A maternal effect rough deal mutation suggests that multiple pathways regulate Drosophila RZZ kinetochore recruitment

Cell division is driven by oscillations in the activity of cyclin-dependent kinases (Cdks) and their associations with regulatory cyclin subunits. Mitotic entry is driven by high Cdk1–Cyclin-B activity, and mitotic exit is dependent on mechanisms that inactivate this complex and dephosphorylate its targets (reviewed in O'Farrell, 2001; Lindqvist et al., 2009; Mochida and Hunt, 2012). The initiation of mitotic exit at the metaphase-anaphase transition is under the control of the spindle assembly checkpoint (SAC), which indirectly regulates the proteolytic degradation of Cyclin B by controlling its ubquitinylation by the E3 ubiquitin ligase known as the anaphase promoting complex (APC/C) (reviewed in Musacchio, 2011; Lara-Gonzalez et al., 2012; Jia et al., 2013). Among the key SAC proteins in this process are Mad1 and Mad2, which form a complex at unattached kinetochores and help generate an inhibitor of the APC/C. Upon proper kinetochore attachment to the spindle, several SAC components, including Mad1, Mad2 and the Rough Deal (Rod)–Zw10–Zwilch (RZZ) complex are removed from the kinetochore by a dynein-dependent process (Howell et al., 2000; Hoffman et al., 2001; Howell et al., 2001; Wojcik et al., 2001; Basto et al., 2004), thus dismantling the SAC apparatus and inactivating the source of the APC/C inhibitor.

RZZ is an important component of the outer kinetochore (Karess, 2005). It is a determinant of Mad1–Mad2 kinetochore levels, as it promotes both Mad1–Mad2 recruitment (Buffin et al., 2005; Kops et al., 2005) and recruitment of dynein and its associated regulatory proteins (Starr et al., 1998; Griffis et al., 2007; Chan et al., 2009). RZZ is also involved in regulating kinetochore–microtubule attachments (Gassmann et al., 2008; Gassmann et al., 2010; Cheerambathur et al., 2013), although how it performs these functions, and how it is itself recruited to kinetochores is poorly understood.

In the early division cycles of Drosophila embryogenesis, nuclei divide synchronously in a syncytium, using a streamlined cell cycle in which S- and M-phase oscillate, driven by stockpiles of maternal mRNAs and proteins (reviewed in Foe et al., 1993; Lee and Orr-Weaver, 2003). During these syncytial cycles, the unfertilized nuclear products of female meiosis remain in M-phase at the periphery of the embryo. These nuclear products, referred to as polar bodies, coalesce to form a ‘starburst’ of condensed chromosomes and maintain this state until they are culled from the embryo during cellularization.

Defective polar body M-phase maintenance has been reported in mutants with either reduced Cdk1 activity or misregulated levels of Cyclin B. For example, the Pan Gu (Png) kinase complex promotes the translation of Cyclin B during female meiosis and syncytial embryogenesis. Females mutant for the png kinase or its regulatory subunits lay eggs with abnormally low Cyclin B levels. As a result, polar bodies do not remain in M-phase but instead re-enter interphase and undergo unregulated rounds of DNA replication, resulting in polyploid interphase-like polar bodies (Fenger et al., 2000; Lee et al., 2001). The SAC, which also regulates levels of Cyclin B, is another proposed mechanism by which polar body M-phase is maintained. Indeed, eggs laid by females homozygous for certain viable mutations of the SAC genes Mps1 (also known as ald in Drosophila) and BubR1 present similar polar body defects (Fischer et al., 2004; Pérez-Mongiovi et al., 2005).

In a screen for new cell-cycle regulators specifically required during the syncytial stages of embryogenesis, we identified a novel, maternal-effect allele of the rough deal (rod) gene, called here rod^Z3. This mutation only slightly perturbs RZZ activity in homozygotes, which are viable and have relatively normal mitoses. However, in eggs laid by rod^Z3 mothers, RZZ is incapable of kinetochore recruitment. As a result, such embryos present decondensed polar bodies, profoundly perturbed syncytial mitoses, and a non-functional SAC. The phenotype of this mutation suggests that aspects of kinetochore assembly may differ in maternal and zygotic Drosophila mitosis.

In a screen for new genes in Drosophila affecting polar body maintenance, we identified a novel allele of rod in a large maternal-effect lethal collection (Koundakjian et al., 2004). Flies homozygous for this allele, called rod^Z3, are viable and appear normal. However, they are completely sterile; the females lay eggs that fail to hatch (Table 1). Examination of 3–5-hour-old eggs revealed a number of early developmental defects. None of the embryos had reached gastrulation and most presented highly abnormal distribution of blastoderm nuclei and unusually large internal nuclei, suggesting problems with the syncytial mitoses (Table 1; Fig. 1A; supplementary material Fig. S1). In contrast, nearly all the wild-type (WT) embryos at this stage had either begun gastrulating (73%) or were cellularizing (22%) (Table 1).

In younger (0–2 h) syncytial stage rod^Z3-derived embryos, the nuclei were frequently unevenly spaced and asynchronously cycling (Fig. 1A,B; Table 1). The majority of the mitotic figures in these embryos were abnormal, including multipolar spindles with multiple centrosomes and poorly focused spindles lacking one or both centrosomes (Fig. 1B). In fact, the overall phenotype of these embryos is similar to that reported for germline clones of zw10-null mutants (Williams and Goldberg, 1994).

In nearly all the eggs derived from rod^Z3 females, the polar body formed a large indistinct mass of interphase-like decondensed chromatin instead of the starburst configuration of condensed chromosomes found in WT embryos (Table 2; Fig. 1A,C). The increased size of these polar bodies and the intensity of their DNA stain compared to WT suggest that they have an increase in DNA content, presumably owing to additional rounds of DNA replication. Expression of the rod^+tC3L9 transgene, which contains the WT rod genomic region, in the rod^Z3 background rescued this polar body defect, the embryonic lethality (Table 1) and the mitotic phenotypes described below, confirming that the mutation in rod was responsible for the phenotypes.

Immunostaining rod^Z3-derived embryos for phosphorylated Histone H3 (PH3), a marker of M-phase chromosome condensation, distinguished two classes of decondensed polar bodies. Roughly 40% of the polar bodies were PH3-negative with rounded, interphase-like chromatin (Fig. 1C; Table 2); the rest were PH3-positive with irregularly-shaped and partially decondensed chromatin. WT polar bodies were PH3-positive as has been reported by others (Fischer et al., 2004; Pérez-Mongiovi et al., 2005). The irregularly-shaped, semi-decondensed and PH3-positive polar bodies of rod^Z3-derived embryos also labeled with Spc25–RFP (a marker of the outer kinetochore, present only during M-phase), indicating that kinetochores were intact. Interestingly, these polar bodies contained more than the expected 12 kinetochore signals (Fig. 3A), suggesting that they had at least partially replicated their DNA and returned to an M-phase-like state. The more rounded interphase-like polar bodies lacked Spc25–RFP (Fig. 3A), consistent with the interpretation that they were in interphase. These results suggest that the polar bodies of rod^Z3-derived embryos might be cycling in and out of M-phase, as has been proposed for other mutants (Fischer et al., 2004; Pérez-Mongiovi et al., 2005).

Decondensed polar bodies have been reported previously in mutants of mps1 and BubR1 (Fischer et al., 2004; Pérez-Mongiovi et al., 2005), two other genes encoding SAC components. rod is the third SAC gene to be associated with this phenotype. To determine whether rod^Z3-derived syncytial embryos could undergo SAC-dependent mitotic arrest in response to spindle damage, syncytial-stage embryos were incubated with or without colchicine for 30 min, and the fraction of mitotic embryos was determined (see Materials and Methods for details). Whereas nearly all WT embryos had become mitotic (Fig. 1D), neither rod^Z3 nor Mad2-null (Mad2^P)-derived embryos (a negative control) displayed any increase in the fraction of mitotic embryos in response to colchicine treatment. These results indicate that the SAC is defective in rod^Z3-derived embryos and suggests that the failure of polar bodies to remain in M-phase is due to the defective SAC.

However, Mps1, BubR1 and the RZZ complex all participate in other mitotic functions that do not involve the SAC (Starr et al., 1998; Ditchfield et al., 2003; Jones et al., 2005; Lampson and Kapoor, 2005; Maure et al., 2007; Jelluma et al., 2008; Hewitt et al., 2010; Santaguida et al., 2010; Sliedrecht et al., 2010; Althoff et al., 2012; Wainman et al., 2012). It is therefore conceivable that the polar body decondensation is not specifically connected to a defect in SAC activity. To confirm the role of the SAC in the maintenance of polar body condensation, we examined embryos laid by Mad2^P females. Drosophila Mad2-null mutants have a defective SAC, but do not exhibit the mitotic defects associated with mutations in other SAC genes (Buffin et al., 2007). We found that most of the polar bodies in Mad2^P-derived embryos were large with semi- or fully decondensed chromatin (Table 2), similar to those of rod^Z3. This result supports the hypothesis that SAC activity is required to maintain polar body condensation, presumably by stabilizing Cyclin B (Fischer et al., 2004; Pérez-Mongiovi et al., 2005). To more rigorously test this latter point, we generated rod^Z3 mutants carrying one copy of a lethal allele of the shattered (shtd) gene encoding APC/C subunit APC1 in Drosophila. We also directly increased Cyclin B levels in rod^Z3 flies by introducing two additional copies of the WT Cyclin B (CycB) gene. Both modifications partially suppressed the polar body phenotype of rod^Z3-derived embryos (Table 2). The frequency of abnormal polar bodies fell to 69% in shtd³/+; rod^Z3-derived embryos, and to 53% in 2×Cyclin B/+; rod^Z3-derived embryos. These results argue that WT rod is required to maintain polar body M-phase through its role in the SAC, which in turn inhibits APC/C-regulated degradation of Cyclin B.

Sequencing the rod^Z3 genomic DNA revealed a point mutation in rod where glycine 1973 was replaced by a glutamic acid residue, predicted to disrupt a short α-helix in the relatively conserved C-terminal region of the 2089-residue Rod protein. Drosophila Rod has an overall predicted structure similar to that described for human Rod (also known a KNTC1) (Çivril et al., 2010) with N-terminal β-folds and a long α-solenoid region extending to the C-terminus. Although not conserved in vertebrate Rod (where it is a tyrosine or a serine residue), the glycine at position 1973 has been conserved since the divergence of mosquitos and Drosophila, over 250 million years (Wiegmann et al., 2011) (Fig. 2A).

The RZZ complex, with an apparent molecular mass of 600–700 kDa, is believed to contain two copies of each of the three subunits Rod, Zw10 and Zwilch, with Rod serving as the scaffold (Scaërou et al., 2001; Williams et al., 2003; Çivril et al., 2010). Immunoblotting revealed that levels of Rod, Zw10 and Zwilch in total extracts of rod^Z3-derived embryos were similar to those in WT embryos (Fig. 2B). More importantly, Zw10 and Zwilch co-immunoprecipitated with Rod equally well in extracts of WT or rod^Z3-derived embryos (Fig. 2B). Fractionating lysates from WT and rod^Z3-derived syncytial embryos by sucrose density gradient centrifugation revealed that the rod^Z3 mutation does not affect the capacity of Rod to form high molecular mass complexes (Fig. 2C). These data argue that the rod^Z3 mutation does not affect RZZ complex formation or steady-state levels.

With the knowledge that the RZZ complex was intact, we next examined its behavior in rod^Z3-derived syncytial embryos by following GFP-tagged Zw10. WT embryos expressing GFP–Zw10 and Spc25–RFP contained RZZ foci at the kinetochores of the polar body and mitotic chromosomes (Fig. 3A,B) as well as along metaphase spindle fibers (Fig. 3B), reflecting the streaming of RZZ from attached kinetochores (Wojcik et al., 2001; Basto et al., 2004). In contrast, no GFP signal was detected on any mitotic structure in rod^Z3-derived embryos (Fig. 3A,B). These data suggest that the rod^Z3 mutation interferes with kinetochore recruitment of the RZZ complex.

Although the absence of RZZ recruitment provided an explanation for the severity of the phenotype in embryos laid by homozygous rod^Z3 mothers, it was also intriguing because the homozygous rod^Z3 offspring of heterozygous parents were apparently normal. A mutation blocking RZZ kinetochore recruitment would normally be a zygotic lethal. Indeed, null mutations in any component of RZZ are homozygous lethal, dying as a consequence of profound defects in chromosome segregation during postembryonic development (Karess and Glover, 1989; Williams et al., 1992; Williams and Goldberg, 1994; Williams et al., 2003; Karess, 2005). We therefore examined RZZ behavior in dividing tissues of homozygous rod^Z3 third-instar larvae (Fig. 3C). Unlike the rod^Z3-derived embryos, rod^Z3 larval neuroblasts displayed substantial but limited recruitment of RZZ to kinetochores during unperturbed mitosis. Average RZZ kinetochore levels (normalized to the Spc25–RFP signal) were approximately one-third that of WT (Fig. 3D). In addition to the reduction of RZZ recruitment to kinetochores, there was little or no RZZ on the spindle fibers and poles of rod^Z3 neuroblasts, suggesting a problem with streaming (see below and Fig. 5).

In WT cells, a brief colchicine or nocodazole treatment to depolymerize kinetochore microtubules greatly increases the levels of several outer kinetochore proteins, including RZZ, Mad1 and Mad2 (Hoffman et al., 2001; Williams and Goldberg, 1994; Basto et al., 2004; Buffin et al., 2005). Strikingly, colchicine treatment did not significantly increase the kinetochore signal of GFP–Zw10 in rod^Z3 neuroblasts; whereas in WT cells, the signal more than doubled (Fig. 3D). This reduced recruitment of RZZ in rod^Z3 neuroblasts was confirmed by quantification of Rod levels by immunostaining with anti-Rod antibody (data not shown).

To assess whether this limited RZZ recruitment is compatible with a functional SAC, we assayed the ability of colchicine-treated rod^Z3 larval brains to accumulate mitotic cells (Fig. 3E). Both rod^Z3 and WT larval brains more than doubled their density of mitotic cells after 1 h incubation in colchicine, whereas, rod-null brains did not. These results indicate that the SAC is properly functioning in rod^Z3 larval neuroblasts.

Examination of karyotypes of mitotic cells in third-instar rod^Z3 larval brains revealed relatively little aneuploidy (Table 3). Live imaging of dividing rod^Z3 neuroblasts (Fig. 3F), however, showed that there was an elevated frequency of anaphases in which some kinetochores migrated poleward more slowly than the others. Indeed, the fraction of anaphases with at least one lagging kinetochore was nearly as high in rod^Z3 as in rod or Mad1-null mutants (Emre et al., 2011), although only rarely was more than one kinetochore involved (Table 3). As is the case for Mad1-null mutants (Emre et al., 2011), these lagging chromatids were in most cases correctly segregated, despite their delay in migration (given that overall aneuploidy levels are low). In several cases, the lagging chromatids appeared to be attached to spindle fibers emanating from both poles (merotelic attachment), which is also a feature associated with mitosis in Mad1-null mutants (Emre et al., 2011).

In summary, RZZ retains a limited capacity to associate with the mitotic apparatus in rod^Z3 larval neuroblasts, but this is apparently adequate for mounting a SAC-mediated mitotic arrest and providing sufficient functionality for maintaining reasonably accurate mitosis. In contrast, RZZ appears to be incapable of kinetochore recruitment in early embryos derived from rod^Z3 mothers. These results largely explain the different mitotic phenotypes of these two tissues.

We next asked whether rod^Z3 affected the recruitment of Mad1 and Mad2 given that kinetochore levels of the Mad1–Mad2 complex depend in part on the presence of functional RZZ (Buffin et al., 2005; Kops et al., 2005). In untreated rod^Z3 neuroblasts, levels of Mad1 (Fig. 4A) and Mad2 (data not shown) were often low, and sometimes undetectable on prometaphase mitotic figures. On average, the Mad1–GFP kinetochore signal in rod^Z3 was not significantly different from that in rod-null cells (Fig. 4B). The low levels of Mad1 might therefore help explain why rod^Z3 displays the elevated rate of lagging chromatids also seen in Mad1 mutants. However, when treated with colchicine, rod^Z3 neuroblasts did accumulate Mad1 and Mad2 at kinetochores (Fig. 4C). Both the rate of this accumulation and the final level attained were reduced; reaching approximately one-third of the levels found in WT after 8–10 min. This degree of Mad1 and Mad2 recruitment must be sufficient to elicit the robust SAC response of colchicine-treated rod^Z3 larval brains (Fig. 3E).

The relatively low levels of Mad1 recruitment after treatment with colchicine might be a quantitative consequence of the reduced RZZ at rod^Z3 kinetochores or a qualitative consequence of the structural change caused by the rod^Z3 mutation (or both). To gain insight into this question, we simultaneously measured the fluorescent signals from GFP–Zw10 and Cherry–Mad1 on the kinetochores of individual colchicine-treated neuroblasts and displayed the values in a scatter plot (Fig. 4D). The plot shows that the highest RZZ and Mad1 signals are (not surprisingly) only found in the WT cells. However, considering only the range of RZZ values overlapping between WT and rod^Z3 (included within the box of Fig. 4D), the distribution of Mad1 values was not significantly different and the linear regression curves that fitted the two data sets had similar slopes. This result indicates that the recruitment levels of Mad1 are ‘normal’ in rod^Z3 cells, given the reduced amount of RZZ on unattached kinetochores.

To determine whether the rod^Z3 mutation is affecting steady-state turnover of Mad1 at unattached kinetochores, we subjected Mad1–GFP to fluorescence recovery after photobleaching (FRAP) analysis (Fig. 4E). The Mad1–GFP that had accumulated on rod^Z3 kinetochores after 15 min in colchicine displayed the same turnover dynamics as that in WT. A fast turnover pool, ∼40% of the total, recovered with a half-life of 10–12 seconds, whereas the remaining 60% was essentially stable, showing little or no recovery. Thus, the steady-state turnover properties of Mad1 are not disrupted in rod^Z3 neuroblasts.

In summary, the primary cause of the reduced kinetochore Mad1 signal in colchicine-treated rod^Z3 neuroblasts is the low levels of RZZ recruitment itself, and not a specific effect of the rod^Z3 mutation on the ability of RZZ to recruit Mad1. Nevertheless, the systematic reduction in the kinetochore Mad1 signal seen in unperturbed mitosis (Fig. 4A) to levels found in rod-null cells, and the fact that Mad1 is slower to accumulate on unattached kinetochores (Fig. 4C), suggests the rod^Z3 mutation might specifically alter the rate of Mad1 accumulation.

Several transient outer kinetochore components, including RZZ and Mad1–Mad2, display a characteristic movement from kinetochores to poles along kinetochore fibers called shedding, stripping or streaming (Howell et al., 2000; Howell et al., 2001; Wojcik et al., 2001; Karess, 2005; Famulski et al., 2011). This dynein-dependent process is believed to contribute to the extinction of the SAC signal following proper kinetochore–microtubule attachment. In live images of GFP–Zw10 expressed in rod^Z3 neuroblasts (Fig. 3C), this streaming appeared to be substantially reduced. Because streaming is highly variable, even in WT, we quantified this apparent difference by scoring mitotic cells into three classes according to their degree of streaming: none (all GFP–Zw10 signal restricted to kinetochores), low (GFP–Zw10 signal mostly on kinetochores) and high (signal on kinetochores, spindle and poles). By these criteria, rod^Z3 neuroblasts consistently failed to show any streaming (Fig. 5A).

Given that streaming is a dynein-dependent process and RZZ is required for the recruitment of the dynein–dynactin complex to kinetochores (Starr et al., 1998; Bader and Vaughan, 2010; Raaijmakers et al., 2013), we asked whether rod^Z3 specifically reduced kinetochore dynein levels. We measured the accumulation of GFP–DLIC2, a dynein subunit, relative to Spc25–RFP, on mutant or WT cells incubated 15 min in colchicine (Fig. 5B,C). In 90% of rod^Z3 cells, GFP–DLIC2 was detectable at kinetochores, averaging about half the signal found in WT. As was the case for kinetochore Mad1–GFP, the range of GFP–DLIC2 levels was very broad and probably reflects the variation in RZZ recruitment levels.

Given that dynein recruitment is only partially reduced, whereas dynein-dependent streaming is entirely absent, we considered the possibility that the Rod^Z3 protein was directly interfering with the ability of RZZ to stream. Consistent with this idea, rod^Z3/+ heterozygous neuroblasts also displayed reduced streaming relative to WT (Fig. 5A), suggesting that rod^Z3 has a dominant effect in disrupting RZZ function. To test this idea, we co-expressed two Rod proteins with different fluorescent tags in a rod-null background: Cherry–Rod+ and either GFP–Rod+ or GFP–Rod^Z3, expressed from a transgene containing the identical mutation as the rod^Z3 allele (Fig. 5D). When two WT transgenes were expressed, both green and red particles of RZZ streamed on kinetochore-attached microtubules, whereas when GFP–Rod^Z3 was co-expressed with Cherry–Rod+ the streaming of both the red (WT) and green (mutant) RZZ was greatly reduced (Fig. 5D). These results are consistent with a partially functional GFP–Rod^Z3 subunit assembling into the multi-subunit RZZ complex and inhibiting its ability to be transported from kinetochores.

In a search for maternal effect mutants affecting the regulation of early embryonic divisions, we found a unique allele of rod, called rod^Z3, which profoundly affects the ability of RZZ to be recruited to kinetochores in the early embryo.

Although previous studies have shown that Cdk1 and Cyclin B activity are required for polar body condensation, we have established that polar bodies maintain Cdk1–Cyclin-B activity through activation of the SAC. We show here that the RZZ complex, like Mps1 and BubR1, localizes to kinetochores in polar bodies and is required to maintain polar body condensation in Drosophila. rod^Z3-derived embryos, like those of Mad2, Mps1 and BubR1 mutants, contain large interphase-like polar bodies.

Pérez-Mongiovi et al. (Pérez-Mongiovi et al., 2005) found evidence that in BubR1^Rev1-derived embryos, polar bodies cycle between periods of condensation and decondensation with accompanying DNA replication. They suggested that these polar body cycles are under the control of the embryonic mitotic oscillator. Because polar bodies in rod^Z3 embryos are not always fully decondensed, it is possible that they undergo similar cycles of DNA replication and condensation. The fact that the semi-decondensed polar bodies have more than the expected 12 kinetochores (labeled with the outer kinetochore protein Spc25; Fig. 3A), is consistent with this idea.

Directly or indirectly increasing Cyclin B levels in rod^Z3-derived embryos partially suppresses the polar body condensation defects. This result argues that Rod and the other SAC components maintain polar body condensation by inhibiting the APC/C-mediated degradation of Cyclin B. The Png kinase complex is also required to maintain Cyclin B and thus polar body condensation (Lee et al., 2001). However, rod and png regulate Cyclin B levels through different mechanisms. During embryogenesis, Png kinase promotes the translation of Cyclin B by regulating the polyadenylation of its mRNA and antagonizing the translational repressor Pumilio (Vardy and Orr-Weaver, 2007). The SAC components, by contrast, recognize the polar body kinetochores as improperly attached to the spindle, and generate the inhibitor of APC/C-mediated proteolytic degradation of Cyclin B.

rod^Z3-derived embryos undergo aberrant syncytial mitosis and a developmental arrest prior to gastrulation. These embryos accumulate asynchronously dividing nuclei with centrosome attachment defects throughout the syncytial divisions. Defects in mitotic spindle function during syncytial embryogenesis can uncouple the nuclear and centrosomes cycles (Archambault and Pinson, 2010). The resulting replication of detached centrosomes and fusion of free centrosomes to neighboring mitoses leads to additional aberrant mitoses. Furthermore, centrosome detachment during syncytial embryogenesis inhibits the nuclear positioning and migration required for morphogenesis, all of which might contribute to the developmental arrest of rod^Z3-derived embryos.

In Drosophila, inactivation of the SAC has only minor consequences on mitosis and on embryonic viability (Buffin et al., 2007). Comparing the embryonic phenotype of rod^Z3 with that of Mad2 nulls is instructive. Both lack a functional SAC in embryos and consequently the polar body fails to arrest in M-phase. However, although 60–70% of Mad2-null embryos hatch and reach adulthood (Buffin et al., 2007), no rod^Z3-derived embryos reach the gastrulation stage. The different phenotypes of Mad2-null and rod^Z3-derived embryos argue strongly that it is not the absence of the SAC per se that is responsible for the lethality of rod^Z3-derived embryos, but rather the combination of defects caused by simultaneously removing the SAC and non-SAC functions of RZZ, the latter regulating dynein and kinetochore–microtubule attachments (Starr et al., 1998; Yang et al., 2007; Gassmann et al., 2008; Gassmann et al., 2010).

The primary defect we have identified in R^Z3ZZ is its failure to associate with syncytial and polar body kinetochores in the early embryo. Most of the other phenotypes can be explained as secondary consequences of this recruitment failure. The striking difference between the maternal and zygotic phenotypes of rod^Z3 can be ascribed to the fact that R^Z3ZZ partially retains its ability to be recruited to kinetochores in dividing larval tissues.

Why is R^Z3ZZ capable of kinetochore recruitment in larval neuroblasts and not in maternally-derived early embryos? One possibility is that the unusual rapidity of syncytial mitosis might render it more sensitive to the structural change in R^Z3ZZ. However, the fact that R^Z3ZZ is also absent from polar bodies argues that this is unlikely to be a complete explanation because the polar body does not rapidly cycle but remains in M-phase. It would, in this scenario, still have enough time to recruit R^Z3ZZ.

We suggest that there are factors involved in recruiting RZZ to kinetochores that subtly differ between the syncytial stage and postembryonic mitoses. For example, there might be two, partially redundant, recruitment pathways for RZZ. One would be employed in postembryonic mitoses, such as in larval neuroblasts, but not in the early embryo. The other recruitment pathway would normally be used in both tissues, but is damaged in rod^Z3. These two ‘pathways’ could be as simple as two different kinetochore-binding domains of RZZ and two corresponding docking sites on kinetochores. This idea might also provide an explanation for the second striking feature of R^Z3ZZ behavior: in larval neuroblasts, it shows no further kinetochore accumulation beyond the basal level when microtubules are depolymerized with colchicine. This supplemental RZZ binding might also depend on the second recruitment pathway, explaining why it is defective in rod^Z3. The mutation in rod^Z3 might thus affect a residue important to RZZ kinetochore binding, about which very little is known, particularly in Drosophila. Kinetochore proteins implicated in RZZ recruitment in vertebrates such as ZWINT (Starr et al., 2000; Famulski et al., 2008) and CENP-I (Matson and Stukenberg, 2014), have not been found in the fly genome (Przewloka et al., 2007).

During unperturbed mitosis, kinetochore Mad1 levels are as low in rod^Z3 as they are in rod-null larval neuroblasts (Fig. 4A,B). We have previously reported that Mad1-null neuroblasts display a relatively mild mitotic phenotype. Despite having a defective SAC, they display little aneuploidy, but have a high incidence of lagging chromatids and merotely (Emre et al., 2011). It is therefore not surprising that we find a similar phenotype in rod^Z3 cells, which have greatly reduced kinetochore Mad1 levels.

The mutation in rod^Z3 does not seem to alter the ratio of Mad1 recruited per unit of RZZ. On unattached kinetochores, given enough time, Mad1 accumulates to levels approximately proportional to the amount of RZZ present (Fig. 4C,D). The recruited Mad1 also displays the same steady-state turnover dynamics as in WT cells (Fig. 4E). However, the time required to reach that maximal recruitment is longer than in WT (Fig. 4C). We speculate that the slower recruitment rate of Mad1 might explain its low levels on kinetochores during unperturbed mitosis. There is little known about the molecular mechanism by which RZZ helps to recruit Mad1. It is not even known if they interact directly, although a fraction of RZZ and Mad1 can be co-immunoprecipitated from mitotic cells (our unpublished work).

Finally, rod^Z3 has a perceptible dominant effect on the dynein-dependent streaming of RZZ. This can best be seen in Fig. 5D, where RFP–Rod^Z3 co-expressed with GFP-Rod+ interferes with the streaming of both tagged proteins from kinetochores. The RZZ complex is believed to contain two copies of each of its three subunits, with Zw10 and Zwilch binding to the N-terminal half of the Rod protein (Starr et al., 1998; Scaërou et al., 2001; Williams et al., 2003; Çivril et al., 2010; Cheerambathur et al., 2013). This stoichiometry means that a given RZZ particle could contain one WT and one mutant copy of RZZ, if both versions are present. Given that the effect of rod^Z3 on streaming is more severe than its effect on dynein recruitment, we suggest that the mutation in rod^Z3 might also perturb an interaction with the dynein motor complex.

In summary, the mutation in rod^Z3 affects three activities of RZZ: its own kinetochore recruitment, its capacity to rapidly recruit Mad1 and its capacity to be removed from attached kinetochores by dynein-dependent streaming. The G to E mutation of rod^Z3 at residue 1973 resides in the most highly conserved portion of the Rod protein, known as Rod_C domain (pfam10493), which is predicted to fold into several short α-helices. This domain is only found in known or suspected homologs of Rod, although no specific function or interacting proteins have been are associated with it. Thus, characterization of rod^Z3 provides a unique opportunity to examine how the Rod C-terminus contributes to RZZ function.

Flies were maintained at 25°C using standard techniques. A y w stock was used as wild type for embryo experiments. The stock bw; rod^Z3 st/TM6B was a gift from Charles Zuker (Columbia University, New York, NY). The rod^Z3 allele had designation Z3-0733 in the Zuker stock collection. We shortened and superscripted it to indicate that it is an allele of rod. The following mutations and transgenes have been previously described: rod-null alleles rod^AG1 and rod^X2 and the transgene rod^+tC3L9 (Scaërou et al., 1999), GFP-Rod (Basto et al., 2004), RFP-Rod (Buffin et al., 2005), the null allele Mad1¹, the transgene Mad1-GFP and Cherry-Mad1 (Emre et al., 2011), the null allele Mad2^P (Buffin et al., 2007), the 2XCyclin B stocks, and transgenes Spc25-GFP and Spc25-RFP (Schittenhelm et al., 2007) (gifts from Christian Lehner, University of Zurich, Zurich, Switzerland), the shtd³ stock (from the Bloomington stock center), transgene GFP-DLIC2 (gift from Jordan Raff, Oxford, UK).The GFP-Zw10 construct will be described elsewhere. Transgenic lines Cherry-Rod, and GFP-Rod^Z3 were obtained by P-element transposition. The transgenes correspond to genomic sequence (including the promoter region) as originally described previously (Basto et al., 2004). Fly transformation was performed by BestGene (Chino Hills, CA).

0–2-h embryos (unless otherwise noted) were collected as described previously (Rothwell and Sullivan, 2000). For Tubulin, Centrosomin and DNA staining, embryos were dechorionated in 50% bleach, and fixed and devitellinized by shaking in a 1∶1 mixture of methanol and heptane. For visualizing RFP or GFP fluorescence during cortical divisions, 0–30-min embryos were collected, aged for 1 h 15 min, dechorionated and fixed and devitellinized as above. For PH3 immunostaining, embryos were dechorionated, fixed in a 1∶1 mixture of 4% formaldehyde (in PBS) and heptane, and devitellinized as described above. For colchicine treatment, 0–1-h embryos were collected, dechorionated and incubated in 1∶1 mixture of 250 µM colchicine (in PBS) and heptane for 30 min. Treated embryos were then fixed with formaldehyde and devitellinized as above. The SAC response in embryos was determined by comparing the percentage of mitotic embryos following a 30 min incubation with or without colchicine. ‘Mitotic’ embryos were defined as those having at least 50% PH3-positive nuclei. For immunostaining, fixed embryos were rehydrated in PBS and incubated in 1 mg/ml RNaseA for 1 h, then incubated in primary antibody overnight at 4°C, washed, then incubated in Cy2- and/or Cy5-conjugated secondary antibodies (Jackson ImmunoResearch) for 3 h at room temperature. Embryos were stained with propidium iodide and mounted in clearing solution (Fenger et al., 2000) or in Prolong-Gold with DAPI (Life Technologies). Embryos were visualized with a Nikon Eclipse 80i microscope equipped with a CoolSNAP ES camera (Photometrics). Statistical analysis of imaging quantifications was performed using the Fisher Exact test. Primary antibodies used were anti-α-Tubulin YL1/2 (AbD Serotec) or anti-Tubulin DM1α (Sigma-Aldrich), anti-Centrosomin (gift from William Theurkauf, University of Massachusetts Medical School, Worcester, MA), anti-PH3 (06-570, Millipore), anti-Rod (Scaërou et al., 1999), anti-Zw10 and anti-Zwilch (gifts from Michael Goldberg, Cornell University, Ithaca, NY).

0–1-h embryos were homogenized in non-denaturing lysis buffer (NDLB) (50 mM Tris-HCl pH 7.4, 300 mM NaCl, 5 mM EDTA, 1% Triton X-100). For immunoprecipitations, 500 µg of embryo lysate was incubated with 5 µl anti-Rod serum (Scaërou et al., 1999) or normal rabbit serum overnight at 4°C. Lysates were then incubated with 25 µl of Protein-G–Sepharose beads (GE Healthcare) for 3 h at 4°C. Beads were pelleted, washed three times in NDLB, and boiled in 30 µl 6× sample buffer. The resulting supernatant and 20 µg of input lysate were analyzed by SDS-PAGE and by immunoblotting using standard techniques. Horseradish peroxidase (HRP)-conjugated secondary antibodies and chemiluminescence were used to detect primary antibodies.

For sucrose gradient analysis, 2 mg of embryo lysate was layered on top of a sucrose gradient column of 5%, 10%, 20% and 30% sucrose, then centrifuged in a L8-70M ultracentrifuge (Beckman) with a SW55Ti rotor for 4 h at 250,000 g and 4°C. A total of 19 fractions were collected and 8% of each fraction was analyzed by SDS-PAGE and immunoblotting.

Immunostaining used the method described previously (Williams et al., 2003). Drosophila third-instar larval brains expressing Spc25–RFP were dissected in 0.7% NaCl and treated for 15 min in 10⁻⁴ M colchicine. After formaldehyde fixation and methanol dehydration, brains were immunostained with a mouse monoclonal anti-RFP (Abcam, ab658556; 1∶500) and rabbit anti-Rod (1∶1000) antibodies. Secondary antibodies were Alexa-Fluor-594-conjugated anti-mouse-IgG and Alexa-Fluor-488-conjugated anti-rabbit-IgG (1∶500; Life Technologies). Images were acquired with a Zeiss Axio Imager.Z1 equipped with a Plan-Apo 100× NA 1.4 oil objective and an apotome module. Acquisition times were 400 ms for Alexa Fluor 594 and 300 ms for Alexa Fluor 488 and images were stacks of five planes at 1-µm intervals. Fluorescence was quantified using ImageJ.

For determination of aneuploidy rates and to assay the SAC response, third-instar larval brains were fixed and stained in aceto-orcein following incubation in colchicine (10⁻⁴ M) for 5, 30 or 60 min, and the mitotic density (the average number of mitotic figures per microscope field with at least one mitosis) was determined as described previously (Buffin et al., 2007).

Larval brains were prepared as described previously (Emre et al., 2011). Fluorescence time-lapse videos were acquired with an Olympus IX-70 inverted microscope, a focused xenon lamp and an OrcaER camera (Hammamatsu), piloted by the Cell-R hardware and software system (Olympus). Acquisition times per frame were 100 ms for GFP–Zw10 and 150 ms for RFP–Rod, and images were obtained as stacks of five planes at 1-µm intervals taken every 15 s with a 60× NA 1.4 oil objective.

Neuroblasts from third instar larval brains were treated 15 min with colchicine at 10⁻⁴ M. FRAP experiments were performed using a confocal microscopy system (Zeiss Axiovert with LSM780) equipped with a Plan-Apo 63× NA 1.4 oil objective (during experiments with Mad1–GFP, Spc25–GFP and GFP–Rod) or a Plan-Apo 40× NA 1.3 oil objective (during experiments with GFP–Zw10). Imaging was controlled by Zeiss confocal software (Zen 2012). A circular area of 0.9 µm (GFP–Zw10) or 4 µm (Mad1–GFP, Spc25–GFP) diameter was bleached with the 488 nm and 514 nm lines of an argon laser at 25 mW. Images were acquired with a temporal resolution of 1 s (for Mad1–GFP) or 5 seconds (for Spc25–GFP and GFP–Zw10) with a GaAsP detector (490–650 nm). Fluorescence intensity in a region of interest was quantified using ImageJ. The exponential kinetics of FRAP was analyzed by nonlinear regression fitting using Origin software.

To test the hypothesis that perdurance of a small amount of WT Rod (protein or mRNA), initially loaded into the cytoplasm of the egg laid by a rod^Z3/+ heterozygous mother, might be responsible for the relatively mild phenotype of homozygous rod^Z3 larvae, we assayed for GFP–Rod+ of strictly maternal origin in third-instar larval brains. From the cross of y w; GFP-Rod/y⁺Cy; +/+ females to y w/Y males, phenotypically y⁺ larvae should not inherit the GFP–Rod transgene. Third-instar larval brains from these y⁺ larvae were examined both by immunoblotting, probing with anti-GFP antibody, and by fluorescence microscopy following colchicine treatment, for evidence of GFP signal on kinetochores. Neither approach detected a positive signal of GFP–Rod.

We thank the ImagoSeine facility, a member of France BioImaging infrastructure, supported by the French National Research Agency [grant number ANR-10-INSB-04, ‘Investments of the Future’].