The checkpoint protein Zw10 connects CAL1-dependent CENP-A centromeric loading and mitosis duration in Drosophila cells

A defining feature of centromeres is the presence of the histone H3 variant CENP-A that replaces H3 in a subset of centromeric nucleosomes. In Drosophila cultured cells CENP-A deposition at centromeres takes place during the metaphase stage of the cell cycle and strictly depends on the presence of its specific chaperone CAL1. How CENP-A loading is restricted to mitosis is unknown. We found that overexpression of CAL1 is associated with increased CENP-A levels at centromeres and uncouples CENP-A loading from mitosis. Moreover, CENP-A levels inversely correlate with mitosis duration suggesting crosstalk of CENP-A loading with the regulatory machinery of mitosis. Mitosis length is influenced by the spindle assembly checkpoint (SAC), and we found that CAL1 interacts with the SAC protein and RZZ complex component Zw10 and thus constitutes the anchor for the recruitment of RZZ. Therefore, CAL1 controls CENP-A incorporation at centromeres both quantitatively and temporally, connecting it to the SAC to ensure mitotic fidelity.


Author summary
Segregation of DNA during mitosis is a highly regulated process necessary to ensure the faithful transmission of genetic material to new daughter cells. Centromeric chromatin, which is defined by the presence of the histone H3 variant CENP-A, mediates the interaction of chromosomes with the spindle apparatus. In most organisms, CENP-A incorporation into centromeric chromatin is uncoupled from DNA replication and depends on specialized loading mechanisms and chaperones. In Drosophila cells, CENP-A loading takes place during mitosis and is mediated by its loading factor CAL1. We show that CAL1 controls the amount as well as the timing of CENP-A incorporation into centromeric chromatin. Moreover, CENP-A loading inversely correlates with the duration of mitosis. Mitosis length is influenced by the spindle assembly checkpoint, which ensures that all centromeres are attached correctly to the microtubule spindle. We identified the checkpoint protein and RZZ  Introduction differences in centromere assembly have been reported between Drosophila cultured cells and embryos. Firstly, CENP-A loading has been shown to take place during mitotic exit in early embryos [9] and prometaphase to early G1 in cultured cells [10,12]. Second, CENP-C incorporates concomitantly to CENP-A in embryos [9] while this time window seems to be larger in cultured cells [10]. We, therefore, set out to determine more precisely the function of CAL1 in CENP-A loading regulation in Drosophila cultured cells. During the course of these studies, we found that overexpression of CAL1 not only increases endogenous and exogenous CENP-A abundance at centromeres, it also uncouples CENP-A loading from mitosis. Strikingly, we discovered a co-dependence of mitotic duration and accurate CENP-A loading that may be coordinated by an interaction of the CENP-A loading machinery with the SAC protein and RZZ subunit Zw10. These data suggest an intricate coordination of the spindle assembly checkpoint, CENP-A loading, and mitotic duration in order to safeguard accurate mitotic progression.

CAL1 levels regulate the amount of CENP-A at centromeres
CAL1 is essential for CENP-A loading but we do not fully understand its role in restricting CENP-A loading to defined cell cycle stages. To investigate how CAL1 influences CENP-A loading, we overexpressed CAL1 in S2 cells from a copper-inducible metallothionein (MT) promoter. Induced CAL1 localized to centromeres and the nucleolus in a majority of transfected cells (84%), similar to endogenous CAL1 [13][14][15][16][17][18][19]. Additional cytoplasmic and/or nucleoplasmic CAL1 was visible in about 16% of cells ( Fig 1A). Augmented total CAL1 protein levels (2.6-fold increase; Fig 1B) led to an increased CAL1 level at centromeres (Fig 1C and  1D) and caused an increase of both endogenous and exogenous, tagged CENP-A (GFP and SNAP) levels ( Fig 1E). Interestingly, endogenous CENP-A levels were downregulated in cell lines that expressed an exogenous CENP-A construct, independent of CAL1 levels (Fig 1E), indicating that cells sense the overall level of CENP-A and either degrade excess CENP-A (possibly by already described proteasomal degradation mechanisms) or cannot stabilize CENP-A due to limiting amounts of CAL1 (and CAL1-dependent mono-ubiquitination of CENP-A) [22,29,31]. Importantly, however, CENP-A localized exclusively to centromeres in CAL1-overexpressing cells, and we did not detect CENP-A at any other chromatin regions on metaphase chromosomes under these conditions ( Fig 1F). This result indicates that CAL1 overexpression does not lead to the misincorporation of CENP-A to non-centromeric sites that has been observed in CENP-A overexpression systems [13][14][15][16][17][18][19]. In contrast to previous observations in embryos where increased centromeric CENP-A levels required both CAL1 and CENP-A overexpression [30], we observed in cells that endogenous centromeric CENP-A levels were increased by up to 2 fold after 24 hours of CAL1 overexpression (Fig 1G and 1H). Furthermore, CENP-C levels remained largely unaffected by CAL1 overexpression (Fig 1G  and 1H). These results show that the abundance of CENP-A at centromeres depends on the available amount of CAL1.
To test whether the increased CENP-A level is caused by increased CENP-A loading we used the SNAP-tag technology in a quench-chase-pulse experiment to label newly synthesized and incorporated CENP-A [5]. At day 1 of CAL1 induction, we quenched existing SNAP--CENP-A molecules with a non-fluorescent ligand (SNAP-Block), washed out unbound Block, cultured the cells for 24 hours to allow approximately one additional cell division, then marked newly synthesized SNAP-CENP-A molecules with a fluorophore, and measured the fluorescent intensity of newly incorporated SNAP-CENP-A at centromeres (Fig 1I). Similar to what we observed for GFP-CENP-A, significantly more SNAP-CENP-A incorporated into  (Fig 1I and 1J). These data confirm that CENP-A centromeric levels are regulated by CAL1 and that CAL1 overexpression specifically increases the incorporation of newly synthesized CENP-A at centromeric chromatin.

Overexpression of CAL1 leads to centromeric CENP-A loading outside of mitosis
CENP-A loading takes place during prometaphase-metaphase in Drosophila S2 cells [10]. To determine if CAL1 overexpression changes the CENP-A loading pattern during mitosis, we measured Fluorescence Recovery After Photobleaching (FRAP) of GFP-CENP-A, on the assumption that fast recovery of the signal at centromeres corresponds to active loading of CENP-A [12]. We partially bleached the GFP-CENP-A signal during prophase and followed the recovery of the signal at centromeres until anaphase. We observed a partial recovery of GFP-CENP-A at centromeres in control cells, which could be due to either turnover of unbleached GFP-CENP-A or to the loading of new molecules. This is in agreement with previous reports that the loading of centromeric CENP-A takes place during mitosis [10] (Fig 2A  and 2B). Surprisingly, the GFP-CENP-A recovery rate decreased in CAL1-overexpressing cells compared to control cells (Fig 2B), suggesting that the maximal level of CENP-A at centromeres has been reached at the time of the bleaching event. One possible explanation could be that CENP-A centromeric levels get replenished at a different cell cycle stage in CAL1-overexpressing cells. Indeed, besides loading in mitosis, CENP-A loading has also been reported in G1 in S2 cells [12]. We, therefore, followed cells through mitosis, partially bleached GFP-CENP-A in early G1 and measured GFP-CENP-A signal intensity for 3 hours. However, we did not detect any recovery of the bleached GFP-CENP-A signal (S1A Fig). We explained these results by the likely absence of any measurable GFP-CENP-A loading in early G1. This also indicates that centromeric CENP-A has a very low turnover, at least during G1.
Since the availability of CAL1 seems to be the crucial determinant for CENP-A incorporation, we tested if CAL1-overexpressing cells incorporate CENP-A into centromeric chromatin outside mitosis. Thus, we arrested cells at the metaphase to anaphase transition using the proteasome inhibitor MG132 for 2 hours, blocked existing SNAP-CENP-A molecules using SNAP-Block, and cultured the cells for an additional 4 hours in the presence of MG132 (S1B Fig) similar to a previously published protocol [10]. The MG132 treatment prevents cells from exiting mitosis, which means that every SNAP-CENP-A positive interphase cell must have acquired SNAP-CENP-A outside of mitosis. While 10% of uninduced pMT-CAL1 cells loaded SNAP-CENP-A, the induction of pMT-CAL1 and resulting overexpression of CAL1 led to incorporation of newly synthesized SNAP-CENP-A in about 40% of the cells, pointing to loading outside of mitosis (S1B Fig). To confirm that overexpressed CAL1, indeed, loads CENP-A to centromeres outside of mitosis, we followed newly synthesized SNAP-CENP-A live by time-lapse microscopy. We blocked existing SNAP-CENP-A with SNAP-Block, washed and added SNAP-640 fluorophores [32] to the culture medium to fluorescently mark newly SNAP N-terminally tagged, under the Copia promoter) expressing cells with or without concomitant pMT-CAL1-V5 induction. Anti-V5 (CAL1-V5), anti-CENP-A (endogenous CENP-A and tagged-CENP-A), and anti-H3 antibodies were used. Fold change of CENP-A levels compared to S2 cell are shown below (N = 4). F. Metaphase chromosomes of pMT-CAL1-V5-overexpressing cells as in A stained with anti-CENP-A antibody. DNA (DAPI) is shown in grey. Scale bar: 1 μm. G. Immunofluorescence of pMT-CAL1-V5/GFP-CENP-A expressing cells as in A stained with anti-CENP-C antibody (red). DNA (DAPI) is shown in grey. H. Quantification of CENP-A and CENP-C signal intensities per centromere as shown in G. I. Timeline of the Quench-Chase-Pulse SNAP-tag experiment after 24 h pMT-CAL1-V5 induction: cells were incubated with SNAP-Block to quench existing SNAP-CENP-A molecules (Q), washed and cultured for 24 h (chase), newly synthesized SNAP-CENP-A molecules were stained with SNAP-SiR647 (P). Representative images of SNAP-CENP-A (red) in non-induced and induced pMT-CAL1-V5 cells. DNA (DAPI) is shown in grey. J. Quantification of I. SNAP-CENP-A signal intensity per centromere shown as % of non-induced control. Scale bar: 2 μm. All graphs show Mean +/-SEM of 3 experiments (n>300 cells), Student's t-test (n.s.: non-significant; � : p<0.05; �� : p<0.01, ��� : p<0.001). https://doi.org/10.1371/journal.pgen.1008380.g001 Zw10-CAL1 interaction connects CENP-A loading and mitosis duration synthesized SNAP-CENP-A molecules. Despite some aggregation of the dye, SNAP-CENP-A loading at centromeres was discernible in both CAL1-overexpressing and control cells ( Fig  2C). In this assay, we only analyzed cells that were actively cycling and eventually entered mitosis during the course of imaging, and determined whether and when they incorporated SNAP-CENP-A. As a reference point, we used the first time point where the intercellular microtubule bridge was visible (cytokinesis). We found that 38% of uninduced control cells had SNAP-CENP-A signals at centromeres by the end of the imaging (Fig 2C-2E; S1 Video). Of those, 91% loaded SNAP-CENP-A during mitosis and 9% just after the cell exited mitosis. In contrast, 81% of cycling CAL1-overexpressing cells had SNAP-CENP-A signals at centromeres (Fig 2C-2E; S2 Video), and of that 31% loaded before mitosis and 51% after the cells exited mitosis. The specificity of the signal was verified by co-staining with anti-CENP-A antibody immediately after block (0 min), or after 20 hours of incubation with SNAP-640 (S1C Fig). No SNAP-CENP-A was detected directly after SNAP-block treatment, confirming the efficient quenching of old SNAP-CENP-A molecules (S1C Fig). After 20 hours newly synthesized SNAP-CENP-A perfectly co-localized with total CENP-A signals at centromeres (S1C Fig). Interestingly, we did not observe any differences in the percentage of cells with SNAP-CENP-A signal at centromeres between control and CAL1-overexpressing cells after 20 h (S1D Fig). However, the intensity of SNAP-CENP-A at centromeres was four times higher in CAL1-overexpressing cells than in control cells (S1E Fig). These results are not consistent with the live analysis of the same cell line (Fig 2C and 2D). This discrepancy is probably due to the masking of low levels of SNAP-CENP-A in the presence of excess SNAP dye in live cell analysis ( Fig 2D). Nevertheless, we can conclude that cells overexpressing CAL1 load CENP-A throughout the cell cycle, while loading outside mitosis does not occur in control cells (Fig 2D  and 2E) suggesting that CAL1 misexpression uncouples CENP-A centromeric loading from mitosis and leads to an overload of CENP-A at centromeres.
We next analyzed CAL1-overexpressing cells live to determine the consequences of increased CENP-A centromeric levels on mitosis progression. We performed time-lapse microscopy experiments in cells overexpressing CAL1 together with H2B-GFP and mCherry-Tubulin to mark the chromatin and mitotic spindle, respectively. Surprisingly, we found that mitosis duration, defined as the time between partial nuclear envelope breakdown (NEB) and anaphase entry, was 10% shorter in CAL1-overexpressing cells as compared to control cells (Fig 2F and 2G; S3 and S4 Videos). We also measured mitosis duration in cells containing GFP-CENP-A and inducible CAL1, which showed similar results (S2A and S2B Fig). Of note, CAL1 overexpression caused a slight but significant increase of lagging chromosomes (S2C Fig). We also scored cells where mitosis was generally defective (multipolar spindles, micronuclei formation) but did not measure an increase in CAL1-overexpressing cells (S2D Fig), suggesting that the observed lagging chromosomes resolved themselves before mitotic exit. Kinetochore proteins (CENP-C, Spc105R, GFP-Zw10) were equally recruited during prometaphase when we compared CAL1-overexpressing cells with non-induced cells (S2E- S2G Fig).
To determine if kinetochores attachment to the spindle microtubules was affected in CAL1-overexpressing cells, we performed an MG132-Taxol assay as described in Maia et al., [33]. This assay allowed us to score the attachment status of each kinetochore, which we found equal in control and CAL1-overexpressing cells (S2H Fig).
Last but not least, we ought to determine if the SAC activity was affected by increased centromeric CENP-A levels. Similar amounts of BubR1 and GFP-Mad2 were recruited in CAL1-overexpressing prometaphase cells when compared to non-induced cells (S3A-S3B Fig an arrested state before reverting to an interphase state without performing cytokinesis indicating that SAC activity is weakened in the presence of excess CENP-A at centromeres (S3E Fig). We concluded that CAL1-overexpressing cells not only incorporated more CENP-A at centromeres, they also progressed faster through mitosis with comparable kinetochore formation and SAC activity. We, therefore, wondered if CENP-A levels regulate mitosis duration.

CENP-A levels correlate with mitosis duration
First, we investigated if elevated CENP-A levels alone (without CAL1 overexpression) would influence mitosis duration. However, upregulating CENP-A to high levels is associated with ectopic CENP-A incorporation into chromosome arms and subsequent defects in mitosis [23]. To avoid these detrimental effects, we induced CENP-A overexpression from a pMT-CENP-A-GFP plasmid with a short and very low CuSO 4 pulse (2 hours; 10 μM CuSO 4 ). Under those conditions, CENP-A-GFP was induced while endogenous CENP-A levels were decreased leading to an overall relative increased CENP-A level (Fig 3A). 22 hours after the CuSO 4 pulse, untagged CENP-A protein was almost undetectable by Western blot, suggesting that cells control the total abundance of CENP-A protein levels to some degree (Fig 3A), probably through proteasomal degradation as reported before [22,23,31]. This compensation of CENP-A protein amount was detectable also in cells constitutively expressing GFP-CENP-A or SNAP-CENP-A from the copia promoter ( Fig 3A). Under these mild induction conditions, non-centromeric CENP-A or CENP-A-GFP were undetectable ( Fig 3B) and an increased amount of CENP-A-GFP and CAL1 localized to centromeres (Fig 3C and 3D and S4A Fig). Using timelapse analysis, we found that cells with increased centromeric CENP-A levels progressed significantly faster through mitosis ( To further test the relationship between centromeric CENP-A abundance and the duration of mitosis we designed a strategy to reduce CENP-A levels without inducing chromosome alignment defects that might arrest cells in mitosis [34]. We performed CENP-A RNAi depletion in GFP-CENP-A-expressing cells, which led to undetectable levels of endogenous CENP-A while small amounts of GFP-CENP-A remained (Fig 4A and 4B). Mitosis duration was significantly extended in partially CENP-A-depleted cells when compared to control cells (Fig 4C and 4D; S7 and S8 Videos), probably due to defects or delays in kinetochore assembly and spindle attachment. Similar observations have been reported in heterozygous CENP-A mutant fly embryos [35], supporting our hypothesis that CENP-A levels control mitotic length. Partial co-depletion of the SAC protein Mad2 and CENP-A abrogates the mitotic delay observed in CENP-A-depleted cells indicating that the SAC is active in these cells (Fig 4D and  4E, S5A Fig). Taken together, these results further suggest that centromeric CENP-A levels influence mitosis duration in a SAC-dependent manner.

CENP-A loading correlates with mitosis duration
CENP-A loading peaks during prometaphase-metaphase in Drosophila S2 cells. Therefore, we hypothesized that the amount of newly synthesized CENP-A that can be incorporated into centromeric chromatin depends on mitosis duration. Mitotic progression is controlled by the action of SAC proteins, which delay entry into anaphase until all chromosomes are properly attached and aligned at the metaphase plate. To shorten mitosis, we depleted the two mitotic checkpoint proteins, BubR1 and Mad2, known to reduce mitosis duration [36][37][38]. These depletions (S5B and S5C Fig) led to a decrease in newly synthesized centromeric SNAP-CENP-A incorporation (Fig 5A and 5B). In contrast, the depletion of Mis12 (S5C Fig), which has been previously shown not to affect the duration of mitosis in Drosophila cultured cells [39], did not alter SNAP-CENP-A amount at centromeres (Fig 5A and 5B). Extending metaphase duration by preventing entry into anaphase through depletion of Spindly (required for silencing of the SAC) [40] or of the APC/C cofactor Cdc27 [41](S5C Fig) did not cause an increase of CENP-A levels at centromeres when compared to control (Fig 5A and 5B). Knockdown efficiencies were determined by immunoblot and/or qPCR (S5B and S5C Fig). We further irreversibly arrested cells in prometaphase by the microtubule-depolymerizing drug Nocodazole or in metaphase by the proteasome inhibitor MG132. These conditions also did not change the amount of newly synthesized SNAP-CENP-A at centromeres (S5D and S5E  Fig). SNAP-CAL1 centromeric levels were similarly affected: an acceleration of mitotic timing by BubR1 or Mad2 RNAi led to a decrease of loading, whereas mitotic arrest by Cdc27 or Zw10-CAL1 interaction connects CENP-A loading and mitosis duration Spindly RNAi did not affect it (S5F Fig). This set of experiments suggests that the amount of new CENP-A molecules that are incorporated into centromeric chromatin is limited at each round of mitosis probably due to CAL1's availability and that the length of mitosis is directly affected by the amount of centromerically loaded CENP-A. This anti-correlation can be seen best when we plotted the recovery rate of GFP-CENP-A measured by FRAP against the mitosis duration for each control cell that we measured ( Fig 5C). This revealed that cells that incorporated GFP-CENP-A quickly progressed through mitosis faster than cells with a low recovery rate. To further illustrate this anti-correlation between CENP-A levels and mitosis duration, we plotted the mitosis duration versus CENP-A total protein levels determined by immunoblotting for all the cell lines used in this study ( Fig 5D). This confirmed a clear anti-correlation of CENP-A levels and mitotic duration and further suggests that mitosis duration correlates with CENP-A centromeric levels rather than with the rate of CENP-A incorporation.

RZZ component Zw10 interacts with centromeric proteins
In order to identify a putative connection between CENP-A loading and mitosis progression, we performed a targeted Yeast-two-hybrid (YTH) screen using either CAL1 or CENP-A as bait and known SAC proteins as prey. These experiments showed that the RZZ component Zw10-CAL1 interaction connects CENP-A loading and mitosis duration Zw10 binds to full-length CAL1 and CENP-A (Fig 6A). To identify the domains of each protein required for the interaction, we repeated the YTH analysis with fragments of CAL1 and Zw10. This analysis revealed a strong interaction of Zw10 with the C-terminus of CAL1 (amino acids 680-979) (S6A Fig). We further identified Zw10 C-terminus (amino acids 482-721) as being sufficient to bind to the CAL1 C-terminus (S6A Fig), the same domain of CAL1 that also binds CENP-C and Roadkill (S6B Fig) [29,30]. No further interactions were detected between the other checkpoint proteins and the centromeric proteins suggesting that Zw10 Zw10-CAL1 interaction connects CENP-A loading and mitosis duration might be the key component that connects the centromere loading machinery to checkpoint proteins and ultimately mitosis duration. Our analysis regarding Bub1 has been inconclusive due to the self-activation of the Bub1 construct in this assay (blue color in the control (-) transformation). The absence of an interaction between CAL1 and Bub1 has, however, already been reported previously [30]. To validate the interaction between Zw10 and centromeric proteins, we expressed and purified GST-Zw10, His-CAL1, and His-CENP-A in bacteria. Pulldown experiments using those proteins confirmed that Zw10 directly binds CAL1 (Fig 6B). We did not detect any interaction between GST-Zw10 and His-CENP-A in this assay (S6C Fig). To further confirm the interaction between CAL1 and Zw10 in cells, we established a cell line expressing GFP-tagged Zw10. Firstly we confirmed that GFP-Zw10 localization in cultured cells (S6D Fig) is in accordance with endogenous protein [42][43][44][45] and transgene expression [46] reported in Drosophila tissues. Following the disassembly of the nuclear envelope at the onset of mitosis, Zw10 is recruited to the kinetochore where it remains until complete chromosome biorientation is reached. Zw10 is then transported along the spindle microtubules in a process mediated by Spindly and dynein, which is required to silence the SAC [40,[47][48][49]. We then performed immunoprecipitation experiments and co-immunoprecipitated CAL1 with GFP-Zw10 (Fig 6C), confirming our pull-down and YTH experiments. CENP-A was not consistently co-immunoprecipitated with Zw10, which we attributed to the fact that CENP-A interacts with CAL1 but probably not directly with Zw10. We concluded from this set of experiments that CAL1 robustly interacts with Zw10, and that this interaction may be involved in coordinating CENP-A centromeric loading with mitosis progression.

Zw10 affects the loading of newly synthesized CENP-A during mitosis
The RZZ complex recruits the MAD proteins to the kinetochore and is thus essential for the SAC [46,50]. Therefore, we sought to determine if Zw10 depletion could recapitulate the effects observed after the depletion of the other SAC proteins (Fig 5A and 5B). Time-lapse microscopy analysis of cells stably expressing H2B-GFP and mCherry-Tubulin confirmed that mitotic timing was significantly shorter in Zw10-depleted cells than in control cells (Fig 6D  and 6E; S9 and S10 Videos), which fits to the Premature Sister Chromatids Separation (PSCS) phenotype described in Zw10 mutant embryos and larval tissues [43,51,52]. And indeed, Zw10 depletion led to a 30% reduction of newly synthesized SNAP-CENP-A (Fig 6F and 6G) and SNAP-CAL1 (S7A and S7B Fig) at centromeres compared to control-depleted cells.
To exclude a potential role of Zw10 in CENP-A stability we performed a SNAP-tag pulsechase experiment in Zw10-depleted cells (S7C and S7D Fig) in which we labeled centromeric SNAP-CENP-A at t0 and measured its remaining levels after 24 hours. This showed that centromeric SNAP-CENP-A was diluted 2-fold after 24 hours as expected after 1 cell cycle both in control and Zw10-depleted cells (S7C and S7D Fig). Furthermore, FRAP analysis revealed that the recovery rate of GFP-CENP-A at centromeres was comparable in control and Zw10-depleted cells (S7E and S7F Fig). Thus, the CENP-A mitotic loading machinery is functional in Zw10-depleted cells suggesting that Zw10 does not influence CENP-A loading directly but possibly through its role in checkpoint activity. in D. Mean +/-SEM, n>200 cells. Student's t-test ( ��� : p<0.001). F. SNAP-CENP-A incorporation in Zw10-depleted cells. After 72 h dsRNA treatment, a Quench-Chase-Pulse experiment (scheme in Fig 1H) was performed to stain newly synthesized SNAP-CENP-A molecules (red). DNA (DAPI) is shown in grey. Scale bar: 2 μm. G. Quantification of F showing the total SNAP-CENP-A centromeric intensity per nucleus as % of control. Mean +/-SEM of 3 experiments (n>300 cells), Student's t-test ( ��� : p<0.001). https://doi.org/10.1371/journal.pgen.1008380.g006 Zw10-CAL1 interaction connects CENP-A loading and mitosis duration

Zw10 mediates the recruitment of the RZZ to the kinetochore independently of Spc105R
How RZZ is recruited to the kinetochore at mitosis onset remains elusive in Drosophila. We first sought to determine if one of the RZZ components is responsible for targeting the whole complex to the kinetochore. We established stable cell lines expressing GFP-Zwilch and GFP-ROD and confirmed that the localization of the tagged proteins is consistent with previous reports [46,51,53,54] (S6E and S6F Fig). Knockdown of RZZ complex components (S8A Fig) led to the expected spindle morphology defects in all cell lines (Fig 7) [43, [54][55][56]. Importantly, depletion of any complex component led to a reduction of the other GFP-RZZ proteins at kinetochores (Fig 7) suggesting that the RZZ components are mutually dependent on each other similar to what has been reported in Drosophila tissues [42,44,54,56,57]. However, levels of GFP-Zw10 at kinetochores were the least affected and the reduction was statistically non-significant indicating that a substantial amount of GFP-Zw10 is still recruited to kinetochores in the absence of its complex partners ROD or Zwilch (Fig 7C). This suggests that Zw10 is the most proximal component of the complex at kinetochores in Drosophila cells. Like all kinetochore proteins tested so far [58], the localization of Zw10 to the kinetochore is abolished when CENP-A (Fig 8B and 8C) or its loading factor CAL1 (Fig 8C and S8B Fig) are depleted, even though total Zw10 protein levels remained largely unchanged (Fig 8A).
In mammalian cells, several pathways have been implicated in the recruitment of the RZZ complex including KNL1, Bub1, and Zwint [59][60][61][62][63]. As no ortholog of Zwint has been identified in Drosophila so far, we tested if the Drosophila KNL1 ortholog Spc105R or Bub1 are required for RZZ recruitment to the kinetochore. Neither depletion of Spc105R nor Bub1 (S8A Fig) affected Zw10 localization (Fig 8C and 8D and S8C Fig). We also tested Mis12, which has been shown to localize to kinetochores throughout the cell cycle [58] and found that it is also not involved in Zw10 recruitment to the kinetochore (Fig 8C and S8A and S8D Fig). We also excluded the KMN network since the recruitment of the KMN network depends on the prior localization of Spc105R and Mis12 at kinetochores [58]. Conversely, depletion of Zw10 (Fig 8A) did not prevent Spc105R (Fig 8E and 8F), Mis12 (S8E and S8F Fig), Ndc80 (Fig  8G and 8H) or BubR1 (S8G and S8H Fig) localization to the kinetochore. Therefore, we propose that the Drosophila checkpoint proteins and outer kinetochore components are assembled through two independent branches (S8I Fig): on the one hand, CAL1 directly recruits the RZZ complex, which is necessary to localize the MAD proteins to the kinetochore [46,50], and on the other hand, CENP-C attracts Mis12-Spc105R, which are required to recruit all other components of the KMN network as well as the Bub proteins [28,58,[64][65][66]. We conclude from these results that there is a direct connection of the SAC and the loading of CENP-A via CAL1 and we speculate that the CAL1-RZZ branch might regulate CENP-A loading and mitotic duration, thereby controlling CENP-A loading in a cell cycle time-dependent manner.

Discussion
In Drosophila cells, CENP-A loading takes place primarily during prometaphase-metaphase [10]. Additional turnover of CENP-A in G1 has been reported leading to the hypothesis that CENP-A could be further incorporated at this stage [12], which we did not observe in our FRAP experiments when we bleached centromeric CENP-A at the end of cytokinesis. However, our FRAP experiments and most importantly our live staining of newly synthesized SNAP-CENP-A confirmed the observations reported by Mellone et al., that the majority of CENP-A loading takes place during mitosis in Drosophila cultured cells [10].  In flies, CENP-A incorporation is controlled by its chaperone CAL1 [17,19,30]. It has been shown previously that co-overexpression of exogenous CENP-A and CAL1 leads to an increase of centromeric CENP-A in embryos [30]. We now show that overexpression of CAL1 alone leads to increased endogenous CENP-A protein levels in Drosophila cultured cells. Ectopic incorporation of CENP-A, however, was never observed suggesting that CAL1 loads CENP-A exclusively to centromeres and that ectopic CENP-A incorporation in flies depends on alternative loading mechanisms similar to what has been suggested in human cells [67]. Importantly, increased centromeric CENP-A levels following CAL1 overexpression correlated with faster mitosis. A similar acceleration of mitotic timing was observed when CENP-A was only mildly overexpressed, revealing a possible link between CENP-A loading and mitotic timing. Indeed, shortening mitosis duration by depleting Mad2 or BubR1 [68] was associated with decreased CENP-A loading. However, just elongating the mitotic time window during which CENP-A can get loaded (Spindly or Cdc27 depletion, or by drug treatment) did not increase the amount of CENP-A incorporated at centromeres, showing that the length of mitosis alone is insufficient to control CENP-A amounts at centromeres. Rather, these experiments showed that only a defined amount of CENP-A can be incorporated at each mitosis probably correlating with CAL1's availability. Indeed, live analysis of CAL1-overexpressing cells allowed us to visualize newly synthesized CENP-A incorporation to centromeres in all stages of the cell cycle. This strongly suggests that CAL1 controls CENP-A incorporation into centromeric chromatin both quantitatively and temporally. How exactly CENP-A levels at centromeres are sensed is unclear but we identified the RZZ-component Zw10 as a new CAL1 interacting partner, which directly connects CENP-A loading to the SAC. It has been proposed that SAC activation is a 2-steps process [35]: at the end of G2-beginning of mitosis, before the kinetochores are assembled, cytosolic Mad1-Mad2 dimers initiate MCC formation, which inhibit APC/ C Cdc20 and determine the timing of mitosis [37,38,[68][69][70][71]. After nuclear envelope disassembly, kinetochore-dependent MCC are generated and regulated by kinetochore-microtubules attachment. Therefore, we suggest the following model (Fig 9): efficient CENP-A loading by CAL1 during mitosis recruits Zw10 up to a threshold, which is sensed by the SAC. Low CENP-A levels at centromeres could lead to more cytosolic Mad2 thereby keeping the timer active longer [35]. Higher CENP-A levels at centromeres during early mitosis would accelerate the recruitment of RZZ and consequently Mad2 to the kinetochores or capture microtubules more efficiently, therefore, releasing the timer and shortening mitosis duration in cells where kinetochores attach properly to the spindle microtubules. Interestingly, Nocodazole treatment did not affect CENP-A loading confirming previous observations that kinetochores attachment to the microtubule spindle does not play a role in CENP-A loading [9,10]. These results are pointing further to an additional function of the SAC independent of the control of microtubule attachment. Interestingly, recent evidence shows that RZZ together with Spindly plays a central role in kinetochore expansion during early mitosis to form a fibrous corona that then compacts upon microtubule capture [72][73][74]. Whether and -if so-how the kinetochore expansion by RZZ and spindly is involved in CENP-A loading needs to be investigated in the future.

Zw10-CAL1 interaction connects CENP-A loading and mitosis duration
Many essential components of the SAC require outer kinetochore components for their localization to centromeric regions. However, several outer components are missing from the Drosophila kinetochore [58,75] and even though Mad1/2 recruitment to the kinetochore depends on the RZZ complex [50], the factors necessary for the localization of the RZZ to kinetochores are unknown. We showed here that RZZ localization to the kinetochores does not require KNL1 Spc105R but depends on the centromeric proteins CAL1 and CENP-A. Therefore, we propose that the Drosophila outer kinetochore and components of the SAC assemble through two independent pathways (S8I Fig): the CENP-C-KMN-Bub1-Bub3/BubR1 branch [28,64,66] or the CAL1-RZZ-Mad1/2 branch. How those two pathways communicate for the formation of MCC complexes remains to be determined. One link may be the KMN complex Model. The model depicts the relationship between the amount of CENP-A at centromeric chromatin at the onset of mitosis and the duration of mitosis. Cells that have a relative large pool of CENP-A at centromeres go through mitosis faster. The precise reason for this decreased mitotic timing is still unknown but we hypothesize that cells with sufficient CENP-A recruit SAC components more quickly or capture microtubules more efficiently to satisfy the SAC faster. This hypothesis is supported by our finding that the CENP-A loading factor CAL1 directly interact with the SAC component and RZZ subunit Zw10. More CAL1 may be able to not only recruit more CENP-A by priming centromeres and loading more efficiently CENP-A but also may recruit the RZZ complex and the SAC more efficiently. Vice versa, an active SAC may give the cell more time to recruit sufficient CENP-A via CAL1 to ensure that the cell does not progress further without having loaded sufficient CENP-A to the centromere. https://doi.org/10.1371/journal.pgen.1008380.g009 Zw10-CAL1 interaction connects CENP-A loading and mitosis duration since Mad2 is diminished in the absence of KMN proteins [39]. Interestingly, Spc105R mutation does not affect SAC function in fly embryos suggesting that flies rely more on the RZZ-Mad1/2 branch to engage the SAC [64].
CENP-A expression [76] and its stability [22,29] together with its dependence on the low abundant and highly specific loading factor CAL1 and the here shown connection to mitotic events are likely interconnected cellular surveillance mechanisms to avoid misincorporation of CENP-A and, therefore, securing genome stability. How CAL1 itself is regulated to obtain such specificity is currently unknown. In conclusion, we show that there is direct crosstalk between the SAC and the maintenance of centromeric chromatin, ensuring mitotic fidelity not only by controlling microtubule attachment but also by regulating the accurate composition of centromeres.
Stable cell lines were generated by co-transfecting the plasmid of interest together with a plasmid carrying the resistance gene for hygromycin with Cellfectin II (Life Technologies), and the selection process was started 2 d after transfection by supplementing the media with 250 μg/ml Hygromycin B.
SNAP quench chase pulse experiments. Cells expressing either SNAP-tagged CENP-A or CAL1 proteins were treated with dsRNA as described in the RNAi section or incubated with CuSO 4 to induce pMT-CAL1-V5 expression. Cells were then collected and incubated in fresh SM containing 10 μM SNAP-Cell Block (New England Biolabs S9106), for 30 min at 25˚C, 400 rpm. Cells were then pelleted, washed 3 times (1 st wash for 30 min, at 25˚C, 400 rpm) and plated in conditional medium. After 24 h, cells were collected and incubated with 4 μM SNAP-Cell TMR-Star (New England Biolabs S9105) or SNAP-SiR647 (New England Biolabs S9102) for 15 min at 25˚C, 400 rpm, protected from light. The cells were then pelleted, washed 3 times with SM (1 st wash for 30 min, at 25˚C, 400 rpm), one time in PBS and prepared for imaging. Cells were settled on a glass coverslip for 10 min, fixed with 4% PFA for 10 min, washed three times in PBS, and counterstained with DAPI before mounting.
SNAP pulse-chase. Cells expressing SNAP-tagged CENP-A proteins were treated with dsRNA as described in the RNAi section. After 3 days, cells were collected and incubated in fresh medium containing 4 μM SNAP-Cell TMR-Star (New England Biolabs, Inc.) for 15 min at 25˚C, 400 rpm, protected from light. Cells were then washed 3 times with SM (1 st wash for 30 min, at 25˚C, 400 rpm) and plated in conditional medium. After 24 h, cells were collected and prepared for imaging as previously described.
SNAP live imaging. To follow the incorporation of newly synthesized SNAP-CENP-A, SNAP-CENP-A was blocked as previously described (see section SNAP quench chase pulse). Cells were then resuspended in conditional medium containing 0.5 μM SNAP-640 dye [32] and placed into an Ibidi imaging chamber before imaging.

Immunofluorescence (IF)
IF was essentially performed as described previously [19]. Cells were harvested, washed once with PBS and placed to settle on a glass coverslip for 10 min before fixation with 4% PFA for 10 min. Cells were washed three times in PBS and permeabilized for 5 min with PBS containing 0.5% Triton X-100. Unspecific binding was prevented by blocking the cells for 30 min with 1% BSA in PBS. Primary antibodies diluted in blocking solution were incubated for 30 min at RT. After three washes with PBS, cells were incubated with the corresponding, fluorescently labeled secondary antibodies (diluted in blocking solution) for 30 min at RT. The secondary antibody incubation and all the following steps were performed while protected from light. After three washes in PBS, DNA was stained for 5 min with DAPI (1 μg/ml in PBS). The cells were then washed two more times with PBS before mounting in Aqua/Polymount medium (Polysciences, Inc.) on a glass slide.

Kinetochore-microtubule interaction assay (MG132-Taxol assay)
After induction (24 h for CAL1-V5-overexpressing cells, 2 h for CENP-A-GFP-overexpressing cells), cells were washed and incubated with 20 μM MG132 for 1 h. Cells were then washed before incubation with 100 nM Taxol in conditional medium for 3 h. Cells were then fixed and stained with DAPI. Cells were scored into 2 categories: either all kinetochores are attached to microtubules (attached) or at least 1 kinetochore lacks microtubule attachment (>1 unattached).

Microscopy and image analysis
All images were acquired on a DeltaVision Core system (Applied Precision) equipped with an Olympus UPlanSApo 100X oil immersion objective (n.a. 1.4) at binning 1x1 (mitotic figures) or 2x2 (interphase). Images were taken as z stacks of 0.3 μm increments to the exception of the kinetochore-microtubule assay where 0.2 μm z stacks were used. All images were deconvolved (20 cycles additive with noise filtering) and maximum-projected using the Applied Precisions soft-WoRx 3.7.1 suite. Image quantification was done using ImageJ software. For determining the number and intensities of centromere/kinetochore dots, a set of plugins developed by the Nikon Imaging Centre, University Heidelberg was used. After marking the nuclear boundaries in the DAPI channel and saving them with the ROI tool, the spots in the channel of interest were enhanced using the DoG spot enhancer plugin. The threshold was adjusted using ''Li" instead of default settings. The number of spots with their mean intensities relative to the nuclear area and the total particle intensity per nucleus were determined using the ROI particle analyzer plugin.
Time-lapse imaging was performed on a DeltaVision Core system (Applied Precision) equipped either with an Olympus 60X/1.42, N Plan Apo oil immersion objective or an Olympus 40X/1.35, UApo/340 oil objective, at binning 2x2 with 3 min (mitosis duration and fidelity measurements) or 15 min (SNAP-CENP-A live imaging) time-lapse series, for 16 h, with z stacks of 1 μm increment.
FRAP was performed on a Zeiss LSM 780 confocal microscope equipped with a 63x/1.4 oil immersion objective. The number of focal planes per time point, as well as time intervals and total duration, were adjusted according to experimental needs. The spacing of focal planes was 1 μm in all experiments. Recovery of the EGFP-CENP-A signal was analyzed using FRAP analysis plugin developed by ZMBH imaging facility. After marking the boundaries of the cell of interest, the threshold allowing visualization of centromeric signal exclusively was defined on a pre-bleach image. The total intensity of the thresholded signal was then measured for each time point and area. Similar measurements were performed on non-bleached cells from the same area to determine the photobleaching level in each time-series. After correction, EGFP-CENP-A signal intensity was plotted against the time, a linear regression curve was fitted.

Statistics
Microscopy data were quantified first separately and then averaged to facilitate the comparison of different treatments and knockdown. For each experiment, the ratio of the sample mean to the control mean was calculated and converted to percentage with the control condition set to 100%. The ratios derived from at least 3 independent experiments were then averaged and are presented on graph bars. When dot plots are presented, at least 3 independent experiments were analyzed and normalized using the highest value obtained as 100. This allowed the pooling of different experiments and the representation of each analyzed cell as dot.
Pulldown assays were performed by incubating GST or GST-Zw10 coupled beads with 100 ng His-Sumo-CAL1 or His-Sumo-CENP-A in interaction buffer (20 mM Tris pH 8, 150 mM NaCl, 0.5 mM EDTA, 10% glycerol, 0.1% NP-40, 1 mM DTT) for 1 h at 4˚C. Beads were then washed 6 times in interaction buffer before resuspension in 1 bead volume 2X sample Laemmli buffer and boiling at 95˚C for 5 min.

WB analysis and quantification
Cell lysates were separated on a 4-15% SDS poly-acrylamide gel, transferred onto a nitrocellulose membrane for 2 h at 100 V (Tris-Glycin buffer containing 0.1% SDS). After blocking in 5% milk in PBST, primary antibodies were incubated O/N at 4˚C in the blocking solution. After washing, secondary antibodies coupled to horseradish peroxidase were added for 1 h at RT before ECL detection (Thermo Fisher Scientific). Immunoblots were quantified using ImageJ.

Reverse transcription and quantitative qPCR
RNA was isolated using Trizol(R) according to the manufacturer's procedures. Whole transcriptome cDNA synthesis was performed using the Quantitect Reverse Transcription kit (Invitrogen), with a combination of oligo (dT) and random hexamer primers in equal proportions. A control reaction with no reverse transcription was always performed in parallel.
qPCR was performed after cDNA synthesis on a LightCycler 480 (Roche) using LightCycler 480 SYBR Green I Master (Roche). All reactions were run in triplicate in a LightCycler 480 multiwell plate. Actin was used as a reference. The level of each targeted gene in the control mock-treated sample was normalized to 1, and compared with the corresponding-depleted samples.

Yeast two-hybrid
Protein-protein interactions were tested using pMM5 and pMM6 fusion constructs and the yeast strain SGY37VIII. The YTH was performed as described previously [80]. In short, interactions were judged based on the activity of β-galactosidase that results in the conversion of X-Gal (5-bromo-4-chloro-3-indolyl ß-D-galactopyranoside) into a blue dye. pMT-CAL1-V5 expression was induced for 24 h in H2B-GFP/mCherry-Tubulin cells. Cells were imaged for 16 h and scored for the accuracy of mitosis: lagging (presence of lagging chromosomes during anaphase that will resolve before cytokinesis)(C) or defective (formation of tripolar spindles, multinucleated cells)(D). Mean +/-SEM n > 200 cells. Student's t-test (n.s.: non-significant). E-G. Amount of kinetochore proteins recruited during mitosis in the presence or absence of CAL1 overexpression. pMT-CAL1-V5 expression was induced for 24 h in GFP-CENP-A/mCherry-Tubulin cells (E-F) or GFP-Zw10 expressing cells (G). Fixed cells were stained with anti-CENP-C (E, red), anti-Spc105R (F, red) or anti-tubulin (G, grey). DNA (DAPI) is shown in blue. Scale bar: 2 μm. Kinetochore signal intensity of the indicated proteins is shown as % of control. Only prometaphase cells were analyzed. Mean +/-SEM of 3 experiments (n > 90 cells). Student's t-test (n.s.: non-significant). H. Kinetochore-microtubule attachment assay. After 24 h pMT-CAL1-V5 induction, GFP-CENP-A/mCherry-Tubulin cells were incubated with 20 μM MG132 for 1 h, then with 100 nM Taxol for 3 h. The fixed cells were scored into 2 categories: either "attached" when each kinetochore was attached to microtubules, or "> 1 unattached" when at least 1 kinetochore was not stably connected to microtubules. The graph shows the percentage of cells in each category as Mean +/-SEM, N = 2. The graph shows the total CAL1 centromeric intensity per nucleus as % of control. Mean +/-SEM of 3 experiments (n>300 cells), Student's t-test ( ��� : p<0.001). B. Mitotic phenotypes of pMT-CENP-A-GFP/mCherry-Tubulin expressing cells. After 2 h pMT-CENP-A-GFP induction cells were imaged for 16 h and scored for their accuracy of mitosis: lagging (presence of lagging chromosomes during anaphase that will resolve before cytokinesis) or defective (formation of tripolar spindles, multinucleated cells). Mean +/-SEM n > 200 cells. Student's t-test (n.s.: non-significant). C. Ndc80 kinetochore levels in pMT-CENP-A-GFP/mCherry-Tubulin mitotic cells. After 2 h pMT-CENP-A-GFP induction, fixed cells were stained with anti-Ndc80 antibody (red). DNA (DAPI) is shown in blue. Scale bar: 2 μm. The graph shows the total Ndc80 kinetochore intensity per cell as % of control. Only prometaphase cells were analyzed. Mean +/-SEM of 3 experiments (n>90 cells), Student's t-test ( ��� : p<0.001). D. Kinetochore-microtubule attachment assay. After 2 h pMT-CENP-A-GFP induction, cells were incubated with 20 μM MG132 for 1 h, and with 100 nM Taxol for 3 h. Cells were then fixed and the DNA counterstained with DAPI. Cells were scored into 2 categories: either "attached" when each kinetochore was attached to microtubules, or "> 1 unattached" when at least 1 kinetochore was not stably connected to microtubules. The graph shows the percentage of cells in each category as Mean +/-SEM, N = 2.  Fig 2F). H2B-GFP/mCherry-tubulin expressing cells containing inducible pMT-CAL1-V5. pMT-CAL1-V5 expression was induced with 100 μM CuSO 4 for 24 h prior to imaging. Imaging was performed for 16 h on a Deltavision microscope (60X, Bin 2), at 25˚C. Time-lapse 3 min. Scale bar: 5 μm.