microtubule attachments. The first could be the N-terminal domain of SPC105R (Nijenhuis et al., 2014; Smith et al., 2019). The N-terminal domain of the SPC105R homolog KNL1 has two known properties: it binds microtubules (Bajaj et al., 2018; Espeut et al., 2012) and it contains two Aurora B phosphorylation sites (Liu et al., 2010; Rosenberg et al., 2011; Welburn et al., 2010). Knowing Aurora B activity can inhibit the conversion of lateral to end-on KT–MT attachments (Kalantzaki et al., 2015; Shrestha et al., 2017), we propose that Aurora B kinase phosphorylation of SPC105R promotes lateral attachments and PP2A reverses these events to allow end-on attachments to occur. The second target of PP2A-B56 could be the Spindly protein, which, when phosphorylated by Polo, allows the Rod–ZW10–Zwilch complex to inhibit conversion of lateral to end-on attachments (Barbosa et al., 2020). Failure to remove Polo-dependent phosphorylation events could result in persistent lateral attachments.

Multiple independent pools of PP2A-B56 regulate meiosis

PP2A-B56 activity has two important functions at the centromeres or kinetochores: protecting cohesion and stabilizing KT–MT attachments. The dramatic disorganization of the spindle and chromosomes in PP2A-B56-depleted oocytes is a consequence of defects in these two functions. PP2A-B56 localization depends on BUBR1 and MEI-S332. This is surprising for two reasons. First, previous studies have suggested that BUBR1 (Kruse et al., 2013; Suijkerbuijk et al., 2012; Xu et al., 2013) and shugoshin/MEI-S332 (Kitajima et al., 2006; Riedel et al., 2006; Tang et al., 2006) regulate two distinct PP2A-B56 pools. Second, the PP2A-B56 loss-of-function phenotypes are much stronger than BUBR1 or MEI-S332 loss of function. To explain these results, we propose there are at least three pools of PP2A-B56 involved in oocyte meiosis.

The first pool is defined by localization of WDB that depends on BUBR1 and MEI-S332. Based on the mild phenotypes of BubR1-RNAi or mei-S332-mutant oocytes, this pool is not essential for cohesion but could regulate end-on attachments. The second pool is defined as being required for cohesion, is independent of BUBR1 and could depend on Dalmatian or be directly recruited by SPC105R using its own short linear motifs (SLiMs; Hertz et al., 2016). The third pool is defined by a role in antagonizing the spindle assembly function of Aurora B. This pool does not depend on kinetochore localization, as shown by the sensitivity of Spc105R-RNAi (Wang et al., 2019) and BubR1-RNAi (this paper) oocyte spindles to BN2 treatment. This pool could be equivalent to the spindle-localized PP2A in C. elegans meiosis (Bel Borja et al., 2020).

An important question for future studies is how different pools of PP2A-B56 are independently regulated and interact. It will be interesting to determine how the functions of the two kinetochore pools are separately regulated. Additionally, it remains to be determined whether the non-kinetochore PP2A-B56 pool regulates kinetochore function, as proposed in mouse oocytes (Touati et al., 2015). If it does, it becomes critical to understand what recruits PP2A to the microtubules of the meiotic spindle. Understanding how PP2A-B56 achieves its multiple functions will require connecting specific functions with its spatial regulation.

MATERIALS AND METHODS

Tissue-specific knockdowns using expression of transgenes and shRNAs

The UAS/Gal4 system was used for tissue-specific expression of transgenes and shRNAs. In most experiments, the transgenes and shRNAs, under UAS control, were expressed using P{w^+mC=matalpha4-GAL-VP16}V37 (matα-GAL4-VP16), which induces expression after premeiotic DNA replication and early pachytene, and persists throughout most stages of meiotic prophase during oocyte development in Drosophila (Radford et al., 2012b; Sugimura and Lilly, 2006). For expression throughout the germline, including the germline mitotic divisions and early meiotic prophase, we used P{w^+mC=GAL4::VP16-nos.UTR}MVD1 (nos-GAL4-VP16). For ubiquitous expression, we used P{tubP-GAL4}LL7. The RNAi lines used in this study are listed in Table S1. We first selected RNAi lines that caused lethality when the shRNA was under the control of P{tubP-GAL4}LL7 and sterility when it was under the control of matα-GAL4. Two RNAi lines were not used because of weak phenotypes. HSM1804 (mts) was fertile with matα-GAL4 and HMS1921 (PP2A-29B) was viable with P{tubP-GAL4}LL7.

The wrd^Δ mutant was generated by FLP-mediated site-specific recombination that removed most of the coding region (Moazzen et al., 2009). The deletion Df(3R)189 was made by imprecise excision of a P-element within the wrd gene and removes all of wrd and a couple genes on each side (Viquez et al., 2006). For knockdown of wrd and wdb, mutations and shRNA were combined to generate Df(3R)189 matα-GAL4/ wrd^Δ; wdb^HMS01864 females, or two shRNAs were combined to generate GL00671/+; matα-GAL4/wdb^HMS01864 females. The sunn mutant females used in this study were sunn^167-109/sunn^147-75 heterozygotes. The mei-S332 mutant females used in this study were mei-S332⁷/mei-S332⁴ heterozygotes.

Generation of shRNA lines and analysis by RT-PCR

Sequences for shRNAs targeting wdb and dmt were designed using DSIR (http://biodev.extra.cea.fr/DSIR/DSIR.html; Vert et al., 2006) and the GPP Web Portal (http://www.broadinstitute.org/rnai/public/seq/search). These were cloned into the pVALIUM22 vector for expression under control of the UASP promoter (Ni et al., 2011).

To measure the knockdown of mRNA in oocytes, Taqman RT-qPCR was used. In order to extract RNA from oocytes of interest, female flies were placed in yeasted vials for ∼3 days. The females were then broken apart in a blender containing 1× phosphate-buffered saline (PBS), and oocytes were filtered through meshes, as described below for cytological analysis of stage-14 oocytes. 50 mg of oocytes was weighed out, and 1 ml of TRIzol reagant was added for RNA extraction according to the manufacturer's instructions (Thermo Fisher Scientific). cDNA was prepared from 2 μg of RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR was performed using a TaqMan Assay (Applied Biosystems) with four replicates per reaction.

Nondisjunction and fertility assay

Drosophila crosses were used to determine the rate of nondisjunction and fertility of certain genotypes. In this cross, female virgin flies with a gene of interest were mated to y w/B^sY males. These males carry a dominant Bar mutation on the Y chromosome. Therefore, the progeny from a cross with y w/B^sY males are Bar males (XY) and wild-type females (XX). Nondisjunction of the sex chromosomes during meiosis results in four different zygotes, two are inviable (OY, XXX) and two are viable with distinguishing phenotypes, Bar females (XXY) and wild-type males (XO). The nondisjunction rate was calculated by: , where NDJ is the number of nondisjunction progeny.

Cytology of stage-14 oocytes and drug treatment

To prepare oocytes for cytology, 0- to 3-day-old females were placed in yeasted vials for 1–2 days, as described previously (Gilliland et al., 2009a). The females were ground in a blender with modified Robb's buffer (Theurkauf and Hawley, 1992) and filtered through a series of meshes to separate the stage-14 oocytes from other body parts. At this stage, oocytes were treated with 50 µM BN2 in modified Robb's buffer for 1 h. The BN2 was dissolved in DMSO to make a 50 mM stock solution, and then 1 µl was added to 999 µl of modified Robb's buffer in a 5 ml tube. Control oocytes were incubated for 1 h in 1 ml of modified Robb's buffer containing 1 µl DMSO. By incubating the 1 ml solution in a 5 ml tube, hypoxia was avoided, which has been shown to affect localization of proteins such as MPS1 and Polo in Drosophila oocytes (Gilliland et al., 2009b).

Oocytes were fixed in 5% formaldehyde and heptane and then washed with 1× PBS (Radford and McKim, 2016). Oocyte membranes were mechanically removed by rolling the oocytes between the frosted side of a glass slide and a coverslip. The rolled oocytes were then rinsed into 15 ml conical tubes containing PBS with 1% Triton X-100 and were rotated for 1.5–2 h. Oocytes were washed in PBS containing 0.05% Triton X-100 and subsequently transferred to a 1.5 ml Eppendorf tube. Oocytes were then blocked in PTB (0.5% BSA and 0.1% Tween-20 in PBS; Theurkauf and Hawley, 1992) for 1 h and then incubated with primary antibodies overnight at 4°C. The next day, the oocytes were washed four times in PTB and the secondary antibody was added. After incubating at room temperature for 4 h, the oocytes were washed in PTB, stained for DNA with Hoechst33342 (10 µg/ml) and then washed twice in PTB.

The primary antibodies for this study were tubulin monoclonal antibody DM1A (1:50) directly conjugated to FITC (Sigma-Aldrich, St Louis, MO), rat anti-Subito (1:150; Jang et al., 2005), rat anti-INCENP (1:600; Wu et al., 2008), rabbit anti-phospho-INCENP (1:1000; Salimian et al., 2011; Wang et al., 2011a), guinea pig anti-MEI-S332 (1:5000; Moore et al., 1998), rabbit anti-CENP-C (1:5000; Heeger et al., 2005), rabbit anti-SPC105R (1:4000; Schittenhelm et al., 2007), rabbit anti-WDB (1:1500; Sathyanarayanan et al., 2004), rabbit anti-DMT (1:1000; Yamada et al., 2017), rat anti-HA (1:50; clone 3F10; Roche), rabbit anti-CID (1:1000; cat. no. 61735; Active Motif), and mouse anti-C(3)G (1:500; Page and Hawley, 2001). These primary antibodies were combined with either a Cy3-, Alexa Fluor 543-, Cy5- or Alexa Fluor 647-conjugated secondary antibody pre-absorbed against a range of mammalian serum proteins (Jackson ImmunoResearch and Thermo Fisher Scientific). This protocol was modified for FISH, as described previously (Radford and McKim, 2016), using probes corresponding to the X chromosome 359-bp repeat labeled with Alexa Fluor 594, second chromosome AACAC repeat labeled with Cy3 and the third chromosome dodeca repeat labeled with Cy5 (IDT). The oocytes were mounted in Slowfade Gold antifade (Thermo Fisher Scientific) and imaged using a Leica TCS SP8 confocal microscope with a 63×, NA 1.4 lens.

Image analysis, quantification and statistical analysis

Leica Confocal Software was used to create maximum intensity projections of complete image stacks of individual or merged channels. Image cropping and addition of scale bars was done in Adobe Photoshop. Centromere foci, SPC105R or WDB intensity, and spindle intensity were measured using Imaris image analysis software (Bitplane). Intensity values are reported relative to the background by dividing the intensity of the each spot by the average intensity of the background. Statistical tests were performed using GraphPad Prism software. Data for chromosome phenotypes, centromere foci or intensities were pooled together, and one-way ANOVA or unpaired t-tests were run. Details of statistical evaluations and the sample sizes are provided in the figure legends.

Peer review history

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.254037