Introduction

Centromere protein F (CENP-F, or mitosin) is a large multi-domain protein (3114 amino-acidic residues) with important roles during the cell cycle and in mitosis. It is overexpressed in several cancer types (https://www.proteinatlas.org/ENSG00000117724-CENPF/pathology). CENP-F mutations have also been identified in individuals affected by primary hereditary microcephaly (MCPH), a neurodevelopmental disorder associated with impaired brain growth and reflecting an insufficient neuronal cell pool (reviewed in1), and with the Stromme syndrome, a ciliopathy with severe microcephaly (OMIM #243605), suggesting additional roles in neurodevelopment.

CENP-F’s functions rely on its ability to interact with multiple protein partners and subcellular structures in a cell cycle-regulated manner (reviewed in2). How CENP-F interaction domains are preferentially used, and what determines the predominant localisation of CENP-F in specific cell cycle windows, is only partly understood. The identification and numbering of CENP-F protein domains has been inconsistent in the literature due to the occurrence of repeated sequences in the CENP-F coding region. Auckland and colleagues3 have recently clarified previous discrepancies between CENP-F sequences reported in the Uniprot and NCBI databases, which have now been resolved. In this study we will refer to the Uniprot nomenclature and numbering (https://www.uniprot.org/uniprotkb/P49454/entry).

In cycling cells CENP-F first appears in late S phase4 and increases during G2, at which time it displays a fully nuclear localisation. The determinants of that localisation were mapped by deletion analysis in the C-terminal region of CENP-F5, within which a classical nuclear localisation signal (c-NLS, 2917-2959) was identified by sequence homology. A CENP-F fragment containing the putative c-NLS (2891-2969) was later shown to interact with purified importin alpha, KPNA2, in vitro6. Other import vectors were not tested. The interaction of CENP-F with importin alpha proved phosphorylation-sensitive in vitro, as phosphorylation of the c-NLS-containing CENP-F fragment by the mitotic cyclin B1/CDK1 kinase complex decreased its affinity for KPNA26. Physiologically, CDK1 is expected to phosphorylate CENP-F from late G2 until metaphase; indeed, CENP-F has been found to be phosphorylated in a proteome-wide study in mitotic cells7. Indeed, CDK1-dependent phosphorylation of CENP-F down-regulated its nuclear import and facilitated its nuclear export in late G2 phase in neural stem cells and HeLa cells8, which represents a key step towards nuclear envelope breakdown (NEB)9.

CENP-F actively participates in the NEB, a process initiated by the nucleoporin RANBP2 and the dynein adaptor bicaudal (BicD2), which together recruit centrosomes to the NE periphery. Therein, a fraction of CENP-F colocalises at nuclear pore complexes via the nucleoporin NUP133 and recruits the dynein motor complex via two adaptors, NDE1 and NDEL, with which CENP-F interacts directly10. Dynein then exerts forces that aid the NE rupture. Interestingly, the interaction of CENP-F with NUP133 also enables dynein-dependent nuclear apical migration in neural stem cells8,11.

After NEB, CENP-F interacts with mitotic microtubules via two binding domains (MTBDs), respectively positioned at the N-terminus (MTBD1, 1-385) and C-terminus (MTBD2, 2927-3114) of the protein. Both CENP-F MTBDs contribute to regulate microtubule stability during kinetochore capture12,13. CENP-F and the dynein complex are recruited to the maturing kinetochore sequentially during mitosis. A CENP-F fraction is targeted to kinetochores via a region (2826–2894) that interacts with the spindle checkpoint kinase Bub114,15. CENP-F is further stabilised at kinetochores by interacting with NUP133, which translocates to mitotic kinetochores after NEB as part of the NUP107/160 nuclear pore subcomplex16. In turn CENP-F recruits motor proteins, including CENP-E and dynein, the latter via the Nde1/Ndel1/Lis1 adaptors that interact directly with CENP-F17,18. CENP-F and the dynein complex regulate microtubule/kinetochore attachments during the search-and-capture process, and CENP-F silencing results in chromosome misalignment and missegregation12,17,19,20,21. Further, CENP-F exerts a specific function at the kinetochore corona via its C-terminal MTBD2: as microtubules form end-on attachments to kinetochores, the corona disassembles in a process that involves the dynein-mediated stripping of cargoes, including the spindle checkpoint protein Mad2, from kinetochores; in the process, CENP-F limits the dynein-mediated stripping of cargoes3, thus regulating the corona protein composition and ensuring proper kinetochore orientation during microtubule attachment. Upon completion of chromosome alignment, a large proportion of CENP-F is degraded22. The timing of CENP-F degradation was precisely mapped using a fluorescently tagged C-terminal fragment of the protein: in that fragment, a KEN box (KEN box 7, 3027-3033) was identified as the target of the APC/CCDC20 ubiquitin ligase complex23. KEN box 7-dependent ubiquitination conveys the CENP-F C-terminal fragment towards proteasomal degradation during anaphase. A residual CENP-F fraction still persists throughout anaphase/telophase and orchestrates the recruitment of mitochondria at the midbody during cytokinesis24. That fraction will finally be degraded after cytokinesis.

In summary, the C-terminal region of CENP-F has a dense organisation of regulatory elements, i.e.: the region driving CENP-F nuclear localisation with the putative NLS, the KEN box required for regulated ubiquitination and degradation, and the MTBD2 domain, regulating microtubule stability and corona composition. Interestingly, viable mutations have been identified in that region in microcephalic individuals: a R2898X substitution causes CENP-F truncation in the region required for nuclear import and is associated with decreased CENP-F protein abundance25, while a R3094X substitution truncates the MTBD226,27. These data indicate a pathogenetic role of mutations in the CENP-F C-terminal region in neurodevelopment.

Importin beta, the major vector of protein nuclear import in interphase, can interact with NLS-containing proteins either directly, or via an importin alpha adaptor, depending on the NLS sequence type in the protein cargo (reviewed in28,29). After NEB, when nucleo-cytoplasmic transport ceases, importin beta localises at the mitotic spindle30. A gene ontology analysis of the importin beta mitotic interactome identified “microtubule-based processes” as one of the most highly represented classifications31,32. Indeed, importin beta orchestrates the spatial distribution of spindle regulatory factors that bear an NLS, often in concert with dynein, thus regulating the mitotic spindle organisation and dynamics (reviewed in33). We previously reported that acute importin beta overexpression from a transiently transfected plasmid resulted in decreaed CENP-F localisation at kinetochores and accumulation at spindle poles34 during mitosis. Inactivation of RANBP2, a direct importin beta interactor, also displaced CENP-F from kinetochores35,36. These data suggest an additional level of control of CENP-F mitotic localisation implicating importin beta.

Here we show that importin beta forms proximity ligation products with CENP-F in mitotic cells that do not require importin alpha. We have assessed the significance of these interactions in engineered cell lines in which importin beta abundance can be modulated, and find that importin beta regulates both CENP-F localisation and stability during mitosis, identifying importin beta as a new player in CENP-F regulation.

Results

CENP-F colocalises and interacts with importin beta at mitotic structures

Earlier findings that acute importin beta overexpression displaced CENP-F from mitotic kinetochores34, raise the question of whether these two proteins interact in mitosis. We addressed that question using proximity ligation assays (PLA). We previously verified the PLA method accuracy for importin beta by visualising its interactions with known mitotic partners31,32,37. At the onset of this work, we also validated the PLA method for CENP-F and some established partners. We preliminarily developed immunofluorescence (IF) assays in both RPE-1 and HeLa cell lines to predict where and when PLA products with known partners should be expected based on the distribution of single components. In these assays, CENP-F was first depicted at microtubule-organising centers during the establishment of the spindle poles, to which a fraction remained associated throughout mitosis (Fig. S1A,B). As the bipolar spindle assembled, CENP-F distributed at microtubules, poles and kinetochores, in parallel with dynactin, the dynein complex activating component (Fig. S1C), with which dynein moves along microtubules38. A CENP-F fraction colocalised with BUB1 at kinetochores from prometaphase to chromosome alignment (Fig. S1D), and with a fraction of NUP133 in metaphase (Fig. S1E). CENP-F substantially decreased in anaphase and telophase in both cell lines (Fig. S1A,B). In PLA assays, CENP-F formed indeed abundant ligation products with BUB1 at kinetochores in early mitotic stages, which decreased at the onset of chromosome segregation after the SAC release and BUB1 detachment from kinetochores (Fig. 1A). In parallel assays, NUP133/CENP-F PLA products localised around the nucleus in G2 cells, then transiently dispersed throughout the cells soon after NEB and concentrated at the microtubule/kinetochore interface in prometaphase and metaphase (Fig. 1B). Thus, the PLA method genuinely detects ligation products between CENP-F and its interactors.

Fig. 1
figure 1

CENP-F forms temporally regulated PLA ligation products with known interactors at mitotic kinetochores in RPE-1 cells. (a) The BUB1/CENP-F pair forms PLA products that are particularly abundant in prometaphase and concentrate in the region of microtubule-chromosome contacts. (b) NUP133 and CENP-F form PLA products that peak during chromosome alignment in metaphase. Scale bar: 10 µm.

Full size image

At this point we tested the CENP-F/importin beta pair in mitotic cells: PLA products were visible at the spindle microtubules and poles, in both RPE-1 (Fig. 2A) and HeLa (data not shown) cell lines. The addition of importazole (IPZ), a specific importin beta inhibitor39, prevented the formation of these products, indicating the specificity of the reaction.

Fig. 2
figure 2

Specificity of proximity ligation products formed by importin beta and mitotic regulators in RPE-1 cells. (a) Importin beta forms PLA products with CENP-F along the spindle, which severely decrease in importazole-treated cells (IPZ), but are not affected by RNA interference-mediated importin alpha down-regulation, indicated as alpha(i). (b) HURP/Importin beta PLA products localise at microtubule plus-ends; IPZ prevents their formation to a highly significant extent, while they are only minimally affected by alpha(i). (c) TPX2/Importin beta PLA products localise predominantly at spindle poles and they significantly decrease in both IPZ-treated and alpha(i) cells. Representative mitoses from three experiments are shown. Insets, 3× magnification of the merged images from control cultures showing the distribution of PLA products with individual mitotic regulators. Scale bar: 10 µm. PLA signals for each protein pair were counted in single cells in late prometaphase and metaphase (between 45 and 75 counted cells per condition) and analysed using the Mann–Whitney test. ns, non-significant; *p < 0.05; ***p < 0.001; ****p < 0.0001.

Full size image

CENP-F/importin beta PLA product formation does not require importin alpha in mitotic cells

Purified importin alpha, the receptor for several NLS signals, was found to bind a purified CENP-F fragment containing the c-NLS in vitro; phosphomimetic mutations in that fragment prevented the interaction6. We asked whether importin alpha was implicated in the formation of CENP-F/importin beta PLA products observed in mitotic cells, when CENP-F is phosphorylated7. IF analysis showed that both importin alpha and beta colocalised with the NE in interphase RPE-1 cells, as expected (Fig. S2A,B); after NEB, instead, they assumed different patterns: importin beta associated with microtubule-organising centers and spindles (Fig. S2B), while importin alpha assumed a largely disperse distribution throughout mitotic cells, with some more intensely stained regions of vesicular appearance (Fig. S2A), consistent with its ability to associate with membranous organelles until NE reformation40,41,42. A similarly disperse importin alpha pattern, contrasting with the spindle-associated importin beta, was observed in HeLa cells (data not shown). During anaphase, both importin types localised back around chromosomes to prepare for NE reassembly (Fig. S2A,B). These different distributions would not favour the formation of trimeric CENP-F/importin alpha/importin beta complexes during mitosis. Indeed, importin alpha and CENP-F formed only rare PLA products that did not have a biologically significant localisation in mitotic cells (Fig. S2C), differently from the spindle-associated CENP-F/importin beta PLA products (Fig. S2D). To assess directly whether importin alpha was involved at all in CENP-F/importin beta mitotic PLA products, we down-regulated importin alpha using specific small interfering RNAs (siRNAs). This did not affect CENP-F levels (Fig. S3). We found that CENP-F/importin beta PLA products formed equally effectively in importin alpha-silenced and in mitotic cells treated with a neutral siRNA (GL2, against the firefly luciferase gene) (Fig. 2A). For comparison, we analysed mitotic PLA products for importin beta and either a direct interactor, i.e. HURP32,43, or a well-established importin alpha cargo, i.e. TPX2, that associates indirectly with importin beta44. In Fig. 2, importin alpha siRNAs affected only weakly importin beta/HURP PLA products (Fig. 2B), while they significantly down-regulated importin beta/TPX2 PLA products (Fig. 2C). IPZ prevented PLA formation for all examined protein pairs. We conclude that importin alpha is not required for the formation of importin beta/CENP-F PLA products in mitotic cells.

Molecular models indicate that importin beta and CENP-F can interact directly

Based on the finding that importin beta/CENP-F ligation products do not require importin alpha, we wondered whether importin beta could potentially bind the canonical NLS (c-NLS) of CENP-F directly. To that end, we searched the Protein Data Bank (PDB) for the presence of experimentally determined 3D structures of importin beta in complex with NLS sequences to be used as templated for molecular modelling. The only importin beta structure satisfying that requirement was that in complex with the non-canonical NLS (nc-NLS) of parathyroid hormone-related protein (PTHrP), which has been determined by X-ray crystallography (PDB identifier: 1M5N45; Resolution: 2.90 Å). Building a model of importin beta with the c-NLS of CENP-F based on this template was complicated by the difficulty of identifying an alignment between the sequences of the c-NLS of CENP-F and the nc-NLS of PTHrP that might reliably reflect the putative binding mode of the c-NLS of CENP-F to importin beta. The scope of building the model was to evaluate the possibility that a direct binding between the two proteins might occur, rather than attempting to predict the actual position of the c-NLS of CENP-F relative to importin beta in a putative complex. We manually aligned the sequences of the CENP-F c-NLS and the PTHrP nc-NLS using their positively charged residues as a guide (Fig. S4A). Figure S4B and C show molecular models of importin beta/CENP-F complex generated based on this alignment. In Fig. S4B, the five positively charged residues of the c-NLS of CENP-F (i.e., four lysines and one arginine) are in contact with the negatively charged surface of importin beta, which is rich in acidic residues46. Comparison of the importin beta-cargo interactions occurring in template structure and in the model (Fig. S4D) indicates that a number of favourable intermolecular interactions, both polar and hydrophobic, occur in the model, several of which involve the basic residues of the c-NLS of CENP-F, and some of which involve the same importin beta residues that contact the nc-NLS of PTHrP in the template structure. Together with the occurrence of multiple favourable interactions, the absence of unfavourable van der Waals interactions (or steric clashes) in the model indicates that CENP-F might indeed directly bind importin beta.

Like NLS sequences, the importin alpha IBB domain comprises a relatively high number of positively charged residues47. In the absence of importin beta, the IBB can fold inward and interact with the NLS-binding region of importin alpha48. Structure superposition of the importin beta fragment in PDB coordinate file 1M5N to the equivalent importin beta residues in PDB coordinate file 8GCN49, where importin beta was experimentally determined by electron microscopy in complex with the importin alpha IBB (Resolution: 3.95 Å), shows that the importin beta region in contact with the nc-NLS of PTHrP in the first structure partly overlaps with the importin beta region that binds the importin alpha IBB in the second structure (Fig. S4A, right panel). The alignment of the IBB sequence to the nc-NLS of PTHrP shown in Fig. S4A (left panel) suggests the possibility of an approximate spatial correspondence between pairs of residues, as the two cargoes cannot be properly structurally aligned. Based on the above, however, it is reasonable to hypothesise that the extended negatively charged region of importin beta can bind several NLS sequences, as well as the IBB.

Importin beta regulates the timing of CENP-F degradation in mitotic cells

To investigate the significance of CENP-F/importin beta mitotic interactions detected by PLA, we used engineered HeLa cells in which importin beta levels can be induced by Doxycycline (Dox) in a controlled manner and monitored using an in-frame GFP tag32,37. Dox addition increased importin beta-GFP abundance by about 2–2.5-fold over the endogenous level (Fig. S5A), reflected by an increase in CENP-F/importin beta PLA products compared to the amount formed by the endogenous proteins (Fig. S5B). We noticed that PLA products persisted in anaphase and telophase in Dox-induced cultures (Fig. S5B), whereas they are down-regulated in control cultures, as CENP-F begins to be degraded at that stage. The persistence of PLA products in late mitosis in importin beta-overexpressing cultures was unexpected and suggested that CENP-F was not effectively degraded in those cultures.

HeLa cells express deregulated importin beta levels compared to non-cancer cells, due to the disruption of E2F/pRB-dependent transcriptional control of the KPNB1 gene in that cell line50. It was relevant to ascertain whether the changes observed in Dox-induced HeLa cells also occurred in non-transformed cells. We therefore generated RPE-1 cell lines with stably integrated, Dox-inducible importin beta-GFP using the same Teton system (see Methods). By time-lapse recording of RPE-1 cells, importin beta-GFP expression was first detected 4–6 h after Dox addition. By 18–20 h, all cells in the population expressed importin beta-GFP (Fig. S6A), with an overall increase of about threefold compared to the endogenous protein by both Western blot (Fig. S6B) and IF assays (Fig. S6C). Importin beta-GFP displayed the canonical localisation at the spindle and around chromatin in late mitosis (Fig. 3). PLA assays in Dox-treated RPE-1 cells confirmed both the increase in CENP-F/importin beta products in metaphase and their persistence in late mitosis, which were abolished by IPZ (Fig. 3A, quantified in Fig. 3B and Fig. S7). IF analysis confirmed CENP-F persistence in late mitosis in both HeLa (Fig. 4A) and RPE-1 cell lines (Fig. 4B) induced with Dox. IPZ addition prevented CENP-F persistence, while the proteasome inhibitor MG132 stabilised it in the absence of Dox. Thus, importin beta influences the timing of CENP-F down-regulation.

Fig. 3
figure 3

Importin beta overexpression increases PLA signals with CENP-F and induces their persistence in late mitotic stages in RPE-1 cell cultures. (a) Importin beta/CENP-F PLA signals during physiological mitosis (NO DOX) and in the presence of importin beta-GFP overexpression (DOX) in RPE-1 Dox-inducible cell cultures. PLA products are abundant until metaphase and decline during anaphase-to-telophase progression, yet persist in Dox-induced cultures. IPZ prevents PLA product formation, with or without Dox induction. Scale bar: 10 µm. (b) CENP-F/importin beta PLA signals were automatically counted in single cells from control (NO DOX), importin beta-overexpressing (DOX) and IPZ-treated (IPZ) cultures. Results from one representative experiment (at least 25 counted cells per stage per condition) are plotted in the graph (*p < 0.05; ****p < 0.0001, Mann–Whitney test). All data obtained in independent experiments are listed in Fig. S7A and displayed in individual graphs for each experiment in Fig. S7B.

Full size image
Fig. 4
figure 4

Importin beta overexpression stabilises CENP-F at spindle poles in late mitosis. (a) Importin beta induction (DOX) stabilises CENP-F signals, in both HeLa Importin beta-GFPdox, and (b) in RPE-1 Importin beta-GFPdox cell lines, compared to non-induced cells (NO DOX). In both cell lines CENP-F signals accumulate at spindle poles and they persist in late mitotic stages. CENP-F similarly persists in cells treated with MG132 (MG), and IPZ counters the increase induced by Dox in both cell lines. Scale bar: 10 µm. (c) The graphs show the distribution of CENP-F signals, in arbitrary units (AU), measured in single cells in metaphase and anaphase taking spindle poles as the region of interest (RoI) of fixed size. The data in the graph were pooled from 5 experiments comparing HeLa Importin beta-GFPdox cell cultures induced with dox, with or without proteasome inhibitor MG132 (indicated as MG), versus non-induced cultures, with or without importazole (IPZ). (d) The graphs show the distribution of CENP-F signals, in arbitrary units (AU), measured in single metaphase and anaphase cells taking spindle poles as the region of interest (RoI) of fixed size as described for (c). The graph shows pooled data from 3 experiments in RPE-1 Importin beta-GFPdox cell cultures induced with dox, with or without proteasome inhibitor (MG), versus non-induced controls, with or without importazole (IPZ). All data obtained in independent experiments are listed in Fig. S8A and displayed in individual graphs for each single experiment in Fig. S8B and C. The distribution of signal intensities was statistically analysed using the Mann–Whitney test. ns, non-significant; *p < 0.05; ***p < 0.001; ****p < 0.0001.

Full size image

Importin beta regulates the subcellular localisation of CENP-F during mitosis

By quantitative IF, CENP-F signals increased in whole cells from both HeLa (Fig. 4A) and RPE-1 (Fig. 4B) cell lines induced with Dox in all stages of mitotic progression. Selecting either the kinetochores, or the spindle poles, as the Region of Interest (RoI), we realised that CENP-F abundance did not increase uniformly throughout the cells, but that it accumulated, with highly statistical significance, in the spindle pole region (quantified in Fig. 4C,D), while kinetochores showed a significantly decreased signal intensity (Fig. S8). IPZ countered the accumulation of CENP-F at poles observed in Dox-induced cells and restored CENP-F at kinetochores. Importin alpha silencing affected neither the original pattern of pole- vs kinetochore-associated CENP-F in control mitoses, nor CENP-F accumulation at poles in importin beta-GFP-expressing cells (data not shown). Thus, the intracellular amount of importin beta, in addition to influencing CENP-F abundance, also regulates its spatial distribution during mitosis.

We previously showed that importin beta overexpression unbalances spindle forces, resulting in spindle pole disruption30, accompanied by the accumulation of importin beta interactors, e.g. RANBP2, RANGAP1, and HURP, at poles and their concomitant decrease at kinetochores34,37 and microtubule plus ends32. We wondered whether the displacement of CENP-F to spindle poles caused by importin beta overexpression also involved proteins that tether CENP-F to kinetochores. We found that CENP-F PLA products both with BUB1 and with NUP133 significantly decreased at kinetochores in Dox-induced compared to non-induced RPE-1 cells (Fig. 5A, quantified in Fig. 5B). Instead, the kinetochore localisation of BUB1 and NUP133 per se remained unperturbed (Fig. 5C): these data suggest that increased importin beta levels engage CENP-F in interactions that render CENP-F less available to interact with kinetochore factors.

Fig. 5
figure 5

Importin beta overexpression decreases the association of CENP-F with kinetochore partners. (a) In RPE-1 Importin beta-GFPdox cell lines, PLA signals for both CENP-F/BUB1 and CENP-F/NUP133 pairs decrease at kinetochores in importin beta-overexpressing (DOX) compared to control (NO DOX) cells. (b) The graphs quantify PLA signals measured in single cells selecting CREST (kinetochores) as the RoI. For each protein pair, PLA signals were quantified in single cells in late prometaphase and metaphase (30–40 counted cells per condition in 2 independent experiments) and analysed using the Mann–Whitney test. *p < 0.05; **p < 0.01. (c) The IF signal intensity of both BUB1 and NUP133 at kinetochores is unaffected by importin beta overexpression. Scale bar: 10 µm.

Full size image

In a normal mitosis, CENP-F recruits dynein adaptors to kinetochores17. To assess the functional consequences of CENP-F spatial redistribution in importin beta-overexpressing mitoses, we also examined the dynactin/dynein complex. Both CENP-F and dynactin have a dynamic localisation during mitosis, yielding some cell-to-cell heterogeneity in the distribution of kinetochore-associated versus pole-associated signals. We assessed dynactin localisation in control and importin beta-overexpressing cells by quantitative immunofluorescence, taking automated measurements either at poles, or at kinetochores, as the Region of interest (RoI), then statistically compared the distribution of signals in control and Dox-treated cultures. We found that the kinetochore fraction of dynactin significantly decreased, while the fraction at poles increased, in importin beta-overexpressing (DOX panels) compared to control cells, parallel to that of CENP-F (Fig. S9A). To gain specific information of the involvement of CENP-F in dynactin redistribution, we repeated similar experiments in cells subjected to CENP-F silencing; CENP-F-specific siRNAs yielded a highly significant decrease of the overall CENP-F cellular content, except for a small residual fraction persisting at spindle poles, while the kinetochores appear to be virtually devoid of CENP-F (Fig. S9B). Dynactin signals were also significantly reduced at kinetochores in these cells (Fig. S9B), likely reflecting an impaired recruitment of dynein adaptors to kinetochores in the absence of CENP-F.

In summary, importin beta overexpression favours CENP-F accumulation at spindle poles and down-regulates its interaction with kinetochore partners, suggesting an antagonistic function of importin beta to that exerted by NUP133 and BUB1 in anchoring CENP-F at kinetochores. This redistribution of CENP-F entails dynein depletion from kinetochores and is accompanied by CENP-F protein stabilisation in late mitosis.

CENP-F is an unstable protein and importin beta stabilises it

The finding that importin beta stabilised CENP-F in late mitosis prompted us to analyse CENP-F turn-over in cycling RPE-1 and HeLa cultures. In a time-course analysis of cell cultures inhibited for protein synthesis by cycloheximide (CHX), Aurora-B, consistent with its being subjected to regulated proteolysis in late mitosis in every cell cycle, was rapidly down-regulated in both cell lines shortly after CHX addition (Fig. 6A,B). CENP-F was also significantly down-regulated within 8 h after CHX administration (Fig. 6A,B), and MG132 addition stabilised its abundance. A short KEN box in the CENP-F C-terminal region (positions 3027-33; see introduction) targets CENP-F for ubiquitination and conveys it to degradation during anaphase23. To assess the involvement of importin beta in the timing of CENP-F degradation, we synchronised cell cultures by inducing arrest at the G2/M transition using cell cycle inhibitors, e.g. STLC, a kinesin Eg5 inhibitor, or RO3306, which inhibits the CDK1 kinase, then released the cells in inhibitor-free medium to resume mitosis (see Methods). RO was used in most experiments and released cells were collected at progressive time-points to enrich samples in late prometaphase-metaphase, and in anaphase-telophase stages. In Western analysis, importin beta-GFP induction effectively prevented the decrease in CENP-F protein level in anaphase-telophase in both cell lines (Fig. 6C,D), similar to MG132.

Fig. 6
figure 6

CENP-F is an unstable protein. CENP-F levels begin to decline between 4 and 8 h after CHX administration in both RPE-1 (a) and HeLa cell lines (b), and continues to gradually decrease, as quantified in the graphs below. The mitotic marker Aurora B is rapidly degraded, while the nucleoporin NUP153 does not vary. Tubulin is a loading control. In the gaphs below the CENP-F signal, normalised to that of NUP153 (average from 3 experiments), is plotted over time in the presence or absence of CHX; values measured in control (CTR) and CHX-treated cultures are compared at each indicated time using the Student’s t test (ns, non-significant; *p < 0.05; ***p < 0.001; ****p < 0.0001); the decline over time in CHX vs CTR is statistically evaluated using the Chi-square for trends test (****p < 0.0001). (c) Importin beta overexpression (DOX) stabilises CENP-F abundance in late mitosis in RPE-1 Importin beta-GFPdox and (d) HeLa Importin beta-GFPdox cell cultures, similar to MG132. Both cell lines were arrested at the G2/M transition with RO3306, then released in RO3306-free medium and harvested at progressive time points after release. Under physiological conditions (-DOX) CENP-F levels decline in ana-telophase (ANA) compared to prometaphase and metaphase (PM-M), yet they remain stable in importin beta-overexpressing cultures (+ DOX). Tubulin and NUP153 are loading controls for RPE-1 Importin beta-GFPdox and HeLa Importin beta-GFPdox, respectively.

Full size image

Microtubules contribute to regulate CENP-F stability

Importin beta has been shown to protect some mitotic factors, e.g. NuSAP1 and HURP, from premature proteasomal degradation during spindle assembly in a process that also involves microtubules51,52. Given the microtubule-regulatory functions exerted by CENP-F, it was interesting to assess the possibility that microtubules in turn play roles in regulating CENP-F abundance. We therefore compared CENP-F in control prometaphase-enriched cultures after RO3306 arrest and release, and in the presence of drugs that inhibit either the spindle dynamics (taxol, TAX, which blocks the dynamic activity of microtubules), or formation (nocodazole, NOC; vinblastine, VBL, both of which inhibit microtubule polymerisation) (Fig. 7A). All microtubule-targeting drugs arrest mitotic progression in prometaphase, when CENP-F abundance is high. We found that CENP-F abundance decreased in Western assays of extracts from microtubule-inhibited prometaphases compared to untreated cells collected by shake-off at round-up (also in prometaphase) from synchronised cultures (Fig. 7B, quantified in the graphs below). By IF analysis, those few CENP-F signals that were still visible in microtubule-inhibited cells localised near tubulin foci nucleated or stabilised at kinetochores (Fig. 7C). During a normal mitotic progression, the distribution of CENP-F evolves from a relaxed localisation in prometaphase, detected as an unstructured staining around kinetochores (Fig. S10B), to a more compact organisation in metaphase, when both sister kinetochores are attached to microtubules that exert the highest tension on them (Fig. S10A; also see53). In microtubule-inhibited cells, in which no tension is exerted onto kinetochores, the rare residual CENP-F signals had a disorganised appearance around the tubulin foci that survived near kinetochores (Fig. S10C). We asked whether microtubules are required as a physical track for importin beta to establish contact with CENP-F and stabilise its abundance. To address that question we used two complementary protocols, inverting the sequence of the treatments: in the first protocol, we induced importin beta overexpression with Dox, then added NOC to inhibit microtubule polymerisation in cells that already overexpressed importin beta; in the reverse sequence, we first inhibited microtubule polymerisation with NOC, then added Dox to induce importin beta in cells with non-polymerised microtubules. Figure 8A shows CENP-F/importin beta PLA products formed under these different conditions. In cells treated with Dox alone, the abundance of CENP-F/importin beta PLA products increased compared to controls, particularly at poles, as expected. NOC administration under all schedules dramatically down-regulated PLA signals. Closer inspection of the images and quantification of residual signals (Fig. 8B) showed that NOC inhibition of microtubules fully abolished the formation of CENP-F/importin beta PLA products (NOC only) and severely down-regulated their abundance in importin beta-overexpressing cells (DOX + NOC schedule) compared to cells treated with Dox alone, indicating that microtubules are required to maintain those products together. In the complementary protocol, importin beta overexpression in cells previously treated with NOC (NOC + DOX schedule) restored some rare PLA products compared to NOC alone; these products were detected near kinetochores, where residual microtubule foci remained, while no significant PLA products were detected in the cytoplasm devoid of microtubules. IPZ abolished the up-regulation of PLA products in NOC + Dox cells, bringing their level back to that seen in NOC alone. In summary, importin beta interacts with, and stabilise, CENP-F, which is otherwise unstable, and the establishment of that interaction needs microtubules as a physical track.

Fig. 7
figure 7

Microtubule inhibitors down-regulate CENP-F abundance. (a) Protocol schematics for Western bot assays in RPE-1 cells (left). Microtubule drugs arrest cells in prometaphase either without a spindle (NOC, VBL), or with non-functional microtubules (TAX). In control cultures, prometaphase-enriched samples were obtained by culturing in RO3306 to arrest the cell cycle at the G2/M transition, then washed-out (W/O) and harvested after 60 min (peak of the prometaphase wave) by mitotic shake off (S/O). In all cultures, parallel samples were treated with MG132. (b) Western blot assays: CENP-F protein levels decrease in the presence of microtubule inhibitors and are stabilised by MG132. The graphs below quantify the CENP-F signal normalised to that of NUP153 (the histograms represent the average from 5 independent experiments): a highly significant decrease was measured in cultures treated with all three MT-targeting drugs compared to controls using the Student’s t test, ***p < 0.001; ****p < 0.0001). (c) In IF assays, NOC and VBL (microtubule polymerisation inhibitors), as well as TAX (microtubule dynamics inhibitor), all yield premature degradation of CENP-F in prometaphase. Residual signals survive near kinetochore-associated tubulin foci (zoom). Scale bar: 10 µm. Bar in the zoom-in boxes: 2 µm.

Full size image
Fig. 8
figure 8

Overexpressed importin beta does not form PLA products with CENP-F in the absence of microtubules. (a) CENP-F/importin beta PLA assays under the indicated conditions: PLA signals were counted in cells with inhibited microtubules (NOC) subsequently induced to overexpress importin beta (NOC + DOX), versus importin beta-induced cells (DOX) subsequently subjected to microtubule inhibition (DOX + NOC): NOC down-regulates CENP-F/importin beta PLA signals; residual PLA signals localise near kinetochores, where microtubule foci are nucleated and/or stabilised. Importin beta overexpression (DOX) partly rescues PLA interactions at residual tubulin, while IPZ abolishes them. (b) Quantification of CENP-F/importin beta PLA signals in single cells in late prometaphase and metaphase (between 32 and 75 counted cells per condition in 2 independent experiments), analysed using the Mann–Whitney test. ns, Non-significant; ****p < 0.0001. Scale bar: 10 µm.

Full size image

Discussion

Here we describe a previously unidentified role of importin beta in regulation of the CENP-F protein during mitosis. CENP-F acts at several steps of mitosis via temporally and spatially regulated interactions with protein partners according to a sequential programme including: NEB, the process of kinetochore search-and-capture by microtubules, the stabilisation of microtubule attachments and kinetochore biorientiation. These steps are essential to ensure faithful chromosome segregation. Additional CENP-F functions have recently been revealed in corona stripping, exerted via dynein and its adaptors3, and in the mitotic surveillance pathway, via 53BP1 and PLK154. All these functions are accurately regulated in space and time. We now show that importin beta is a novel regulator of CENP-F and orchestrates a dual level of control, in regulating both the balance of CENP-F localisation at spindle poles versus kinetochores, and the timing of CENP-F degradation.

The PLA method provides spatial and temporal information on CENP-F interactions with partners that localise it during mitosis. The method faithfully depicted the stage-specific distribution of PLA signals formed by CENP-F and known kinetochore partners, BUB1 and NUP133, prior to, and during, microtubule attachments, respectively. Applying the PLA method to the CENP-F/importin beta pair visualised ligation products at spindle microtubules and poles, but not at kinetochores, a complementary pattern to that seen for the CENP-F/BUB1 and CENP-F/NUP133 pairs.

CENP-F is entirely nuclear in interphase, indicating that it is an effective cargo for import complexes. Deletion mapping studies to delimit the region required for CENP-F nuclear import5,55 identified an element bearing homology with a classical bipartite NLS. The actual binding of that element was tested in vitro with purified importin alpha and was found to be down-regulated by CDK1-dependent phosphorylation6, an event that, in a normal cell cycle, would take place from late G2 until metaphase. The in vitro binding assays6 predict therefore that importin alpha/CENP-F interactions are prevented or severely disfavoured in mitotic cells, due to phosphorylation. Furthermore, CDK1 also phosphorylates importin alpha itself at mitotic onset: this facilitates the disassembly of interphase import complexes, enabling the release and activity of NLS-containing spindle regulatory factors56. In addition to the reduced affinity of importin alpha for phosphorylated CENP-F, the largely diffuse distribution of importin alpha in mitotic cells would further disfavour their association during mitosis, in contrast with importin beta, which associates with the mitotic spindle. Consistent with these predictions, the CENP-F/importin alpha pair formed only rare, delocalised PLA products, while abundant CENP-F/importin beta PLA signals were depicted at the spindle. These products formed independent on importin alpha, similar to HURP, a direct importin beta interactor32,43. The molecular model hypothesised in this work further supports the notion that the importin beta region involved in the interaction with the CENP-F c-NLS and the importin alpha IBB partly overlap, reinforcing the idea that mitotic PLA products reflect the occurrence of direct importin beta/CENP-F interactions.

Using Dox-inducible importin beta engineered cell lines, we found that increasing importin beta abundance affected both the localisation and the timing of CENP-F/importin beta PLA product formation. Indeed, importin beta accumulated CENP-F at spindle poles, concomitantly decreasing the kinetochore-associated fraction in prometaphase and metaphase. IPZ, which inhibits importin beta, prevented the formation of CENP-F/importin beta PLA products and restored CENP-F abundance at kinetochores. Importin beta is therefore a novel player in the spatial localisation of CENP-F during mitosis, capable to regulate the pool of pole-associated versus kinetochore-associated CENP-F during spindle formation and microtubule attachment to kinetochores. CENP-F was reported to be continuously transported from kinetochores to spindle poles via dynein during mitosis57,58. Here we find that both CENP-F silencing, and CENP-F mitotic relocalisation induced by importin beta overexpression, deplete the kinetochore pool of dynein and accumulate it at poles. These findings indicate a mutual regulation of dynein and CENP-F localisation at the mitotic apparatus and show that importin beta levels modulate this process.

During metaphase-to-anaphase progression, the CENP-F signal declines concomitant with BUB1 detachment from microtubule-attached kinetochores and with importin beta recruitment back around chromosomes; CENP-F/importin beta PLA products decline in parallel. These data hinted at the possibility that the loss of interaction with importin beta rendered CENP-F available to the ubiquitination machinery. Indeed, importin beta overexpression caused the persistence of PLA products with CENP-F and concomitantly counteracted CENP-F down-regulation in late mitosis. Experiments designed to manipulate both the state of microtubules, and importin beta levels, in RPE-1 cells also elicited a function of microtubules in preserving the stability of CENP-F, particularly in the vicinity of kinetochores. Microtubule-targeting drugs reversed the stabilising effect of importin beta over CENP-F. These data suggest that intact microtubules are required as a platform over which importin beta can interact with CENP-F and protect it from degradation. A model summarising these observations is proposed in Fig. 9. Thus, a mutual interplay appears to take place between regulation of microtubule dynamics by CENP-F and CENP-F stabilisation by microtubules observed here. Interestingly, microtubules protect HURP and NuSAP151,52, and now CENP-F, from unscheduled degradation, ultimately determining a self-reinforcing ability of microtubules to stabilise their own regulators. That mechanism ceases concomitant with chromosome decondensation and importin beta repositioning around chromatin (Fig. 9). Both the NLS and KEN7 are embedded in the MTBD2 of CENP-F. It may be speculated that CENP-F interaction with importin beta via the NLS, and with microtubules via MTBD2, prevent the APC/C complex from binding to and ubiquitinating the KEN7 box. Additionally, but not necessarily mutually exclusively, importin beta may act primarily by accumulating CENP-F at spindle poles, away from the kinetochore sites where the APC/C is active. When importin beta detaches from CENP-F at anaphase, KEN7 box would become available to ubiquitination.

Fig. 9
figure 9

Summary model: spatial and temporal control of CENP-F by importin beta and MTs. Upper panel: in a normal mitosis, importin beta interacts with CENP-F along the spindle microtubules (MTs) and contributes to regulate the balance between pole- and kinetochore (KT)-associated CENP-F during spindle organisation (metaphase panel). In anaphase, when MTs reorganise in the central spindle and importin beta begins to relocalise at sites where the nuclear envelope (NE) will reform, CENP-F becomes accessible to the ubiquitin/proteasome system (rightmost panel, zoom) via recognition of the KEN box by APC/Ccdc20/cdh123. Bottom panel: importin beta overexpression disrupts the spatial control of CENP-F in metaphase, accumulating it at poles, and also causes its persistence in late mitosis. In anaphase, overexpressed importin beta molecules, in addition to localising at the reforming NE, continue in part to interact with CENP-F via the NLS, possibly hindering the recognition of the adjacent KEN box (rightmost panel, zoom) and inhibiting subsequent degradation. Importin beta may also stabilise CENP-F by displacing it at poles, away from sites where the APC/Ccdc20/cdh1 is active. Functional MTs provide physical tracks for importin beta and CENP-F interactions, and inhibiting their polymerisation or dynamics reverses the stabilising effect of importin beta (Created in BioRender).

Full size image

Together, the present results may offer clues towards understanding the pathogenetic effects of rare MCPH mutations affecting the integrity of the C-terminal region of CENP-F: the data presented here suggest C-terminal truncated CENP-F proteins may be defective both in mitotic localisation and in controlled proteolysis, with possible detrimental effects on the mitotic division of neural precursors and in proper organisation of the brain during development.

Methods

Cell cultures and synchronisation

Human HeLa cells (CCL-2, American Tissue Culture Collection) and HeLa importin beta-GFPdox cells with stably integrated importin beta-GFP37, were grown in DMEM (Dulbecco’s Modified Eagle Medium) supplemented with 10% fetal bovine serum, 2% L-glutamine, 2.5% HEPES and 2% penicillin/streptomycin at 37 °C in 5% CO2. Human hTERT RPE-1 cells (in the text, RPE-1) immortalised with hTERT (CRL-4000, American Tissue Culture Collection) were grown in DMEM/F12 (Dulbecco’s Modified Eagle Medium F-12) supplemented as above. Cell cycle synchronisation in mitosis was achieved by adding the CDK1 inhibitor RO3306 (9 µM, Sigma Aldrich SML0569, 20 h) to arrest cells at the G2/M transition, then releasing in drug-free medium to progress through mitosis. Round-up was monitored under an inverted microscope and detaching cells were collected at various time points by mechanical shake-off.

Generation of stable RPE-1 cell lines expressing importin beta-GFP

The inducible epB-Bsd-TT-importin-β-EGFP vector for importin-β-EGFP expression in human cells was described37. Briefly, the vector derives from a piggyBac backbone and carries a tetracycline-responsive promoter element (TRE Tight) followed by a multicloning site linked to an EGFP tag59, within which the sequence encoding human importin-β–EGFP30 was inserted; the puromycin resistance gene in the original epB-Puro-TT was replaced with a blasticidin resistance gene. HeLa cells were generated as reported in37. RPE-1 importin beta-GFPdox cells were generated using the same procedure, i.e. co-trasfecting the epB-Bsd-TT-importin-β-EGFP vector together with the hypb7 plasmid (encoding the transposase gene) using Lipofectamine (Invitrogen 11668019). 24 h after transfection, the medium was replaced with fresh DMEM supplemented with 9 µg/ml blasticidine-S hydrochloride (Sigma Aldrich 15205). After about 2 weeks, single blasticidine-S-resistant foci were manually picked and expanded. Importin beta-GFP expression was tested after administration of 1 µg/ml Doxycyline hyclate at various time-points (Dox, Santa Cruz Biotechnology sc-204734).

RNA interference (RNAi)

The following small interfering (si)RNA duplexes were used: CENP-F: 5' AAGAGAAGACCCCAAGUCAUC 3’ (custom made by Sigma Aldrich; same sequence as in14); KPNA2 (importin alpha-1) (Santa Cruz Biotechnology, sc-35741); GL2 (against the firefly luciferase, with no target in mammalian cells, negative control): 5′-CGUACGCGGAAUACUUCGAtt-3′ (custom made at Ambion by Life Technologies). Final siRNA concentrations were 120 nM for CENP-F and 20 nM for importin alpha. GL2 was used at the same concentration. siRNAs were diluted in serum-free OptiMem (Invitrogen 31985062) and transfected using Lipofectamine RNAiMAX (Invitrogen 13778030).

Immunofluorescence (IF)

Cells were grown on poly-L-lysine (Sigma Aldrich P4832) coated coverslips and were fixed as follows: (a) 3.7% paraformaldehyde (PFA)/30 mM sucrose/PBS (10 min), permeabilisation in 0.1% Triton X-100 (5 min) and incubation in 0.1 M glycine (10 min); (b) incubation in PHEM (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl2, pH 6.9) containing 0.3% Triton X-100 (6 min), followed by PFA fixation as above. Blocking and incubation with antibodies (in PBS containing 0.05% Tween-20/3% bovine serum albumin) were at room T°. Primary antibodies are listed in Supplementary Table 1. Secondary antibodies were conjugated to fluorescein isothiocyanate (FITC), Cy3, Alexa-Fluor 647 (Cy5) or 7-amino-4-methylcoumarin-3-acetic acid (AMCA) (all from Jackson Immunoresearch Laboratories). DNA was stained with 0.1 µg/mL 4,6-diamidino-2-phenylindole (DAPI, Sigma Aldrich D9542). Coverslips were mounted in Vectashield (Vector Laboratories).

Proximity ligation assay (PLA)

PLA was performed in PFA- or PHEM-fixed cells using primary antibodies directed against two proteins of interest (Supplementary Table 1; incubation was for 1 h, room T°). Cells were then incubated with secondary antibodies conjugated to complementary single-stranded oligonucleotide tails (Duolink In Situ PLA Probes, including Anti-Rabbit PLUS, DUO92002, and Anti-Mouse MINUS, DUO92004, Sigma Aldrich, diluted 1:5 in PBS containing 0.05% Tween-20 and 3% bovine serum albumin) in a pre-heated humidity chamber (1 h, 37 °C). Interaction or close proximity between the proteins of interest enables the complementary oligonucleotides to anneal. Subsequent ligation (30 min) and amplification (15 min or higher depending on the analysed protein pairs) were at 37 °C following the Olink Bioscience protocol using Duolink Detection kit DUO92007 orange (Sigma-Aldrich), containing a fluorescent detection probe for the ligated products (excitation wavelength 554 nm, emission wavelength 576 nm), using a Cy3 filter. PLA signals were localised by co-staining for one of the following: DAPI (DNA), CREST (kinetochores), antibodies to mitotic markers as indicated in the text, or cyclin B1 for G2 phase-cells. PLA signals were quantified using the “spot count” function of NIS-Elements AR 5.02 software (Nikon) in regions of interest (RoI), or in whole cells.

Time-lapse videorecording

Cells were seeded in 8-well µ-slide (80821, IbiTreat; Ibidi). Recording began after addition of Dox (1 µg/ml) in cultures kept at 37 °C in a T°- and CO2-controlled stage incubator (Okolab) and continued for 24 h under a Ti Eclipse automated inverted microscope (Nikon) equipped with a DS-Qi1MC camera, an Intensilight C-HGFIE lamp and NIS-Elements 5.11 software (Nikon). Images were taken in phase contrast (60x, 0.7 NA) or under an immersion oil (60x, 1.4 NA) objective: phase every 15 min, GFP fluorescence every 60 min.

Fluorescence microscopy and image acquisition

Single cell images were acquired either under a Nikon Eclipse 90i microscope using an immersion oil 100× objective (NA 1.3) and equipped with a Qicam Fast 1394 CCD camera (Qimaging) followed by deconvolution, or a Nikon Ti2 confocal spinning disk microscope (implemented with Crest X-Light V3 module from CrestOptics) equipped with the Kinetix sCMOS camera (Teledyne Photometrics), a 60× (immersion oil, NA 1.4) objective and CELESTA lasers (Lumencor). Field images were taken under the Nikon Eclipse 90i using a 40× objective (NA 0.75). Image acquisition and analyses were performed using NIS-Elements AR software modules (Nikon). 3D deconvolution (0.4 μm z-serial optical sections) was performed on single cell images using the AutoQuant deconvolution module. Image projections from z-stacks were created using the Maximum Intensity Projection (MIP) for quantitative analyses. IF signals were quantitatively analysed on .nd2 format files with automated external background correction, and the sum intensity of signals within automatically selected RoIs of fixed size was measured. PLA signals were automatically counted on 3D-acquired images, then processed by activating the “spot detection” and “count objects” tools of NIS-Element AR software (Nikon). All figures represent MIP images. Figures were processed with Adobe Photoshop CS 8.0.

Cell extract preparation and Western immunoblotting (WB)

Cells were lysed in RIPA buffer (50 mM Tris–HCl pH 8, 150 mM NaCl, 1% NP40, 1 mM EGTA, 1 mM EDTA, 0.1% SDS, 0.25% sodium deoxycholate) supplemented with protease (Roche 05892791001) and phosphatase (PhoSTOP, 04906837001, Roche) inhibitors. To prepare anaphase cell-enriched extract, cells were synchronised in mitosis as described and collected by mechanical shake-off at progressive time-points, defined by analysing mitotic figures on parallel slides from the same cultures. Whole cell extracts were separated through SDS-PAGE and transferred to nitrocellulose filters (Protran BA83, Whatman). Blocking and antibody incubation were in TBS (10 mM Tris–HCl pH 7.4, 150 mM NaCl) containing 0.1% Tween 20 and 5% low-fat milk (1 h, room T°). Primary antibodies are listed in Supplementary Table 1. HRP-conjugated antibodies (Santa Cruz Biotechnology and BIORAD) were revealed using the ECL system (Amersham by Cytiva) on Hyperfilm-ECL films (GE Healthcare) or using Chemiluminescent Imaging system Azure 300 (Azure Biosystems,).

Drug treatments and microtubule-targeting protocols

Importazole (Sigma Aldrich SML0341, 50 µM) and MG132 (Santa Cruz sc-201270, 10 µM) were added to cultures 2 h before fixing the cells. Microtubule-targeting drugs included: nocodazole (Sigma Aldrich M1404, 0.2 µg/ml), taxol (Paclitaxel, Selleck, S1150, 100 nM) and vinblastine (Sigma Aldrich V1377, 10 nM) and were added for 14 h. In the combination NOC + DOX, nocodazole was added for 9 h alone followed by Dox addition. When used in the combination DOX + NOC, Dox was administered for 9 h alone and nocodazole was then added.

Structure analysis and molecular modelling

Structure visualisation, superposition and analysis, and picture generation, were performed with PyMOL (PyMOL Molecular Graphics System, Version 3.0 Schrödinger, LLC, https://www.pymol.org/). The experimentally determined 3D structures analysed here were downloaded from the Protein Data Bank (PDB), including: importin beta in complex with the non-classical NLS (nc-NLS) of parathyroid hormone-related protein (PTHrP; PDB ID: 1M5N45), and importin beta in complex with the IBB of importin alpha (PDB ID: 8gcn49). The sequences of the c-NLS of CENP-F, nc-NLS of PTHrP and IBB of importin alpha were manually aligned using their positively charged residues as a guide and, for PTHrP and IBB, their spatial position relative to importin beta in the respective experimental structure. A 3D structural model of the importin beta/c-NLS CENP-F complex was built using as a template the crystal structure of importin beta/PTHrP complex. The Modeller 10.6 software60 was used to generate up to 10 models, and PROSA-web to select the best 3D structure of the importin beta and the c-NLS element of CENP-F. APBS61 was used to calculate the electrostatic surface of importin beta. The in-house developed FACE2FACE software (https://face2face.ibpm.cnr.it/) was used to analyse the interactions of importin beta with cargoes. The output of the program reports, for each residue, both the secondary structure, calculated by DSSP62, and the residue surface contributing to the interface, defined as the difference between the solvent accessible surface area of the residue in the free protein and in the complex. Solvent accessible surface areas were calculated using Naccess63.

Statistical analysis

Data were analysed using GraphPad Prism 9. The following tests were employed: Mann–Whitney non-parametric test, or Dunn’s test, to compare continuous values, which in most experiments had a non-normal distribution across samples; the distribution of values measured in single cells is represented in dot plots with the superimposition of bars representing mean values and standard deviation; Student’s t test, to compare mean values between two independent groups; Chi-square for trends test, to evaluate the association between a nominal variable with two levels (say CTR vs CHX) and an ordinal variable (say the length of exposure time).