Distinct Roles of RZZ and Bub1-KNL1 in Mitotic Checkpoint Signaling and Kinetochore Expansion
SUMMARY
The Mad1-Mad2 heterodimer is the catalytic hub of the spindle assembly checkpoint (SAC), which controls M phase progression through a multi- subunit anaphase inhibitor, the mitotic checkpoint complex (MCC) [1, 2]. During interphase, Mad1- Mad2 generates MCC at nuclear pores [3]. After nuclear envelope breakdown (NEBD), kineto- chore-associated Mad1-Mad2 catalyzes MCC assembly until all chromosomes achieve bipolar attachment [1, 2]. Mad1-Mad2 and other factors are also incorporated into the fibrous corona, a phospho-dependent expansion of the outer kinet- ochore that precedes microtubule attachment [4–6]. The factor(s) involved in targeting Mad1- Mad2 to kinetochores in higher eukaryotes remain controversial [7–12], and the specific phosphory- lation event(s) that trigger corona formation remain elusive [5, 13]. We used genome editing to eliminate Bub1, KNL1, and the Rod-Zw10- Zwilch (RZZ) complex in human cells. We show that RZZ’s sole role in SAC activation is to tether Mad1-Mad2 to kinetochores. Separately, Mps1 ki- nase triggers fibrous corona formation by phos- phorylating two N-terminal sites on Rod. In contrast, Bub1 and KNL1 activate kinetochore- bound Mad1-Mad2 to produce a ‘‘wait anaphase’’ signal but are not required for corona formation. We also show that clonal lines isolated after BUB1 disruption recover Bub1 expression and SAC function through nonsense-associated alternative splicing (NAS). Our study reveals a fundamental division of labor in the mammalian SAC and highlights a transcriptional response to nonsense mutations that can reduce or eliminate penetrance in genome editing experi- ments.
RESULTS
To analyze RZZ’s roles in fibrous corona assembly and SAC signaling, we used adeno-associated virus (AAV) and CRISPR/ Cas9 to modify both alleles of KNTC1 (Rod) in HCT116 cells, a diploid human colorectal cell line (Figures S1A–S1C). KNTC1HF/— (hypomorph-flox) cells expressed Rod at ~20% of the wild-type level (Figure S1D) and exited mitosis prematurely when microtubule polymerization (nocodazole, 99 ± 6 min SEM) or spindle bipolarity (S-trityl-L-cysteine [STLC], 193 ± 9 min) were inhibited. In contrast, wild-type cells never exited mitosis during the 16-hr time lapse (Figure 1A). We obtained viable KNTC1–/– clones after expressing Cre recombinase that were as SAC defective as KNTC1HF/— cells (Figures 1A and S1E). Early escape from spindle-poison-induced mitotic arrest was also observed in KNTC1–/– human retinal pigment epithelial (RPE) cells and KNTC1, ZW10, and ZWILCH KO HeLa cells (Figures 1B, 1C, and S1F–S1I). On the other hand, untreated RZZ-null cells had longer and more heterogenous mitotic timing, suggesting frequent but transient SAC activation (Fig- ure 1D). Consistently, inhibition of the SAC kinase Mps1 caused KNTC1–/– cells to exit mitosis with wild-type kinetics (Figure 1D). These observations suggest that RZZ maintains (but does not initiate) SAC signaling at unattached or improperly attached ki- netochores in multiple human cell types.To understand RZZ’s impact on mitotic chromosome and SAC signaling dynamics, we expressed and imaged H2B-mCherry and FLAG-GFP-TEV-S peptide (FLAP)-Mad1 using spinning disk confocal microscopy.
Mad1 first localized at kinetochores at nuclear envelope breakdown (NEBD) and then dissociated as chromosomes congressed at the metaphase plate (Figure 1E and Video S1; n = 10 cells). Congression was less efficient in KNTC1–/– cells, consistent with the lack of Spindly and dynein at kinetochores [15] (Figures S1F–S1I), but Mad1 was still targeted to misaligned chromosomes as effectively as in wild-type cells 25 cells per condition per experiment (N = 2). p values were computed using Kruskal-Wallis and Dunn’s multiple comparisons tests. Error bars throughout the paper indicate SEM unless stated otherwise.(B)Wild-type and KNTC1–/– RPE1 cells (Figures S1F–S1I) were treated with nocodazole, STLC, or taxol and were followed using DIC optics. Cell rounding (mitotic entry) and cortical blebbing and flattening (mitotic exit) were used as landmarks.(C)Clonal HeLa KNTC1, ZW10, and ZWILCH knockouts [14] were treated with nocodazole and were followed as in (B).(D)Mitotic timing in unperturbed wild-type and RZZ-deficient HeLa, RPE, and HCT116 cells. Where indicated, Mps1 kinase was inhibited with reversine.(E and F) Wild-type and KNTC1–/– HCT116 cells expressing H2B-mCherry and FLAP-Mad1 were filmed during unperturbed mitosis using spinning disk confocal microscopy. Insets show enlarged views of FLAP-Mad1 recruitment to and dissocia- tion from kinetochores. Scale bars throughout the paper are 10 mm unless stated otherwise. See also Videos S1 and S2.(G)Quantification of FLAP-Mad1 at misaligned chromosomes in (E) and (F).(H)Cells in (E) and (F) were filmed in the presence of nocodazole (n = 6 for wild-type and n = 14 for KNTC1–/–).(I and J) Wild-type and KNTC1-null RPE cells were treated with nocodazole and MG132 for 30 min or4 hr before fixation for IFM. Mad1/CREST fluorescence intensity ratios were determined for at least 100 kinetochores in five cells per condition (N = 3).(K) Wild-type and KNTC1-null RPE cells were treated with nocodazole in the presence or absence of hesperadin (hesp) to inhibit Aurora B kinase.
Mitotic duration was determined from 30 cells per condition.See also Figure S1.(Figures 1F and 1G and Video S2; n = 14 cells). We conclude that early mitotic cells can recruit Mad1-Mad2 to kinetochores and inhibit anaphase onset in the absence of the RZZ complex.Next, we analyzed Mad1 dynamics in cells undergoing noco- dazole-induced SAC arrest. Mad1 initially localized to kineto- chores at NEBD in both wild-type and KNTC1–/– cells, but this localization was not persistently maintained in the absence of RZZ (Figure 1H). To confirm this result for endogenous Mad1, we treated wild-type and KNTC1–/– RPE cells with nocodazole and MG132 (to block mitotic exit) for 30 min or 4 hr and then fixed and analyzed them by immunofluorescence microscopy (IFM). In wild-type cells, Mad1 formed large crescents that were stable over time, whereas it formed compact foci in KNTC1–/– cells that were eventually lost from kinetochores (Figures 1I and 1J). Suppression of early mitotic Mad1-Mad2 recruitment by treat-ment with Aurora B inhibitors [16, 17] eliminated the residual SAC response in KNTC1–/– cells (Tmitosis = 39 ± 10 min; Fig- ure 1K). These results reveal a temporal switch from RZZ-inde- pendent to RZZ-dependent recruitment of Mad1-Mad2 during chronic SAC signaling.Mps1 Promotes Kinetochore Expansion by Phosphorylating the N Terminus of RodMad1-Mad2 and RZZ localize to the fibrous corona, a phospho- dependent expansion of the outermost kinetochore layer that per- sists until end-on microtubule attachments are formed [5, 6, 18].
Kinetochore expansion is thought to accelerate mitotic ‘‘search and capture’’ by promoting lateral microtubule attachment [4] and to enhance SAC signaling [5]. RZZ is closely related to endo- membrane coatomers that form oligomeric lattices [19, 20] and ismost likely a ‘‘building block’’ of the corona itself [13, 15, 21, 22]. Consistent with these proposals, two other corona-associated proteins (CENP-E [23] and CENP-F [24]) did not form crescents in KNTC1–/– cells (Figures 2A, 2B, and S2A). To ensure that these results reflected loss of kinetochore expansion and not proteinmislocalization, we performed correlative light-electron micro-Serial sectioning revealed circumferential expansion of trilaminar plates and fibrous material in wild-type cells (n = 14 kineto- chores), whereas the kinetochores of KNTC1–/– cells appeared as compact discs (n = 15; Figure S2B) [13]. We conclude that the RZZ complex is required for fibrous corona formation.In parallel, we looked for mitotic kinases that might activate RZZ for kinetochore expansion. We found that CENP-E kinet- ochore crescents become compact after treating cells with an Mps1 inhibitor, but not after treatment with an Aurora B inhib- itor (Figures 2C and 2D) [13]. Through global phosphoproteomic screening, we identified two Mps1-modified sites at the N terminus of Rod (T13 and S15), up- stream of its b-propeller domain [27]. To test the function of these sites, we ex- pressed wild-type (WT) and nonphos- phorylatable (2A) versions of Rod inT-Rex FLP-in HeLa cells as LAP (EGFP-TEV-S-peptide) fusions (Figure S2C). Both LAP-RodWT and LAP-Rod2A were incorpo- rated into the full RZZ complex based on co-immunoprecipita- tion assays (Figure S2D). We then disrupted the KNTC1 locus in these cells using CRISPR/Cas9 and isolated transgene-com-plemented clones.
Although LAP-RodWT and LAP-Rod2A bothendogenous Rod, only LAP-RodWT formed crescents (Figures 2E and 2F). Thus far the only post-translational modification known to be required for crescent formation is C-terminal farne- sylation of Spindly, which enables its kinetochore recruitment via interaction with Rod’s b-propeller domain [15, 21, 25, 26]. How- ever Rod2A recruited Spindly and other corona-associated pro- teins in proportion to its own reduced abundance (Figure 2F). Despite having lower levels of Mad1-Mad2, the compact kineto- chores in Rod2A cells sustained mitotic arrest in nocodazole as effectively as those in RodWT cells (Figure 2G). We conclude that Rod’s N-terminal phosphorylation is required for fibrous corona formation but not SAC signaling.Mad1-Mad2 Requires a Non- receptor Activity of Bub1 to Inhibit AnaphaseBub1 is required for kinetochore expan- sion in Xenopus egg extracts [5], but its role in fibrous corona formation in human cells has not been examined. Bub1’s role in the SAC also remains controversial, with inconsistent results across studies [7–12]. To test Bub1’s contribution to these aspects of kinetochore structure and function, we deleted BUB1 in RPE cells via doxycycline-inducible CRISPR/ Cas9 [14]. BUB1–/– cells treated with no- codazole formed kinetochore crescents containing Rod, CENP-E, and Mad1, but not CENP-F [12] (Figures 3A and S3A). To ensure complete depletion and avoid postmitotic arrest [14], we deleted BUB1 or its kinetochore scaffold KNL1 [1, 2] inp53-deficient RPE cells. KNL1–/– cells formed crescents with normal levels of RZZ but slightly less Mad1 (22% reduction; Figures 3B and S3B). However deletion of BUB1 or KNL1 decreased the period of nocodazole-induced mitotic arrest by 76% and 93% (median Tmitosis = 130 min for KNL1–/– cells and 460 min for BUB1–/– cells, versus 1,935 min for control cells), indicating that that SAC signaling was functionally compromised.Although kinetochores in BUB1–/– and KNL1–/– cells have high levels of Mad1-Mad2, we could not exclude the possibility that a small but functionally important pool was missing.
There- fore, we tested the consequences of combining deletion ofBUB1 or KNTC1 with expression of a constitutively kineto- chore-bound form of Mad1 (Mis12-Mad1) that is refractory to SAC silencing at metaphase [28] (Figures 3D–3J). Mis12- Mad1 expression triggered a mitotic arrest in KNTC1–/– cells that was even longer (median Tmitosis = 1170 min) than that observed in wild-type cells expressing Mis12-Mad1 (median Tmitosis = 780 min; Figure 3I). This hyperactive response most likely reflects RZZ’s role in stripping Mad1 and other SAC mediators from metaphase kinetochores via dynein-dependent transport [29–31]. In contrast BUB1–/– cells had a much weaker response to Mis12-Mad1 kinetochore tethering (median Tmitosis = 130 min; Figure 3J). We next combined Mis12-Mad1 expression with nocodazole treatment to eliminate dynein- dependent stripping and engage upstream SAC signaling. This regimen further extended the mitotic arrest in KNTC1–/– cells (median Tmitosis = 1,355 min, versus 1,560 min for wild-type cells) but accelerated mitotic exit in BUB1–/– cells (median Tmitosis = 240 min) relative to nocodazole treat- ment alone (median Tmitosis = 460 min; Figures 3I and 3J). We conclude that Mad1-Mad2 kinetochore tethering can bypass RZZ, but not Bub1, with respect to SAC signaling. Our findings suggest that RZZ’s crucial and most likely sole function in SAC activation is to maintain Mad1-Mad2 at kinetochores, whereas RZZ-dependent corona formation is not required. In contrast Bub1 is not required for RZZ to localize at kinetochores, form the fibrous corona, or recruit Mad1-Mad2. However, Mad1-Mad2still requires a non-receptor activity of Bub1 to generate a ‘‘wait anaphase’’ signal.
The SAC defect that we observed after acute BUB1 disruption is consistent with studies in Bub1 conditional-knockout mouse embryonic fibroblasts [9, 32], but not with recent studies in BUB1-disrupted human cell clones [33–35] (Figure 4A). To un- derstand the basis of this discrepancy, we isolated 13 clones after acute disruption of BUB1 in p53-deficient RPE cells. All clones exhibited a partial (3%–30%) recovery of Bub1 expres- sion, kinetochore localization, and H2A kinase activity as judged by IFM with antibodies that recognize Bub1’s N terminus and T120-phosphorylated H2A (Figures 4B–4E). We performed RT-PCR and sequencing on five clones (Figure 4F and Data S1). Full-length BUB1 transcripts harbored exon 4 indels that induce frameshift and early termination (Figure S4A). We also observed shorter transcripts that skipped part or all of exon 4 and/or uti- lized cryptic splice sites (Figures S4B–S4F). A number of alterna- tively spliced transcripts encoded BUB1 open reading frames (ORFs) with short N-terminal deletions or insertions, thus explaining Bub1 re-expression (Figures S4C–S4F). Eleven of 13 clones exhibited partial or complete recovery of SAC function relative to acute deletion of BUB1 (Figure 4G). Among the five clones analyzed by RT-PCR and sequencing, clone 12 was fully SAC proficient and had the highest rate of in-frame transcripts (6 of 36), whereas clone 21 had intermediate SAC function and a lesser rate (3 of 31). No in-frame transcripts were identified in clone 8 (0 of 18) and clone 24 (0 of 21), which were the most SAC defective (Figure 4G). Taken together, these results suggest that nonsense-associated alternative splicing (NAS) [36] attenuates and in some cases suppresses the effects of null mutations in BUB1. Our findings demonstrate how genome editing can trigger both acute loss of function and compensatory changes in mRNA structure that result in pheno- copying of unedited cells.
DISCUSSION
In cells treated with spindle poisons, Mad1-Mad2 remains bound to kinetochores and catalyzes MCC production for 1,000 min or more, thus extending mitosis at least 30-fold. How (and why) this sustained response occurs is not well understood. In yeast, Bub1 is the sole receptor for Mad1-Mad2 and required for the SAC [37], but models for Mad1-Mad2 regulation in mammalian cells differ considerably [7–12]. In our studies, acute BUB1 or KNL1 deletion led to SAC failure despite high levels of Mad1- Mad2 at kinetochores. Furthermore BUB1–/– cells were largely unresponsive to Mis12-Mad1, which is constitutively tethered to kinetochores and cannot be silenced at metaphase. Similar results were obtained independently using a Ndc80-Mad1 fusion [35]. These data strongly suggest a non-receptor function of Bub1-KNL1 that is required for kinetochore-bound Mad1- Mad2 to inhibit anaphase. One possibility is that Bub1 functions as a co-catalyst in MCC assembly by recruiting Cdc20 to kinetochores [7, 38]. Asking whether this occurs in vivo will require SAC-independent methods for synchronizing BUB1–/– and KNL1–/– cells in mitosis in sufficient quantity and purity for biochemical studies of MCC assembly [39] or tools for rapidly eliminating Bub1 and KNL1 after SAC-dependent synchronization.
Our studies also shed light on RZZ’s role in the SAC. By tracking Mad1 dynamics with high temporal resolution, we demonstrate that kinetochores in early mitotic cells can recruit Mad1-Mad2 without RZZ and can delay anaphase onset by 100–300 min, thus mitigating the impact of less efficient chromo- some congression in RZZ-null cells. However kinetochores with attachment defects that persist beyond this timeframe require RZZ to recruit Mad1-Mad2 and maintain SAC arrest. Expression of Mis12-Mad1 reinstated long-term arrest in KNTC1–/– cells, suggesting that RZZ’s crucial and perhaps only role in the SAC is to tether Mad1-Mad2 to kinetochores. It has been proposed that RZZ mediates SAC signaling at unattached, but not tension-less, kinetochores [10]. However, KNTC1–/– cells challenged with spindle poisons that block attachment (nocodazole) or permit attachment without tension (STLC and taxol) escaped SAC ar- rest with similar kinetics (Figure 1B).
RZZ is also implicated in formation of the fibrous corona, a structural expansion of the outer kinetochore that precedes microtubule attachment [4, 6, 18]. Kinetochore expansion de- pends on mitotic kinases [5, 13], but relevant phosphorylation events are not known. We identified two Mps1-regulated phos- phosites just upstream of Rod’s b-propeller domain [27] that are required for kinetochore expansion, but not SAC arrest. Rod and Spindly not only interact via this domain [15, 21], but also inhibit their own assembly into polymers [13, 40]. Together, these findings suggest that phosphorylation alleviates a struc- tural barrier to Spindly-RZZ polymerization.
In this study and others [3, 14, 39, 41, 42], we used genome editing to delete or disrupt exons of genes involved in cell division. Normally this results in a ‘‘knockout’’ because of nonsense-mediated decay (NMD), a pathway that degrades mRNAs with premature termination codons (PTCs) [43, 44]. However PTCs can also trigger NAS, a less-well-understood pathway in which splicing rules are relaxed to bypass the PTC and restore expression of near-full length ORFs [36, 45, 46]. NAS could also explain why BUB1 exon-2-disrupted HeLa cells manifest a clear SAC defect after BUB1 exon-8-specific RNAi [35]. In conclusion, Empesertib NMD and NAS have opposite effects on the expressivity and penetrance of nonsense mutations and should not be overlooked in the design, analysis, and interpretation of genome editing experiments.