## Introduction

During mitosis, a eukaryotic cell divides into two genetically identical daughter cells. To achieve this, the duplicated chromosomes in the parent cell must be equally distributed into the daughter cells. The spindle assembly checkpoint (SAC) serves as a surveillance mechanism to ensure that duplicated chromosomes are stably attached to spindle microtubules through an adapter structure named the kinetochore. Kinetochores lacking end-on microtubule attachment activate the SAC to prevent premature anaphase onset and avoid chromosome missegregation. The effector molecule generated upon SAC activation is the Mitotic Checkpoint Complex (MCC). The MCC consists of two subcomplexes: BUBR1:BUB3 and CDC20:MAD21,2. It inhibits the E3 ubiquitin ligase Anaphase-Promoting Complex/Cyclosome (APC/C)3,4,5. APC/C ubiquitinates Cyclin B1, a key mitosis regulator, thereby targeting it for proteasome-mediated degradation6,7,8. Inhibition of the APC/C suppresses the degradation of Cyclin B1, which in turn delays anaphase onset.

## Results

### Structural modeling predicts that the MAD1:MAD2 complex may assume a folded conformation

To gather possible clues, we used AlphaFold225,26 to predict how the structurally known segments of the MAD1 C-terminal region may be arranged. This analysis predicted the existence of folded conformations of MAD1, which are enabled by a flexible hinge spanning residues 582 and 600, in addition to an extended conformation (Fig. 1A, right panel). We reasoned that the folded MAD1 conformation would permit the phosphorylated C-terminal RWD domains of MAD1 to approach the reaction center of the MAD1:MAD2 template complex where O-MAD2 is expected to undergo the conformational switch and bind CDC20 (Fig. 1B). Interestingly, the primary sequence of the hinge region is not conserved from yeast to human (Supplementary Fig. 1A), but an interruption of the coiled-coil with a disordered hinge appears to be very common (Supplementary Fig. 1B). According to AlphaFold2’s predictions, the flexibility of the hinge region enables MAD1 to assume a spectrum of conformations, from fully extended to folded (Fig. 1A)25,26.

### Assessment of the in vivo conformation of the MAD1:MAD2 complex using Förster resonance energy transfer (FRET)

Because of the modest FRET efficiency observed, we performed several control experiments to establish the efficacy of our FLIM acquisition and analysis protocol. We transiently expressed fusions of either mNG, mScarlet-I, or a tandem mNG-mScarlet-I tag to the C-terminus of NUP50, which is a component of the nuclear basket34 (Supplementary Fig. 3A). The fluorescent protein fused to NUP50 should experience a similar micro-environment as MAD1-mNG. Consistent with this, the NUP50-mNG fluorescence lifetime was indistinguishable from the MAD1-mNG lifetime. As expected, the lifetime of the mNG in NUP50-mNG-mScarlet-I was significantly lower (2.78 ± 0.03 and 2.16 ± 0.14 ns, respectively, Supplementary Fig. 3D). These values indicate a FRET efficiency for NUP50-mNG-mScarlet-I of 22%, which is comparable to the reported FRET efficiency values spanning 30–40% for a cytoplasmic EGFP-mCherry tandem FRET pair35. Importantly, the lifetime of MAD1-mNG in the presence of NUP50-mScarlet-I was only slightly lower than the lifetime of MAD1-mNG alone (2.77 ± 0.02 vs. 2.79 ± 0.03, respectively), indicating negligible FRET (Supplementary Fig. 3D). These data validate our FLIM acquisition and analysis protocols.

### MAD1’s hinge is important to the rate of MCC assembly in vitro

To test the role of the structural flexibility of MAD1 in the assembly of the MCC, we purified recombinant MAD1:MAD2 and MAD1ΔL:MAD2. Importantly, these complexes appeared stable and properly folded (Supplementary Fig. 4A). We compared their MCC assembly activity in vitro using the previously established MCC FRET-sensor-based assays (Fig. 2A)10,16. In this assay, MCC assembly is monitored by quantifying the FRET intensity signal arising from the close proximity between mTurquoise2-BUBR1 (the donor fluorophore) and MAD2-TAMRA (the acceptor fluorophore, see Fig. 2A).

Deletion of MAD1’s hinge caused a moderate but reproducible decrease in the rate of MCC assembly compared to the wild-type (Fig. 2B), indicating that the hinge is important to maximize the rate of MCC assembly in vitro. The rate difference between MAD1:MAD2 and MAD1ΔL:MAD2 relied on the presence of BUB1:BUB3 (Fig. 2C). More specifically, the rate difference required a functionally intact BUB1:BUB3 complex to interact with MAD1:MAD2, because the BUB1ΔCM1 mutant that prevents this interaction erased the difference (Supplementary Fig. 4B).

## Discussion

In the aforementioned parallel study40, the purified MAD1:MAD2 complex was shown to exhibit a folded conformation in vitro. Here, we showed that the MAD1:MAD2 complex assumes such a folded conformation also in vivo. Our data indicate that the structural flexibility is enabled by a flexible hinge in the C-terminus of MAD1, whose secondary structure—rather than the primary sequence—is conserved. This hinge is important for MCC assembly in vitro and SAC signaling in vivo, and we provide evidence that it can be replaced with similarly flexible but different sequences, implying that the hinge is unlikely to mediate hitherto unknown physical interactions with other proteins. Thus, collectively, the structural flexibility of MAD1 appears to be important to the SAC signaling activity.

Whether MAD1 switches between an extended conformation and the folded conformation at a physiologically meaningful rate in vivo, and whether this switching cycle correlates with the formation of a CDC20:MAD2 complex is currently unclear. The distribution of conformations of the two proline residues (P585 and P596) in the hinge may be under active, energy-consuming regulation in the cell, but assessing this will require further analyses. We note that no MAD1-interacting protein with peptidylprolyl cis-trans isomerase activity has been identified in the PrePPI database as of March 202241,42. It remains unknown whether the proline residues simply serve to break the coiled-coil or play a more complex role in promoting the folding of MAD1.

## Methods

Wide-field, z-stack fluorescence imaging for the quantification of the localization of MAD1-mNG, MAD1ΔL-mNG, and MAD2mScarlet-I at signaling kinetochores was the same as described previously44. AlphaFold2 structural predictions were conducted using the ColabFold advanced algorithm25. All ColabFold parameters were set to their default values except for “max recycles” (which was set to 6) and “tol” (which was set to 0.1).

### Theoretical end-to-end root-mean-square distance (RMSD) of a flexible unstructured peptide

First, we model a flexible peptide with n amino acid residues as a 3-D random walk (without considering steric hindrance and restrictions imposed by the Ramachandran plot). We denote the displacement of residue number i + 1 relative to residue number i as a random vector ri, i = 1, 2,…, n − 1. The end-to-end displacement, D, can be expressed as

$${{{{{\bf{D}}}}}}=\sum _ {i=1}^{n-1}{{{{{{\bf{r}}}}}}}_{i}$$
(1)

The RMSD is therefore

$$\sqrt{\left\langle {\left|{{{{{\bf{D}}}}}}\right|}^{2}\right\rangle }=\sqrt{{\sum }_{i=1}^{n-1}\left\langle {\left|{{{{{{\bf{r}}}}}}}_{i}\right|}^{2}\right\rangle+{\sum }_{i\ne j}\left\langle {{{{{{\bf{r}}}}}}}_{i}\cdot {{{{{{\bf{r}}}}}}}_{j}\right\rangle }$$
(2)

For a 3-D random walk, the random vectors representing each step are independent of each other. Therefore, ij,

$$\left\langle {{{{{{\bf{r}}}}}}}_{i}\cdot {{{{{{\bf{r}}}}}}}_{j}\right\rangle=0$$
(3)

Suppose that the contour length of each amino acid residue is universal (|ri | = r, i = 1, 2,…, n − 1; we take r = 0.37 nm here45), we have

$$\sqrt{\left\langle {\left|{{{{{\bf{D}}}}}}\right|}^{2}\right\rangle }=r\sqrt{n-1}=\frac{L}{\sqrt{n-1}}$$
(4)

wherein L = (n − 1)r is the contour length of the peptide.

Next, we model the same peptide using a worm-like chain model23,24. This model considers the peptide as a continuous worm-like chain rather than a discrete, step-by-step walk. The end-to-end RMSD is

$$\sqrt{\left\langle {\left|{{{{{\bf{D}}}}}}\right|}^{2}\right\rangle }=\sqrt{2{pL}\left[1-\frac{p}{L}(1-{e})^{-\frac{L}{p}}\right]}$$
(5)

wherein p is the persistence length (we take p = 0.3–0.7 nm here23,24), a metric for the stiffness of the chain.

### Purification of recombinant proteins

Wild-type or mutant constructs of MAD1:MAD2, MAD2, MPS1, BUB1:BUB3, CDC20, and BUBR1:BUB3 are of human origin. The constructs of MBP-MAD1ΔL:MAD2 and MBP-MAD1AL11:MAD2 are cloned via site-directed mutagenesis from the MBP-MAD1:MAD2 wild-type construct described previously10,16. All recombinant proteins used in this study have been expressed and purified according to the protocols described previously10,16.

### Low-angle metal shadowing and electron microscopy

MBP-MAD1:MAD2 or MBP-MAD1ΔL:MAD2 was diluted 1: 1 with a spraying buffer (200 mM ammonium acetate and 60% glycerol) to a final concentration of 0.5–1.0 μM and air-sprayed onto freshly cleaved mica pieces (V1 quality, Plano GmbH). Specimens were mounted and dried in a MED020 high-vacuum metal coater (Bal-tec). A 1-nm platinum layer and a 7-nm carbon support layer were subsequently evaporated onto the rotating specimen at angles of 6–7° and 45°, respectively. Pt/C replicas were released from the mica on water, captured by freshly glow-discharged 400-mesh Pd/Cu grids (Plano GmbH), and visualized using a LaB6-equipped JEM-1400 transmission electron microscope (JEOL) operated at 120 kV. Images were recorded at a nominal magnification of 60,000× on a 4k × 4k CCD camera F416 (TVIPS).

### FRET assay with the MCC FRET sensor

The MCC FRET sensor has been described previously10,16. The catalysts preparation consisted of 2 μM MBP-MAD1(wild-type or mutant):MAD2 and 2 μM BUB1 (wild-type or mutant):BUB3, which were separately incubated with 500 nM MPS1 in the assay buffer [10 mM HEPES (pH 7.5), 150 mM NaCl, 2.5% glycerol, and 10 mM β-mercaptoenthanol] supplemented with 1 mM ATP and 10 mM MgCl2 for 16 h at 4 °C. All assays were performed using a 100 nM final concentration of all proteins, except for CDC20, which was added at 250 nM. The fluorophores MAD2-TAMRA and mTurquoise2-BUBR1(1-571):BUB3 were added before measurements started. All measurements were performed on a CLARIOstar plate reader (BMG Labtech), using UV-Star 96-well plates (Greiner). The reactions had a final volume of 100 μL in the assay buffer. The excitation light and emitted fluorescence were filtered by a 430–10 nm excitation filter, an LP 504 nm dichroic mirror, and a 590–20 nm emission filter. The plate reader read at a 60-s interval for 120 min (6 mm focal height, 200 flashes, gain 1200) and mix the reactions for 5 s at 500 revolutions per minute after each measurement.

### Flow cytometry

The complete genotype of the mad2Δ S. cerevisiae strain (AJY4951) is leu2Δ−1, trp1Δ63, ura3-52, his3Δ200, lys2-8Δ1, mad2Δ::TRP1. The complete genotype of the Mad2GFP-expressing S. cerevisiae strain (AJY5041, constructed for this study) is leu2Δ0, met15Δ0, ura3Δ0, mad2Δ::KAN, Mad2101::GFP (HIS3).

Yeast strains were grown to mid-log phase, and then 15 μg/mL nocodazole was added to the media. Sample aliquots containing 2 × 106 cells were collected 0, 1, 2, 3, and 4 h after the addition of nocodazole. Samples were fixed by 70% ethanol and then stored at 4 °C overnight. On day 2, samples were washed with the RNase buffer [10 mM Tris (pH 8.0), 15 mM NaCl] and treated with 170 ng/mL bovine pancreatic RNase (Millipore Sigma) at 37 °C for one day in the RNase buffer. On day 3, samples were washed again, resuspended in PBS, and stored at 4 °C. The samples were treated with 5 mg/mL propidium iodide (Millipore Sigma) for 2 h at room temperature and subject to flow cytometry on an LSRFortessa Cell Analyzer (BD Biosciences). Approximately 10,000 cells were analyzed from each sample.

Data were analyzed using FlowJo (BD). Cells were first gated based on the area of the forward scatter and side scatter peak, followed by the area of the fluorescence peak (610 nm). Exemplary plots depicting the gating are included in the Source Data file.

### RNA interference

The two siRNAs targeting the 3’-UTR of MAD1 (siMAD1’s)46 were applied to unsynchronized cells at a concentration of 40 nM each for 2 days before imaging or cell-harvesting unless specified otherwise. The sense-strand sequence of siCDC20 was 5’-GGAGCUCAUCUCAGGCCAU-3’47, which was applied at a concentration of 40 nM for 2 days before FLIM or cell-harvesting. The sense-strand sequence of siMAD2 was 5’-GGAAGAGUCGGGACCACAGUU-3’48, which was applied at a concentration of 40 nM for 1 day before imaging or cell-harvesting. Desalted double-stranded siRNAs modified by double-deoxythymidine overhangs at 3’-ends of both strands were synthesized by Sigma. AllStars Negative Control siRNA (QIAGEN) is used as the control siRNA (siCtrl) and applied at the same dosage and time as the corresponding experimental group(s). All siRNAs were transfected into the cells via Lipofectamine RNAiMAX (Invitrogen).

### Fluorescence lifetime imaging microscopy (FLIM)

All FLIM data were collected on an ISS Alba v5 Laser Scanning Microscope, connected to an Olympus IX81 inverted microscope equipped with a UPLSAPO60XW objective. A Fianium WL-SC-400-8 laser with an acousto-optic tunable filter was used to generate excitation pulses at a wavelength of 488 nm and a frequency of about 20 MHz. The excitation light was further filtered by a ZT405/488/561/640rpc (Chroma Technology) quadband dichroic mirror. The emission light of the green channel was redirected by a 562 longpass dichroic mirror (FF562-Di03, Semrock), filtered by an FF01-531/40-25 filter (Semrock), and finally detected by an SPCM-AQRH-15 avalanche photodiode (Excelitas Technologies). The time-correlated single photon counting module to register detected photon events to excitation pulses was SPC-830 (Becker & Hickl). Data acquisition was facilitated by VistaVision (ISS).

A multi-component exponential fit is intrinsically flexible49. The fluorescence decay of mNeonGreen alone is multi-exponential50. Furthermore, the FRET efficiency between MAD1-mNG and MAD2mScarlet-I may be variable depending on the conformation of the MAD1:MAD2 complex as well as the presence of possible unlabeled MAD1 and MAD2 molecules in our heterozygous genome-edited cell lines. With these complications in mind, we used a two-component exponential decay model to fit the FLIM data:

$$I\left(t\right)=C\left[\alpha {e}^{-\frac{t}{{\tau }_{1}}}+\left(1-\alpha \right){e}^{-\frac{t}{{\tau }_{2}}}\right] \circledast {{{{{\rm{IRF}}}}}}(t)+D$$
(6)

In the equation above, D is the background signal offset. τ1 and τ2 are the lifetimes of the two exponential components. IRF(t) is the instrument response function. The IRF was determined by determining the photon arrival histogram for the donor channel using a 500 nM Rose Bengal solution in 5.6 M potassium iodide illuminated with the excitation laser. IRF drift51,52 was corrected by translating the IRF along the time axis by up to 2 ps to eliminate exponential components with unrealistic short lifetimes. The MATLAB nonlinear optimization function “fmincon” was used to find the best parameter set that fit the FLIM data. For more details, please refer to the data analysis toolkit available at https://github.com/CreLox/FluorescenceLifetime.

To demonstrate how fluorescence-lifetime measurements can quantify the FRET efficiency, consider a large number of donor fluorophore molecules with a lifetime of τ0. In the absence of acceptor fluorophores, the exponential decay D0 of donor fluorescence after pulsed excitation at time zero is

$${D}_{0}\left(t\right)=C{e}^{-\frac{t}{{\tau }_{0}}}$$
(7)

The total donor fluorescence intensity is

$${S}_{0}={\int }_{0}^{+{{\infty }}}{D}_{0}\left(t\right){dt}=C{\tau }_{0}$$
(8)

wherein C is a constant determined by the total number and properties of fluorophores, as well as the imaging setup. Without altering any of these conditions, in the presence of acceptor fluorophores and FRET, the possibility that an excited fluorophore stays excited (has not relaxed to the ground state either through the fluorescence-emitting route or the FRET-quenching route) at time t is

$${P=e}^{-\left(\frac{1}{{\tau }_{0}}+\frac{1}{\tau {\prime} }\right)t}$$
(9)

wherein τ' is the time constant of FRET (although an excited fluorophore can only relax through one route, the two stochastic processes—fluorescence-emitting and FRET-quenching—are independent). Therefore, in the presence of acceptor fluorophores and FRET, the new decay dynamics of the donor fluorescence are

$${D\left(t\right)={Ce}}^{-\left(\frac{1}{{\tau }_{0}}+\frac{1}{\tau {\prime} }\right)t}=C{e}^{-\frac{{\tau }_{0}+\tau {\prime} }{{\tau }_{0}\tau {\prime} }t}$$
(10)

The effective lifetime of the donor fluorophore (which can be measured through FLIM) becomes

$$\tau=\frac{{\tau }_{0}\tau {\prime} }{{\tau }_{0}+\tau {\prime} }$$
(11)

and the total donor fluorescence intensity becomes S = . Therefore, the FRET efficiency

$$\frac{{S}_{0}-S}{{S}_{0}}=\frac{{\tau }_{0}-\tau }{{\tau }_{0}}$$
(12)

Because the fluorescence lifetime in the absence of quenching is an intrinsic property of a mature fluorescent protein under a certain temperature53, the equation above greatly simplifies experiments to measure the FRET efficiency.

### Time-lapse live-cell imaging in knockdown-rescue mitotic duration assays

Time-lapse live-cell imaging was performed on an ImageXpress Nano Automated Imaging System (Molecular Devices). A SOLA Light Engine (Lumencor) served as the excitation light source. Cells were plated on 24-well cell imaging plates (with black walls and glass bottom, Eppendorf) and treated with siRNAs and 100 nM nocodazole accordingly. Humidified 5% CO2 was supplied to the environment chamber maintained at 37 °C.

### Immunoprecipitation (IP) using mNeonGreen-Trap

HeLa-A12 cells integrated with the Tet-On expression cassette of either mNeonGreen, MAD1-mNG, or MAD1ΔL-mNG were induced to express the ectopic exogenous protein by 0.1 μg/mL doxycycline (for 2 days until being harvested) and arrested at mitosis using the thymidine–nocodazole synchronization protocol (see the next subsection on immunoblotting). Cells were harvested by mitotic shake-off, washed once by PBS, pelleted down by centrifugation at 200–500×g for 3 min, snap-frozen in liquid nitrogen, and stored at −80 °C before the IP experiment.

On the day of the IP experiment, cells were thawed on ice and lysed in the IP lysis buffer [75 mM HEPES-HCl (pH 7.5 at 4 °C), 150 mM KCl, 10% (by volume) glycerol, 1.5 mM MgCl2, 1.5 mM EGTA, and 1% (by mass) CHAPS] supplemented with 1 mM PMSF, the cOmplete EDTA-free Protease Inhibitor Cocktail, Phosphatase Inhibitor Cocktail IV (RPI), 1 mM Na4P2O7, 0.1 mM Na3VO4, 5 mM NaF, and 2 mM sodium β-glycerophosphate before usage. For 1 mg of wet cell pellet, 40 μL of 4 °C IP lysis buffer was added, yielding a total protein concentration of about 5.6 mg/mL (if cells were lysed completely). Resuspended cells were rotated for 30 min at 4 °C and then centrifuged at 18,000×g for 20 min at 4 °C. In total, 600 μL of supernatant was subsequently cleared to reduce non-specific binding by adding 50 μL of equilibrated control agarose beads (ChromoTek) and rotating for 45 min at 4 °C. The mixture was centrifuged at 2000×g for 5 min at 4 °C. In all, 580 μL of cleared supernatant was then mixed with 30 μL of equilibrated mNeonGreen-Trap Agarose (nta-20, ChromoTek) and rotated for 1 h at 4 °C. These beads were then pelleted down at 2000×g for 5 min at 4 °C to remove the supernatant. The beads were further washed four times (rotated for 5 min at 4 °C and then pelleted down at 2000×g for 5 min at 4 °C) using 1 mL of the IP wash buffer [75 mM HEPES-HCl (pH 7.5 at 4 °C), 150 mM KCl, 10% (by volume) glycerol, 1.5 mM MgCl2, and 1.5 mM EGTA] each time. The beads were transferred to a fresh tube before the last wash to avoid the non-specific binding of proteins to the wall of the tube. Finally, 2× Laemmli sample buffer supplemented with β-mercaptoethanol was added to the beads. Samples were boiled in a boiling water bath for 10 min and analyzed by SDS-PAGE and immunoblotting.

### Immunoblotting

To acquire unsynchronized HeLa-A12 cells, asynchronous cells were either scrapped or trypsinized off the surface of dishes. To acquire mitotic HeLa-A12 cells, cells were first synchronized in G1/S with 2.5 mM thymidine and then arrested in mitosis with 330 nM of nocodazole for 16 h. This procedure is referred to as the thymidine–nocodazole synchronization protocol.

Harvested cells were then washed once by PBS, pelleted down, and chilled on ice. Lysis was performed by directly mixing with 2× Laemmli sample buffer supplemented by β-mercaptoethanol at a ratio of 1 μL per 0.1 mg of cell pellets. Lysates were boiled immediately afterward for 10 min in a water bath and then chilled on ice. In all, 8 μL of supernatant was loaded onto each lane of a 15-well, 0.75-mm SDS-PAGE mini gel.

Primary antibodies (and their working dilution factors by volume) used included anti-BUBR1 (Bethyl Laboratories A300-995A-M, 1:1000), anti-BUB1 (Abcam ab9000), anti-CDC20 (Santa Cruz Biotechnology sc-5296 for Fig. 2F and sc-13162, 1:200 for others), anti-MAD2 (Bethyl Laboratories A300-301A-M, 1:330), anti-GAPDH (Proteintech 60004-1-Ig, 1:5000), anti-MAD1 (GeneTex GTX109519, 1:2000 for Supplementary Fig. 3E and PLA0092, 1:1000 for others), anti-mNeonGreen (Cell Signaling Technology 53061S, 1:100), and anti-BUB3 (Sigma-Aldrich B7811, 1:500).

### Amino acid sequences used for multiple sequence alignment and coiled-coil prediction

The following UniProt accession codes were used to retrieve Mad1 amino acid sequences for multiple sequence alignments or prediction of coiled-coil domains: human: Q9Y6D9-1, mouse: Q9WTX8-1, zebrafish: D9IWE2, African clawed frog: Q6GPD1, budding yeast: P40957, fission yeast: P87169.

### Statistics and reproducibility

In Fig. 1F, ordinary one-way ANOVA was used to compare the mean lifetime of the donor–acceptor cases with the donor-only case. Left: F = 14.84, right: F = 21.81, P < 0.0001 in both cases. The P values displayed in the Fig. were obtained using Dunnett’s test in GraphPad Prism.

### Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.