KIF2A stabilizes intercellular bridge microtubules to maintain mouse embryonic stem cell cytokinesis

Microtubules (MTs) are dynamic filaments that are part of the cytoskeleton and that are involved in many cellular processes, including cell division, differentiation, and migration (Goodson and Jonasson, 2018). MTs are hollow tubes made up of α- and β-tubulin heterodimers, which are arranged in a head-to-tail manner into protofilaments, 13 of which fold up to form an MT. This arrangement causes an intrinsic MT polarity, exposing β-tubulin at one end (called the plus-end) and α-tubulin at the other (the minus-end). Tubulin dimers assemble into MTs in a GTP-bound form, and GTP hydrolysis occurs once dimers are incorporated into the lattice. Isolated MTs exhibit growth and shrinkage at both ends, as well as conversion from growth to shrinkage (catastrophe) or from shrinkage to growth (rescue). The rate of growth is higher at the plus-end resulting in a “GTP cap” arising from the excess of unhydrolyzed GTP. It has been proposed that GTP hydrolysis in MTs is accompanied by changes in the longitudinal distance between dimers in a protofilament (interdimer distance), with GDP-MT lattices displaying a compacted state and MTs made up of the slowly hydrolysable GTP analog GMPCPP, which mimics the GTP state at growing MT ends, having an expanded conformation (Zhang et al., 2018). These MT conformations may attract different sets of MT-associated proteins (MAPs), and conversely, MAPs may affect the compacted or expanded state, and thereby alter MT behavior (Shen and Ori-McKenney, 2024; Siahaan et al., 2022).

The behavior of MTs is modulated not only by MAPs, but also by posttranslational modifications (PTMs). Acetylation at lysine 40 (K40) of alpha-tubulin is a long-known PTM that occurs in the lumen of the MT (Janke and Montagnac, 2017). MT acetylation levels increase as MTs age, and acetylation appears to protect MTs from mechanical wear-down and breakage (Portran et al., 2017). In order to acetylate MTs, αTAT1 (the enzyme that acetylates alpha-tubulin) has to enter the lumen of MTs to access the luminal K40. Interestingly, αTAT1 has been shown to act preferentially on MTs with an expanded MT lattice (Shen and Ori-McKenney, 2024).

MAPs are categorized into subclasses based on the MT localization pattern or on structural or functional similarity. Examples of the first category are the plus-end tracking proteins (+TIPs), which accumulate at the plus-ends of growing MTs (Galjart, 2010), and the minus-end targeting proteins (−TIPs), which accumulate at MT minus-ends (Akhmanova and Steinmetz, 2019). An example of the second category is the kinesin superfamily, consisting of MAPs with an ATPase domain, which is generally used as a motor to propel the proteins over MTs in a directional manner, resulting in transport of cargo associated with the kinesin (Hirokawa et al., 2009). Kinesin-13 family members have their ATPase domain in the middle of the protein instead of within the N- or C-terminal segments, and are atypical as they do not use ATP hydrolysis for motility, but to depolymerize MTs (Desai et al., 1999). There are four kinesin-13 proteins in mammals, i.e., KIF2A, KIF2B, KIF2C (or MCAK), and KIF24 (Walczak et al., 2013).

The mitotic spindle is an MT-based structure essential for chromosome segregation (Glotzer, 2009). It consists of two spindle poles from which various types of MTs emanate, including astral MTs, which contact the plasma membrane, kinetochore MTs, which connect the chromosomes to the spindle poles, and interpolar MTs, which connect the spindle poles themselves. Maintenance of spindle size is critical for chromosome congression and is controlled by MT flux, a continuous poleward movement of kinetochore MTs (Barisic et al., 2021). It has been shown that KIF2A depolymerizes MT minus-ends near spindle poles and plays an important role in regulating poleward flux (Rogers et al., 2004; Sun et al., 2021). During anaphase, when chromosomes are pulled apart, the network of overlapping interpolar MTs rearranges to form central spindle or midzone MTs between the segregated chromosomes (Douglas and Mishima, 2010). Subsequently, during telophase, daughter nuclei are formed. Cytokinesis, the final stage of cell division, divides the contents of a cell into two daughter cells (Green et al., 2012). Cytokinesis is characterized by constriction of the cell membrane and formation of an intercellular bridge (ICB), which contains the network of midzone MTs (henceforth termed ICB MTs) and the midbody or Flemming body, a dense proteinaceous structure. Once chromosomes and intracellular material have been correctly divided over the daughter cells, abscission occurs, which physically separates the daughter cells. During abscission, adaptor proteins such as CEP55 recruit the Endosomal Sorting Complex Required for Transport (ESCRT)-III machinery, which guides membrane constriction and scission (Mierzwa and Gerlich, 2014).

Mouse embryonic stem cells (mESCs) are defined by two properties: pluripotency (the ability to give rise to all somatic lineages and the germline) and self-renewal (proliferation while preserving pluripotency) (Kinoshita and Smith, 2018; Martello and Smith, 2014). Whereas in the mouse embryo, pluripotent cells exist for approximately 1 day, self-renewal is essential for pluripotency maintenance in cultured mESCs. Interestingly, cytokinesis takes several hours in mESCs, which is unusually long and has been coupled to pluripotency maintenance (Chaigne et al., 2020). The mechanisms that control cytokinesis duration remain largely undefined.

Here, we investigate the function of KIF2A in pluripotent mESCs. We find that KIF2A has a dual role. Consistent with studies in differentiated cells (Rogers et al., 2004; Sun et al., 2021), KIF2A functions as a depolymerase during metaphase in mESCs. Surprisingly however, we find that KIF2A converts from a depolymerase to an MT-stabilizing factor during cytokinesis, where it is required for the maintenance of MTs at the ICB and for cytokinesis duration. We propose that KIF2A-mediated MT stabilization prolongs cytokinesis to maintain pluripotency of naïve mESCs.

To examine the cellular localization and function of KIF2A, we engineered various Kif2a knock-in mESC lines using CRISPR-Cas9-based genome editing. We inserted four consecutive tags just before the stop codon: eGFP, FKBP12^F36V (which is part of the dTAG-13 degradation system for rapid proteasomal depletion of target proteins [Nabet et al., 2018]), a cleavage site for TEV protease, and a Strep-tag II (StrepII) for affinity pull-downs (Fig. S1 A). We generated homozygous mESC lines producing endogenous KIF2A-eGFP-FKBP12^F36V-TEV-StrepII (abbreviated as KIF2A-GFTS). In addition, we generated Kif2a knockout (Kif2a^KO) mESCs, by deleting a large part of the Kif2a gene, as well as Kif2a^GTS knock-in mESC lines, containing the same tags as Kif2a^GFTS mESCs, but without FKBP12^F36V (Fig. S1 A). Western blot (WB) analysis demonstrated the presence of KIF2A-eGFP or KIF2A-GFTS fusion proteins of the right size, and the absence of KIF2A in homozygous Kif2a^KO mESCs (Fig. 1 A). We noted that the level of KIF2A-GFTS in Kif2a^GFTS mESCs was somewhat lower compared with that of KIF2A in wild-type (WT) mESCs (Fig. 1 A). However, treatment of mESCs with the proteasome inhibitor MG132 did not lead to changes in the level of KIF2A-GFTS (Fig. S1 B), indicating that there is no leaky proteasome-mediated degradation of FKBP12^F36V-containing KIF2A in the absence of dTAG-13. Furthermore, Kif2a mRNA levels in WT mESCs were comparable to Kif2a^GFTS mRNA levels (Fig. S1 C), indicating that the insertion of multiple tags in the Kif2a locus does not affect mRNA stability. We therefore hypothesize that the lower amount of KIF2A-GFTS in Kif2a^GFTS mESCs is due to a reduced translation efficiency.

To assess how fast KIF2A-GFTS is degraded in the presence of dTAG-13, we performed a time course depletion experiment. We could not detect KIF2A on a WB 2 h after the addition of dTAG-13 to the medium of mESCs (Fig. 1 B). We subsequently treated Kif2a^GFTS mESCs with dTAG-13 for 72 h. This did not lead to significant cell proliferation changes (Fig. S1 D), nor did we observe cell cycle defects (Fig. S1 E) or deviations in the mitotic index (Fig. S1 F). Thus, in contrast to the essential function of murine KIF2A at birth and during postnatal development (Homma et al., 2003, 2018; Ruiz-Reig et al., 2022), depletion of KIF2A in pluripotent mESCs is not detrimental.

To obtain an overview of KIF2A-GFTS localization in mESCs, we fixed cells and visualized KIF2A-GFTS together with other structures using specific antibodies (Fig. 1, C–E and Fig. S1 G). This immunofluorescence (IF) approach revealed that KIF2A-GFTS was fairly inconspicuous during interphase, although we did observe occasional MT staining. At the onset of mitosis (prophase), KIF2A-GFTS accumulated near spindle pole centrosomes and this localization was maintained in metaphase (Fig. 1 C). In anaphase and telophase, KIF2A-GFTS was still near centrosomes, but the signal was weaker (Fig. 1 C). During cytokinesis, KIF2A-GFTS became highly enriched at the ICB, as shown by costaining with antibodies against acetylated tubulin and the midbody marker MKLP1 (Zhu et al., 2005) (Fig. 1, C and D). The KIF2A-GFTS signal overlapped with anti-KIF2A antibody staining in non-treated mESCs (Fig. S1 G), and dTAG-13–mediated depletion of KIF2A-GFTS resulted in a disappearance of green fluorescence (Fig. 1 E), demonstrating that the signal observed in non-treated cells represents KIF2A-GFTS. Thus, in mESCs, KIF2A-GFTS first accumulates near the spindle poles close to MT minus-ends, while later a strong accumulation is observed on ICB MTs, where KIF2A appears to be bound along the length of the MTs, and hence on the lattice of these MTs.

To concurrently analyze the dynamic behavior of KIF2A-GFTS and MTs during cytokinesis in naïve mESCs, we added SiR-tubulin (Lukinavicius et al., 2014) to mESCs and used light sheet fluorescence microscopy (LSFM) for visualization, as this method is well suited for long-term fluorescence time-lapse imaging of relatively thick 3D samples, and also reduces phototoxicity and photobleaching as compared to confocal microscopy (Tomer et al., 2013). We acquired z-stacks of between 30 and 60 μm (z-step size of 1 μm) and imaged colonies every 10 min for 16 h.

We detected ICB MTs throughout mESC colonies (examples of a z-slice through, and 3D projection of, KIF2A-GFTS and SiR-tubulin localization in an mESC colony are shown in Video 1 and Video 2, respectively). In time-lapse experiments, we observed dynamic behavior of KIF2A-GFTS and MTs (Fig. 2 A and Video 3). The onset of cytokinesis, 20 min after the last metaphase image, was defined by an intensely fluorescent bundle of MTs of ∼5 μm in length (Fig. 2 A). During cytokinesis, the length of the ICB MT bundle first increased and then decreased (Fig. 2 A and Video 3). KIF2A-GFTS became visible at these ICB MTs after SiR-tubulin and accumulated gradually (Fig. 2 A and Video 3). In contrast, at the end of cytokinesis KIF2A-GFTS remained visible longer than SiR-tubulin (Fig. 2 A and Video 3). Quantification of cytokinesis duration using both fluorescent dyes revealed a longer duration with KIF2A-GFTS compared with SiR-tubulin (Fig. 2 B, duration with KIF2A-GFTS: 6.1 ± 2.2 h [367 ± 133 min], duration with SiR-tubulin: 4.9 ± 1.9 h [292 ± 195 min]).

We observed a considerable fluorescence decay of the SiR-tubulin signal during imaging (Fig. S2 A), presumably because SiR-tubulin, which is a Taxol-derivative (Lukinavicius et al., 2014), is actively pumped out of mESCs over the course of the imaging experiment. Despite this decrease in signal intensity, we were able to accurately measure ICB MTs throughout cytokinesis (see, e.g., Video 3). KIF2A-GFTS signal also decreased during the 16-h time lapse, but to a lesser extent (Fig. S2 B). In an attempt to enhance the MT signal in live-imaging experiments, we increased SiR-tubulin concentration. However, this led to a longer cytokinesis (Fig. 2 C). Thus, stabilizing MTs with SiR-tubulin lengthens cytokinesis duration.

The SiR-tubulin signal was always detected throughout the ICB, including the midbody, whereas KIF2A-GFTS eluded this central domain until the end of cytokinesis (Fig. 2, A and D). We furthermore observed asymmetric accumulation of KIF2A-GFTS and MTs on ICBs, with one of the future daughter mESCs having more signal than the other (Fig. 2, A, D, and E; and Video 3). Examination of fluorescence signal in individual z-sections showed that the asymmetry was not an artifact of maximum intensity projections (MIPs) (Fig. 2 E). Investigation of fluorescence intensities of SiR-tubulin and KIF2A-GFTS during the early phases of cytokinesis revealed a threefold increase in signal intensity of SiR-tubulin on spindle and ICB MTs over the cytoplasm, whereas KIF2A-GFTS fluorescence intensity on ICB MTs was approximately sevenfold higher than cytoplasmic fluorescence intensity (Fig. S2 C). These data indicate that most KIF2A-GFTS molecules are associated with ICB MTs and that the affinity of KIF2A for ICB MTs is high.

To analyze the dynamic behavior of KIF2A-GFTS in mESCs, we carried out fluorescence recovery after photobleaching (FRAP). We bleached KIF2A-GFTS both at metaphase spindles and on ICB MTs, and observed a faster initial recovery at metaphase spindles compared with ICB MTs (Fig. 2 F). At the end of the FRAP experiment, the fraction of fluorescent KIF2A-GFTS molecules on ICB MTs was ∼40% (Fig. 2 F). Given the low amount of KIF2A-GFTS present in the cytoplasm (Fig. S2 C), exchange of cytoplasmic with MT-bound KIF2A-GFTS is limited. To examine the mechanism underlying the partial recovery of KIF2A-GFTS on ICB MTs, we selected ICB MTs of similar length and KIF2A-GFTS intensity profile, and then analyzed fluorescence dynamics on the complete MT structure throughout the FRAP experiment. We observed no obvious fluorescence loss in the non-bleached “arm” (ICB MTs in one of the future daughter mESCs) in the first phase of the recovery (Fig. 2 G), indicating there is no diffusion of KIF2A-GFTS from one daughter mESC to the other. Instead, the limited recovery of KIF2A-GFTS that we observed on the bleached arm (Fig. 2, F, and G) appeared to be partly due to diffusion from the cytoplasmic side of the ICB MTs (Fig. S2 D). This diffusion coincided with a decrease in KIF2A fluorescence next to the bleached region (Fig. S2 D). These results indicate that KIF2A is tightly bound to the lattice of ICB MTs but can undergo lateral movement within the bundle of MTs. Despite its relatively immobile behavior in the time window of the FRAP studies (∼1 min), KIF2A-GFTS displays dynamic behavior within the time frame of the LSFM-based experiments (∼10 min).

ICB MT bundles of ∼5 μm in length first appear 20 min after metaphase, increase further in length in the first hours of cytokinesis, and subsequently decrease in length and disappear (Fig. 2). To better understand ICB MT organization in mESCs, we examined the localization and dynamic behavior of several MAPs. We first analyzed PRC1, which is a key regulator of cytokinesis that cross-links antiparallel MTs at the midzone at the onset of cytokinesis (Asthana et al., 2021). We generated a mScarlet-PRC1 knock-in (Prc1^mS) within the Kif2a^GFTS mESC line (Fig. S2, E and F) and performed dual-color time-lapse imaging using LSFM. We observed a sudden mScarlet-PRC1 accumulation at the midbody region at the onset of cytokinesis (Fig. 3, A and B). This behavior is in stark contrast to other cell types, where PRC1 first enriches on anaphase MTs and subsequently moves to the midbody (Asthana et al., 2021). The abrupt localization of mScarlet-PRC1 at the midbody upon the start of cytokinesis coincided with the appearance of the thick 5-μm-long MT bundle (Fig. 2), indicating that in mESCs, the region of antiparallel midzone MTs is quite narrow from the beginning of cytokinesis. mScarlet-PRC1 remained associated with the midbody throughout cytokinesis, with its fluorescence intensity declining in time (Fig. 3 B). The mScarlet-PRC1–positive region at the onset of cytokinesis did not contain KIF2A-GFTS (Fig. 3 C), consistent with our observations that KIF2A largely avoids the midbody at this stage (Fig. 2 and Video 3).

We examined ICB MT organization in mESCs further using a transgenic line expressing EB3-GFP, a marker for the ends of growing MTs (Stepanova et al., 2003), and a knock-in line expressing GFP-CLASP2, an MT growth-promoting +TIP (Yu et al., 2016). In fixed cells, we observed EB3-GFP in comet-like structures, representing the ends of growing MTs (Fig. 3, D and E). Whereas a bright EB3-GFP signal was detected on individual MTs in the cytoplasm, in most of the ICB MT bundles, no EB3-GFP was present (Fig. 3 D). However, a subset of ICB MTs did contain fluorescence, with EB3-GFP localizing at the ends of the bundles (Fig. 3 E).

To further analyze dynamic EB3-GFP behavior in relation to ICB MTs, we concurrently imaged SiR-tubulin and EB3-GFP in live mESCs, this time using confocal microscopy for fast time-lapse acquisitions. Consistent with results in fixed mESCs, we observed EB3-GFP comets (Fig. 3 F), which moved in time (Video 4), suggesting that growth of individual MTs occurs throughout mESCs. While the majority of ICB MT bundles in the live-imaging experiments were EB3-GFP–negative (see an example in Fig. 3 G, upper panel), a subset did contain EB3-GFP signal, which was most often present at the ends of the ICB MTs (see an example in Fig. 3 G, lower panel). On rare occasions, EB3-GFP signal within ICB MTs was detected (see an example in Fig. 3 F). Notably, we did not observe EB3-GFP comets moving through ICB MTs (Video 4), nor did we detect comets originating from ICB MT ends, even when EB3-GFP signal was high (Fig. 3 G and Video 4). These results suggest that the EB3-GFP–positive domains at the ends of ICB MTs are distinct entities. The EB3-GFP–positive ICB MT bundles were thicker in their midbody region compared with bundles lacking EB3-GFP (see an example in Fig. 3 G). Quantification of ICB MT length revealed that EB3-GFP–positive bundles were smaller than ICB MT bundles lacking EB3-GFP (Fig. 3 H). Interestingly, the average length of the EB3-GFP–positive bundles was like that of ICB MTs at the onset of cytokinesis (Fig. 2). Taken together, our data suggest that EB3-GFP temporarily accumulates at the ends of ICB MT bundles as they are formed at the onset of cytokinesis.

We next looked at CLASP2 localization on ICB MTs. As described previously (Klaus et al., 2022), GFP-CLASP2 was mainly enriched at the Golgi of naïve mESCs (indicated by asterisks in Fig. S2 G). We also detected a relatively weak GFP-CLASP2 signal in the midbody region of ICB MTs (indicated by white arrows in Fig. S2 G), and occasionally, GFP-CLASP2 accumulated within the ICB MT bundle itself (indicated by yellow arrows in Fig. S2 G). Combined, our results suggest that ICB MTs are stable structures, whose longitudinal extension during the first phase of cytokinesis is accompanied by a persistent GFP-CLASP2 localization in the midbody region and a transient EB3-GFP localization at ICB ends.

We next examined the role of KIF2A in mESCs, first focusing on metaphase, and subsequently on cytokinesis. We used two models, i.e., WT versus Kif2a^KO (KO) mESCs, or Kif2a^GFTS mESCs that were either treated for 24 h with dTAG-13 or not treated.

With respect to metaphase, analysis of fixed mESCs revealed that KIF2A depletion leads to an increased pole-to-pole distance during metaphase in both KIF2A depletion models (Fig. S3 A). Examination of live mESCs using LSFM showed that dTAG-13–treated Kif2a^GFTS mESCs had a significant increase in the transition time from metaphase to the start of cytokinesis (Fig. S3 B). Thus, the absence of KIF2A increases spindle length and causes a mild metaphase transition delay, which is consistent with RNAi-based KIF2A knockdown results in other cell types (Gaetz and Kapoor, 2004; Jang et al., 2008; Rogers et al., 2004), and with the notion, supported by in vitro experiments (Henkin et al., 2023), that during metaphase, KIF2A localizes near MT minus-ends on spindle poles and acts as an MT depolymerase. Interestingly, when we added high levels of SiR-tubulin (200 nM) to dTAG-treated Kif2a^GFTS mESCs, ∼50% of observed spindles were multipolar, in contrast to non-treated Kif2a^GFTS mESCs (or mESCs treated with 20 nM dTAG) that had normal bipolar spindles (Fig. S3 C). These data reveal a synergistic effect; i.e., the addition of an MT stabilizer to cells that lack an MT depolymerase aggravates problems in metaphase. This reinforces the notion that KIF2A is an MT depolymerase during metaphase.

We subsequently analyzed the role of KIF2A in cytokinesis. Both Kif2a^KO mESCs (Fig. 4 A) and dTAG13-treated Kif2a^GFTS mESCs (Fig. 4 B) displayed a significantly shorter cytokinesis duration, compared with WT or non-treated Kif2a^GFTS mESCs, respectively. The addition of dTAG-13 to WT mESCs did not affect cytokinesis duration (Fig. 4 A), and the size of mESC colonies did not influence duration times (Fig. S3 D). The fact that two control mESC lines display similar cytokinesis duration times, whereas two KIF2A depletion models show shorter cytokinesis durations, strongly suggests that KIF2A regulates cytokinesis duration in pluripotent mESCs. Cytokinesis lasted 5.4 ± 1.1 h (324 ± 68 min) in WT mESCs and 3.8 ± 0.79 h (227 ± 48 min) in Kif2a^KO mESCs, whereas it lasted 4.9 ± 1.9 h (292 ± 115 min) in non-treated Kif2a^GFTS mESCs (same SiR-tubulin data as reported in Fig. 2 B), and 4.0 ± 2.1 h (238 ± 124 min) in dTAG-13–treated Kif2a^GFTS mESCs. Similar to control mESCs (see Fig. 2 C), we were able to lengthen cytokinesis duration in KIF2A-depleted mESCs by increasing SiR-tubulin concentration (Fig. 4 C). Thus, MT stabilization by SiR-tubulin rescues the cytokinesis phenotype of KIF2A-depleted mESCs, in line with the view that one MT stabilizer (SiR-tubulin) substitutes for another (KIF2A).

To analyze the dynamic behavior of ICB MTs and KIF2A-GFTS in more detail, we developed a visualization tool, which yielded time–length plots (Fig. 4, D and E). In these kymographs, ICB MTs in each frame of the time-lapse movie are represented by a line, and these are plotted in time with an imaging interval of 10 min. Fluorescence intensities are represented as heatmaps for each time point (Fig. 4, D and E; see Videos 5 and 6 for the corresponding time-lapse experiments). Consistent with the data presented in Fig. 2, the kymographs revealed ICB arm-specific accumulation of MTs and KIF2A-GFTS, with the two signals often colocalizing within an arm (see green arrows in Fig. 4 D), although we also detected independent accumulations (see yellow and blue arrows in Fig. 4 D; note that in this example, the accumulation of SiR-tubulin [yellow arrow] is followed 10 min later by an increase in KIF2A-GFTS signal [blue arrow], suggesting positional dependence).

By averaging ICB MT length across individual kymographs, we found that in non-treated Kif2a^GFTS mESCs, length increased by about 60% in the first hours of cytokinesis, followed by a gradual decrease afterward (Fig. 4 F). Thus, cytokinesis in pluripotent mESCs has two distinct phases, one where the net ICB MT length increases and one where the ICB MT length decreases. The ICB MT length increase was accompanied by an increase in KIF2A-GFTS signal in time; however, KIF2A-GFTS accumulation continued after the maximum ICB MT peak length was attained, and peaked later (Fig. 4, D and F). Strikingly, whereas at the start of cytokinesis the ICB MT length was similar in mESCs lacking KIF2A-GFTS and control mESCs, we did not observe a similar ICB MT length increase in KIF2A-depleted mESCs in the first phase of cytokinesis (Fig. 4, E and F). We next measured both the length of MT bundles in the ICB and the thickness of these bundles within the midbody region. We examined a fixed time point in control and KIF2A-depleted mESCs, i.e., 100 min after metaphase onset, when the effect of KIF2A depletion is the strongest. Consistent with the kymograph-based results (Fig. 4 F), ICB MT length was significantly decreased in KIF2A-depleted mESCs compared with controls (Fig. 4 G). In contrast, the thickness, or width, of ICB MT bundles was not different from control mESCs (Fig. 4 H). These results indicate that a similar number of MTs are present in the midbody, where MT plus-ends overlap, but that MTs in the arms of KIF2A-depleted mESCs are shorter. In agreement with these data, we observed a similar SiR-tubulin fluorescence intensity in the midbody region in dTAG-13–treated and non-treated mESCs at the onset of the time-lapse experiment (Fig. S3 E); however, we observed less SiR-tubulin throughout the time-lapse experiment in ICB MTs of dTAG-13–treated mESCs (Fig. S3 F). Collectively, our LSFM-based studies in live mESCs indicate that KIF2A localizes in a dynamic fashion along ICB MTs and serves to increase MT length and not to decrease it, contrary to its expected function as a depolymerase.

Our live-imaging data suggest that KIF2A prolongs cytokinesis by stabilizing ICB MTs. To strengthen these results, we analyzed MT mass in fixed mESCs stained with antibodies against alpha-tubulin. We used antigen retrieval to uncover the tubulin epitope in ICB MTs, since in fixed mESCs without antigen retrieval, we hardly observed ICB MT staining, in contrast to the prominent staining of other MT-based structures (Fig. 1 C). Despite antigen retrieval, we still did not detect tubulin signal inside the midbody of fixed mESCs (Fig. 5, A and B), even though MTs are present, as shown using the SiR-tubulin marker. Nevertheless, in cells lacking KIF2A the level of tubulin was clearly decreased in the arms of ICB MTs (Fig. 5, A and B). These results suggest that there are less MTs in the ICB of KIF2A-depleted mESCs compared with WT cells, consistent with our live-imaging data using SiR-tubulin. Taken together, our data strongly support the hypothesis that KIF2A stabilizes ICB MTs.

Since ICB MTs are tightly packed bundles that are constantly bending and moving inside an mESC colony (Video 7), we reasoned that they might undergo changes in interdimer distance, as well as mechanical, friction-induced damage, and we therefore tested the level of acetylated tubulin in ICB MTs. In contrast to total tubulin, we detected almost twice as much acetylated tubulin in KIF2A-depleted mESCs compared with control cells (Fig. 5, C and D). These results suggest that the presence of KIF2A at ICB MTs prevents excessive ICB MT acetylation in mESCs. In line with this view, we often observed that the distribution of acetylated tubulin on the ICB arms was lower adjacent to the midbody (MKLP1 peak), where KIF2A-GFTS accumulated more strongly (Fig. 1 D). Conversely, the acetylated tubulin signal was higher in distal regions of ICB MTs where KIF2A-GFTS accumulation was less pronounced (Fig. 1 D). In conclusion, KIF2A maintains ICB MT mass and its loss increases acetylation of ICB MTs.

Early in vitro experiments performed with a monomeric “minimal” KIF2A protein containing only the ATPase domain and the neighboring neck region showed that this minimal domain binds to two consecutive tubulins in a protofilament, with the neck of KIF2A binding to one tubulin dimer and the ATPase domain to the other (Trofimova et al., 2018). This minimal domain was shown to be sufficient to depolymerize MTs (Ogawa et al., 2017; Trofimova et al., 2018). These experiments also revealed that dissociation of KIF2A from MTs is ATP hydrolysis–dependent and accompanied by MT depolymerization. More recent in vitro MT reconstitution assays using purified full-length KIF2A revealed that KIF2A is an autonomous −TIP and depolymerase (Henkin et al., 2023).

We have shown that in mESCs, KIF2A alters its function and binding mode. To explain these cellular data, we examined the in vitro behavior of KIF2A. We produced and purified human KIF2A fused to eGFP (Fig. S4 A), confirmed the purity of the protein by mass spectrometric analysis (Fig. S4, B and C; and Table S1), and used GFP-KIF2A in in vitro MT reconstitution assays. We performed experiments both in the presence of ATP (i.e., active KIF2A), when GFP-KIF2A can go through cycles of ATP hydrolysis, and in the absence of ATP (i.e., inactive KIF2A), when KIF2A cannot cycle. At 3 nM GFP-KIF2A + ATP, we observed intermittent accumulation of GFP-KIF2A at MT minus-ends (which were distinguished because they grow slower than the plus-end) and weak accumulation on the MT lattice (Fig. 6, A and B; and Video 8). At 6 nM + ATP, we observed a continuous accumulation of GFP-KIF2A at MT minus-ends (Fig. 6, C and D). These data are similar to a recent publication (Henkin et al., 2023) and show that in the presence of ATP, GFP-KIF2A is an autonomous −TIP.

When we removed ATP from the reconstitution mix, we observed a strongly increased accumulation on the newly formed MT lattice as compared to experiments with ATP (Fig. 6, E and F). Importantly, KIF2A avoided binding to the GMPCPP-derived MT seed, and no accumulation on the minus-end was detected (Fig. 6, E and F). We conclude that in the presence of ATP, KIF2A is an autonomous −TIP, whereas in the absence of ATP, KIF2A prefers the MT lattice. Furthermore, our data suggest that KIF2A prefers compacted GDP-MTs over expanded GTP-MTs. To provide more evidence for a preference of KIF2A for the compacted MT state, we incubated purified GFP-KIF2A with MTs in the presence of 10 μM Taxol (paclitaxel, PTX), which increases the MT lattice spacing (Alushin et al., 2014), or with Tau, which stimulates formation of compacted MTs (Siahaan et al., 2022). We detected much less GFP-KIF2A on PTX-stabilized MTs and an increased accumulation on Tau-stabilized MTs (Fig. 7, A and B), validating our hypothesis. To examine whether KIF2A also displays preferential MT binding in cells, we treated KIF2A-GFTS–expressing mESCs with PTX for 6 h and subsequently examined KIF2A-GFTS localization. Strikingly, we observed reduced KIF2A accumulation on ICB MTs of PTX-treated mESCs (Fig. 7, C and D). Combined, these results indicate that KIF2A prefers the compacted MT lattice both in vitro and in cells.

Our in vitro studies showed that inactive KIF2A prefers the MT lattice over the minus-end, and this is similar to the binding of KIF2A to ICB MTs. We therefore examined whether KIF2A activity is also regulated on ICB MTs. The Aurora kinases A and B can phosphorylate and inactivate KIF2A (Jang et al., 2009; Knowlton et al., 2009). To test whether Aurora kinases control KIF2A function during cytokinesis in mESCs, we treated cells for 20 min with ZM447439, a selective Aurora kinase inhibitor (Ditchfield et al., 2003). We then fixed mESCs and stained them with antibodies against acetylated tubulin and MKLP1 to investigate ICB MTs. ZM447439 treatment of WT mESCs caused a significant shortening of ICB MTs as compared to non-treated cells (Fig. 7, E and F), indicating that inactivation of Aurora kinases shortens ICB MTs. Strikingly, ICB MTs of Kif2a^KO mESCs were not affected by the inhibitor (Fig. 7, E and F), suggesting that the shortening of ICB MTs observed in WT mESCs is largely due to the activation of the MT depolymerase activity of KIF2A. Similar to our results with WT and Kif2a^KO mESCs, ICB MTs of Kif2a^GFTS mESCs were also shorter after ZM447439 treatment, whereas ICB MTs of ZM447439- and dTAG-treated Kif2aGFTS mESCs were not affected (Fig. 7 G). These data strongly suggest that KIF2A is maintained in an inactive state at ICB MTs by one or both of the Aurora kinases and that this is essential for ICB MT length maintenance.

As cytokinesis duration has been linked to pluripotency exit (Chaigne et al., 2020), and KIF2A-depleted mESCs have a shorter duration, we set out to investigate whether KIF2A plays a role in pluripotency maintenance. mESCs are maintained in a naïve pluripotent state in defined medium containing the two inhibitors CHIR99021 and PD0325901 (referred to as 2i) and leukemia inhibitory factor (LIF) (Ying et al., 2008). We first performed a colony formation assay, an established test of pluripotency exit (Mulas et al., 2019), and found that Kif2a^KO mESCs were approximately half as efficient in reverting to pluripotency compared with WT mESCs after 24 h of 2i/LIF removal (Fig. 8, A and B). This was also true when exit was slowed down by the addition of IWP2, a Wnt inhibitor (ten Berge et al., 2011) (Fig. 8, A and B). These results suggest that KIF2A depletion causes a premature exit from pluripotency.

In a second experiment, we examined the naïve-to-primed pluripotency transition. We cultured WT and Kif2a^KO mESCs either in naïve medium or in primed medium containing FGF2 and activin A (Hayashi et al., 2011), and examined the localization of KLF4, a marker of naïve mESCs, and OCT6, a marker of primed mESCs (Mulas et al., 2019). In both WT and Kif2a^KO mESCs, KLF4 signal decreased and OCT6 signal increased upon culture in primed medium (Fig. S5 A). However, we detected significantly more OCT6-positive cells in primed KIF2A-depleted mESCs compared with WT (Fig. S5 B), consistent with the view that KIF2A depletion speeds up the transition from naïve to primed.

We next performed RNA sequencing (RNA-Seq) to compare the transcriptomes of WT and KIF2A-depleted mESCs in an unbiased manner (see Table S2 for normalized RNA-Seq reads). For this analysis, we used two independently edited Kif2a^KO lines (KO1 and KO2) and a WT mESC line. Principal component analysis showed clustering of the triplicates and segregation of WT and KIF2A-depleted samples (Fig. S5 C), attesting to the quality of the experiment. A differential gene expression analysis using DESeq2 yielded 123 upregulated and 111 downregulated mRNAs in KO mESCs as compared to WT (Fig. 8 C). We confirmed downregulation of two transcripts, Myc and Actb, which are expressed at low and high levels, respectively, by RT-PCR (Table S2). With the exception of Kif2a itself, the deregulation of the other mRNAs was mild (Fig. 8 C), in line with the relatively mild phenotype observed in KIF2A-depleted mESCs. Metascape analysis (Zhou et al., 2019) of the downregulated genes revealed several pathways and terms enriched in KO mESCs, including “Mechanisms associated with pluripotency,” as well as pathways related to RNA biology (Fig. 8 D). The latter is intriguing given the potential RNA-binding capacity of KIF2A in mESCs (Mallam et al., 2019).

To further analyze the transcriptome after KIF2A depletion, we performed gene set enrichment analysis (GSEA), which investigates the differential expression of complete sets of genes involved in specific pathways and/or processes (Subramanian et al., 2005). We first assembled a set of established pluripotency and primed factors (Acampora et al., 2013; Chen et al., 2008; Graf et al., 2017; Kalkan et al., 2017; Martello and Smith, 2014; Mulas et al., 2019; Soochit et al., 2021) (see Table S3 for a description of all genes present in this set). GSEA on this local “pluripotency” set revealed that down- but not upregulated genes contributed to the core enrichment (Fig. 8 E and Table S3). Interestingly, Rex1 (Zfp42), whose downregulation correlates with naïve pluripotency exit (Kalkan et al., 2017), was among the downregulated factors in the KO mESCs. Thus, unbiased analysis of the transcriptome through Metascape and GSEA shows that a number of pluripotency-related mRNAs are downregulated in Kif2a^KO mESCs, indicating that maintenance of a normal cytokinesis duration by KIF2A is required to keep the transcriptome robustly pluripotent. Interestingly, Otx2, which encodes a marker for primed mESCs (Acampora et al., 2013; Mulas et al., 2019), was upregulated in KIF2A-depleted mESCs (Fig. 8 E and Table S3), consistent with the view that these mESCs are biased toward pluripotency exit.

Finally, we performed GSEA on a locally assembled set of genes encoding proteins involved in centrosome regulation and MT nucleation, including all tubulin isotypes, the gamma-tubulin ring complex (γTuRC) and associating proteins (Bohler et al., 2021; Jackson, 2014; Yan et al., 2014), proteins involved in abscission (Carlton et al., 2012; Chaigne et al., 2020; Paine et al., 2023), and Cep170, an established KIF2A interaction partner (Zhang et al., 2019) (see Table S3 for a description of all genes present in this set). GSEA on this “centrosome and MT nucleation” set revealed deviations of the majority of the mRNAs in the two KO lines compared with WT. Notably, several mRNAs upregulated in Kif2a^KO mESCs contributed to the core enrichment (Fig. S5 D and Table S3). These included PPdcd6ip, or Alix, which regulates recruitment of ESCRT-III components to the ICB (Chaigne et al., 2020), as well as the protease Capn7 and its binding partner, the ESCRT-III protein Ist1, which function together to regulate abscission (Paine et al., 2023). Thus, cytokinesis duration in KIF2A-depleted mESCs might be shortened because of the deregulation of factors involved in MT homeostasis and abscission.

Here, we combined cellular and in vitro approaches to gain insight into the function of the kinesin-13 member KIF2A in naïve mESCs. We find that in metaphase, KIF2A localizes near spindle poles, and our observations in KIF2A-depleted mESCs are consistent with KIF2A acting as a minus-end MT depolymerase at metaphase spindle poles. However, during cytokinesis, the dynamic behavior of KIF2A changes, with KIF2A localizing over the entire length of ICB MTs, suggesting strong binding to the lattice. FRAP results support the view of a differential binding mode of KIF2A near centrosomes (minus-end binding) compared with ICB MTs (lattice binding). The dual binding behavior of KIF2A in mESCs translates into a dual function, with KIF2A unexpectedly converting from an MT depolymerase in metaphase to a net MT stabilizer during cytokinesis. Our data show that both active and inactive KIF2A have a function in mESCs. As an MT stabilizer, KIF2A prolongs the duration of cytokinesis, as does the Taxol-based derivative SiR-tubulin. Thus, two independent lines of evidence suggest that MT stabilization at the ICB controls cytokinesis duration.

Our in vitro studies suggest that in the presence of ATP, KIF2A is an autonomous −TIP. These results are similar to a recent report (Henkin et al., 2023). In contrast, in the absence of ATP, KIF2A prefers the MT lattice over the minus-end, indicating that by inactivating KIF2A, a switch in its MT binding mode is induced. We show that selective inhibition of the Aurora kinases with ZM447439 results in the shortening of ICB MTs in WT but not in KIF2A-depleted mESCs. These results indicate that KIF2A is locally inactivated on ICB MTs by Aurora kinases, where it serves to bind MT lattices and stabilize MTs. Furthermore, our results suggest that KIF2A, which acts downstream of the Aurora kinases to control ICB MT length in cytokinesis, is the major MT depolymerase that is inactivated by the Aurora kinases during cytokinesis. Of note, although the cellular experiments with the Aurora kinase inhibitor ZM447439 are convincing, the exact mechanism by which KIF2A is inactivated, and whether both kinases are involved or only one, remains to be investigated.

Our in vitro experiments also indicate that KIF2A avoids the GMPCPP seed. Since GMPCPP is a GTP analog that maintains the MT lattice in an expanded state (Zhang et al., 2018), these data suggest that KIF2A prefers the compacted MT lattice and not the expanded one. Reconstitution experiments with PTX, which promotes MT expansion (Alushin et al., 2014), or with Tau, which stimulates MT compaction (Siahaan et al., 2022), strengthen this hypothesis, as KIF2A binds MTs less efficiently in the presence of PTX, whereas the addition of Tau stimulates KIF2A binding. Since KIF2A binding to ICB MTs is also impaired when mESCs are cultured in the presence of PTX, we propose that KIF2A prefers the compacted state both in vitro and on ICB MTs in mESCs.

Using LSFM, we were able to acquire high-resolution time-lapse movies in mESC colonies over long periods of time. Imaging revealed that in naïve mESCs, cytokinesis lasts 5–6 h, which encompasses about one third of their total cell cycle time (Soochit et al., 2021). We found that the ICB MT arms of future daughter cells display asymmetric bouts of KIF2A-GFTS and SiR-tubulin, indicating that arms act independently. These asymmetric bouts might represent temporarily altered MT lattice behavior in one daughter cell, for example, in response to mechanical forces. Despite these bouts, our experiments show that ICB MTs are stable, long-lived structures. LSFM imaging furthermore revealed two phases in cytokinesis, an initial one wherein ICB MT length increases, and a second phase wherein length decreases. Concomitant with the ICB MT length increase, we observed an increase in KIF2A-GFTS accumulation on ICB MTs. Based on our FRAP results, we propose that KIF2A binds the lattice of ICB MTs with high affinity and thereby maintains ICB MT stability. This is particularly important for the first phase of cytokinesis, as this phase is most affected in KIF2A-depleted mESCs. KIF2A might act simply by preventing the binding of factors that negatively regulate MT stability, such as severases. In addition, tight binding of KIF2A to ICB MT lattices in mESCs could help to maintain the compacted state, which may also repel certain MAPs. This behavior is similar to that of Tau, a MAP with intrinsically disordered domains, which was shown to form cohesive structures around neuronal MTs, thereby altering interdimer spacing (Siahaan et al., 2022). αTAT1 acts preferentially on MTs with an expanded MT lattice (Shen and Ori-McKenney, 2024), and the presence of more acetylated tubulin in ICB MTs of KIF2A-depleted mESCs could indicate that in the absence of KIF2A ICB, MTs become more expanded and hence better substrates for αTAT1. The increased acetylation of ICB MTs in KIF2A-depleted mESCs might in turn serve as a backup mechanism for MT stabilization (Portran et al., 2017). This would explain why the cytokinesis phenotype of KIF2A-depleted mESCs is relatively mild.

ICB MTs undergo lengthening in the first few hours of cytokinesis. We detected CLASP2 in the midbody region, where, similar to its paralog CLASP1 (Liu et al., 2009; Mani et al., 2021), it may interact with PRC1 to control MT growth in the overlap zone. The occasional GFP-CLASP2 signal observed inside bundles would be consistent with CLASP2-mediated MT repair occurring within ICB MTs (Aher et al., 2020). We did not detect EB3-positive MT growth events along ICB MTs or in the midbody region, indicating that ICB MT lengthening in the first phase of cytokinesis does not require EB3, in contrast to the growth of individual interphase MTs. Interestingly, although we observed EB3-GFP–positive domains at the ends of a subset of ICB MT bundles, we did not detect comet-like structures emanating from these EB3 domains. In vitro reconstitution experiments have shown that EB proteins autonomously bind the minus-ends of growing MTs (Bieling et al., 2007; Maurer et al., 2011). One possibility therefore is that EB3-GFP accumulates at minus-ends of newly formed ICB MT bundles. As ICB MT bundles elongate and become thinner, the EB3-GFP signal disappears while KIF2A accumulates. Thus, despite their long lifetime the ICB MTs of naïve mESCs are different from the long-lived MT bridges observed in the early embryo in vivo, in which dynamic MT plus-ends were found to be present (Zenker et al., 2017).

We have shown that in naïve mESCs, KIF2A positively regulates ICB MTs, thereby prolonging cytokinesis and postponing abscission. Accordingly, KIF2A-depleted mESCs show premature pluripotency exit, coinciding with deregulated levels of mRNAs encoding pluripotency factors and proteins involved in MT homeostasis and abscission. Our data establish a link between MT stabilization in cytokinesis and a pluripotent mESC transcriptome. Recently, the midbody was shown to be involved in local mRNA translation and to harbor mRNAs encoding proteins involved in cytokinesis, cell fate, and pluripotency (Park et al., 2023). KIF23 was proposed to function in midbody RNA localization by binding mRNAs with a disordered domain and coupling RNAs to MTs (Park et al., 2023). Interestingly, KIF2A also possesses disordered domains and has been shown to bind to RNA in mESCs (Mallam et al., 2019). It is possible that KIF2A not only stabilizes ICB MTs, but is also involved in localizing mRNAs to ICB MTs, and that these mRNAs are deregulated in KIF2A-depleted mESCs because of defects in ICB MTs. Future efforts will be directed at examining a possible RNA-related function of KIF2A.

We used the following primary antibodies in this study (dilutions and applications are also listed): mouse anti-acetylated tubulin (T6793, IF staining, 1:1,000; Sigma-Aldrich), mouse anti-beta-tubulin (T8328, WB, 1:1,000; Sigma-Aldrich), rabbit anti-beta-tubulin (ab6046, IF, 1:200; Abcam), rat anti-alpha-tubulin (ab6160, WB, 1:4,000, IF 1:1,000; Abcam), mouse anti-gamma-tubulin (T6557, IF, 1:300; Sigma-Aldrich), rabbit anti-KIF2A (NB500-180, WB, 1:1,000; Novus Biologicals), rabbit anti-MKLP1 (ab174304, IF, 1:500; Abcam), rabbit anti-PRC1 (ab51248, WB, 1:1,000; Abcam), goat anti-KLF4 (AF3158, IF, 1:200; R&D), mouse anti-OCT6 (a kind gift from Derk ten Berge [Department of Cell Biology, Erasmus University Medical Center, Rotterdam, The Netherlands], IF, 1:40), mouse anti-GAPDH (KT186, WB, 1:2,000; Absea). We used the following secondary antibodies: goat anti-mouse IgG Alexa Fluor 594 (A11005; Invitrogen), goat anti-rabbit IgG Alexa Fluor 647 (A21244; Invitrogen), goat anti-mouse IgG Alexa Fluor 647 (A21236; Invitrogen), goat anti-rabbit IgG Alexa Fluor 594 (A11012; Invitrogen), donkey anti-mouse IgG Alexa Fluor 594 and donkey anti-goat IgG Alexa Fluor 488 (# 715-585-151 and # 705-545-147, respectively; both from Jackson ImmunoResearch), IRDye 680RD goat anti-mouse IgG (926-68070; LI-COR), IRDye 800RD goat anti-mouse IgG (926-32210; LI-COR), IRDye 680RD goat anti-rabbit IgG (926-68071; LI-COR), IRDye 800RD goat anti-rabbit IgG (926-32211; LI-COR). Secondary Alexa antibodies were used at 1:1,000, and secondary LI-COR antibodies were used at 1:15,000.

For RNA isolation, cells were collected and resuspended in 1 ml TRIzol (Sigma-Aldrich) and incubated for 5 min at 30°C. 200 μl phenol–chloroform (Sigma-Aldrich) was added, and samples were incubated on a shaker for 3 min at 30°C. Samples were subsequently centrifuged for 15 min at 12,000 rpm at 4°C in a tabletop centrifuge. The aqueous phase (top layer) was collected, and 250 μl 100% ethanol (Sigma-Aldrich) was added. Samples were then transferred to RNeasy spin columns (Qiagen) and processed according to the manufacturer’s instructions.

For qRT-PCR, cDNA was made from 1 to 3 µg RNA per sample using the Superscript IV reverse transcriptase kit and oligo d(t) primers according to the manufacturer’s instructions. cDNA was diluted twice prior to performing qRT-PCR using platinum Taq polymerase and SYBR Green (S9430; Sigma-Aldrich). Primer sequences are provided in Table S2.

SDS-PAGE was carried out according to standard procedures using the mini-PROTEAN system (Bio-Rad). After electrophoresis, gels were either fixed and stained with Coomassie Brilliant Blue R-250 or blotted onto a PVDF (Millipore) membrane through wet transfer for 2 h at 4°C. The membrane was then blocked with 5% skim milk (Sigma-Aldrich) in phosphate-buffered saline (PBS) and 0.1% Tween-20 (PBS-T), for 30 min, and incubated overnight (O/N) at 4°C with primary antibody. Membranes were washed three times with PBS-T and incubated for 45 min with secondary antibody. After three washes with PBS-T, membranes were imaged using the Odyssey CLx (LI-COR).

HEK293T cells were maintained in DMEM (Gibco) with 10% fetal bovine serum (Capricorn Scientific) and 1% penicillin/streptomycin (P/S; Sigma-Aldrich) at 37°C, 5% CO₂. HEK293T cells were transfected with X-tremeGENE HP DNA transfection agent (Roche) with a 1:2 DNA:reagent ratio and incubated for 24–48 h before continuing with other experiments.

Experiments with mESCs were performed in specified media, which contained either two inhibitors (2i), i.e., 1 μM PD0325901 (04-0006; Stemgent) and 3 μM CHIR99021 (4423-10; Tocris), as well as serum and LIF, or just serum and LIF. For next-generation sequencing experiments, mESCs in 2i/L were used. For fluorescence microscopy experiments, live mESCs were grown on 0.2% gelatin-coated plates or coverslips, or in chambers, in serum-free N2B27 medium (1:1 DMEM/F12-GlutaMAX [Gibco]:Neurobasal [Gibco] medium, 1x N-2 supplement [17502001; Gibco], 1x B27 minus vitamin A [12587010; Gibco], 5 μM β-mercaptoethanol, 12.5 ng/ml insulin [19278; Sigma-Aldrich], 1% P/S, 1000 U/ml LIF), supplemented with 2i. For IF and WB experiments, cells were plated on 0.2% gelatin-coated plates or coverslips in medium containing serum and LIF.

Transfection of mESCs for gene targeting experiments (see below) was performed with Lipofectamine 2000 (11668030; Thermo Fisher Scientific) with a DNA:reagent ratio of 1:4, at 37°C, 5% CO₂ on a shaker at 150 rpm to prevent cell attachment. After 15-min incubation, cells were plated on iMEFs. To deplete KIF2A-GFTS from cells, mESCs were treated with a concentration of 50 nM dTAG-13 (6605; Tocris), unless specified otherwise. The mESCs were treated with dTAG-13 for the indicated times.

For the pluripotency exit and colony formation assay (abbreviated as clonogenicity assay), 200,000 cells growing in naïve (N2B27 medium plus 2i/L) conditions were plated on a 6-well plate in exit medium. This was either N2B27 medium without 2i/L or N2B27 medium with LIF + 1 μM IWP2 (72124; StemCell Technologies), as indicated. After 24 h of exit, cells were trypsinized into a single-cell suspension, and replated on 24-well plates in triplicate at 300 cells per well in complete N2B27 medium containing 2i/L. As a positive control, cells were maintained in naïve conditions without exit, and replated like the rest. After 5 days, cells were fixed for 2 min in 4% paraformaldehyde (PFA) (pH 7.4), washed once with PBS, and then stained using the alkaline phosphatase staining kit from Millipore (scr004) according to the manufacturer’s instructions. Then, cells were washed once with Milli-Q water and air-dried in the dark O/N. The entire well was imaged using an Olympus dissecting microscope, and all alkaline phosphatase–positive colonies were counted. The number of colonies formed by naïve cells was set to 100%.

For the priming assay, mESCs were induced for priming with N2B27 medium supplemented with LIF, 20 ng/ml activin A (120-14e; PeproTech), 12 ng/ml bFGF (450-33; PeproTech), and 2 µM IWP2 (S7085; Selleck Chemicals). Cells were cultured for 36 h in this primed medium.

For the proliferation assay, 100,000 cells were plated in triplicate for each condition and cell line in 6-well dishes. Cell counting was performed after 24, 48, and 96 in triplicate. The cell suspension was diluted 1:1 with trypan blue, and counting was performed on the Countess automated cell counter (Thermo Fisher Scientific).

For the cell cycle analysis, 200,000–300,000 cells were plated in duplicate in a 6-well plate for each cell line. The next day, either dTAG-13 or DMSO was added. After 24 h of incubation, the medium and the cells were collected, washed, and resuspended in Mg²⁺ and Ca²⁺-free PBS. For fixation, ice-cold 100% ethanol was added dropwise while vortexing to a final concentration of 72.5% and incubated for 1–2 h at 4°C. Cells were equilibrated to room temperature (RT) and washed with 1% BSA in PBS. The cells were resuspended in staining buffer (PBS containing 1% BSA, 50 μg/ml RNase A, 50 μg/ml propidium iodide [P3566; Thermo Fisher Scientific]) and incubated at 37°C for 15 min. Cells were analyzed using BD Biosciences Fortessa Flow Cytometer and subsequently quantified in FlowJo. The FSC-A/SSC-A, FSC-W/FSC-H, and SSC-W/SSC-A were used to identify single cells. The propidium iodide signal was measured using the 561-nm yellow–green laser in combination with the 610/20 band-pass collection filter. The percentage of cells in the G1, S, or G2/M phase was normalized to WT for each phase in each experiment.

To generate knock-in mESC lines expressing various KIF2A fusion proteins or knockout mESCs in which a large part of the Kif2a gene was deleted, short guide RNAs (sgRNAs) were designed using the CHOPCHOP online tool (https://chopchop.cbu.uib.no/) to target the last exon of the Kif2a locus (knock-in lines and knockout line, 5′-ACG​CCA​ACT​TAG​AGG​GCT​CG-3′) or the second exon of Kif2a (knockout line only, 5′-AGG​CCG​AAT​ACA​CCA​AGC​AA-3′). sgRNAs were ligated into the pSpCas9(BB)-2A-Puro plasmid (PX459, #62988; Addgene). The pUC8 vector (Sigma-Aldrich) was used as a backbone for cloning of homology arms (∼500 base pairs) and inserts. Human FKBP12 (324 base pairs) was ordered as a gBlock (IDT DNA), and eGFP was amplified from a previously published plasmid (Yu et al., 2016). The F36V mutation of FKBP12 together with the TEV-linker-StrepII sequence was introduced by PCR. The Gibson assembly was used to ligate and clone all sequences. All plasmids were sequence-verified.

mESCs were transfected with PX459 and pUC8 plasmids as described above. 24 h after transfection, cells were selected with 1 μg/ml puromycin (Sigma-Aldrich) for 2 days. After recovery of 3–4 days, putative knock-in mESCs were sorted using flow cytometry and plated at low density on a 10-cm dish (putative knockout mESCs were not sorted but plated as such). Single colonies were picked and grown in 96-well plates. DNA was extracted from individual clones with lysis buffer (50 mM Tris-HCl, pH 8.0, 5 mM EDTA, pH 8.0, 0.5% SDS, 0.3 mg/ml Proteïnase K). PCR was performed with primers located outside the homology arms (sequences available upon request). After verification by PCR, we performed Sanger sequencing and WB analysis to validate correct homologous recombination.

The same strategy as described above was used to generate the PRC1 knock-in mESC lines. A sgRNA with the sequence 5′-GGT​CTC​GAT​TCT​GGA​CAG​CTT​GG-3′ was designed to target the first exon of the mouse Prc1 locus and ligated into the PX459 plasmid using Bbs1 overhangs. A pUC8 vector with about 500 base pairs of homology with the Prc1 locus and the 2xFLAG-mScarlet insert was cloned. Kif2a^GFTS mESCs were transfected with the PX459 and pUC8 plasmids as described above. Clones were verified by PCR, Sanger sequencing, and WB to validate correct homologous recombination.

For generating GFP-CLASP2 knock-in mESCs, the sgRNA 5′-CGC​AGA​AGT​ACT​CGG​CGC​CGC​GG-3′, targeting the ATG of the alpha-isoform of Clasp2, was used. Homology arms, corresponding to 700 bp upstream and downstream, respectively, of the sgRNA cut site, were amplified from genomic DNA of mESCs. The StrepII-TEV-GFP sequence was synthesized (Integrated DNA Technologies) and cloned into a pUC8 vector together with left and right homology arms using the Gibson assembly (NEB). For targeting, 600,000 ES cells were reverse-transfected with 1 μg of guide RNA and 2 μg of homology template vector. Cells were then plated on 10-cm dishes with iMEFs in ES medium with serum and LIF. 24 h after transfection, medium was refreshed and puromycin was added at 1 μg/ml for 48 h. Cells were sorted after 5 days for GFP expression using the FACSAria III system. After sorting, cells were seeded on iMEFs in 10-cm dishes at low density. After a week, single colonies were picked and maintained in 96-well plates, and screened by PCR and WB.

To generate a transgenic mESC line expressing EB3-GFP, the EB3-GFP construct (Stepanova et al., 2003) was cloned downstream of a chicken beta-actin (CAG) promoter, linearized with Pvu1 (NEB), and transfected into mouse ESCs cultured under serum + LIF conditions. Transfection was performed in 6-cm dishes using 7.5 μg of linearized DNA and 15 μg of FuGENE HD transfection reagent (Promega) according to the manufacturer’s instructions. 24 h later, cells were expanded into 3 × 10 cm dishes, and puromycin (1 μg/ml; Sigma-Aldrich) was added. Cells were kept under selection for 5–7 days before picking clones. Selected clones were expanded in medium containing serum and 2i/LIF and verified for EB3 expression using a confocal spinning disk microscope.

KIF2A (2,118 bp) was amplified from HEK293T cDNA, which was synthesized from HEK293T RNA using the SuperScript IV Transcriptase kit (18090050; Invitrogen). Primers used for the KIF2A amplification are available on request. eGFP was amplified from a previously published plasmid (Yu et al., 2016). The StrepII-TEV sequence (126 bp) was ordered as gBlock fragment from IDT DNA with overhangs for insertion into the pcDNA3 backbone and ligation to eGFP. The TEV sequence is followed by a linker sequence that translates to GGSGG. The complete StrepII-TEV-eGFP-KIF2A construct was cloned into pcDNA3 (Invitrogen) by the Gibson assembly (NEB). Constructs were verified after cloning by Sanger sequencing.

StrepII-TEV-eGFP-KIF2A was transfected in HEK293T as described above. After 48 h, cells were washed with cold PBS, collected, and lysed with lysis buffer (50 mM HEPES, pH 7.4, 300 mM NaCl, 0.5% Triton X-100 [Sigma-Aldrich], 1x cOmplete protease inhibitor cocktail [Roche]) for 20 min on ice. The lysate was incubated with Strep-Tactin Sepharose beads (Cytiva) for 1 h at 4°C and washed three times with lysis buffer without protease inhibitor at 4°C. Beads were resuspended in elution buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM MgCl₂, 1 mM EGTA, 1 mM DTT, 2.5 mM d-desthiobiotin [Merck], 0.05% Triton X-100) and rotated on a rotor at 4°C for 2 h. The eluate was collected and aliquoted in assay-sized aliquots, flash-frozen, and stored at −80°C. KIF2A concentration was determined with a bovine serum albumin (BSA; Sigma-Aldrich) standard curve on a Coomassie R-250–stained SDS-PAGE gel.

In vitro MT reconstitution assays were carried out according to previously published procedures (Bieling et al., 2007; Leslie and Galjart, 2013), with some modifications. Fluorescently labeled MT seeds were generated by mixing 62% porcine brain tubulin (T240B; Cytoskeleton), 16% X-rhodamine tubulin (TL620M-A; Cytoskeleton) or Hi-Lyte 647-tubulin (TL670M-A; Cytoskeleton), and 22% biotin-tubulin (T333PA; Cytoskeleton) with 0.7 mM GMPCPP (NU-405S; Jena Biosciences). Mixtures were incubated at 37°C for 40 min to allow polymerization. The mixture was afterward centrifuged for 10 min at 100,000 × g in a Beckman tabletop Airfuge, the supernatant was removed, and the pellet was gently resuspended in 30 μl warm MRB80 buffer (80 mM PIPES-KOH, pH 6.8, 1 mM EGTA, 4 mM MgCl₂). MT seed samples were divided into 1 μl aliquots and flash-frozen. Oxygen scavenger (OS) mix was prepared by dissolving 5 mg of catalase (c9322; Sigma-Aldrich), 10 mg of glucose oxidase (G7141; Sigma-Aldrich), and 15 mg of DTT in 500 μl of MRB80 buffer. The solution was then flash-frozen in 1 μl aliquots.

Chambers were prepared 1 day before the reconstitution assay. Glass slides (Epredia) and 18 × 18 mm coverslips (VWR) were treated O/N with 1 M KOH in 100% ethanol. Slides and coverslips were washed six times and stored in autoclaved water (Milli-Q). On the day of the assay, glass slides and coverslips were dried with filtered air and a coverslip was mounted onto the glass slide with double-sided tape to create a flow chamber with a width of 5 mm and a volume of 10 μl.

The flow chamber was functionalized with 1 chamber volume of PLL-PEG-biotin (SuSoS) for 5 min. The excess was washed away with 5 chamber volumes of warm MRB80 buffer. Two chamber volumes of NeutrAvidin (A2666, 1 mg/ml in PBS; Invitrogen) were then added for 3 min, and afterward, chambers were again washed with 5 chamber volumes of warm MRB80. Seeds were diluted in warm MRB80, flowed into the chamber, and incubated for 5 min. Unbound seeds were washed away with 2 chamber volumes of warm MRB80, and next 3 chamber volumes of 5 mg/ml κ-casein (C0406; Sigma-Aldrich) dissolved in MRB80 were added to block aspecific binding. A protein mixture of 20 μl that was centrifuged for 8 min at 100,000 × g to remove insoluble material was injected into the chamber, and the flow chamber was immediately imaged afterward. The protein mixture contained 15 μM porcine brain tubulin, 1.25 μM X-rhodamine tubulin, 0.5 mg/ml κ-casein, 1.25 mM GTP (Cytoskeleton), 0.15% (vol/vol) methylcellulose (Sigma-Aldrich), and 75 mM KCl (Sigma-Aldrich). For these experiments, 0.5 μl of the OS mix and 25 mM D-(+)-glucose solution in water (G8270; Sigma-Aldrich) were added to the reaction mix. ATP (BSA04001; Cytoskeleton) was added in a concentration of 0.5 mM. Purified Tau protein from bovine brain (#TA01; Cytoskeleton) was used at a final concentration of 0.9 µM in the reconstitution mix. PTX (#T1912; Sigma-Aldrich) was added at a final concentration of 10 µM.

MT in vitro reconstitution assays were imaged using a Nikon Ti-Eclipse inverted microscope with a total internal reflection fluorescence (TIRF) unit, equipped with a PLAN APO TIRF 100×/1.49 NA oil objective and a QuantEM 512C 512 × 512 pixel 16-bit camera (Photometrics). The chamber was heated to 30°C with a stage-top incubator and objective heating (Tokai Hit). Dual-color imaging was performed with a DV2 Beamsplitter Green/Red (MAG Biosystems). The 491- and 561-nm lasers were used to image eGFP and X-Rhodamine, respectively. A single frame of the 647-tubulin–labeled seed was acquired prior to the start of the time lapse with the 633-nm laser, also using the DV2 beamsplitter. Stream acquisition was used with 500-ms exposure with MetaMorph imaging software.

mESCs were grown on coverslips coated with 0.2% gelatin. dTAG-13 was added to the cells 24 h before fixation, unless otherwise indicated. ZM447439 (2458, 5 µM; Tocris) was added for 20 min, and Taxol (PTX [5 nM, #T1912; Sigma-Aldrich]) was added to cells for 6 h prior to fixation.

Cells were fixed either for 10 min with 100% methanol at −20°C, followed by 10 min with 4% PFA (Sigma-Aldrich) at RT, or for 10 min at RT with 4% PFA only, and then washed three times with PBS. Cells were then permeabilized with PBS containing 0.15% Triton X-100 (Sigma-Aldrich) for 10 min, washed with PBS, and incubated with blocking buffer (PBS, 0.1% Tween-20, 1% BSA [Sigma-Aldrich]) for 30–60 min on a shaker (100 rpm). Cells were subsequently incubated with the indicated antibodies in blocking buffer, either for 1 h at RT or O/N at 4°C. Cells were washed three times, for 10 min each time, in wash buffer (PBS, 0.1% Tween-20) and incubated with secondary antibodies for 1 h at RT. Cells were washed three times in wash buffer, once in 70% ethanol, and once in 100% ethanol, and finally air-dried. Cells were mounted on glass slides with ProLong Gold Antifade with DAPI (P36931; Invitrogen).

For antigen retrieval (alpha-tubulin IF), fixed coverslips were immersed in 10 mM sodium citrate buffer, pH 6.0, with 0.05% Tween-20, which had been heated to 95°C. Coverslips were left in this hot solution for 30–45 min until they had cooled to RT. Coverslips were washed twice in wash buffer, blocking solution was added, and the rest of the procedure was performed as described above.

Standard confocal imaging was performed using the Leica SP8 AOBS microscope equipped with a PLAN APO 63×/1.4 NA oil objective. DAPI, eGFP or Alexa 488, Alexa 594, and Alexa 647 were imaged using a 405-, 488-, 561-, and 633-nm laser, respectively. Detection was either by a photomultiplier tube or by a hybrid detector and a scCMOS camera (DFC9000 GTC DLS, 2,048 × 2,048 pixels; Leica). All images were acquired at a resolution of 1,024 × 1,024 pixels.

For imaging EB3-GFP–expressing mESCs, cells were seeded in 2i/L medium on µ-Dish 35-mm dishes (ibidi) coated with 0.2% gelatin. SiR-tubulin was added for 6 h at 40 nM and washed out with warm medium prior to imaging. Cells were imaged on a Leica SP8 AOBS microscope as described at 37°C, 5% CO_2. Images were acquired at 512 × 512 pixel resolution, with a z-step size of 700 nm, and 3–4 z-slices were acquired per time lapse at a frame rate of 2.2–3.2 s per frame.

FRAP experiments were largely performed as described previously (Soochit et al., 2021), using a Nikon Ti-Eclipse inverted microscope with a CSU-X1 (Yokogawa) spinning disk equipped with a PLAN APO TIRF 100×/1.49 NA oil objective and a QuantEM 512C 512 × 512 pixel 16-bit camera (Photometrics). Kif2a^GFTS mESCs were plated on 0.2% gelatin-coated coverslips the day before the FRAP experiment. Cells were heated to 37°C with a stage-top incubator and objective heating (Tokai Hit). Photobleaching was performed using the iLAS FRAP module (Roper Scientific) integrated into MetaMorph, with the 491-nm laser set at 100%. Region of interests (ROIs) were selected and bleached for 90 ms with 10 repetitions. Subsequent images were acquired with 35% laser power for 1 min, 2 frames per second with a 400-ms exposure. 11 images were acquired before bleaching.

For long-term imaging, 20,000 WT, Kif2a^GFTS, or Kif2a^KO mESCs were plated 1 day before imaging on a TruLive3D dish (ibidi, Bruker) that was plasma-cleaned for 3 min with Glow Plasma System (Glow Research) and coated with 0.2% gelatin for 15 min. On the day of the experiment, SiR-tubulin (Spirochrome) was added to the medium 6 h prior to imaging. Afterward, cells were washed once with PBS and regular medium without dye was added. The TruLive3D dish was inserted into the InVi SPIM sample holder and placed into the imaging chamber. The environmental control was set at 37°C, 5% CO₂, and 87% humidity, and lid heating was enabled. dTAG-13 treatment was performed for 24 h.

The InVi SPIM fluorescence light sheet microscope (Luxendo; Bruker) was equipped with a Nikon CFI Plan Fluor 10×/0.3 NA illumination water objective and a Nikon CFI Apo 25×/1.1 NA detection water objective. For imaging of KIF2A-GFTS, the 488-nm laser was set at 8% and combined with the 497- to 554-nm band-pass filter. SiR-tubulin was imaged with 10–15% 633-nm laser and the 656-nm long-pass filter. Beams were split with 560-nm dichroic switch. 3D images were acquired with the Bessel light sheet with a thickness of 40, a line mode of 200 pixels, and a magnification of 62.5×. Images were acquired every 10 min for a total of 16 h with a 90-ms exposure, 10-ms delay, and a z-step size of 1 μm with 2× high-speed sCMOS cameras (Orca Flash 4.0 V3 2,048 × 2,048 pixels; Hamamatsu). For the Kif2a^GFTS cells, all samples were imaged with the 488- and 633-nm laser, even when KIF2A-GFTS was depleted by dTAG-13 treatment. This way we could verify KIF2A-GFTS depletion in the dTAG-13–treated cells and keep experiments comparable.

Image analysis was performed on IF stainings in fixed mESCs, standard fluorescence–based microscopy experiments in live mESCs, LSFM-based experiments in live mESCs, and in vitro MT reconstitution experiments.

IF stainings in fixed mESCs were used to analyze fluorescence intensities of selected proteins, either directly (e.g., KIF2A-GFTS) or via antibody stainings, and to investigate various phenotypes in mESCs, including the distance between spindle poles in metaphase, the length of ICB MTs after treatment with selected compounds, and the mitotic index. Quantification of fluorescence intensities was performed on unprocessed MIP images. To calculate fluorescence intensities of KIF2A and either alpha-tubulin or acetylated tubulin at the ICB in WT or KIF2A-depleted mESCs, intensity profiles, including that of MKLP1, were determined by line plots tracing the ICBs manually. The intensity profiles were subsequently aligned based on the midbody, which was either a local minimum or the MKLP1 local maximum. Data for each experiment were normalized. Quantification of fluorescence intensities of KIF2A and acetylated tubulin at the ICB in WT mESCs that were either not treated or treated with PTX was performed by tracing the ICBs with segmented lines, determining fluorescence intensities over the ICB using the Plot Profile command in Fiji (Schindelin et al., 2012), deducting background intensities, and then calculating the average fluorescence intensity per MT. Quantification of ICB MT length in mESCs that were either not treated or treated with ZM447439 was performed by tracing the fluorescence intensity of acetylated tubulin at the ICBs with the segmented line tool in Fiji. To measure OCT6-positive cells, we generated MIPs of all z-stacks, and manually counted nuclei in the DAPI channel for each image. We then thresholded the OCT6 signal to the same level in both WT and KO images and counted all nuclei that exceeded the threshold as positive. For the mitotic index experiment, MIPs of confocal images of fixed, DAPI-stained mESCs were made in Fiji. Nuclei were counted using Cellpose3 (Stringer and Pachitariu, 2024, Preprint) with cyto3 settings, followed by manual correction. The different phases of mitosis were manually identified and counted. Finally, the distance between spindle poles in metaphase was determined manually in Fiji with line plots, taking acetylated tubulin as a spindle marker.

Standard fluorescence–based microscopy experiments in live mESCs were used to analyze FRAP of KIF2A-GFTS, and the localization of EB3-GFP in combination with SiR-tubulin. FRAP analysis was performed using either the EasyFRAP-web application (https://easyfrap.vmnet.upatras.gr/) (Koulouras et al., 2018) or a simplified protocol. For the EasyFRAP-web application in each experiment, the fluorescence intensity over time for three ROIs was determined, i.e., the ROI that was bleached (ICB or metaphase spindle pole), an area inside the same cell but outside the bleached ROI, and an area outside the cell. For the simplified protocol, we selected ICB structures of similar length and determined fluorescence intensity over time using Fiji. We examined three ROIs, i.e., the ROI that was bleached, an ROI on the non-bleached arm of the same ICB MT structure, and a background ROI. We deducted background fluorescence intensity from the other ROIs. In both approaches, a full-scale normalization was performed to normalize for differences in starting intensity, total fluorescence, and bleaching depth. For the line analysis, we drew a line with a pixel width of 10 over the complete ICB MT network and determined fluorescence intensity over time for the line. With respect to the EB3-GFP experiments, we first identified whether ICB MTs (defined by SiR-tubulin signal) were EB3-GFP–positive or EB3-GFP–negative, and then measured ICB MT length using Fiji, by placing a segmented line over the SiR-tubulin signal and measuring its length.

LSFM was used to analyze the dynamic behavior, localization, and fluorescence intensities in time of KIF2A-GFTS, SiR-tubulin, and mScarlet-PRC1 in engineered mESCs. For LSFM image analysis, acquired data were visualized by opening stacks and cropping movies in x, y, z and time (t) using the BigDataProcessor plugin (Tischer et al., 2021). When necessary, channels were aligned by manual transformation in x and/or y. The surface area of colonies was determined by taking the image nearest to the surface of the TruLive3D dish, and outlining the colony edge using fluorescence signal and the segmented line command in Fiji. Of most movies, we generated MIPs to facilitate subsequent image analysis. Metaphase-to-cytokinesis duration in live mESCs was manually determined based on mitotic MT structures visualized by SiR-tubulin. Time was measured from the start of metaphase, i.e., formation of the metaphase spindle, until the formation of ICB MTs. The number of bipolar or abnormal/multipolar spindles was calculated over the first 25 frames (i.e., 250 min). The duration of cytokinesis was also manually determined. We traced either KIF2A-GFTS or SiR-tubulin signal, from the formation of ICB MTs until the disappearance of signal (which we defined as the end of cytokinesis), by following the ICB MTs in time. The fluorescence intensity of KIF2A-GFTS and SiR-tubulin in single MIPs on ICB MTs at the onset of cytokinesis, or at t = 50, 250, and 290 (i.e., 50, 250, or 290 min after appearance of metaphase [t = 0]) was determined using Fiji, by drawing a segmented line with a width of 12 pixels over the respective signals and using the Plot Profile command. The length and width of ICB MTs were also determined manually in Fiji with line plots (line width = 3). In this case, we traced SiR-tubulin signal over the ICB MTs at t = 100 (i.e., 100 min after appearance of metaphase [t = 0]) in multiple experiments. The fluorescence intensity decay of SiR-tubulin and KIF2A-GFTS over time was measured at selected time points during a complete 16-h LSFM time-lapse experiment either by placing a circular ROI over a centrosomal area at the onset of metaphase when the SiR-tubulin signal should be high, or by placing an ROI over ICB MTs at the moment when KIF2A was most abundant. Other ROIs were placed to measure background signals. Intensities minus background were calculated using the Analyse tool in Fiji.

The fluorescence intensity of mScarlet-PRC1 in Kif2a^GFTS knock-in mESCs during cytokinesis was determined using Fiji, by drawing a segmented line with a width of 10 pixels over the ICB MTs (defined by KIF2A-GFTS). We then used the Plot Profile command to generate fluorescence intensity profiles and took the average value of the three most intense mScarlet-PRC1 pixels. The fluorescence intensity of KIF2A-GFTS and mScarlet-PRC1 in single MIPs on ICB MTs was determined using Fiji, by drawing a segmented line with a width of 10 pixels over the respective signals and using the Plot Profile command.

For a detailed kymograph analysis of KIF2A-GFTS and SiR-tubulin dynamics, a subset of 10 cropped time-lapse movies was selected in dTAG-13–treated and non-treated mESCs (i.e., 20 time-lapse movies in total). Lines were manually annotated in each frame based on the SiR-tubulin channel, allowing us to extract the length and line intensity profile per time frame. To account for the thickness of the line profiles, 17 pixels in width were averaged (the annotated pixel plus 8 pixels on each side). Background intensity correction was applied by subtracting from the mean background intensity value per time frame.

In vitro reconstitution experiments were used to investigate fluorescence intensities of GFP-KIF2A, X-rhodamine tubulin, or Hi-Lyte 647-tubulin. Intensities were determined by placing a segmented line with pixel width 3 over the length of the MT and using the Plot Profile command in Fiji. Individual graphs are shown in Fig. 6. In Fig. 7 A, the average fluorescence intensity of individual MTs is shown, which was calculated by placing a segmented line with pixel width 3 over the length of an MT and using the Plot Profile command in Fiji to obtain values. We then took the fluorescence intensity minus background signal over individual MTs and subsequently calculated the average intensity for that MT.

SDS-PAGE gel lanes were cut into 2-mm slices and subjected to in-gel reduction with DTT, alkylation with iodoacetamide, and digestion with trypsin (sequencing grade; Promega), as described previously (Schwertman et al., 2012). Nanoflow liquid chromatography–tandem mass spectrometry (nLC-MS/MS) was performed on an EASY-nLC coupled to an Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific), operating in a positive mode. Peptides were separated on a ReproSil C18 reversed-phase column (Dr Maisch; 15 cm × 50 μm) using a linear gradient of 0–80% acetonitrile (in 0.1% formic acid) during 90 min at a rate of 200 nl/min. The elution was directly sprayed into the electrospray ionization source of the mass spectrometer. Spectra were acquired in continuum mode; fragmentation of the peptides was performed in data-dependent mode by HCD.

Raw mass spectrometry data were analyzed with the MaxQuant software suite ([Cox et al., 2009]; version 2.1.3.0) as described previously (Schwertman et al., 2012) with the additional options “LFQ” and “iBAQ” selected. The false discovery rate of 0.01 for proteins and peptides and a minimum peptide length of 7 amino acids were set. The Andromeda search engine was used to search the MS/MS spectra against the UniProt database (taxonomy: Homo sapiens, release May 2022) concatenated with the reversed versions of all sequences. A maximum of two missed cleavages was allowed. The peptide tolerance was set to 10 ppm, and the fragment ion tolerance was set to 0.6 Da for HCD spectra. The enzyme specificity was set to trypsin, and cysteine carbamidomethylation was set as a fixed modification. Both the PSM and protein FDR were set to 0.01. In case the identified peptides of two proteins were the same or the identified peptides of one protein included all peptides of another protein, these proteins were combined by MaxQuant and reported as one protein group. Before further statistical analysis, known contaminants and reverse hits were removed.

To prepare RNA samples, we plated Kif2a^KO and WT mESC lines on 0.2% gelatin-coated 6-cm plates 2 days before RNA preparation. Cells were then collected, and RNA was isolated as described under “Standard molecular biology methods.” Total RNA was checked on an Agilent 2100 Bioanalyzer using an RNA nanoassay. All samples had a RIN value >9.2. RNA-Seq libraries were prepared according to the Illumina TruSeq Stranded mRNA Library Prep kit and sequenced according to the Illumina TruSeq Rapid v2 protocol on an Illumina NextSeq 2000 sequencer, generating paired-end clusters of 50 bases in length. Illumina adapter and poly-A sequences were removed from the reads, and the trimmed reads were then aligned to the GRCm38 mouse genome reference using HISAT2 (Kim et al., 2019).

Downstream analysis was performed using SeqMonk (https://www.bioinformatics.babraham.ac.uk/projects/seqmonk), followed by DESeq2 (Love et al., 2014), Metascape (Zhou et al., 2019), principal component analysis (performed with the 300 most variant genes) using pcaExplorer (Marini and Binder, 2019), or GSEA (Mootha et al., 2003; Subramanian et al., 2005). For GSEA, we assembled two local datasets, called pluripotency or centrosome and MT nucleation. The pluripotency set actually consisted of established pluripotency and primed factors (Acampora et al., 2013; Chen et al., 2008; Graf et al., 2017; Kalkan et al., 2017; Martello and Smith, 2014; Mulas et al., 2019; Soochit et al., 2021). The centrosome and MT nucleation set consisted of mRNA encoding proteins involved in centrosome regulation and MT nucleation, including all tubulin isotypes, the γTuRC and associating proteins (Böhler et al., 2021; Jackson, 2014; Yan et al., 2014), proteins involved in abscission (Chmp4b, Pdcd6ip, Cep55, Capn7, and Ist1 [Carlton et al., 2012; Chaigne et al., 2020; Paine et al., 2023]), and Cep170, an established KIF2A interaction partner (Zhang et al., 2019). Table S3 lists all genes present in the two datasets.

Independent Student’s t tests (also known as two-sample t test) were performed when comparing the means of two samples. Analysis of statistical variance (ANOVA) was carried out when comparing multiple groups, which was followed either by a Tukey honest significant difference post hoc test or by Dunnett’s post hoc test in case of comparison of all groups with one control group. Statistical tests were performed in R. Repeated-measures ANOVA was performed in Prism using the time points without missing values (GraphPad; Dotmatics). Each experiment (N) was carried out at least twice (exact N and sample sizes [n] are found in the figure legends). P values are shown in the figures.

Fig. S1 shows features of mESCs engineered in the Kif2a locus. Fig. S2 shows time-dependent decay of fluorescence intensities. Fig. S3 shows effects of KIF2A depletion on mitotic parameters. Fig. S4 shows in vitro behavior of KIF2A. Fig. S5 shows effects on pluripotency in control and KIF2A-depleted mESCs. Table S1 shows mass spectrometry data. Table S2 lists RNA data. Table S3 lists GSEA local datasets. Video 1 shows 3D localization of KIF2A-GFTS and SiR-tubulin. Video 2 shows 3D projection of KIF2A-GFTS and SiR-tubulin localization. Video 3 shows time-lapse imaging of KIF2A-GFTS and SiR-tubulin in mESCs. Video 4 shows time-lapse imaging of EB3-GFP and SiR-tubulin in mESCs. Video 5 shows imaging-based analysis of KIF2A-GFTS and SiR-tubulin in mESCs. Video 6 shows imaging-based analysis of SiR-tubulin in KIF2A-depleted mESCs. Video 7 shows time-lapse imaging of KIF2A-GFTS and SiR-tubulin in mESCs. Video 8 shows in vitro reconstitution of GFP-KIF2A behavior on MTs.

The RNA-Seq data underlying Fig. 8, C–E; Fig. S5, C and D; and Table S2 have been deposited to the Gene Expression Omnibus repository with the dataset identifier GSE230038. Primary data, reagents, and cell lines are available from the corresponding authors upon reasonable request.

We thank Jan Sakoltchik for making the GFP-CLASP2 knock-in mESC line and Emiel van Genderen for his advice on the pluripotency exit experiment.

This work was supported by grants from the Netherlands Organisation for Scientific Research (ZonMW TOP 40-00812-98-17045). H. Kabbech and I. Smal were funded by the Dutch Research Council through the Building Blocks of Life Research Program (GENOMETRACK project, Grant No. 737.016.014).

Author contributions: L. Stockmann: conceptualization, data curation, formal analysis, investigation, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing. H. Kabbech: data curation, formal analysis, and visualization. G.-J. Kremers: formal analysis, methodology, resources, and writing—review and editing. B. van Herk: investigation and methodology. B. Dille: investigation and methodology. M. van den Hout: formal analysis, software, and writing—review and editing. W.F.J. van IJcken: resources, supervision, and writing—review and editing. D.H.W. Dekkers: investigation. J.A.A. Demmers: formal analysis, investigation, and resources. I. Smal: formal analysis, methodology, and software. D. Huylebroeck: conceptualization, funding acquisition, supervision, and writing—review and editing. S. Basu: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing. N. Galjart: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.