Regulation of Cenp-F localization to nuclear pores and kinetochores

ABSTRACT

In metazoans, the assembly of kinetochores on centrometric chromatin and the dismantling of nuclear pore complexes are processes that have to be tightly coordinated to ensure the proper assembly of the mitotic spindle and a successful mitosis. It is therefore noteworthy that these two macromolecular assemblies share a subset of constituents. One of these multifaceted components is Cenp-F, a protein implicated in cancer and developmental pathologies. During the cell cycle, Cenp-F localizes in multiple cellular structures including the nuclear envelope in late G2/early prophase and kinetochores throughout mitosis. We recently characterized the molecular determinants of Cenp-F interaction with Nup133, a structural nuclear pore constituent. In parallel with two other independent studies, we further elucidated the mechanisms governing Cenp-F kinetochore recruitment that mainly relies on its interaction with Bub1, with redundant contribution of Cenp-E upon acute microtubule depolymerisation. Here we synthesize the current literature regarding the dual location of Cenp-F at nuclear pores and kinetochores and extend our discussion to the regulation of these NPC and kinetochore localizations by mitotic kinase and spindle microtubules.

Introduction

Intricate relationships between kinetochores and nuclear pore constituents

The kinetochores are elaborated structures that assemble on the centromeric chromatin to connect chromosomes to the microtubules of the mitotic spindle. Thanks to the coordinated action of hundred of proteins, the kinetochores provide a dynamic interaction with the spindle microtubules and a quality control mechanism, the mitotic checkpoint, that ensures that each chromosome is properly attached to spindle microtubules before anaphase onset. These two parameters allow for the faithful segregation of chromosomes between the daughter cells in mitosis (reviewed in [1–4]). For metazoans that undergo open mitosis, spindle assembly further requires the dismantling of the nuclear envelope (NE) and of its constituents, including the nuclear pore complexes (NPCs) (reviewed in [5,6]). The NPCs are, like kinetochores, huge macromolecular assemblies, composed of hundreds of proteins called nucleoporins (or Nups) that form transport channels embedded in the NE (reviewed in [7,8]). While NPCs are critically required to allow bidirectional trafficking between the nucleoplasm and the cytoplasm during interphase, some nucleoporins are also localized at kinetochores where they participate to proper cell division (reviewed in [9]). This is notably the case for the constituents of the Nup107–160 complex (also known as Y-complex for its peculiar shape) one of the major structural units of the NPC that is composed of nine different Nups, including Nup133 [10–12]. Conversely, some kinetochore proteins interact with nucleoporins and are thus recruited to the NPCs, either throughout the entire interphase (as the case for the checkpoint proteins, Mad1 and Mad2, that are tethered to the nuclear side of the NPC [13–15]) or only in G2, prior to nuclear envelope breakdown (NEBD) (reviewed in [16]).

Cenp-F, a multifaceted “kinetochore” constituent with cell cycle regulated NE localization

One of the few “kinetochore” proteins displaying NE localization in G2 is Cenp-F (centromeric protein F), a large microtubule binding protein also known as mitosin in human and Lek1 in mouse. Proteins with homology to Cenp-F have been identified in organisms from yeast to human [17–21]. In mammalian cells, Cenp-F was shown to interact with a variety of partners (Figure 1) that may account for its implication in multiple cellular processes ranging from cell division to gene regulation, vesicular transport, cell morphogenesis and ciliogenesis, (reviewed in [20,22,23] see also [24–26]). Strong interests were raised for this multifunctional protein because its overexpression has been correlated to various forms of cancer whereas its inactivation or mutations were shown to contribute to diverse developmentally-related disorders (reviewed in [22,23] see also [25–28]).

Figure 1.

Functional organization of Cenp-F and homodimerization properties of its minimal domains of interaction with Nup133 and Bub1.

Schematic representation of Cenp-F full-length (top) and magnification of its C-terminal domain reporting the known binding domains based on data collected for human and mouse Cenp-F (adapted from [42]). The human (hCenp-F) and mouse Cenp-F (mCenp-F) amino acid (aa) numbering and domains nomenclature are written in black and grey, respectively. Note that two hCenp-F aa sequences co-exist in the literature: the sequence annotated in NCBI (NP_057427.3, used in [42]) is 3114 aa long while the one available in UniProt (UniProtKB – P49454, used in [44,47]) is 3210 aa long. These two human sequences differ by a 96 aa-long insertion (represented by a black triangle above the full-length Cenp-F schematics). The 115 aa-long segment found in human but not in mouse Cenp-F (NCBI reference sequence NP_001074832.2) is represented by a grey triangle. Binding sites for microtubules [29,30,58], Cenp-E [65], NudE/NudEL [39], Nup133 [42] and Bub1 [44] that overlap with the previously identified kinetochore core “KT core” binding site [43], nuclear localization signal (NLS) [17,19,47], Rb [50] and Miro [32] binding domains, KEN7 degradation motif (KEN) [37] and farnesylation site (F) [34] are indicated. The continuous vertical lines indicate the positions of the leucines that constitute the three leucine zippers present in the C-terminal domain of Cenp-F. The dashed vertical lines within the NLS indicate the positions of the three aa that were reported to be phosphorylated by Cdk1 in vitro [47]. The yellow crosses below Nup133 and Bub1 binding sites indicate the position of the mutations specifically inhibiting the interaction of Cenp-F with each of these two partners. The different Cenp-F constructs that were tested for homodimerization and/or interaction with Nup13 and Bub1 are reported in the lower part of the picture. The table on the right synthesizes the results. “SEC” stands for size-exclusion chromatography, “SM” for SEC-MALS (size-exclusion chromatography-multi-angle light scattering), “Y2H” for yeast-2-hybrid and “CL” for in vivo crosslinking experiments. The red asterisks indicate the positions of the L2668G/L2696G mutations affecting mCenp-F homodimerization. Note that although these two mutations are positioned around the first leucine zipper, in the binding site for Nup133, they also impair the homodimerization of the larger mCenp-F construct that encompasses the binding site for Bub1 and the second leucine zipper, as revealed by crosslink experiments.

Cenp-F expression and localization are highly regulated during the cell cycle [18,19,31]: while not detectable in G1 cells, Cenp-F expression increases and reaches its maximal in G2 when it is mainly localized in the nucleoplasm, with a minor pool present at the centrosomes and mitochondria [32,33]. In late G2/early prophase, prior NE breakdown (NEBD), Cenp-F becomes detectable at the NE [34–36], just before being recruited to kinetochores [18,19]. Electron microscopy experiments further revealed that Cenp-F localizes on the outermost domain of the kinetochores, the fibrous corona. The latter can expand into crescent shaped structures around unattached kinetochores, as notably reflected by Cenp-F localization upon prolonged treatments with nocodazole, a drug that depolymerises microtubules [18,19]. Cenp-F remains localized at kinetochores until early anaphase, a time point at which it relocalizes to the spindle midzone [18,19,31]. Finally, Cenp-F is degraded after anaphase, a process that involves a C-terminal KEN-box [37]. The large size and the multiplicity of Cenp-F localizations and partners along the cell cycle have long hampered the fine dissection of its functions. Here we analyse and discuss recent studies in which we and others have dissected the molecular determinants contributing to the timely mitotic recruitment of Cenp-F to the nuclear envelope and kinetochores.

Cenp-F nuclear envelope localization

A key albeit redundant function for Cenp-F in the recruitment of cytoplasmic dynein/dynactin to the NE in late G2/prophase

We previously reported that in late G2/early prophase Cenp-F initiates a molecular chain on the cytoplasmic side of the NE, that begins with the recruitment of NudE/NudEL. These two proteins are also localized via Cenp-F to kinetochores, where they subsequently recruit the cytosolic minus-end molecular motor dynein and its regulator dynactin [38,39]. Likewise, NudE/NudEL recruits dynein and dynactin to the NE in prophase [36]. This molecular network, along with another independent pathway involving the RanBP2/Nup358 nucleoporin and BicD2 [40], ensures the strength and the timing of dynein recruitment to the NE. Once anchored at the NE, dynein is critical to move and position the centrosome in proximity of the NE prior NEBD in HeLa cells. These two dynein-anchoring pathways also participate to the so-called “interkinetic movements” of the nuclei from radial glial progenitor cells in the rat brain. This cell type spans the developing cortex and displays cell cycle-dependent nuclear migration that, during G2, is directed from the apical towards the ventricular surface of the cortex where the centrosomes of these progenitor cells are positioned. In these cells, G2 nuclei behave as cargoes for dynein. Inhibition of NE recruitment of dynein blocks the migration of G2 nuclei along microtubules towards the ventricular surface in G2, and consequently prevents mitotic entry [41].

Refining the molecular determinants of Cenp-F – Nup133 interaction

Our previous study revealed that in late G2/early prophase, prior NEBD, Cenp-F localization at the NE largely relies on its interaction with Nup133 [36]. Using a combination of in silico predictions and yeast-2-hybrid (Y2H) assays followed by in vivo studies, we recently refined the molecular details underlying Cenp-F – Nup133 interaction.

The binding site of Nup133 for Cenp-F was mapped on a conserved α-helix (Nup133-α1, aa 87–99 of mNup133, aa 88–100 of hNup133) that protrudes outside of the N-terminal β-propeller domain of Nup133 [42]. Based on in silico modeling, mutations impairing mCenp-F – mNup133 interaction were designed within mNup133-α1 (V89D/M92D/T96D and E93R) and validated by Y2H assays. This Nup133-α1 helix interfaces Cenp-F on a short polypeptide encompassing a leucine zipper (aa 2655–2723 of mCenp-F, aa 2799–2854 of hCenp-F¹) that overlaps with its “kinetochore-core” (“KT-core”) binding domain [43] (Figure 1). In silico predictions, validated by biochemical and in vivo studies, indicated that this minimal Cenp-F peptide, as well as segments of Cenp-F spanning its “KT-core” and including a second leucine zipper, homodimerize in a parallel coiled-coil structure [19,42,44]. Importantly, homodimerization is essential for Cenp-F interaction with Nup133 because it is the dimer that provides the contact surface for Nup133-α1 helix [42] (Figure 1). Functional studies in HeLa cells revealed that insertion within GFP-hCenp-F of the L2797E/I2799E mutations that specifically impair Cenp-F interaction with Nup133 but not its homodimerization, largely prevented its NE localization in the majority of late G2 cells. However, some of the transfected cells displayed a faint GFP signal at the NE [42]. This residual NE localization of GFP-Cenp-F L2797E/I2799E is probably due to its homodimerization with endogenous Cenp-F. The L2797E/I2799E mutation, despite specifically impairing Cenp-F – Nup133 interaction, is thus not expected to cause a dominant phenotype. Future functional studies in cells or animals will therefore require either ectopic expression of this Cenp-F mutant allele along with depletion of endogenous Cenp-F or the genetic engineering of both copies of endogenous Cenp-F.

We previously hypothesized that mitotic phosphorylation of Nup133 and/or Cenp-F might regulate their interaction and thereby restrict the recruitment of Cenp-F to the NE to late G2/prophase cells [36]. However, having refined the binding interface between Nup133 and Cenp-F, we noticed that none of the so-far-identified phosphorylated residues are directly located within the defined interaction domains of these two proteins. Moreover, we observed that a C-terminal domain of Cenp-F fused to GFP (GFP-mCenp-F-C2, that comprises the binding site for Nup133, its “KT-core” domain and its nuclear localization sequence -NLS, Figure 1), is recruited in a Nup133-dependent manner to nuclear bodies (called GLFG bodies for the presence of the GLFG-repeat-containing nucleoporin Nup98 [45]). Because this localization to nuclear bodies takes place throughout interphase, this indicated that, unlike anticipated, Cenp-F – Nup133 interaction does not directly relies on cell-cycle-dependent post-translational modifications affecting their binding interface [42]. Rather, it suggested a crucial role of other cell-cycle dependent processes in the regulation of Cenp-F – Nup133 interaction.

Spatiotemporal regulation of Cenp-F intranuclear/NE localization

Because the so far established functions of Cenp-F at the NE rely on the recruitment of cytoplasmic dynein, the current (albeit not demonstrated) model posits that Cenp-F is exported from the nucleus in late G2, and that its cytoplasmic accumulation would then enable its subsequent recruitment to the cytosolic face of the NPCs (Figure 2). Phosphorylation events, while not directly affecting the Nup133-Cenp-F interface, were shown to underlie the cell cycle-dependent relocalization of Cenp-F from the nuclear interior to the NE. Indeed, Baffet et al. showed that upon treatment with roscovitine, an inhibitor of cyclin-dependent kinase 1 (Cdk1) activity, Cenp-F retains a nucleoplasmic localization in late G2/early prophase and never localizes on the NE [35]. However, the Cdk1 substrate(s) involved in this process were unknown.

Figure 2.

Potential mechanisms by which Cdk1 may enable the NE localization of Cenp-F in late G2/early prophase.

Schematic representation of nuclear pores embedded in the NE, with the Y-complex (including Nup133) highlighted in red. (A) In late G2/early prophase, homodimeric Cenp-F localizes to the nuclear pore complexes thank to its interaction with Nup133 [36,42], with possible contribution of its farnesylation [34]. (B) Cdk1-dependent phosphorylation of Cenp-F-NLS may inhibit its nuclear import thereby favoring Cenp-F – Nup133 interaction on the cytoplasmic face of the NE [47]. (C) Cdk1-dependent phosphorylation may enable the release of Cenp-F from intranuclear sites (possibly on the chromatin). (D) NES-dependent nuclear export of Cenp-F may be favoured by Cdk1. (E) Cdk1-dependent phosphorylation of nucleoporins may alter NPC organization and thereby allow Cenp-F – Nup133 interaction. Inset on the bottom right: Unlike endogenous Cenp-F, the mCenp-F-C2 domain (see Figure 1) localizes in a diffuse manner in the nucleoplasm and to GLFG bodies in a Nup133-dependent manner during interphase and was never detected at the NE.

A subsequent study revealed that Cdk1 can phosphorylate in vitro three residues that are located within the bipartite classical NLS of hCenp-F (T2946, T2949 and S2952, based on NCBI sequence [46,47], see also [19] and [17] for initial descriptions of Cenp-F NLSs) (Figure 1). In turn, these phosphorylation events were reported to inhibit the binding of Cenp-F NLS to karyopherin α, the nuclear import adaptor for classical NLSs (reviewed in [48]) (Figure 2(b)). However, a GFP-GST-Cenp-F-NLS construct bearing phosphomimetic mutations (T2946D, T2949D and S2952D) was still imported in the nucleus, albeit slightly less efficiently than the corresponding wild type construct [47]. It was proposed that these phosphomimetic amino acids might be less potent inhibitors of interaction as compared to the actual phosphorylated residues. In addition, the authors hypothesized that a mild decrease in the efficiency of nuclear import of Cenp-F, if combined with active nuclear export, may lead to its cytoplasmic accumulation [47]. In the future, characterization of a T2946A/T2949A/S2952A mutant preventing phosphorylation of full length Cenp-F should help to decipher the contribution of Cdk1-dependent nuclear import inhibition to Cenp-F NE localization.

In addition, Cdk1-dependent phosphorylation events may also regulate other targets that would affect the compartmentalization of Cenp-F and ultimately lead to its NE localization in late G2 (Figure 2): as already proposed by Zhu in 1999 [43], and consistent with the presence of a ATF4 and Rb-binding domain within its C-terminal domain [49,50], (Figure 1), Cenp-F might be bound to chromatin during interphase. This may restrict its diffusion within the nucleoplasm, possibly explaining its exclusion from the nucleoli and GLFG bodies [42]. Phosphorylation events may thus enable the timely release of Cenp-F from intranuclear anchors, a pre-requisite for its subsequent nuclear export (Figure 2(c)). Nuclear export of Cenp-F could itself be regulated by the accessibility or activity of a nuclear export signal (NES) (Figure 2(d)). Of note, the GFP-mCenp-F-C2 domain, despite being able to interact with Nup133 and to diffuse within the nucleus, was never detected at the NE [42]. Because the NES predicted by Loftus and colleagues [47] is absent from mCenp-F-C2, the fact that this fragment cannot be actively exported from the nucleus may explain why its recruitment to the cytoplasmic face of the NE is prevented. One should however note that this mCenp-F-C2 domain also lacks the C-terminal CAAX box that enables farnesylation of Cenp-F, a modification that was shown to also contribute to the NE localization of Cenp-F [34] (Figure 2(a)). Whether Cenp-F farnesylation contributes to its association with NE lipids or facilitates its interaction with proteins from the NPC, and whether this step might be regulated by Cdk1-dependent mechanisms has not been investigated so far. Finally, Cdk1 also plays a key function in the phosphorylation-dependent disassembly of the NPCs in mitosis (reviewed in [5]). One can thus imagine that a partial phosphorylation of the NPCs prior NEBD might favor a transiently relaxed conformation of the NPCs that, in turn, would allow or favor Cenp-F – Nup133 interaction (Figure 2(e)).

Of note, the reason why neither GFP-Cenp-F-C2 nor Cenp-F full-length would interact with the fraction of Nup133 molecules that localizes on the nucleoplasmic side of the NPCs remains unknown. In view of the inherent asymmetry of the NPCs, that feature distinct cytoplasmic and nuclear filamentous appendages (Figure 2), one may hypothesize that asymmetric nucleoporins sterically hinder the α1-helix of Nup133 on the nuclear side of the NPCs. Alternatively, this region of Nup133 might be differentially post-translationally modified on the nucleoplasmic versus the cytoplasmic side of the NPCs. These assumptions, as well as the hypothesis that Cenp-F NE localization is restricted to the cytoplasmic side of the NPCs deserve to be experimentally verified.

Cenp-F kinetochore localization

Cenp-F features a multiplicity of domains and partners and still debated function(s) at kinetochores

Shortly after being detected at the NE, Cenp-F begins to be recruited to kinetochores in early prophase. Cenp-F kinetochore localization was first shown to rely on its “KT core” domain (aa 2756–2901 of hCenp-F, that encompasses two leucine zippers) and more specifically to be dependent of a very conserved cysteine located in the second leucine zipper of this domain (C2865 in human and C2748 in mouse [19,43]). An internal repeated region (aa 2094–2488 in hCenp-F) that encompasses the NudE/NudEL binding site of Cenp-F was shown to facilitate the kinetochore localization of the “KT core” domain [39,43]. However this internal repeated sequence is not sufficient to target Cenp-F to kinetochores. Cenp-F-NLS was also shown to indirectly contribute to its kinetochore recruitment; indeed a mutant form of Cenp-F lacking the NLS and therefore localizing in the cytoplasm prior to NEBD is not efficiently recruited to kinetochores after NEBD [43]. It was thus hypothesized that Cenp-F proper localization in metaphase requires its early recruitment to kinetochores before NEBD. Finally, inhibition of Cenp-F farnesylation was shown to affect Cenp-F kinetochore localization, possibly because this post-translational modification contributes to protein-protein interactions [34,44].

Functional studies performed in DLD-1 or HeLa human cell lines further revealed that RNAi-mediated depletion of several kinetochore constituents including Bub1 and Cenp-E [51], RanBP2/Nup358 [52], CENP-I [53], Zwint-1 [54], or Sgt1 [55] impair, to various extent, the kinetochore localization of Cenp-F [51,56]. On the other hand, such knockdown experiments have also revealed the yet controversial contribution of Cenp-F to kinetochore-microtubule attachment and/or to the robust activation of the spindle assembly checkpoint [17,44,57–61]. Nonetheless, no major viability or proliferation defects were observed upon Cenp-F knockout in mouse embryonic fibroblasts, HeLa or HAP1 cells [62–64] possibly reflecting redundancy in its functions leading to cell adaptation.

Bub1 recruits Cenp-F to kinetochores in early prophase

Although as detailed above many proteins were shown to contribute to Cenp-F kinetochore localization, its direct kinetochore tether(s) has long remained unknown. Among the candidates already localized at kinetochores in early prophase and thus susceptible to explain the early stages of Cenp-F recruitment to this structure, Bub1 was the only protein suggested to directly interact with Cenp-F [65]. However, the experimental demonstration and the molecular description of this interaction were missing. Recently, three independent studies [42,44,64] demonstrated that Bub1 is the direct partner responsible for Cenp-F kinetochore localization. Bub1 was shown to establish a direct contact with Cenp-F through its C-terminal kinase domain [44,64]. Furthermore, deletion of the very last 21 aa of Bub1, that form a conserved α-helix, almost completely abolished Cenp-F kinetochore localization whereas inhibition of Bub1 kinase activity did not [64]. Conversely, Cenp-F was shown to contact Bub1 via its “KT core” domain [42,44]. Both Y2H assays and size-exclusion chromatography experiments further revealed that mutation of the cysteine located within the leucine zipper of the “KT-core” domain and previously reported to be important for Cenp-F kinetochore localization [43] leads to a strong inhibition of Cenp-F – Bub1 interaction [42,44].

While the dimerization of the Cenp-F segment that binds Bub1 is well established [19,42,44] (Figure 1), its importance to generate a binding interface for Bub1 remains to be determined. Indeed, using size-exclusion chromatography, Ciossani et al. observed that a short domain of Cenp-F, encompassing only the second leucine zipper of its “KT core” domain, neither dimerizes nor interacts with Bub1 [44]. However, as discussed by the authors, this result may simply reflect the independent requirement of residues located upstream of this second leucine zipper to stabilize the short dimer and to create the binding interface for Bub1. On the other hand, we showed using crosslink experiments in HeLa cells that the C2 domain of Cenp-F (that includes its “KT-core” domain) is able to dimerize and that mutation of two leucines (L2668G – L2696G), located around its first leucine zipper, impairs its homodimerization (Figure 1). However, these mutations did not prevent Cenp-F – Bub1 interaction in the Y2H assay. Nonetheless GFP-mCenp-F-C2 L2668G – L2696G was never detected at kinetochores, suggesting that, at least in vivo, Cenp-F homodimerization is an essential condition for GFP-mCenp-F-C2 – kinetochore interaction to happen [42]. A possible explanation to this discrepancy might be that, although Bub1 is able to interact with monomeric Cenp-F, a Cenp-F dimer needs to be bound by two molecules of Bub1 to be efficiently recruited to kinetochores in prophase. Alternatively, the different outcome of the in vivo and Y2H studies might reflect the overexpression of the proteins in the Y2H assay. In this context, a local homodimerization of the second leucine zipper of Cenp-F may allow a residual Cenp-F – Bub1 interaction (see Figure 1 for the position of the L2668G – L2696G mutations in comparison to the position of the leucine zipper involved in Bub1 interaction). In contrast, the local environment of mitotic cells could challenge the residual homodimerization of GFP-mCenp-F-C2 protein and therefore abolish its interaction with Bub1 and consequently its kinetochore localization. In the future, defining mutations within the second leucine zipper preventing Cenp-F local dimerization may help to solve this issue.

Contribution of Cenp-E to Cenp-F kinetochore localization

Although Bub1 depletion completely prevents Cenp-F kinetochore localization [51], the C2865S mutation that impairs Cenp-F – Bub1 interaction only caused a partial displacement of GFP-hCenp-F C2865S from kinetochores as compared to GFP-hCenp-F wild type, even when HeLa cells were depleted of endogenous Cenp-F [42]. This result suggests that Bub1 is not the only partner contributing to Cenp-F kinetochore recruitment. Johnson and colleagues previously reported that in DLD-1 cells depletion of Cenp-E, which localizes at kinetochores from prometaphase on (that is at a later stage than Cenp-F [66]), impaired the kinetochore localization of endogenous Cenp-F [51]. Cenp-E is a probable paralog of Cenp-F that however contains a motor domain promoting plus end-directed motion [44,67,68]. In our study we showed that depletion of Cenp-E nearly entirely abolishes GFP-hCenp-F C2865S kinetochore localization while not affecting the localization of GFP-hCenp-F [42]. These results thus point to Cenp-E contributions to Cenp-F kinetochore localization, albeit to various extent possibly due to the analysis of endogenous versus transfected and GFP-tagged Cenp-F. This is however in contrast with the studies of Raaijmakers and colleagues and from Ciossani and colleagues [44,64]. Indeed using Bub1 deletion mutants lacking the binding site for Cenp-F (i.e., deleting the entire kinase domain or just the last 21 aa of Bub1) these two groups came to the conclusion that preventing the direct interaction between Bub1 and Cenp-F is sufficient to abolish Cenp-F kinetochore localization. Moreover, these Bub1 mutants did not detectably affect Cenp-E localisation at kinetochores. While all these studies were performed on nocodazole-treated cells to depolymerize microtubules and increase the fraction of cells in mitosis, we noticed that our study used the highest nocodazole concentration. This suggests that the contribution of Cenp-E to Cenp-F kinetochore localization might become detectable only upon acute microtubule depolymerization. Interestingly, while the recruitment of Cenp-F and Cenp-E to kinetochores was shown to be independent from the assembly of the Rod/Zwilch/ZW10 (RZZ) complex, the main constituents of the expandable kinetochore corona ([44,69], Preprint [70]:) this RZZ complex is required for the assembly of Cenp-F and Cenp-E in crescent-like structures when cells are treated with nocodazole ([69], Preprint [70]:, see also Figure 2 in [44]). Ciossani and colleagues [44] also observed that depletion of Cenp-E somehow affects Cenp-F crescent formation (detected in their Figure 2(h); A. Musacchio, personal communication). Together these data suggest that Cenp-E might have a minor role in Cenp-F kinetochore recruitment or stabilization in prometaphase [44,71] that would become substantial only when Cenp-F – Bub1 interaction is inhibited and upon treatment with high nocodazole concentrations [42].

In line with these findings, Sacristan and colleagues recently reported that in HeLa cells treated with nocodazole, depletion of the RZZ complex leads to a decreased level of Hec1, a component of the Ndc80 complex that localizes to the outer kinetochores [69]. Importantly, we also observed that nocodazole-treated HeLa cells display fainter Hec1 staining when Cenp-E is depleted (see raw data from Table S2 in [42]). These results indicate that upon acute microtubule depolymerisation, proteins localizing on the external layers of kinetochores contribute to the recruitment or stabilization of proteins localizing to more inner layers. A similar and complex interplay might be taking place also in the case of Cenp-E-dependent Cenp-F kinetochore localization.

Concluding remarks

Here we have summarized and discussed the outcome of various studies revealing how Nup133 recruits Cenp-F on the nuclear envelope in late G2/early prophase and how Bub1, along with Cenp-E, localizes Cenp-F to kinetochores in mitosis.

As mentioned in the introduction, the dual localization of Cenp-F at NPCs and kinetochores is not unique, and it is noteworthy that the Y-complex, including Nup133, is recruited to mitotic kinetochores [11]. This localization was shown to mainly relies on the Ndc80 complex with however some contribution of the Cenp-F – Nup133 interaction, notably for the recruitment of the Y-complex to the expanded crescent upon nocodazole treatment [12]. Our recently refined description of Cenp-F – Nup133 interaction [42], when combined in the future with the characterization of the yet-unknown Ndc80-dependent tethers of the Y-complex to kinetochores, should be instrumental to further dissect the kinetochore function(s) of these nucleoporins (reviewed in [16] see also [72]).

To unravel Cenp-F mitotic functions, we and others tried to identify Cenp-F partners contributing to its kinetochore recruitment. These studies have revealed the complexity of the Cenp-F kinetochore localization that is modulated by the mitotic environment. As discussed above, interaction networks at kinetochores are critically affected by the presence/absence of microtubules. In addition, Ciossani et al. reported that the kinase activity of Aurora B, Mps1 and Plk1 is required for the efficient kinetochore localization of Cenp-F [44]. Although the authors speculated that this effect might be indirect and possibly mediated by Bub1, this observation highlights protein phosphorylation as another regulation level controlling Cenp-F kinetochore localization.

In view of the multiplicity of proteins and post-translational modifications contributing to Cenp-F recruitment/stabilization at kinetochores, we believe that functional studies critically require the precise molecular description of Cenp-F interface with its interaction partners. For instance, modeling approaches, similar to those used to dissect Cenp-F – Nup133α1 helix interaction, could be used to refine the interface between Cenp-F and the 21aa-long C-terminal α-helix of Bub1. Such studies might further help to characterize other Bub1 partners underlying the recently identified role of this short α-helix in chromosome congression [64].

Although our understanding of the Cenp-F interaction network at kinetochores it is still incomplete, the recent studies discussed here have now shed some light on the Cenp-F complex localization pattern that will be instrumental to solve some of the controversy regarding its functions.

Funding Statement

This work was supported by the Centre national de la recherche scientifique (CNRS), the Fondation pour la Recherche Médicale [DEQ20150734355, “Equipe FRM 2015”]; Labex Who Am I? (ANR-11-LABX-0071; Idex ANR-11-IDEX-0005-02)[Transitional post-doc fellowship to AB]; Ligue Nationale contre le Cancer [Fourth year PhD fellowship to AB]; Ministère de l’Enseignement Supérieur et de la Recherche [PhD fellowship to AB].

Acknowledgments

We are grateful to B. Palancade for critical reading of the manuscript. We also acknowledge R. Gassman and A. Musacchio for fruitful discussions.

Disclosure statement

No potential conflict of interest was reported by the authors.