KIF2C condensation concentrates PLK1 and phosphorylated BRCA2 on kinetochore microtubules in mitosis

Abstract

During mitosis, the microtubule depolymerase KIF2C, the tumor suppressor BRCA2, and the kinase PLK1 contribute to the control of kinetochore-microtubule attachments. Both KIF2C and BRCA2 are phosphorylated by PLK1, and BRCA2 phosphorylated at T207 (BRCA2-pT207) serves as a docking site for PLK1. Reducing this interaction results in unstable microtubule-kinetochore attachments. Here we identified that KIF2C also directly interacts with BRCA2-pT207. Indeed, the N-terminal domain of KIF2C adopts a Tudor/PWWP/MBT fold that unexpectedly binds to phosphorylated motifs. Using an optogenetic platform, we found that KIF2C forms membrane-less organelles that assemble through interactions mediated by this phospho-binding domain. KIF2C condensation does not depend on BRCA2-pT207 but requires active Aurora B and PLK1 kinases. Moreover, it concentrates PLK1 and BRCA2-pT207 in an Aurora B-dependent manner. Finally, KIF2C depolymerase activity promotes the formation of KIF2C condensates, but strikingly, KIF2C condensates exclude tubulin: they are located on microtubules, especially at their extremities. Altogether, our results suggest that, during the attachment of kinetochores to microtubules, the assembly of KIF2C condensates amplifies PLK1 and KIF2C catalytic activities and spatially concentrates BRCA2-pT207 at the extremities of microtubules. We propose that this novel and highly regulated mechanism contributes to the control of microtubule-kinetochore attachments, chromosome alignment, and stability.

Graphical Abstract

Graphical Abstract.

Introduction

During mitosis, the proper distribution of chromosomes into daughter cells is a highly controlled process that is essential for genome integrity. Accurate chromosome segregation relies on the bi-oriented attachment of chromosomes to the mitotic spindle. This is ensured through binding of the kinetochores to microtubules oriented towards each spindle pole. During early phases of mitosis, initial capture of kinetochores by microtubules is asynchronous and stochastic, and erroneous attachments are frequent. Persistence of these errors is a common cause of chromosome mis-segregation, associated chromosomal instability and aneuploidy, as observed in solid tumors [1–3]. Erroneous attachments are converted to bi-oriented attachments to ensure faithful chromosome segregation [4]. The efficiency of correction depends on the rate of turnover of kinetochore-microtubule attachments: erroneous attachments must be released to enable the formation of new, correct attachments. Cancer cells with chromosomal instability have excessively stable kinetochore-microtubule attachments [5]. Strategies that destabilize these attachments promote faithful chromosome segregation in cancer cells by increasing the rate of correction of attachment errors. The microtubule depolymerase KIF2C (alias mitotic centromere-associated kinesin, MCAK) plays a key role at centromeres/kinetochores by increasing kinetochore-microtubule attachment turnover, thus facilitating error correction prior to anaphase [6, 7]. Overexpression of KIF2C reduces the rate of chromosome mis-segregation [5, 8]. Inversely, depleting KIF2C is sufficient to promote chromosome segregation defects to levels comparable to those of cancer cells with chromosomal instability [9].

Kinases and phosphatases also contribute to proper chromosome segregation by generating switching phospho-sites that are fine-tuned to correct attachment errors [10]. At the kinetochore-microtubule interface, phosphorylation events are inhibitory to the attachment process, because they electrostatically interfere with microtubule binding [11]. Phosphorylation of NDC80 by Aurora B thus inhibits the microtubule attachment process. Conversely, Polo-like kinase 1 (PLK1) stabilizes kinetochore-microtubules attachments through phosphorylation of different substrates [12, 13]. In particular, it phosphorylates residues S676 and T680 of BUBR1, which enhances recruitment of the B56 subunit of Protein Phosphatase 2A (PP2A-B56), thereby ensuring the correct strength and dynamics of the kinetochore-microtubule attachments [14–17]. Mitotic regulation by AuroraB-PLK1-PP2A operates only at specific subcellular localization and at certain times, but how this is controlled remains unclear.

Assembly of membrane-less organelles through multivalent intermolecular interactions was recently established as a mechanism that creates high local concentrations, favoring prompt catalysis with precise spatial-temporal control. Increasing evidence suggests that this mechanism contributes to the regulation of microtubule structure and dynamics [18]. Indeed, several proteins interacting with tubulin form condensates in vitro and recruit tubulin within these condensates. Tubulin condensation is implicated in spindle apparatus assembly [19], nucleation of acentrosomal and branched microtubules [20, 21] and centrosome maturation [22, 23]. Microtubule dynamics is also fine-tuned by Tip-Interacting Proteins. Condensation of these proteins on microtubules, and more specifically at the tip of microtubules, was recently reported in vitro and in cells. This process can lead to high local concentrations of tubulin and favor microtubule growth [24, 25]. The Tip-Interacting EB1 protein phase-separates in vitro and in cells, and EB1 condensates contain KIF2C [26, 27]. These condensates could be involved in the development of a tension across the spindle microtubule and kinetochore [26].

The tumor suppressor protein BRCA2 is present at the kinetochore in mitosis [28] and contributes to the stability of the kinetochore-microtubule attachments [28]. We and others demonstrated that BRCA2 is phosphorylated by PLK1 [28–30]. We identified four PLK1 phospho-sites in the N-terminal region of BRCA2, including S193 and T207 that are conserved from mammals to fishes [28, 31 ,32]. Phosphorylation of T207 triggers a direct interaction between BRCA2 (specifically BRCA2-pT207) and PLK1, and the assembly of a complex between BRCA2, PLK1, BUBR1, and PP2A. This complex is essential for the stability of the kinetochore–microtubule interactions and the proper alignment of chromosomes [28]. Reduced binding of phosphorylated BRCA2 to PLK1, as observed in BRCA2 breast cancer variants S206C and T207A, results in erroneous chromosome segregation and aneuploidy. In this study, we identified a direct, PLK1-dependent, interaction between BRCA2-pT207 and KIF2C, which is mediated by a previously unrecognized KIF2C phospho-peptide binding domain. We showed that, in mitotic cells, KIF2C assembles into biomolecular condensates that are located adjacent to microtubules and kinetochores. These condensates depend on the phospho-binding capacity of KIF2C and on the PLK1 kinase activity. They are enriched in activated PLK1 as well as in phosphorylated BRCA2. However, they exclude tubulin. We propose that KIF2C condensates locally amplify the activities of PLK1 and KIF2C and recruit BRCA2-pT207 at the kinetochore in mitosis.

Materials and methods

Cell lines and synchronization

Flp-In T-REx 293 cell lines were grown under standard conditions (37°C, 5% CO2) in Dulbecco's modified Eagle's medium (Merck-Sigma-Aldrich, D5796) containing 10% fetal bovine serum (FBS). Parental cells were selected with 100 μg/ml Zeocin (ThermoFisher Scientific R25001) and 15 μg/ml Blasticidin (InvivoGen, ant-bl) and Flp-In 293 T-REx derived stable cell lines were maintained with 5 μg/ml Blasticidin (InvivoGen, ant-bl) and 50 μg/ml Hygromycin B (Sigma-Aldrich, H3274). DLD1+/+ cells were grown in RPMI media supplemented with 2 mM L-Glutamine (prepared in the culture facility of CBMSO) and 10% FBS (BioWest S1810-500). DLD1 BRCA2 -/- cells stably expressing EGFP-MBP-BRCA2 WT or EGFP-MBP-BRCA2 T207A (generated in [28]) were grown in RPMI media supplemented with 2 mM L-Glutamine, 10% FBS, 1 mg/ml G418 (Enzo Life Sciences ALX-380–013-G005) and 0.1 mg/ml Hygromycin B (Thermo Scientific 10 687 010). All cells were grown in a humidified incubator at 37°C and 5% CO₂. For synchronization of cells in mitosis, nocodazole (100–300 ng/ml, Sigma-Aldrich) was added to the growth media and the cells were cultured for 14 h before harvesting.

Plasmids

Oligonucleotides used for plasmid construction are listed in Supplementary Table S1. For expression in bacteria, optimized genes coding for human 6His-AviTag-BRCA2_167-260 and 8His-TEVsite-BRCA2_48–218 were synthetized by Genscript and cloned in a pETM13 vector. In addition, optimized genes coding for human 8His-GB1-TEVsite-KIF2C_1-79, its K52E/K54E mutant, and 8His-GB1-TEVsite-KIF2A_1-77 were synthetized by Genscript and cloned in a pET-22b vector. For expression in mammalian cells, we first obtained pcDNA5-FRT-Hygro plasmids coding for HsKIF2C-2–725-V145A-mCherry-AtCRY2-1-498 and TurboID-HsKIF2C-2–725-V145A-mCherry-AtCRY2-1-498. These plasmids were mutated to generate the list of constructs displayed in Supplementary Table S2. The EGFP-MBP-tagged BRCA2 cDNA was previously described [28].

Cloning

Primers used for plasmid construction are listed in the KRT. For pcDNA5_FRT_TurboID constructs, KIF2C cDNA was amplified with primers 1 and 2 (See Supplementary Table S1) using Phusion High-Fidelity DNA Polymerase. The amplified sequence was inserted into the KpnI site of pCDNA5_FRT_TO_TurboID-mCherry-Cry2 (Addgene 166 504). pcDNA5_FRT_opto-KIF2C-cDNA was obtained by removing the TurboID from the pcDNA5_FRT_TurboID-opto-KIF2C construct using the primers 3 and 4. Mutations in KIF2C were generated using the QuickChangeMulti Site-Directed Mutagenesis Kit (Agilent, 00 514) with the primers listed in Supplementary Table S1.

Generation of stable cell lines

Flp-In 293 T-REx cells are seeded to reach 80–90% confluence on the day of transfection. pcDNA5_FRT_TO expression plasmids were mixed with pOG44 encoding the Flp recombinase at a 1:7 ratio in opti-MEM. For a single transfection in a 6 well plate, 500 ng of the expression plasmid was mixed with 3.5 μg of pOG44 in 250 uL opti-MEM. Additionally, 8 uL Lipofectamine 2000 Transfection Reagent was added to 250 uL opti-MEM. After an incubation period of 5 min at room temperature, both solutions were mixed and incubated for a further 15 min at room temperature. The mixture was then pipetted dropwise onto the cells. The medium was changed after 6 h. At 48 h post- transfection, the cells were transferred to a 100mm petri dish, and 24 hours later, the selection was performed by adding 5 μg/mL Blasticidin and 50 μg/mL Hygromycin B. Clones were pooled, and the cells were examined for the expression of the construct by immunoblotting and fluorescence microscopy.

Production of the recombinant Avi-tagged BRCA2_167-260 peptide

A recombinant 6His-Avi-BRCA2_167-260 construct was designed that contains the 15 amino acid sequence GLNDIFEAQKIEWHE, so-called Avi-Tag [33]. It was produced in E. coli BL21 (DE3) Star cells in either its unlabeled form for pulldown experiments or its ¹⁵N-labeled form for NMR experiments. Cells were grown in M9 minimal medium containing 0.5 g/l ¹⁵NH₄Cl when ¹⁵N labeling was needed. The bacterial culture was induced at an OD₄₀₀ of 0.6–0.8 with 1 mM IPTG and was further incubated during 4 h at 37°C. Harvested cells were resuspended in 50 mM Tris-HCl pH 8.0, 50 mM NaCl, 5 mM DTT, 2 mM EDTA, 1 mM PMSF, 1 mM ATP, 5 mM MgSO₄, 500 μg of lysozyme, 0.5 μL of benzonase (E1014; Millipore). They were sonicated on ice 2.5 min in total with 1s ON/1s OFF cycle of sonication (50% amplitude), and the lysate was clarified by centrifugation during 15 min at 15 000 g at 4°C. The soluble fraction was loaded on a Ni-NTA poly-histidine-affinity column (5 mL HisTrap Excel, GE Healthcare) at a 2 mL/min flow rate. The column was washed with a solution containing 50 mM Tris-HCl at pH 8.0, 50 mM NaCl, and 1 mM DTT. The sample was then eluted with an imidazole gradient over 45 mL, the buffer containing 50 mM Tris-HCl at pH 8.0, 50 mM NaCl, 1 M imidazole, and 1 mM DTT. The sample was boiled 5 min at 95°C and spun down 5 min at 16 000 g at 4°C. It was concentrated using Novagen concentrators with 3.5 kDa cut off membranes centrifuged at 5000 g and later injected on a gel filtration column (Superdex 16/600 HiLoad 75 pg) equilibrated with a Dulbecco’s phosphate buffered saline (DPBS) (D1408; Sigma) at pH 7.2. Fractions were pooled, 1 mM of fresh DTT was added and the sample was concentrated using a 3 kDa cut off concentrator centrifuged at 5000 g. EDTA-free protease inhibitors (cOmplete EDTA-free, Sigma-Aldrich) were added at a final 1X concentration. The quality of the purified proteins was analyzed by SDS-PAGE and the protein concentrations were determined by spectrophotometry using the absorbance at 280 nm.

Production of the recombinant BRCA2_48-218 peptide

¹⁵N-labeled 8His-TEVsite-BRCA2_48-218 was produced by transforming E. coli BL21 (DE3) Star cells, cultured in M9 minimal medium containing 0.5 g/l ¹⁵NH₄Cl, and supplemented with 100 μg/ml kanamycin, at 37°C. Expression was induced at an OD₄₀₀ of 0.6–0.8 with 1 mM IPTG and cells were harvested after 3–4 h of incubation by centrifugation. The bacterial pellet was resuspended in 35 mL of a solution containing 50 mM Tris-HCl at pH 8.0, 50 mM NaCl, 5 mM DTT, 2 mM EDTA, 1 mM PMSF, 1 mM ATP, 5 mM MgSO₄, 500 μg lysozyme, and 0.5 μL benzonase (E1014; Millipore). Cells were sonicated on ice 2.5 min in total with 1s ON/1s OFF cycle of sonication (50% amplitude), and the lysate was clarified by centrifugation during 15 min at 15 000 g at 4°C. The soluble fraction was loaded on a Ni-NTA histidine-affinity column (5 mL HisTrap Excel, GE Healthcare) using a 2 mL/min flow rate. The column was washed with a solution containing 50 mM Tris HCl at pH 8.0, 50 mM NaCl, 1 mM DTT. The sample was eluted using an imidazole gradient over 45 mL, the final buffer containing 50 mM Tris HCl at pH 8.0, 50 mM NaCl, 1 M imidazole, 1 mM DTT. Fractions of interest were pooled, and 0.4 mg of TEV protease and 2 mM of fresh DTT were added to the sample, which was incubated 1 h at RT. Then, the sample was boiled 5 min at 95°C and spun down 5 min at 16 000 g at 4°C. It was concentrated using Novagen concentrators with 3.5 kDa cut off membranes centrifuged at 5000 x g, and injected on a gel filtration column (Superdex 16/600 HiLoad 75 pg) equilibrated with DPBS (D1408; Sigma) at pH 7.2. Fractions were pooled, supplemented with 1 mM fresh DTT, and concentrated using a 3.5 kDa cut off concentrator centrifuged at 5000 g. EDTA-free protease inhibitors (cOmplete EDTA-free, Sigma-Aldrich) were added at a final 1X concentration. The quality of the purified proteins was analyzed by SDS-PAGE and the protein concentrations were determined by spectrophotometry using the absorbance at 280 nm.

Expression and purification of KIF2C and KIF2A N-terminal domains

The constructs 8His-GB1-TEVsite-KIF2C_1-79 (WT and K52E/K54E) and 8His-GB1-TEVsite-KIF2A_1-77KIF2C_1-79 were expressed in E. coli BL21 (DE3) Star cells supplemented with 100 μg/ml ampicillin. Unlabeled proteins were produced by growing bacteria in 2x Luria broth (LB) media. ¹⁵N and/or ¹³C-labeled proteins were produced by growing the bacteria in M9 minimal medium containing 0.5 g/l ¹⁵NH4Cl and 2 g/l ¹³C-glucose. Expression was induced at an OD₄₀₀ of 0.6–0.8 with 1 mM IPTG and cells were harvested after overnight of incubation at 20°C by centrifugation. The bacterial pellet was resuspended in 35 mL of a solution containing 50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 0.5% glycerol, and 30 μL Triton X-100. Cells are sonicated on ice 2.5 min in total with 1s ON/9s OFF cycle of sonication (60% amplitude) at 10°C. 5 mM MgSO₄, 0.5 μL benzonase (E1014; Millipore), and 2 mM EDTA were added to the solution and incubated for 20 min at 4°C. Then, the lysate was clarified by centrifugation during 15 min at 15 000 g at 4°C. The soluble fraction was loaded on a Ni-NTA histidine-affinity column (5 mL HisTrap Excel, GE Healthcare) using a 1.5 mL/min flow rate. The column was washed with a solution containing 50 mM Tris HCl at pH 8.0, 150 mM NaCl, and 5 mM EDTA. Then, the sample was eluted using an imidazole gradient over 45 mL, the final buffer containing 50 mM Tris HCl at pH 8.0, 150 mM NaCl, and 1 M imidazole. The fractions of interest were pooled and dialyzed in a solution containing 50 mM Tris HCl at pH 8.0, 150 mM NaCl, and 5 mM EDTA overnight. To cleave the tag, 0.4 mg of TEV protease, 1X of EDTA-free protease inhibitors (cOmplete EDTA-free) and 2 mM of fresh DTT were added to the sample, which was incubated 1 h at room temperature. The sample was then loaded on a HisTrap column and the tag-free proteins were collected in the flow through. The sample was concentrated using a Novagen concentrator with a 3.5 kDa cut off membrane centrifuged at 5000 g, and injected on a gel filtration column (Superdex 16/600 HiLoad 200 pg) equilibrated with a buffer containing 50 mM Tris-HCl, 50 mM NaCl at pH 7.5 for non-labeled samples or DPBS (D1408; Sigma) at pH 7.0 for ¹⁵N and/or ¹³C-labeled samples. Fractions are pooled, concentrated using 3 kDa cut off concentrators centrifuged at 4500 x g. The quality of the purified proteins was analyzed by SDS-PAGE 15% and the protein concentrations were determined by spectrophotometry using the absorbance at 280 nm.

Search for BRCA2 partners by pull-down and mass spectrometry

Identification of phospho-dependent BRCA2 partners was carried out following our recently published protocol [34]. After production and purification of the 6His-Avi-BRCA2_167-260 construct, half of the peptide sample was phosphorylated by adding PLK1 (provided by the recombinant protein platform of Institut Curie, Paris) at a molar ratio of about 200:1, as published [35, 36]. Biotinylation of 6His-Avi-BRCA2_167-260, either non-phosphorylated or phosphorylated, was performed by incubating the peptide at 100 μM in a solution containing 2 mM ATP, 600 μM biotin, 5 mM MgCl2, 1 mM DTT, and 1X protease inhibitors, together with 0.7 μM of the BirA enzyme (produced and purified in the lab) in a buffer containing 50 mM HEPES at pH 7.0, 150 mM NaCl, 1 mM EDTA. The peptide was then incubated for 1 h and 30 min at RT and injected on a gel filtration column (Superdex 16/60 HiLoad 75 pg) previously equilibrated with a solution containing 50 mM HEPES pH 7.0, 75 mM NaCl, and 1 mM EDTA. The fractions of interest were pooled, supplemented with 1 mM fresh DTT, and the sample was concentrated using a 3 kDa cut off concentrator centrifuged at 5000 g. EDTA-free protease inhibitors (cOmplete EDTA-free; Sigma-Aldrich) were added at a final 1X concentration and the sample was flash frozen using liquid nitrogen.

From here, all conditions were performed with five replicates treated at the same time in order to favor the reproducibility of the experiment. Two μg of recombinant biotinylated 6His-Avi-BRCA2_167-260 in PBS at pH 7.5, 1 mM DTT, 1X protease inhibitors were added to 50 μL of Streptavidin-coated magnetic beads (Streptavidin Mag-Beads; Genscript) in a final volume of 100 μL. Beads were incubated for 1 h at RT on a rotating wheel and were washed three times using 500 μL of PBS. They were then washed two times with 500 μL of a solution containing 50 mM HEPES at pH 7.2, 150 mM NaCl, 2 mM EDTA, 10 mM NaF, 0.5 mM PMSF, 1 X protease inhibitors, 1 mM DTT, and 1X PhosphoSTOP mixed with 800 μg of lysed cells extracts (HEK293 cells synchronized or not with nocodazole) in 20 mM HEPES at pH 7.6, 150 mM NaCl, 0.1% NP40, 2 mM EGTA, 1.5 mM MgCl2, 50 mM NaF, 10% glycerol, 20 mM b-glycerophosphate, 1 mM DTT, and 1X protease inhibitors. The beads were incubated for 2 h at RT on a rotating wheel. They were washed three times with 100 μL of a solution containing 50 mM HEPES at pH 7.2, 150 mM NaCl, 2 mM EDTA, 10 mM NaF, 0.5 mM PMSF, and 1 mM DTT and washed twice without resuspension of the beads with 500 μL of a solution containing 50 mM ammonium bicarbonate at pH 8.2. The beads were then kept on ice in 500 μL of the buffer containing 50 mM ammonium bicarbonate at pH 8.2, and analyzed by the mass spectrometry platform of Institut Curie (Paris). Further data processing was achieved on the website of the platform using the tool MYPROMS.

NMR experiments

Most experiments were carried out at 10°C on 600 and 700 MHz Bruker spectrometers equipped with a triple resonance cryoprobe. Spectra were processed in Topspin 4.3.0 and analyzed with CcpNmr Analysis 2.4.2 software. NMR chemical shift assignments of the BRCA2 fragments used in this study are published [31]. To assign the NMR chemical shifts of the KIF2C and KIF2A N-terminal domains, standard 3D triple resonance NMR experiments (HNCO, HNCACO, CBCA(CO)HN, HNCA and HNCACB) were carried out on samples of proteins at 300 μM in DPBS pH 7.0, 2 mM DTT, 5% D₂O, 50 mM DSS. Analysis of these experiments was performed, and the secondary structure of the KIF2C and KIF2A N-terminal domains was deduced from the resulting assigned backbone ¹H, ¹⁵N, and ¹³C chemical shifts using the website https://st-protein02.chem.au.dk/ncSPC/cgi-bin/selection_screen_ncSPC.py.

To check the phosphorylation states of the ¹⁵N-labeled BRCA2 fragments, we carried out ¹H-¹⁵N SOFAST-HMQC experiments on samples of proteins at 150 μM in DPBS pH 7.2, 2 mM DTT, 5% D₂O, 50 mM DSS, mixed with 3 μM PLK1, 5 mM MgCl₂, 2 mM ATP, 1X EDTA-free protease inhibitors.

To identify protein-protein interaction sites, we carried out ¹H-¹⁵N SOFAST-HMQC experiments (1536 × 200 timepoints, 256 scans, 50 ms of interscan delay) on 150 μL of ¹⁵N-labeled non-phosphorylated and phosphorylated BRCA2_167-270 (and BRCA2_48-218 in Fig. 2C) mixed with non-labeled KIF2C_1-79 and KIF2A_1-77 at molar ratios of 1:0, 1:1, in DPBS pH 7.2. In addition, we performed ¹H-¹⁵N SOFAST-HMQC experiments (1536 × 160 timepoints, 200 scans, 40 ms of interscan delay) on 150 μl of ¹⁵N-labeled KIF2C_1-79 (WT or K52E/K54E) and KIF2A_1-77 mixed with non-phosphorylated and phosphorylated BRCA2 peptides (pT207) at molar ratios of 1:0, 1:0.5, 1:1, 1:2, in DPBS pH 7.0. The peptides used for all experiments were synthesized by GenScript (Piscataway, NY) (Supplementary Table S3).

Figure 2.

BRCA2 pT207 binds to the N-terminal domains of KIF2C and KIF2A. (A) NMR analysis of the interaction between BRCA2_167-260 phosphorylated by PLK1 (F1^P) and the N-terminal domain of KIF2C (KIF2C Nt). 2D NMR ¹H-¹⁵N SO-FAST HMQC spectra of F1^P, where each peak represents the chemical environment of a BRCA2 residue, were recorded before (black) and after (red) addition of KIF2C Nt (ratio 1:1). Superimposition of these spectra revealed that, upon binding, several peaks disappeared, while other peaks shifted (marked with arrows). The associated graph displays the peak intensity decrease upon addition of KIF2C Nt as a function of the BRCA2 residue number. (B) NMR analysis of the interaction between BRCA2_167-260 phosphorylated by PLK1 (F1^P) and the N-terminal domain of KIF2A (KIF2A Nt). Here again, the 2D NMR ¹H-¹⁵N SO-FAST HMQC spectra of phosphorylated Avi-tagged BRCA2_167-260 recorded before (black) and after (magenta) addition of the N-terminal domain of KIF2A (KIF2A Nt; ratio 1:1) were superimposed. The peaks affected by the interaction are labeled. (C) NMR characterization of the KIF2C-bound state of phosphorylated BRCA2_48-218. On the left, 2D NMR ¹H-¹⁵N SO-FAST HMQC spectra acquired on phosphorylated BRCA2_48-218 in the presence and absence of the N-terminal domain of KIF2C (ratio 1:1) were superimposed; several peaks, including those corresponding to S205, pT207, L209, and V11, disappeared after addition of KIF2C. On the right, ¹⁵N CEST analysis performed on phosphorylated BRCA2_48-218 in the presence of the KIF2C N-terminal domain (ratio 1:0.15); the major and minor dips observed for residues S205, pT207, L209, and V211 indicate the ¹⁵N chemical shifts in the free and bound states, respectively.

To measure the chemical shifts of the BRCA2 peptide bound state, we recorded ¹⁵N D-CEST experiments at 10°C on a 800 MHz Bruker spectrometer [37]. The NMR sample consisted of ¹⁵N-labeled BRCA2_167-260 and BRCA2_48-218 at 200 μM in the presence of unlabeled KIF2C_1-79 at a molar ratio 10:1, in DPBS pH 7.2. We used effective DANTE B1 fields of 10, 20, and 40 Hz with the position of the RF field varied over 500, 800 and 1500 Hz (sw_CEST), respectively, using step sizes of 20.8, 36.4 and 65.2 Hz. The D‐CEST period, T_Ex, was set to 300 ms; each 2D plane was recorded with 16 transients per FID. The prescan delay was 1.5 s, and (795 115) complex points in (t2, t1) were recorded, to give a net acquisition time of ∼2 h per spectrum. Total measurement time for D-CEST experiments was ∼2 d per B1 field, so altogether ∼6 d. All NMR exchange data were fitted simultaneously using the program ChemEx.

Isothermal titration calorimetry

Isothermal titration calorimetry (ITC) measurements were performed with the KIF2C_1-79 or KIF2A_1-77 protein at 8–10 μM in the cell and the BRCA2 peptide (Supplementary Table S3) at 80–100 μM in the syringe. The buffer was 50 mM Tris-HCl, pH 7.5, 50 mM NaCl. The experiments were carried out on a VP-ITC instrument (Malvern), using automatic injections of 8 or 10 μl at 20°C. The peptides used for all experiments were synthesized by GenScript (Piscataway, NY) (Supplementary Table S3). Control experiments were carried out by injecting peptides into the cell filled with buffer, to estimate the heat of dilution. The titration curves were analyzed using the program Origin 7.0 (OriginLab) and fitted to a one-site binding model.

Size-Exclusion Chromatography coupled to multi-angle light scattering (SEC-MALS)

SEC-MALS was used to measure the molecular mass of the N-terminal domain of KIF2C in solution. KIF2C N-terminal domain was loaded on an Agilent Technologies HPLC system with a BIOSEC 3–300 column (4,6 × 300mm) (flow rate at 200 μl/min). The light scattering was measured with a 3 angles MALS system from WYATT company and the RI (Refractive Index) to analyze the difference of refraction index is a VISCOTEK (MALVERN) equipment. The chromatography buffer was 25 mM Tris-HCl buffer (pH 7.0), 50 mM NaCl. A calibration was performed with BSA as a standard. Data were analyzed using the ASTRA software. To represent the data, the normalized absorbance at 280 nm was overlaid with the molar mass (Da), and both parameters were plotted as a function of the elution volume.

AlphaFold calculations

AlphaFold models of full-length monomeric proteins were obtained from the AlphaFold Protein Structure Database [38, 39]. Models of protein-peptide complexes were computed using the server of the Integrative Bioinformatics platform of I2BC (https://bioi2.i2bc.paris-saclay.fr). For each complex, a series of 15 models were calculated and the models were analyzed using the delivered heatmap and lDDT plots, as well as the pTM and ipTM scores [38, 40].

Cell extracts, GFP-TRAP, and western blotting for identifying BRCA2-KIF2C interaction in cells

For analysis of the complex between BRCA2 and KIF2C in mitosis, DLD1 BRCA2-/- cells stably expressing EGFP-MBP-BRCA2 WT were synchronized with 0.1 μg/ml nocodazole (Sigma Aldrich) for 15 h and 30 min. The mitotic population was enriched by shake off. Cell pellets were harvested and lysed with extraction buffer A (20 mM HEPES (pH 7.5), 150 mM NaCl, 0.1% NP₄0, 2 mM EGTA, 1.5 mM MgCl2, 50 mM NaF, 10% glycerol, 1 mM Na₃VO₄, 20 mM ß-glycerophosphate, 1 mM DTT and EDTA-free Protease Inhibitor Cocktail). Samples were pre-cleared by centrifugation at maximum speed for 30 min. GFP-TRAP agarose beads (ChromoTek) were equilibrated with extraction buffer A before being incubated with mitotic lysates for 1.5 h at 4°C. The beads were then washed 4 times with extraction buffer A and 3 times with extraction buffer A containing 250 mM NaCl. Then, beads were heated at 95°C for 5 min in 3X SDS-PAGE Laemmli buffer and spun again. The supernatant containing the eluted proteins were then migrated on a 4–15% gradient pre-cast SDS-PAGE (Bio-Rad), transferred onto nitrocellulose membranes (Amersham) and finally blotted with the following antibodies: anti-mouse BRCA2 (1:1000, OP95) and anti-mouse KIF2C (1:1000).

Metaphase spreads coupled with immunofluorescence

To study the localization of BRCA2-pT207 and KIF2C at the kinetochores, cells stably expressing either BRCA2 WT or T207A were synchronized using 0.1μg/ml nocodazole for 4 h to enrich the mitotic population. Dividing cells were collected by shake off, pelleted, and subjected to a hypotonic shock with 50 mM KCl for 20 min. Cells were then spun onto Poly-D-lysine (Sigma Aldrich) precoated coverslips for 5 min at 500 g. Then cells were fixed and permeabilized simultaneously with 4% PFA, 0.5% Triton, 50mM NaF, 20mM ß -glycerophosphate and 1mM Na₃VO₄ for 20 min at RT. After 3 washes of PBS-Tween 0.2%, 5 min each, coverslips were blocked with PBS-BSA 4% for 30 min at RT before being incubated with the following antibodies overnight at 4°C: rabbit anti-BRCA2-pT207 (as previously described [28] (1:500, Genscript), mouse anti-KIF2C (1:100) and human anti-CREST (1:100). Coverslips were washed twice for 5 min each with PBS-Triton 0.1% and incubated for 2h at RT with the appropriate Alexa Fluor secondary antibody: donkey anti-rabbit Alexa-488 (1:1000), donkey anti-mouse Alexa-488 (1:1000), donkey anti-rabbit Alexa-594 (1:1000) or goat anti-human Alexa-555 (1:1000). Primary and secondary antibodies were diluted with PBS-Tween 0.1%, 5% BSA. After 2 washes of the secondary antibodies with PBS-Triton 0.1% and a final wash with PBS, coverslips were stained with DAPI (1μg/ml, Merck, Cat. # 268 298) and mounted using Prolong Glass antifade (P-36984, Thermo Fisher Scientific). Images were acquired with a Laser Scanning Confocal Microscope LSM900 coupled to an upright Axio Imager 2 Microscope (Zeiss). Image acquisition was done with a 100x oil objective.

Optogenetic activation of KIF2C condensation

Cells were plated at around 70% confluency in DMEM. Expression of Opto-KIF2C was induced with 1 μg/ml doxycycline for 16 h. In the case of studying cells in mitosis, 100 ng/ml nocodazole was added for 16 h. To activate the light, the samples were transferred into a custom-made illumination box containing an array of 24 LEDs (488 nm) delivering 10 mW/cm². Cry2 oligomerization was induced using 5 min of light-dark cycles (4 s light followed by 10 s dark).

Pulldown of biotinylated proteins

TurboID Flp-In T-Rex 293 cell lines stably transfected with opto-KIF2C recombinant protein and grown to 75% confluence were incubated with 6 ng/ml of doxycycline for 16 hours. The next day, the cells were incubated with 500 mM of biotin for 15 min. Cells were then washed with PBS and lysed with lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% NP-40, 0.2% SDS, 0.5% sodium deoxycholate) supplemented with 1X complete protease inhibitor, 1X phosphatase inhibitor and 250U benzonase. Lysed cells were incubated on a rotating wheel for 1 hour at 4°C prior sonication on ice (40% amplitude, 3 cycles 10 sec sonication- 2 sec resting). After centrifugation (7750 rcf.) for 30 min at 4°C, the cleared supernatant was transferred to a new tube and the total protein concentration was determined using the Bradford protein assay. For each condition, 1 mg of proteins was incubated with 50 μl of streptavidin-Agarose beads on a rotating wheel at 4°C for 3 hours. After 1 min centrifugation (400 rcf.), the beads were washed sequentially with 1 ml of lysis buffer, 1 ml wash buffer 1 (2% SDS in H₂O), 1 ml wash buffer 2 (0.2% sodium deoxycholate, 1% Triton X-100, 500 mM NaCl, 1 mM EDTA, and 50 mM HEPES pH 7.5), 1 ml Wash Buffer 3 (250 mM LiCl, 0.5% NP-40, 0.5% sodium deoxycholate, 1 mM EDTA, 500 mM NaCl and 10 mM Tris pH 8) and 1 ml Wash Buffer 4 (50 mM Tris pH 7.5 and 50 mM NaCl). For Western blot analysis of KIF2C partners enriched in optogenetic KIF2C condensates, cells were simultaneously incubated with 500 mM of biotin and exposed to blue light for 15 min of light-dark cycles (4 s light followed by 30 s dark). Biotin proximity labeling of light-induced KIF2C partners was performed using streptavidin-coated beads as described previously. Bound proteins were eluted from the agarose beads with 80 μl of 2X Laemmli sample buffer and incubated at 95°C for 10 min. 5 μg of the lysates were used for Western blot analysis and probed by immunoblotting to detect proteins that are associated with KIF2C clusters.

Western blotting for TurboID constructs

Constructs used for western blotting are listed in Supplementary Table S2. Whole cell extracts were prepared by lysing cells in RIPA buffer (50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 1% NP-40, 1% deoxycholate, 0.1% SDS) for 30 min on ice. After centrifugation, the supernatant was collected and analyzed for protein amount using the Quick Start Bradford protein assay kit. After addition of Laemmli Sample buffer 2X (Biorad, C161-0737), the sample was heated for 5 minutes at 95°C. According to the manufacturer's instructions (BioRad), 40 μg of protein samples were resolved on precast SDS-PAGE gels (4–15% and 10%) and transferred to a nitrocellulose membrane using the Bio-Rad Trans-Blot Turbo transfer device. Membranes were saturated with 10% non-fat milk diluted in TBS-0.2% Tween 20 (TBS-T), incubated with primary antibodies (Supplementary Table S4) overnight at 4°C and with secondary anti-mouse-HRP (1:1000) or anti-rabbit-HRP (1:1000) antibodies for 1h. Blots were developed with ECL according to the manufacturer's instructions.

Immunofluorescence staining and Fluorescent Microscopy Imaging

Cells grown on coverslips were fixed with PBS/4% paraformaldehyde (PFA) for 15 min at RT followed by a 10 min permeabilization step in PBS/0.2% Triton X-100-PBS and blocked in PBS/3% BSA for 30 min. An intermediate wash with PBS 1x for 5 min was used to completely remove buffers. For immunofluorescence staining, primary antibodies (Supplementary Table S4) were diluted in blocking solution and incubated for 1 h at RT, after which cells were washed using 1X PBS. Next, the corresponding secondary antibodies coupled to the fluorochrome were diluted in a blocking solution and also incubated for 1 h at RT. DNA was stained with Hoechst 33 258 (Invitrogen, Cat H21491) during 5 min at RT. Coverslips were then mounted onto glass slides using Prolong Gold Antifade Reagent (Invitrogen, Cat P36930). The finished coverslips were stored at 4°C. Images were acquired on a LEICA SP8X inverted confocal laser scanning microscope equipped with a 63x HC Plan Apochromat CS2 oil-immersion objective (NA: 1.4) (Leica) with hybrid GaAsP detectors (Hamamatsu). For staining with three channels, a white light laser was used at 405, 488 and 561 nm wavelengths to sequentially excite DAPI, GFP and mCherry respectively and their fluorescence was collected in line accumulation mode (4-line accumulation) through the following bandpasses: 415–461, 495–550 and 575–800 nm. For staining using four channels, a white light laser was used at 405, 488, 561 and 633 nm wavelengths to sequentially excite DAPI, GFP, mCherry and far red respectively and their fluorescence was collected in frame accumulation mode (4-frame accumulation) through the following bandpasses: 415–461, 495–550, 575–630 and 695–800 nm.

FRAP experiments

For FRAP experiments, we used a Leica TCS SP8X system equipped with a 63x HC Plan Apochromat CS2 oil-immersion objective (NA: 1.4) (Leica), 488 nm. Cells were preseeded in a μ-Slide 8 Well high Glass Botton 170 μm (Ibidi, 80 807). They were incubated for 12 h in the presence of 1 μg/ml doxycycline to induce opto-KIF2C expression and nocodazole was added to block cells in mitosis at 37°C. The whole living cells were illuminated to primary activate foci for one cycle of light (15 s followed by dark for 30 s), and then a plane about 0.5 μm thick showing a large number of foci was exposed during 20 min with light-dark cycles (8 s light followed by 30 s dark). To maintain cell viability, all experiments were carried out at 37°C and 5% CO₂. FRAP acquisitions were taken using 3D imaging and a 3 Airy pinhole, because the condensates move very fast and we did not want to lose them or to confuse the recovery with a movement on itself. The bleaching events were performed in a circular region with a diameter of 500 nm and monitored over time for fluorescence recovery. In order not to bleach the entire condensate, we targeted its boundaries. The laser was set at 2% for bleaching and 20% for imaging with a scanning speed of 400 Hz. The FRAP sequence was composed of a short pre-bleach sequence of 3 images, the bleaching event during 2.61 s, and 1 post-bleach sequence repeated 100–300 times. To collect the images, we used a PMT with a gain fixed at 400 and a bandpass 576–800 (not to lose fluorescence). Since we only had one marking and no auto-fluorescence, FRAP curves were independently corrected and processed to obtain a double normalization as follow: the mean intensity of every bleached region was measured and the background intensity was subtracted by measuring a region outside the cell. Acquisition-related bleaching correction was performed by dividing values by the whole cell. Then, to display the recovery curves from 0 to 1, normalization was performed using the average of the pre-bleached signal and the 1^st post-bleached value. As the recovery curves display a biphasic aspect, the mean curve was fitted by a double exponential function to extract both half-time recovery and the mobile fraction of pooled recovery curves.

Structured illumination microscopy

SIM experiments were performed on a ZEISS Elyra 7 – Lattice SIM equipped with a 63x Plan-Apochromat (N.A. 1.4) oil immersion objective and coupled with a PCO.edge sCMOS camera (pixel size: 6.5 μm; bit depth 16 bit). For Alexa 568 imaging, excitation was performed with a 561 nm laser (100 mW), and fluorescence collected using a dual band emission filter (BP 495–550 nm, BP 570–620 nm). For Alexa 488 imaging, excitation was performed with a 488 nm laser (100 mW, OPSL), and fluorescence collected using a triple band emission filter (BP 420–480 nm, BP 495–525 nm, LP 650 nm). For DAPI imaging, excitation was performed with a 405 nm laser (50 mW), and fluorescence collected using a triple band emission filter (BP 420–480 nm, BP 495–525 nm, LP 650 nm). For all the channels the illumination was structured as a lattice pattern (G4, 32 μm) and 13 phases were shifted for each plane and a z-step of 273 nm was used to generate 3D-SIM acquisitions (leap mode). SIM processing was performed on ZEN Black (ZEISS, version 16.0) and to correct chromatic aberrations, alignment procedure (ZEN Black) was applied on both channels after measurements on multispectral calibration beads.

Image and statistical analysis

Image treatment and analysis were performed using Imaris version 10.1.0 and FIJI version 2.0.0 software [41]. The plugin ezcolocalization for FIJI software was used to evaluate colocalization in 2D and 3D [42]. To quantify colocalization, Pearson correlation coefficients were calculated. DAPI was used as a mask in all experiments except for colocalization between KIF2C and BRCA2-pT207. In the case of colocalization between KIF2C and pBRCA2-pT207, a mask based on KIF2C condensates was used. The analysis of the FRAP experiment was carried out using a plugin for FIJI developed at I2BC. All graphs were generated using GraphPad Prism or SuperPlotsOfData by Huygens (https://huygens.science.uva.nl/SuperPlotsOfData/). Data were plotted as means ± standard deviation (SD). For the sake of clarity, the data distribution was displayed as a half-violin plot on the right-hand side of the dotplots. Statistical significance of differences was calculated on the basis of the total number of analysed cells, with an unpaired two-tailed t-test of Mann–Whitney by using GraphPad Prism version 10.1.1. In all cases, significance thresholds were: ns: non-significant; ∗: p < 0.05; ∗∗: p < 0.01; ∗∗∗: p < 0.001.

Time-lapse video microscopy

To observe the fusion events in interphase and mitosis, cells expressing either KIF2C-mCherry or Opto-KIF2C WT were illuminated under the microscope using a protocol similar to that described for FRAP. A cage incubator (OKO Lab) was used to control environmental conditions during the whole experiment (37°C and 5% CO₂). Images were collected in 3–5 z-stacks with a step of 0.3–0.5 μm every 5 s. The whole system was driven by MetaMorph software version 7.10 (Molecular Devices). Further 3D visualization and measurement of the volume of the condensates were performed using Imaris software version 10.1.0 with a PSF correction of 0.6785. In the case of KIF2C-mCherry, the 3D movie acquisition was performed using a Nipkow Spinning Disk confocal module (Yokogawa CSU-X1-A1) mounted on a Nikon Eclipse Ti E inverted microscope equipped with a 100 X APO TIRF oil immersion objective (NA: 1.49), whereas in the case of Opto-KIF2C, it was performed using a Leica TCS SP8X system equipped with a 63x HC Plan Apochromat CS2 oil-immersion objective (NA: 1.4).

Results

BRCA2 phosphorylated at T207 interacts with the microtubule depolymerases MCAK/KIF2C and KIF2A

We searched for phospho-dependent partners of BRCA2 in mitosis. We focused on partners of BRCA2 phosphorylated by PLK1 at two highly conserved sites: S193 and T207. Therefore, we produced and purified two batches of recombinant Avi-tagged BRCA2_167-260, and phosphorylated one of these batches with PLK1. We verified using Nuclear Magnetic Resonance (NMR) that S193 and T207 were totally phosphorylated (Fig. 1A). The less conserved T219 and T226 were also phosphorylated, as reported [28]. Following our recently published protocol [34], we then (i) biotinylated the two batches (phosphorylated and non-phosphorylated) using BirA, (ii) attached the resulting peptides on streptavidin-coated magnetic beads and (iii) incubated the beads in mitotic human cell extracts. Proteins interacting with either non-phosphorylated or phosphorylated BRCA2_167-260 were identified and quantified using mass spectrometry. The resulting plot showed that PLK1 was the most enriched protein on the phospho-peptide coated beads (Fig. 1B, left panel). We performed a similar experiment using BRCA2_167-260 mutated at T207. By comparing the lists of proteins binding to the phospho-peptides BRCA2_167-260 WT and T207A, we observed that PLK1 interacts with BRCA2_167-260 only in the presence of phosphorylated T207 (pT207) (Fig. 1B, right panel). This is fully consistent with our previously published crystal structure of a complex between the Polo-Box Domain of PLK1 and a peptide fragment of BRCA2 phosphorylated at T207 [28].

Figure 1.

BRCA2 phosphorylated by the mitotic kinase PLK1 interacts with the microtubule depolymerases KIF2C and KIF2A. (A) NMR characterization of BRCA2_167-270 before (F1; black) and after phosphorylation by PLK1 (F1^P; red). Superposition of the 2D NMR ¹H-¹⁵N SO-FAST HMQC spectra of F1 and F1^P, where each peak represents the chemical environment of a BRCA2 residue, identified phosphorylated residues; peaks corresponding to the four phosphorylated residues are labeled. (B) Volcano plots showing proteins from mitotic HeLa cell extracts that were identified by mass spectrometry as binding to (left) BRCA2_167-270 phosphorylated by PLK1 (F1^P) versus non-phosphorylated BRCA2_167-270 (F1), and (right) BRCA2_167-270 phosphorylated by PLK1 (F1^P) versus phosphorylated BRCA2_167-270 variant containing the T207A mutation (F1 T207A^P). (C) Domain organization of KIF2C and KIF2A, with residues of KIF2C mutated in this study indicated by colored circles: K52 and K54 in the N-terminal domain, G495 in the catalytic site, and S715 involved in KIF2C dimerization. (D) AlphaFold model of KIF2C, colored as in (C).

We also identified two other proteins from the Volcano plot analysis: the microtubule depolymerases MCAK/KIF2C and KIF2A (Fig. 1B). Interaction with these proteins also involves pT207: peptides corresponding to KIF2C and KIF2A were almost twice more enriched on beads with phosphorylated BRCA2_167-260 WT as compared to T207A. All three BRCA2 binding partners, i.e. PLK1, KIF2C and KIF2A, were not detected when beads were incubated in asynchronous human cell extracts (Supplementary Fig. S1A), probably because of the lower expression levels of these proteins outside of mitosis (as shown in the Cyclebase 3.0 database).

Phosphorylated BRCA2_167-260 binds directly to the N-terminal domains of KIF2C and KIF2A

The sequences of the two depolymerases KIF2C and KIF2A are 53% identical (Supplementary Fig. S1B). As predicted by AlphaFold, KIF2C and KIF2A exhibit an uncharacterized N-terminal folded domain, a large disordered region, a neck and motor domain that is responsible for the catalytic activity of the enzymes [43, 44], and an α-helical C-terminal region that controls the oligomerization of KIF2C [45] (Fig. 1C and D; Supplementary Fig. S1B). The motor domain of KIF2C interacts with tubulin [46] and dimerizes upon binding to a C-terminal peptide of KIF2C centered on S715 (Supplementary Fig. S1C; [47]). We tested direct binding of phosphorylated BRCA2_167-260 (F1^P) to the different domains of KIF2C.

Therefore, we superimposed 2D NMR ¹H-¹⁵N spectra of this construct recorded in the absence and presence of KIF2C domains. We could not detect any interaction between phosphorylated BRCA2_167-260 and the neck and motor domain of KIF2C in our conditions (Supplementary Fig. S2A), whereas we observed a clear interaction between phosphorylated BRCA2_167-260 and the N-terminal domain of KIF2C (KIF2C Nt) (Fig. 2A). NMR ¹H-¹⁵N peaks corresponding to the BRCA2 region T203-R212 showed a strong (larger than 60%) decrease in intensity upon addition of KIF2C Nt, demonstrating that this region interacts directly with KIF2C Nt. We also showed that phosphorylated BRCA2_167-260 interacts with the N-terminal domain of KIF2A (KIF2A Nt) through the same BRCA2 region (Fig. 2B). To further identify BRCA2 residues whose chemical environment is most affected upon binding and confirm the central role of phosphorylated T207, we performed an NMR ¹⁵N CEST experiment. We used a BRCA2_48-218 construct that contains only two PLK1 phosphorylated sites (S193 and T207), hence avoiding overlap between NMR peaks of phosphorylated residues. By superimposing 2D NMR ¹H-¹⁵N spectra of this construct recorded in the absence and presence of KIF2C Nt, we observed that several NMR peaks, including that of S205, pT207, L209 and V211, disappeared upon addition of KIF2C Nt (Fig. 2C). Furthermore, the ¹⁵N CEST experiment revealed that the ¹⁵N frequency difference between BRCA2 free and bound states is particularly large for these 4 residues, demonstrating that they form the core of the binding motif (Fig. 2D). Finally, we did not detect any interaction between non-phosphorylated BRCA2_167-260 and the N-terminal domain of either KIF2C or KIF2A (Supplementary Fig. S2B and C). Altogether, we concluded that BRCA2 interacts with KIF2C and KIF2A N-terminal domains through a S₂₀₅S(pT)VLIV₂₁₁ motif centered on phosphorylated T207.

The N-terminal domains of KIF2C and KIF2A exhibit a conserved and positively charged cavity that is responsible for phospho-peptide binding

We then searched for the BRCA2 binding sites in the KIF2C and KIF2A N-terminal domains, which share 43% of identity and 68% of similarity (Supplementary Fig. S1B). KIF2C Nt is monomeric, as shown by Size-Exclusion Chromatography coupled to Multi-Angle Light Scattering (SEC-MALS) (Fig. 3A). We measured by ITC that the Nt domains of KIF2C and KIF2A bind to BRCA2_{194-214(pT207)} with K_d of 0.48 ± 0.04 and 0.60 ± 0.10 μM, respectively (Fig. 3B; Supplementary Fig. S2D–F; Table 1). We also confirmed by ITC that this binding is phospho-dependent (Supplementary Fig. S2E).

Figure 3.

A conserved and positively charged cavity in the N-terminal domains of KIF2C and KIF2A is responsible for phosphorylated BRCA2 peptide binding. (A) Determination of the oligomeric state of the N-terminal domain of KIF2C (KIF2C Nt) by SEC-MALS. (B) Affinity and stoichiometry of the interaction between KIF2C/KIF2A Nt and BRCA2_{194-214(pT207)}, as measured by ITC. (C) Identification of the KIF2C Nt residues interacting with BRCA2_{194-214(pT207)} by NMR. Superposition of the 2D NMR ¹H-¹⁵N SO-FAST HMQC spectra of KIF2C Nt recorded upon addition of increasing concentrations of BRCA2_{194-214(pT207)} identified the peaks that shifted upon binding. These shifts were quantified by calculating the NMR CSP as a function of the KIF2C residue number. Residues with CSP values larger than 0.2 ppm are colored in green on an AlphaFold model of the N-terminal domain of KIF2C. (D) Identification of the KIF2A Nt residues interacting with BRCA2_{194-214(pT207)} by NMR, performed as in (C). Here again, residues with CSP values larger than 0.2 ppm are colored in green on an AlphaFold model of the N-terminal domain of KIF2A. G51 is marked by a star because it is not reported in the graph (poor estimation of its CSP). (E) Representation of the surface properties of the N-terminal domain of KIF2C. On the left, surface conservation from human to fish is shown, with identical residues in the sequence alignment of Supplementary Fig. S3A in cyan. On the right, the electrostatic potential at the surface of the KIF2C N-terminal domain is presented (blue—positively charged—to red negatively charged). (H) Models of the complexes between KIF2C/KIF2A Nt (colored as in panels C and D) and BRCA2_194-214, as calculated by AlphaFold. Only BRCA2 residues from T203 to N213 are displayed for clarity. These residues are colored as a function of their pLDDT score (red to white: high-to-low precision on residue position). The corresponding heat maps (representing the error in the relative positioning of the residues of the complex) are shown on the left, and the interface scores (ipTM, from 0 - low confidence - to 1 - high confidence) are indicated.

Table 1.

Summary of the ITC data obtained by adding synthetic phosphorylated BRCA2 peptides to KIF2C/KIF2A recombinant proteins

	Kd (μM)	N	ΔH (kcal/mol)	ΔG (kcal/mol)	-TΔS (kcal/mol)
BRCA2194-214(pT207) vs KIF2C Nt expt 1	0.48 ± 0.04	0.62	−19.1	−19.8	−0.7
BRCA2194-214(pT207) vs KIF2C Nt expt 2	1.30 ± 0.80	0.77	−12.1	−12.4	−0.3
BRCA2194-214(pT207) vs KIF2A Nt expt 1	0.60 ± 0.10	0.51	−15.7	−16.2	−0.5
BRCA2194-214(pT207) vs KIF2A Nt expt 2	0.80 ± 0.10	0.47	−18.1	−18.8	−0.7

To identify KIF2C and KIF2A residues involved in BRCA2 binding, we assigned the 2D NMR ¹H-¹⁵N HSQC peaks of both N-terminal domains (Supplementary Fig. S3A–D), and we measured the changes in chemical shifts (chemical shift perturbations, or CSP) of these peaks upon addition of BRCA2_{194-214(pT207)} (Fig. 3C and D; Supplementary Fig. S3E and F). We identified 11 KIF2C residues and 12 KIF2A residues with CSP values larger than 0.2 ppm. AlphaFold 2 predicts that the N-terminal domains of KIF2C and KIF2A adopt a β-barrel structure. Half of the KIF2C and KIF2A residues interacting with phosphorylated BRCA2 are located in the C-terminal β-strand of the β-barrel structure (Fig. 3C and D; Supplementary Fig. S3G and H). KIF2C Nt is significantly less conserved in evolution (from mammals to fishes) than KIF2A Nt (Supplementary Fig. S3A and B). In the case of KIF2C, we could identify a conserved site at the surface of the N-terminal domain (Fig. 3E). The BRCA2 binding site as defined by NMR strongly overlaps with the KIF2C Nt conserved site, which is also positively charged (Fig. 3E). We performed a similar chemical shift analysis upon addition of non-phosphorylated BRCA2_194-214 and did not detect any binding between KIF2C or KIF2A Nt and this peptide (Supplementary Fig. S3I).

Using AlphaFold 2, we further calculated 15 models of the complexes between KIF2C or KIF2A Nt and BRCA2_194-214. 10 and 14 models have medium interface scores (0.7 > ipTM > 0.5), respectively. All these models showed a β-sheet formed by the KIF2C β-strand A50-D57 (or KIF2A β-strand D48-D55) and the BRCA2 peptide (Fig. 3F). In this β-sheet, KIF2C K52 and K54 (or KIF2A K50 and K52) are located in front of phosphorylated T207. These lysines were consistently identified as part of the BRCA2 binding site by NMR (Fig. 3C and D). We also produced a mutant construct of the KIF2C N-terminal domain in which K52 and K54 were both mutated into glutamic acid. NMR analysis showed that mutations do not affect the overall structure of the domain, but completely disrupt the binding to BRCA2_{194-214(pT207)} (Supplementary Fig. S3J and K). Finally, we noticed that, based on AlphaFold 2 predictions, the phospho-peptide binding domains of KIF2C and KIF2A adopt the same barrel-like fold as Tudor/PWWP/MBT domains, which bind to methylated peptides [48–50]. We experimentally found that KIF2C and KIF2A bind to phosphorylated peptides through a cavity that is located as the methyl-binding cavity in Tudor/PWWP/MBT domains (Supplementary Fig. S3L). However, this cavity is positively charged in the case of KIF2C and KIF2A, whereas it is negatively charged in the case of Tudor/PWWP/MBT domains.

The N-terminal domain of KIF2C mediates the formation of KIF2C condensates in cells

Endogenous KIF2C forms foci in mitotic cells, and some of these foci colocalize with kinetochores / centromeres [51]. Moreover, its partner EB1, which is a microtubule plus-end regulator, assembles into condensates that contain KIF2C [26, 27]. KIF2C exhibits an N-terminal phosphorylated peptide-binding domain, a large disordered region and a coiled coil oligomerization domain. It is phosphorylated by various mitotic kinases such as CDK1, Aurora A, Aurora B and PLK1 [45, 52 , 53–55]. On these bases, we hypothesized that, during mitosis, KIF2C could self-assemble through phospho-dependent intermolecular interactions at the kinetochores / centromeres. To investigate this, we induced the expression of two KIF2C constructs with doxycycline in Flp-In T-Rex 293 cells: either KIF2C-mCherry or KIF2C-mCherry-Cry2 (further named Opto-KIF2C), in which KIF2C-mCherry is fused at its C-terminal to the light-responsive oligomerization domain of Arabidopsis thaliana cryptochrome 2 (Cry2) (Fig. 4A; Supplementary Fig. S4A) [56–58]. Such an optogenetic system enables the precise control of biomolecular condensates in space and time, and thus the characterization of molecular events associated with condensation. We observed that, in the absence of blue light, both KIF2C-mCherry and Opto-KIF2C form foci in these cells (Fig. 4B, line 1; Supplementary Fig. S4A; Supplementary Movies S1 and S2); moreover, after addition of nocodazole, the foci of Opto-KIF2C seem brighter and/or larger (Fig. 4B, line 2). These foci were too small to characterize their dynamics by fluorescence recovery after photobleaching (FRAP). However, using time-lapse 3D imaging, we found that foci formed by KIF2C-mCherry are able to fuse in living cells, supporting that they are membrane-less condensates (Supplementary Fig. S4B).

Figure 4.

KIF2C forms nuclear membrane-less condensates in cells. (A) KIF2C construct used for optogenetic experiments ⁴⁵. (B) Representative fluorescence images obtained after expression of Opto-KIF2C induced by doxycycline in different conditions. In lines 3 and 4, cells were illuminated with blue light (Light ON) during 5 min of light-dark cycles (4 s light followed by 10 s dark). In lines 2 and 4, nocodazole was added to the cell culture (Noc (+)). DNA was stained with Hoechst 33258. Scale bars: 5 μm. (C) SIM image obtained with the Opto-KIF2C construct on a mitotic cell in Light ON and Noc (-) conditions, and quantification of the foci area from a set of SIM images acquired after a 5-min illumination. The total number of analyzed nuclei was: 147 (n = 2; non-mitotic cells) and 215 (n = 2; mitotic cells) in Light ON Noc (-) conditions and 256 (n = 2) in Light ON Noc (+) conditions. The P-value calculated between non-mitotic and mitotic cells in Noc (-) conditions is 0.4270, whereas the P-value calculated for mitotic cells in the absence (Noc (-)) and presence (Noc (+)) of nocodazole is lower than 0.0001. More representative images are shown in Supplementary Fig. S4H. Scale bars: 5 μm. (D) Time-lapse microscopy images recorded after addition of nocodazole (Noc (+)) and activation of Opto-KIF2C foci (light ON) in living cells. Time 0: after a first cycle of 15 s light and 30 s dark. Times 5 and 20 (in min): after further illumination by cycles of 8 s light and 30 s dark. Image on the right: merged fluorescence and DIC images acquired after 20 min. Scale bar: 10 μm. (E) Line scan showing the increase in size and fluorescence intensity of a selected condensate after 20 min of illumination. (F) Violin plot quantification of the condensate area during blue light activation. (G) Merged fluorescence and DIC images of a representative cell analyzed by FRAP. Scale bar: 5 μm. (H) Quantification of the FRAP data from six different experiments. The red curve represents the mean recovery in those experiments. (I) Time-lapse 3D microscopy images of activated Opto-KIF2C condensates (Light ON) in mitosis (left) and interphase (right). See also Supplementary Movie S3. Scale bars: 1 μm. Time in seconds.

To further characterize these foci, we stimulated condensation by illuminating cells that express Opto-KIF2C in a time-controlled manner (Supplementary Fig. S4C). Upon exposure to an array of blue light LEDs during 5 min of light-dark cycles (4 s light followed by 10 s dark), the number of KIF2C foci per nucleus increased significantly (Supplementary Fig. S4D), and the foci seemed even brighter and/or larger (Fig. 4B, lines 3–4). We quantified their fluorescence intensity and observed that the mean fluorescence intensity of the foci increased with illumination (Supplementary Fig. S4E), while the intensity of the free protein in the nucleus decreased (Supplementary Fig. S4F), suggesting that the free KIF2C protein is recruited to the foci upon illumination. The resulting foci were significantly brighter than those detected in cells expressing KIF2C-mCherry (Supplementary Fig. S4G). Using structured illumination microscopy (SIM), we measured the dimensions of the Opto-KIF2C foci in both non-mitotic and mitotic cells after exposure for 5 min of 488 nm light, and found an average foci area of about 0.1 μm². The addition of nocodazole further increased the area of the condensates significantly, reaching about 0.15 μm² (Fig. 4C; Supplementary Fig. S4H).

We then performed fluorescence recovery after photobleaching (FRAP) measurements just after illumination. To properly detect fluorescence recovery, we added nocodazole, illuminated the living cells for one cycle of 15 s light followed by 30 s dark, and then exposed a plane of about 0.5 μm thick showing a large number of foci with light-dark cycles of 8 s light followed by 30 s dark. We observed that, upon blue light exposure, the size and fluorescence signal of KIF2C nuclear foci are enhanced in this plane (Fig. 4D and E). Several foci reached an area of about 1 μm² after 20 min of 8 s / 30 s cycling (Fig. 4F). In these conditions, we could perform a FRAP analysis of their dynamics by illuminating half of one of the foci and measuring the fluorescence recovery of the protein within it (Fig. 4G and H). Although we observed variability between points, mostly due to the different sizes of the foci and the difficulty to bleach part of such highly mobile objects, we found that fluorescence in these foci recovers at 50% in average with a time constant of about 300 s, which could correspond to the recovery time of a membrane-less condensate in cells. We verified that this recovery was not due to condensate disassembly: indeed, the size and fluorescence intensity of non-bleached foci only weakly decreased during the FRAP experiment (Supplementary Fig. S4I). We further studied the dynamics of these condensates by searching for foci fusion events in movies of interphase and mitotic living cells recorded in the absence of nocodazole. As with the analysis of KIF2C-mCherry foci, we performed 3D imaging to distinguish between a fusion and a superposition of condensates. We observed that Opto-KIF2C condensates fused and continued to grow from surrounding molecules (Fig. 4I; Supplementary Movie S3). The condensates resulting from fusion exhibited a volume equal to the sum of the volumes of the two initial condensates and regained a circular shape. The rapidly evolving morphology of these condensates confirmed that they are membrane-less cellular sub-compartments.

To characterize the molecular mechanisms of KIF2C condensate assembly, we designed 5 mutants: (i) a variant deleted from its N-terminal domain (Opto-KIF2C_ΔNt), (ii) K52E/K54E, which lost its phospho-binding capacity (Opto-KIF2C_K52E/K54E), (iii) S715A, which lost a mitotic PLK1 phosphorylation site (Opto-KIF2C_S715A), (iv) S715E (Opto-KIF2C_S715E), which mimics phosphorylation of S715 that disrupts a KIF2C intermolecular interaction (Supplementary Fig. S1C), and (v) G495A, which is catalytically inactive (Opto-KIF2C_G495A) (Fig. 5A). We tested whether these mutants formed condensates in mitotic cells upon blue light exposure: Opto-KIF2C_ΔNt and Opto-KIF2C_K52E/K54E completely lost their ability to form foci, whereas the other mutants could still assemble into condensates to different degrees (Fig. 5B and C; Supplementary Fig. S5A). In the presence of nocodazole, S715E strongly reduced the number of KIF2C condensates, whereas G495A had only a mild impact, and S715A had no impact at all. The same results were obtained in the absence of nocodazole, both for mitotic and non-mitotic cells, except that G495A significantly decreased the number of Opto-KIF2C foci when microtubules were present in the cells (Fig. 5B and C; Supplementary Fig. S5A–C). In parallel, we tested the impact of PLK1 activity on KIF2C foci formation. We observed that supplementation with a PLK1 inhibitor significantly reduced the number of KIF2C nuclear foci (Fig. 5D and E). Even the (small) number of foci formed by KIF2C S715E decreased upon addition of a PLK1 inhibitor (Supplementary Fig. S5D and E). Altogether, our results demonstrated that the lysines of the KIF2C N-terminal domain involved in phospho-peptide binding are key to the assembly of KIF2C condensates in cells. Phosphorylation by PLK1 is also necessary to establish these condensates. Mutation S715E, reported to disrupt a KIF2C intermolecular interaction, decreases KIF2C condensate formation.

Figure 5.

KIF2C assembly into condensates depends on its phosphorylated peptide binding capacity as well as the activity of the mitotic kinase PLK1. (A) Constructs used for optogenetic experiments, containing KIF2C either WT, deleted from its N-terminal domain, or with point mutations: K52E/K54E, G495A, S715A, and S715E. (B) Representative fluorescence images obtained after induction by doxycycline of the expression of Opto-KIF2C mutants and illumination of the cells with blue light (light ON). In the two upper lines, nocodazole was added to the cells (Noc (+)) to observe mostly mitotic cells. In the two lower lines, mitotic cells were analyzed in Noc (-) conditions. Scale bars: m. (C) Quantification of the number of foci per cell as measured on images recorded as in panel (B). The total number of analyzed nuclei was: in Noc (+) conditions, WT - 86 (n = 3); and K52EK54E (n = 2); G495A-51 (n = 2); S715A-88 (n = 3); S715E-86 (n = 3) (P-values: S715E- lower than 0.0001, G495A-0.0394 and S715A-0.6132); in Noc (-) conditions : WT – 54 (n = 3); and K52EK54E (n = 2); G495A-37 (n = 2); S715A-49 (n = 3); S715E-47 (n = 3) (P-values: G495A and S715E-lower than 0.0001, S715A0.5219). (D) Impact of PLK1 inhibition on Opto-KIF2C WT foci formation (light ON Noc (-)). (E) Quantification of the number of foci per cell as measured on the images recorded as in panel (D). The total number of analyzed nuclei was 131 (n = 3) for the control and 61 (n = 2) with the PLK1 inhibitor (P-value lower than 0.0001).

KIF2C condensates contain BRCA2 phosphorylated at T207 but are next to microtubules and kinetochores

To further describe KIF2C condensates, we searched for KIF2C partners recruited within these condensates. As we revealed that, in vitro, KIF2C binds to BRCA2-pT207, we first verified that KIF2C binds to BRCA2-pT207 in mitotic cells. We performed GFP-trap pulldowns using DLD1 BRCA2 deficient cells stably expressing GFP-MBP-BRCA2. We observed that GFP-MBP-BRCA2 co-immunoprecipitated with endogenous KIF2C in mitosis (Fig. 6A). Using these same DLD1 cells expressing GFP-MBP-BRCA2, we further showed that KIF2C and BRCA2-pT207 co-localized at the kinetochores in metaphase chromosome spreads (Fig. 6B and C; Supplementary Fig. S6). In cells expressing GFP-MBP-BRCA2 T207A, we observed that KIF2C also colocalized with CREST (Fig. 6D and E), suggesting that the localization of KIF2C at the kinetochore does not depend on BRCA2-pT207. In Flp-In T-Rex 293 cells, we found that Opto-KIF2C and BRCA2-pT207 co-localized in condensates assembled upon blue light exposure (Fig. 6F and G). We noticed that Opto-KIF2C S715A completely lost its ability to co-localize with BRCA2-pT207. It was previously reported that KIF2C S715 is phosphorylated by PLK1 in nocodazole conditions [59]. We concluded that the phosphorylation of S715 might be important for BRCA2-pT207 to be localized within Opto-KIF2C condensates.

Figure 6.

KIF2C condensates concentrate phosphorylated BRCA2. (A) GFP-trap pull-down experiments performed in mitotic DLD1 BRCA2 deficient cells stably expressing GFP-MBP-BRCA2, showing that GFP-MBP-BRCA2 forms a complex with endogenous KIF2C. The parental DLD1 cell line (DLD1++) was used as a negative control. (B) Colocalization of KIF2C, BRCA2-pT207, and the centromere marker CREST observed on metaphase chromosome spreads from DLD1+/+ cells. (C) Quantification of the colocalization of KIF2C, BRCA2-pT207, and CREST in metaphase chromosome spreads, analyzed as shown in panel (B). The total number of analyzed centromeres were 278 for all conditions (n = 3). (D) Co-localization of KIF2C and CREST observed on metaphase chromosome spreads from DLD1+/+ cells expressing either BRCA2 WT or T207A. (E) Quantification of the co-localization of KIF2C and CREST in metaphase chromosome spreads, analyzed as shown in panel (D). The total number of analyzed centromeres was: BRCA2 WT- 744 (n = 3); BRCA2 T207A - 825 (n = 3) (P-value: 0.7205). (F) Representative immunofluorescence images of Opto-KIF2C WT or S715A, BRCA2-pT207, and DNA obtained in Light ON and either Noc (+) or Noc (-) conditions. Line scans on the right show co-localization. (G) Quantification, using Pearson's correlation coefficient, of the co-localization between the indicated proteins (Opto-KIF2C WT or S715A) and BRCA2-pT207, as observed in (F) after addition of nocodazole (Noc (+)). Median shown in red. The total number of analyzed nuclei was: KIF2C WT-99 (n = 3); KIF2C S715A-67 (n = 2). The P-value is lower than 0.0001.

We also tested if BubR1 phosphorylated at T680 (BuBR1-pT680), a known partner of BRCA2-pT207 at the kinetochore [28], was present within Opto-KIF2C condensates. Using SIM, we found that in Flp-In T-Rex 293 cells, both KIF2C and BuBR1-pT680 were located near the centromere markers CREST and CENPE (Fig. 7A and B; Supplementary Fig. S7A and B). Condensates of Opto-KIF2C WT, as well as S715A and S715E, were observed adjacent to foci of BubR1-pT680, in the outer region of the kinetochore.

Figure 7.

KIF2C condensates are found next to microtubules at the kinetochores in cells. (A) Representative SIM images of Opto-KIF2C, CREST, and DAPI in Light ON and Noc (-) conditions. The line scan on the right shows that KIF2C foci are next to centromeres, as marked by CREST. Scale bars: 5 μm. (B) Representative SIM images of Opto-KIF2C, BUBR1 phosphorylated on T680, and CENPE in Light ON and Noc (+) conditions. Scale bars: 5 μm. (C) Representative immunofluorescence images of Opto-KIF2C WT, tubulin, and DNA in Light ON conditions (in the presence (Noc (+)) or absence (Noc (-)) of nocodazole). Line scans on the right show that KIF2C foci are next to tubulin. (D) SIM images of Opto-KIF2C, tubulin, and DAPI in Light ON and Noc (-) conditions. See also Supplementary Movie S4.

We finally searched for tubulin in the condensates, because it is the best characterized KIF2C partner: it directly interacts with the KIF2C motor domain [46]. We found that Opto-KIF2C condensates did not co-localize with α-tubulin (Fig. 7C). Opto-KIF2C condensates were positioned adjacent to microtubules, as confirmed using SIM microscopy, and after addition of nocodazole, they excluded α-tubulin (Fig. 7D; Supplementary Movie S4; Supplementary Fig. S7C). Mutating S715 into either A or E did not change the positioning of condensates of Opto-KIF2C next to microtubules (Supplementary Fig. S7D). KIF2C-mCherry (devoid of Cry2) also formed condensates positioned on microtubules (Supplementary Fig. S4A). Thus, our data showed that KIF2C interacts with α-tubulin only at the periphery of the condensates.

Functional consequences of KIF2C phosphorylation

To investigate the function of KIF2C condensates, we searched for proteins that localized in the vicinity of KIF2C only after condensate assembly. To do so, we tagged the N-terminus of our Opto-KIF2C construct with TurboID, an optimized biotin ligase that executes promiscuous biotinylation of nearby proteins within minutes (Fig. 8A; [60]). We induced TurboID-Opto-KIF2C condensation using blue light in the presence of biotin in the cell culture medium and then purified biotinylated proteins with streptavidin-coated beads. We found that, upon 15 min of blue light exposure of nocodazole-arrested cells, PLK1, PLK1-pT210 and γ-tubulin were specifically enriched within KIF2C condensates (Fig. 8B; Supplementary Fig. S8A–D). Addition of a PLK1 inhibitor significantly altered their proximity with Opto-KIF2C (Fig. 8B; Supplementary Fig. S8A, C, and D). In contrast, BRCA2 and BuBR1-pT680 were identified in the vicinity of Opto-KIF2C, but condensation did not favor these proximities (BRCA2-pT207 was difficult to detect in these conditions) (Supplementary Fig. S8A, B, and D). Using immunofluorescence microscopy, we found that Opto-KIF2C and PLK1, as well as activated PLK1 (PLK1-pT210), form similar patterns in all conditions (Fig. 8C; Supplementary Fig. S9A). When quantifying their co-localizations in nocodazole conditions, we obtained a particularly high Pearson's correlation coefficient for Opto-KIF2C and PLK1-pT210 (Supplementary Fig. S9B and C). Following the increase in the number of Opto-KIF2C foci upon blue light exposure (Supplementary Fig. S4D), the numbers of PLK1 and PLK1-pT210 foci significantly increased upon illumination (Fig. 8D). We did not observe any strong impact of mutating KIF2C S715 on the co-localization of Opto-KIF2C with PLK1 and PLK1-pT210 (Supplementary Fig. S9B and C). In the case of S715E, where there are fewer Opto-KIF2C foci, we were still able to detect a small but significant decrease in Opto-KIF2C colocalization with PLK1-pT210.

Figure 8.

Activated PLK1 is enriched within KIF2C condensates. (A) KIF2C construct used for TurbolD experiments. (B) Identification of TurbolD-Opto-KIF2C partners in nocodazole-arrested cells by Western Blot. Partners of TurbolD-Opto-KIF2C were biotinylated in different doxycycline and light conditions and purified using streptavidin-coated beads. PLK1, PLK1-pT210 and tubulin were identified as enriched in KIF2C condensates. (C) Representative immunofluorescence images of Opto-KIF2C WT, PLK1 or PLK1-pT210 and DNA in nocodazole (Noc (+)) and either Light OFF or Light ON conditions. Scale bars: 5 μm. (D) Quantification of the number of PLK1 and PLK1-pT210 foci per cell as measured on the images recorded as in panel (C). The total number of observed nuclei for counting PLK1/PLK1-pT210 foci was 79/79 (n = 3) and 100/79 (n = 3) in Light OFF and ON conditions, respectively (P-values: PLK1-0.0098 and PLK1-pT210-0.0057). (E) Representative immunofluorescence images obtained as in (C) but upon addition of the Aurora B inhibitor ZM447439. Scale bars: 5 μm. (F) Quantification of the number of Opto-KIF2C foci observed in panels (C) and (E). The total number of analyzed nuclei was: Light OFF - ZM447439-80 (n = 3); Light OFF + ZM447439 (n = 3); Light ON - ZM447439-91 (n = 3); Light ON + ZM447439 - 92 (n = 3). The P-value between the distributions of foci numbers under Light OFF and Light ON conditions is lower than 0.0001, whereas the P-value between Light ON and Light ON + ZM447439 conditions is 0.6951. (G) Model for PLK1 regulation in KIF2C condensates. KIF2C condensation is Aurora-B and PLK1-dependent, and through an Aurora B-dependent mechanism, triggers local concentration of PLK1, as well as further phosphorylation of proteins involved in the control of kinetochore-microtubule attachment, including KIF2C-S715 and BRCA2-T207. As KIF2C condensates are next to microtubules, the KIF2C depolymerase activity, enhanced by phosphorylation of S715, might be spatially limited to the periphery of the condensates.

To elucidate the role of BRCA2-pT207 in the colocalization of KIF2C and PLK1, we detected endogenous KIF2C and PLK1 in mitotic DLD1 BRCA2 deficient cells expressing either BRCA2 WT or T207A. Quantification of their colocalization revealed that mutation of BRCA2 T207 into alanine does not perturb the colocalization of KIF2C and PLK1 (Supplementary Fig. S10A and B). As Aurora B regulates KIF2C at the mitotic centromere [53] and is required for PLK1 activation at kinetochores [61], we also checked if the Aurora B inhibitor ZM447439 affected KIF2C condensation and co-localization with PLK1. We observed that, in mitotic cells, the addition of the Aurora B inhibitor completely abolished the formation of KIF2C condensates formed without light (Fig. 8E), whereas it either did not change (Noc (+); Fig. 8E and F) or mildly decreased (Noc (-); Supplementary Fig. S9A and D) the number of optogenetic KIF2C condensates formed upon blue light exposure. Thus, the Aurora B activity is essential for KIF2C condensation, and artificially favoring KIF2C condensation bypasses the requirement for Aurora B activity. We also observed that, upon addition of ZM447439, PLK1 was less abundant in the vicinity of KIF2C condensates in both light OFF and ON conditions (Supplementary Fig. S9E). Using immunofluorescence, we were unable to detect PLK1 and PLK1-pT210 within Opto-KIF2C condensates in mitotic cells, nor in the absence (Supplementary Fig. S9A) or presence of nocodazole (Fig. 8E). Similarly, we could not detect BRCA2-pT207 in KIF2C condensates upon addition of the Aurora B inhibitor (Fig. 8E). We concluded that Aurora B not only facilitates KIF2C condensate formation, but also promotes the accumulation of PLK1, PLK1-pT210 and BRCA2-pT207 in these condensates (Fig. 8G).

Discussion

During mitosis, endogenous KIF2C accumulates in foci, some of which colocalize with kinetochores / centromeres [51]. The activity of KIF2C at the kinetochores is regulated by Aurora B and PLK1 through multiple feedback loops [62]. The kinase cascades converge on activation of PLK1, which phosphorylates KIF2C on S715, thus promoting its microtubule depolymerase activity; this event is essential to ensure the timely correction of aberrant kinetochore-microtubule attachments and a proper chromosome segregation [6, 7, 59]. Moreover, PLK1 phosphorylates BRCA2 at T207, which triggers direct binding between PLK1 and BRCA2, as well as the assembly of a complex between BRCA2, PLK1, PP2A and BUBR1 phosphorylated by PLK1 at T680 [28]. This complex plays a critical role in stabilizing kinetochore-microtubule attachments and promoting accurate chromosome alignment. Here, we revealed that KIF2C is able to form membrane-less organelles (Fig. 4) located at the kinetochore-microtubule junctions (Fig. 7). It exhibits an N-terminal phosphorylated peptide-binding domain (Figs 1–3), which is essential for the formation of KIF2C condensates (Fig. 5). Both Aurora B and PLK1 activities are required for KIF2C condensation (Figs 5 and 8). KIF2C directly binds to and colocalizes with BRCA2-pT207 (Figs 2, 3, and 6). Upon illumination, Opto-KIF2C specifically concentrates PLK1 within the condensates (Fig. 8). Aurora B inhibition prevents the colocalization of both BRCA2-pT207 and PLK1 in KIF2C condensates. Our findings suggest that KIF2C condensation enhances PLK1 activity at the kinetochores.

KIF2C contains a newly described phosphorylated peptide binding domain

While searching for new partners of phosphorylated BRCA2 in mitosis, we found out that the microtubule depolymerases KIF2C and KIF2A exhibit an N-terminal domain binding to BRCA2 phosphorylated at T207 (Fig. 2). AlphaFold 2 predicts that this domain has the same barrel-like fold as Tudor/PWWP/MBT domains, which bind to methylated peptides [48–50]. All these domains interact with their targets through the same cavity. However, this cavity is positively charged when binding to phosphorylated peptides and negatively charged when binding to methylated peptides. AlphaFold 2 also predicts that the KIF2C (or KIF2A) β-barrel forms an antiparallel β-sheet with the BRCA2 peptide, the side chains of K52 and K54 of KIF2C (K50 and K52 of KIF2A) being close to that of T207 (Fig. 3). These models suggest that one of these lysines interacts with phosphorylated T207 through a salt-bridge. We verified that AlphaFold 3, which can take into account phosphorylation at T207, provides similar models. Supporting these models, we found that mutating K52 and K54 completely abolishes both KIF2C binding to phosphorylated BRCA2 in vitro and KIF2C condensate assembly in cells. Our data indicate that binding of the N-terminal domain of KIF2C to either its own phosphorylated motifs [63] or the phosphorylated motifs present in its partners is essential for condensate formation.

KIF2C forms membrane-less organelles through interactions regulated by Aurora B and PLK1 kinases

Endogenous KIF2C forms foci in mitosis [51]. Here, we observed that KIF2C-mCherry and Opto-KIF2C form condensates in cell nuclei even without activation of our optogenetic system (Fig. 4 A and B; Supplementary Fig. S4A and B). However, these condensates are so small that they were difficult to observe by immunofluorescence. They became significantly brighter and/or larger upon addition of nocodazole, as well as upon illumination in the case of Opto-KIF2C. We quantified the impact of illumination on the fluorescence intensity and size of the Opto-KIF2C condensates. We confirmed that light exposure triggered further recruitment of surrounding KIF2C molecules into the condensates that got slightly brighter (Supplementary Fig. S4C and E) and larger (Fig. 4F). We further characterized the dynamic properties within the Opto-KIF2C condensates using FRAP: the fluorescence recovery time was about 300 s, as also reported for other membrane-less organelles in cells [64].

By illuminating cells, we could show that the mechanism of KIF2C condensation and the content of KIF2C condensates are similar in both the absence and presence of nocodazole (Fig. 6B and C). The kinase activities of Aurora B and PLK1 contribute to KIF2C condensation, as inhibition of these kinases significantly decreased the assembly of KIF2C foci (Figs 5D and E and 8E and F). It was previously reported that KIF2C is phosphorylated by Aurora B and PLK1 at several sites, including S715 [59]. We identified that S715 is involved in the condensation mechanism, but its phosphorylation might decrease condensation. Indeed, when KIF2C S715 was replaced by an alanine and thus could not be phosphorylated, condensation was efficient, but when this residue was replaced by the phosphomimetic glutamic acid, condensation was reduced (Fig. 5A–C; Supplementary Fig. S5A–C). Thus, our data showed that phosphorylation by Aurora B and PLK1 of either KIF2C at sites other than S715 or its partners promotes KIF2C condensation [63, 65].

KIF2C condensates might locally amplify the activity of the PLK1 kinase

We observed that KIF2C condensates colocalized with activated PLK1 (PLK1-pT210) and BRCA2 phosphorylated by PLK1 (BRCA2-pT207) (Figs 6F, 7B, and 8C). We also found that BRCA2-pT207 is present at the kinetochore of metaphase chromosomes, and verified that the associated IF signal is significantly (even if only partially) reduced in cells expressing BRCA2 T207A (Supplementary Fig. S6). This finding is consistent with our previous observation that BRCA2 is at the kinetochore of mitotic cells and is phosphorylated by PLK1 in mitosis [28]. The co-localization of KIF2C and BRCA2-pT207 in Flp-In T-Rex 293 cells is abolished when KIF2C is mutated at S715 (variant S715A; Fig. 6G) and after addition of the Aurora B inhibitor ZM447439 (Fig. 8C and E), two situations in which PLK1 phospho-sites are decreased. Supporting this colocalization, we demonstrated that BRCA2 is indeed in the KIF2C condensates in mitosis using our TurboID setup, the biotinylated proteins (proximal to KIF2C) being identified by combining affinity chromatography and a WB using an anti-BRCA2 antibody (Supplementary Fig. S8A and B).

Our biotin-proximity labeling and immunofluorescence approaches showed that KIF2C condensation triggers the recruitment of PLK1 (Fig. 8B). KIF2C might directly bind to PLK1, as previously suggested [45]. It might also recruit a partner that binds specifically to PLK1. Both KIF2C and PLK1 binds to BRCA2-pT207. NMR and X-ray crystallography analyses showed that they bind to the same residues of BRCA2. Thus, they compete for BRCA2-pT207 binding. As both KIF2C and BRCA2 are oligomeric, they might form a network of direct interactions together with PLK1. BRCA2-pT207 is not necessary for the recruitment of PLK1 in KIF2C foci (Supplementary Fig. S10A and B). Thus, BRCA2-pT207 might be a consequence of the presence of PLK1 in KIF2C foci. Strikingly, a lack of KIF2C phosphorylation at S715, as observed in the KIF2C S715A mutant, does not impair PLK1 recruitment at KIF2C foci (Supplementary Fig. S9B and C), but impairs co-localization with BRCA2 phosphorylated at T207 (Fig. 6F and G). We propose that condensates of KIF2C phosphorylated at S715 amplify the phosphorylation of KIF2C partners, including BRCA2, by the kinase PLK1.

During mitosis, Aurora B is responsible for KIF2C localization at the centromere via the phosphorylation of Sgo2 [53, 66]. We observed that, in absence of blue light, Aurora B is essential for KIF2C condensation (Fig. 8E and F). As Aurora B activates PLK1, we propose that this activation is necessary for KIF2C condensate formation. However, exposure to blue light bypasses the requirement for active Aurora B, promoting KIF2C condensation even in the presence of Aurora B inhibitor (Fig. 8E and F). As PLK1 is essential for light-induced condensate formation, we believe that optogenetic KIF2C condensates locally enhance PLK1 activity, thereby bypassing Aurora B activity. Finally, we showed that Aurora B not only facilitates KIF2C condensate formation, but also promotes the accumulation of PLK1, PLK1-pT210 and BRCA2-pT207 in these condensates (Fig. 8G).

KIF2C depolymerase activity is restrained to the periphery of the condensates

It was previously reported by others that, in prometaphase chromosomes, KIF2C and tubulin localizations are close but distinct [53], and KIF2C is enriched at the plus ends of polymerizing microtubules [67]. We consistently found that KIF2C foci are next to microtubules, but do not contain tubulin (Fig. 7C and D; Supplementary Movie S4). This is particularly clear after addition of nocodazole, because tubulin staining is mainly diffuse, and the few tubulin foci are all located at the surface of KIF2C foci. In the absence of nocodazole, KIF2C foci are frequently found at the extremities of microtubules. The depolymerase activity of KIF2C is thus restricted to the interface between condensates and microtubules. It was further observed that the localization of KIF2C is dynamic next to the centromeres, as tension develops across sister kinetochores [53, 61]. Our SIM images revealed that KIF2C condensates are located next to CREST- and CENPE-marked centromeres (Fig. 7A and B). Finally, we observed that the catalytic activity of KIF2C facilitates its condensation in the presence of microtubules (Fig. 5A–C; Supplementary Fig. S5A–C). Moreover, KIF2C condensation in turn might regulate its depolymerase activity. Indeed, KIF2C condensates contain PLK1 that phosphorylates KIF2C S715, and this phosphorylation increases KIF2C affinity for microtubules, thus enhancing its microtubule depolymerase activity [45, 59]. Altogether, our data suggest that, in agreement with previous reports [26, 27], KIF2C condensates form in the vicinity of both microtubules and kinetochores, and that the enzymatic activity of KIF2C is regulated by condensation: indeed, it can be enhanced by condensation, and it can only occur at the periphery of KIF2C condensates.

Conclusion

Understanding the temporal regulation of the key mitotic regulator PLK1 is essential, as PLK1 is an important oncogene in cancer initiation, progression, and drug resistance. The microtubule depolymerase KIF2C, which facilitates proper microtubule-kinetochore attachment, forms foci in mitosis. Using a recently published optogenetic system, we deciphered how KIF2C assembles into foci. We showed that KIF2C uses a novel N-terminal domain to bind to phospho-sites, including BRCA2 phosphorylated at T207 by PLK1. The KIF2C foci formed through these phospho-dependent interactions are membrane-less organelles that both fuse and grow from surrounding KIF2C molecules. These organelles concentrate PLK1 and co-localize with BRCA2 phosphorylated at T207, which suggests that they favor BRCA2 phosphorylation by PLK1. This, in turn, may enhance the partitioning of PLK1 within KIF2C condensates, as a BRCA2 peptide centered at T207 binds directly to the Polo-Box Domain of PLK1. Moreover, KIF2C depolymerase activity promotes KIF2C condensate formation, and phosphorylation of KIF2C by PLK1 enhances its depolymerase activity. Our findings cannot support a direct causal relationship between KIF2C condensation and the local increase in the activities of the enzymes localized in KIF2C condensates. However, they strongly suggest that assembly of these condensates amplifies both PLK1 and KIF2C catalytic activities, thus controlling kinetochore-microtubule attachment and favoring proper chromosome segregation during mitosis.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

The authors declare no competing interests.

Funding

This work was supported by the CNRS and the CEA-Saclay, by the French Infrastructure for Integrated Structural Biology (https://frisbi.eu/, grant number ANR-10-INSB-05-01, Acronym FRISBI), by INSERM (PCSI 2022 grant BRCAPS coordinated by S.Z.J.), by ANR (grant number ANR-21-CEA13-0030 coordinated by Au.C.) and by ARC (ARC 2021 PJA3 grant N°248989 coordinated by S.Z.J.). A.S. was funded by the CEA, Synchrotron SOLEIL, and FRM (grant N°FDT202304016927). R.C. was supported by a PSL University Fellowship. J.C.C. was supported by a grant from Asociacion Española contra el Cancer (AECC) PRYGN234480CARR to Au.C. The Constantinou lab was supported by the French National Cancer Institute INCa (PLBIO 2021), by the French Agence Nationale de la Recherche ANR (AAPG2021 and AAPG2023), and by the Fondation MSD AVENIR. The present work has benefited from Imagerie‐Gif core facility supported by I’Agence Nationale de la Recherche (FBI ANR-24-INBS-0005 (BIOGEN); SPS ANR-17-EUR-0007, EUR SPS-GSR). Financial support from the IR INFRANALYTICS FR2054 for conducting the research is also gratefully acknowledged.

Data availability

All experimental data and 3D models are available upon request to the corresponding authors.