The nuclear F-actin interactome of Xenopus oocytes reveals an actin-bundling kinesin that is essential for meiotic cytokinesis

Abstract

Nuclei of Xenopus laevis oocytes grow 100 000-fold larger in volume than a typical somatic nucleus and require an unusual intranuclear F-actin scaffold for mechanical stability. We now developed a method for mapping F-actin interactomes and identified a comprehensive set of F-actin binders from the oocyte nuclei. Unexpectedly, the most prominent interactor was a novel kinesin termed NabKin (Nuclear and meiotic actin-bundling Kinesin). NabKin not only binds microtubules but also F-actin structures, such as the intranuclear actin bundles in prophase and the contractile actomyosin ring during cytokinesis. The interaction between NabKin and F-actin is negatively regulated by Importin-β and is responsive to spatial information provided by RanGTP. Disconnecting NabKin from F-actin during meiosis caused cytokinesis failure and egg polyploidy. We also found actin-bundling activity in Nabkin’s somatic paralogue KIF14, which was previously shown to be essential for somatic cell division. Our data are consistent with the notion that NabKin/KIF14 directly link microtubules with F-actin and that such link is essential for cytokinesis.

Introduction

Keeping the chromosome number of a species constant over generations is a critical aspect of genetic stability. Before merging to a zygote, female and male germ cells therefore half their chromosome sets in meiosis—a process comprising two consecutive cell divisions without intervening DNA replication.

While sperm and egg contribute equal numbers of chromosomes, it is the egg that donates most of the other materials needed by the developing early embryo. Oocytes (the precursors of eggs) therefore stockpile proteins, RNAs and lipids while being arrested in prophase of the first meiotic division cycle. As a consequence of such stockpiling, oocytes grow to a tremendous size, in the case of X. laevis to ∼1.3 mm in diameter. Nuclei of these oocytes reach such gigantic dimensions (diameter ∼400 μm) that they require a nuclear F-actin structure for mechanical stability (Bohnsack et al, 2006). This structure can assemble only because the Exportin 6 pathway, which normally depletes actin from nuclei (Stüven et al, 2003), is disabled in Xenopus oocytes (Bohnsack et al, 2006).

A wealth of information is available about the regulation of the various cytoplasmic F-actin assemblies: specific sets of accessory proteins nucleate filaments, organise protofilaments into higher-order structures, or mediate a crosstalk with other cellular activities (Pollard and Cooper, 2009). In contrast, we know very little about the actin structure inside oocyte nuclei. Its interactors are largely unknown and it was even questioned whether this structure is really based on F-actin (Gall, 2006; Jockusch et al, 2006).

The maturation of the oocyte into a fertilisable egg involves dramatic rearrangements of actin and microtubule structures. It is initiated by a progesterone trigger, followed by nuclear envelope breakdown (NEBD), which makes the chromosomes accessible for spindle formation. In Xenopus oocytes, a disk-shaped transient microtubule array (Huchon et al, 1981; Gard, 1992), which also involves F-actin (Gard et al, 1995), is used for initial chromosome collection. The transient microtubule array compacts and evolves to the first meiotic spindle (MI spindle). This spindle is then anchored to the oocyte cortex and rotated into an orthogonal position (Gard et al, 1995).

Homologous chromosomes are segregated and one set is discarded into a membrane extrusion that eventually forms a small extra cell called ‘polar body (PB)’. The membrane extrusion occurs after anaphase onset through Cdc42-dependent F-actin polymerisation (Zhang et al, 2008; Leblanc et al, 2011) at the site of spindle attachment to the cortex. A RhoA-controlled, contractile F-actin/myosin II ring then encompasses the membrane extrusion (Zhang et al, 2008) and drives segregation of the PB similar to somatic cell division (Green et al, 2011; Liu, 2012; Maddox et al, 2012). The extreme asymmetry of this division preserves the valuable cytoplasm of the oocyte.

The chromosomes remaining in the egg are assembled into the second meiotic (MII) spindle and the egg awaits fertilisation in metaphase II arrest. Thereafter, sister chromatids are segregated and a second PB is extruded. The now reduced (haploid) chromosome set forms the female pronucleus, which eventually fuses with the male complement to yield the first nucleus of the embryo.

In this study, we developed a new method for identifying actin interactors and applied it to compartment-specific extracts from Xenopus oocytes. This revealed a comprehensive set of nuclear actin binders, including several conventional F-actin interactors known to nucleate or crosslink F-actin. These are likely to contribute to mechanical stability of the giant oocyte nuclei. We also identified a rather unexpected F-actin binder, namely a novel kinesin, which we named NabKin (Nuclear and meiotic actin-bundling Kinesin). As kinesins usually bind to and move along microtubules, NabKin was a prime candidate for connecting actin and microtubules following NEBD. During meiosis, NabKin indeed colocalised with microtubule and F-actin structures, such as the transient microtubule array, meiotic spindle, nascent PBs or the contractile actomyosin ring. The NabKin–actin interaction is favoured by RanGTP, which also controls several other meiotic processes (Kalab et al, 2011). Intriguingly, cytokinesis failed when the NabKin–actin interaction was blocked. KIF14, the somatic paralogue of NabKin, is known to be required for mitotic cytokinesis and to bind the central spindle (Carleton et al, 2006; Gruneberg et al, 2006). We also identified an actin-binding domain in KIF14 and show that a fraction of KIF14 localises to the actomyosin ring during mitotic cytokinesis. Taken together, our results suggest that these kinesins coordinate interactions between the actin and microtubule cytoskeleton, and thereby contribute to faithful cytokinesis.

Results

A matrix for identifying highly substoichiometric F-actin interactors

To identify factors that organise and maintain the F-actin network inside the Xenopus oocyte nucleus, we developed a novel method based on phalloidin (Löw and Wieland, 1974). The highly specific binding of phalloidin to F-actin has been extensively utilised to visualise F-actin structures in situ. However, immobilised phalloidin is not commonly used for biochemical isolation of F-actin and F-actin-binding proteins. Possible reasons are the lack of a convenient coupling group at the toxin and the typically low abundance of individual interactors in an F-actin network, which severely complicates their detection in a background of proteins that non-specifically bind to a stationary phase.

We solved these problems by a four-step chemical synthesis, which couples phalloidin through an ∼300 Å polyethylene glycol (PEG) linker to silica-coated superparamagnetic beads, followed by passivation of the remaining silica surface with a dense PEG layer (Figure 1A and Supplementary Figure S1). This yielded an affinity matrix that combines high capacity for F-actin with an exceptionally low non-specific background binding (Figure 1C). As an example for the application of this new matrix, we have studied the composition of nuclear F-actin structures in Xenopus oocytes. The methodology, however, is suitable for a far wider range of applications, including comparison of F-actin interactomes between species, tissues, cells, cellular fractions, cell cycle positions or biochemical states.

Figure 1.

Compartment-specific F-actin interactomes of Xenopus oocytes. (A) Scheme of the phalloidin-based affinity matrix. Phalloidin is shown in blue, the PEG–silane-based linker in green, the solid phase in black and the passivating PEG coat in orange. The control matrix contains a methoxy-PEG group instead of phalloidin. See Supplementary Figure S1 for more details. (B) Scheme for preparing clean nuclear and cytoplasmic oocyte extracts. Oocytes were punctured with forceps at the animal pole and nuclei were isolated into aqueous buffer by application of mild pressure to the oocyte’s equatorial plane. (C) F-actin complexes isolated from nuclear and cytoplasmic extracts, analysed by SDS–PAGE/Coomassie staining. See Table I for complete list of identified nuclear F-actin interactors. (D) Nuclear and cytoplasmic fractions were immunoblotted with antibodies raised against identified nuclear F-actin interactors. Tubulin, eEF1A and Arp2 served as cytoplasmic markers; nucleoplasmin is a nuclear marker. Since the loads had been normalised to the compartments’ volumes (nucleus ∼50 nl, yolk-free cytoplasm ∼500 nl), the blots directly compare nuclear and cytoplasmic concentrations. (E) Immunoblot analysis of proteins bound from a nuclear extract to phalloidin beads. NabKin and Supervillin bound strongly to the matrix; binding of nucleoplasmin was undetectable. (F) Immunoblot analysis of NabKin expression in various oocyte and embryo developmental stages. eEF1A served as loading control.

The predominant nuclear actin binder from Xenopus oocytes is a kinesin

We then dissected Xenopus oocytes (∼300 oocytes per experiment) individually into clean cytoplasmic and nuclear fractions (Figure 1B), solubilised F-actin along with associated proteins by a Latrunculin A treatment, removed any insoluble material by centrifugation and, finally, induced repolymerisation of actin by adding phalloidin beads, Mg²⁺ and an ATP-regenerating system. Mass spectrometric analysis of the phalloidin-bound fraction (Figure 1C) revealed that the most abundant cytoplasmic F-actin interactors were expected factors, namely the filament-severing and -capping protein gelsolin (Ankenbauer et al, 1988) and subunits of the Arp2/3 nucleator complex (Mullins et al, 1998). Myosins were not detected in this actin-bound fraction, because binding was performed in the presence of a regenerating system for ATP, which counteracts a stable myosin binding to actin (La Cruz and Ostap, 2004). Following ATP depletion by apyrase, however, a massive myosin II band and several less-abundant myosins were recovered (Supplementary Figure S2).

The pattern of nuclear F-actin interactors differed strikingly from the cytoplasmic one (Figures 1C and D). Gelsolin and the Arp2/3 complex were virtually absent. Instead, a kinesin turned out to be the most abundant F-actin interactor (Figure 1C and Table I). This was indeed surprising, because canonical kinesins are motor proteins that interact with microtubules, but not with actin. For reasons detailed below, we named this novel kinesin NabKin (Nuclear and meiotic actin-bundling Kinesin).

Table 1. Nuclear F-actin interactors of Xenopus oocytes.

F-actin binders, isolated with phalloidin beads from nuclear extracts, were separated by SDS–PAGE (see Figure 1C). The entire lane was then subjected to protein identification by mass spectrometry. The numbers of sequenced and assigned peptides from four independent experiments are given. Listed specific F-actin binders were identified in at least three independent phalloidin-binding experiments, but were completely absent in purification experiments with control beads. Copurified actin is not listed. The last column indicates which proteins had been independently identified from individually excised bands.

*GenBank accession number.

The other nuclear F-actin interactors (Table I) comprise two major groups: DNA-binding proteins (such as Msh2, MCM3 and RuvB-like 1) and factors that are known to organise F-actin in the cytoplasm by filament nucleation and severing (Inverted formin-2; Chhabra and Higgs, 2006), filament capping (Capping proteins alpha+beta; Isenberg et al, 1980), filament side-binding (Drebrin; Imamura et al, 1992) and filament cross-linking (Filamin-A and Supervillin; Wang et al, 1975; Pestonjamasp et al, 1997). We did not detect any ATP-sensitive myosins in the nuclear actin fraction.

To further characterise these nuclear F-actin interactors, we raised antibodies against the respective Xenopus orthologues. Immunoblotting of manually microdissected oocytes verified the nuclear localisation of these F-actin interactors and indicated significant nuclear enrichment for most of them (Figure 1D and Supplementary Figure S8). A nuclear localisation had previously been observed only for Supervillin (Wulfkuhle et al, 1999) and Capping protein alpha (Ankenbauer et al, 1989). For the remaining proteins, the deviation from the cytoplasmic localisation in other cell types might be explained by frog-specific features, oocyte-specific splice variants or post-translational modification of the proteins.

NabKin binds and bundles F-actin filaments

The initial identification of NabKin was difficult, because it had not been listed in any protein database. We therefore assembled the protein sequence from X. tropicalis genomic and X. laevis cDNA data. It revealed identity to a novel kinesin-3 family member with a motor domain comprising all sequence features expected for a kinesin (Supplementary Figure S3). Immunoblotting showed a strong enrichment of NabKin on phalloidin beads, validating our initial mass spectrometry-based identification (Figure 1E). Likewise, the nuclear localisation of NabKin was confirmed by cell microdissection (Figure 1D) and immunofluorescence microscopy (Figures 2E and 4A).

Figure 2.

NTEs of NabKin and KIF14 bundle F-actin. (A) A simplified neighbour-joining tree of the kinesins-3 family, including the paralogues NabKin and KIF14. (B) Predicted domain organisation of NabKin and KIF14 (Xl, Xenopus laevis; Xt, Xenopus tropicalis; Hs, Homo sapiens). In contrast to other kinesin-3 family members, NabKin and the KIF14s possess unusually long NTEs. The NabKin and KIF14–NTEs share no obvious sequence similarity (indicated by distinct colouring). (C) Differential centrifugation assay was used for distinguishing actin-bundling and actin-binding factors from non-interacting proteins. Candidate proteins (3 μM) were mixed with actin (10 μM) in F-actin buffer and tested for mutual interaction as outlined in the scheme on the left. Fractions obtained were analysed by SDS–PAGE and Coomassie staining. Red asterisks mark the F-actin-bundling NTEs and the green asterisk marks Drebrin, which binds, but does not bundle F-actin (Sekino et al, 2007). The image was cut at black lines and rearranged for clarity. The NabKin–NTE had been fused to an N-terminal His₁₄-SUMO-tag in the experiments shown in 2C+D; KIF14–NTEs (XtKIF14^aa1-341 and Hs KIF14^aa1-345) were untagged. (D) Electron micrographs of negatively stained F-actin filaments obtained in the absence or presence of indicated NTEs. Scale bars=200 nm. (E) NabKin colocalises with nuclear F-actin in Xenopus oocyte cryostat sections. Analysis was by confocal laser-scanning microscopy (CLSM). Dashed line marks the nuclear envelope, separating nuclear (N) and cytoplasmic (Cy) compartment. The F-actin-free intranuclear regions probably represent the nucleoli. The white square marks the area depicted in the lower panel. Scale bars=10 μm.

At this point, it was still unclear whether the actin interaction of NabKin was direct or mediated by auxiliary factors. Any direct interaction would require an actin-binding domain, yet the NabKin sequence lacks homology to previously characterised actin-binding motifs. We noticed, however, an unusual N-terminal extension (NTE) of the motor domain (Figure 2B). When tested for interaction with purified actin, the NabKin–NTE (XlNabKin^aa1-268) indeed cosedimented with actin filaments (Figure 2C). Moreover, the NTE induced the formation of rapidly sedimenting higher-order actin structures, which electron microscopy identified as F-actin bundles (Figures 2C and D). Conversely, this predicts that the NTE preferentially binds actin structures that contain bundled actin. Finally, we asked if NabKin binds F-actin also in its native context, and observed a striking colocalisation with the actin bundles present in the oocyte nucleus (Figure 2E).

An F-actin-bundling ligand should contact at least two actin filaments simultaneously. We see at least two (not mutually exclusive) possibilities of how the (presumably non-globular) NabKin–NTE becomes such multivalent ligand. First, NabKin’s actin-bundling activity maps to a 49-residue stretch (NabKin^163-212; Supplementary Figure S5A–E), which appears long enough for harbouring multiple F-actin-binding peptides. Second, the NabKin–NTE itself might multimerise. However, given that the isolated NabKin–NTE is monomeric (Supplementary Figure S4A), such multimerisation should be induced by F-actin.

The NabKin paralogue KIF14 also contains an actin-bundling NTE

NabKin is present in stage II–VI oocytes and eggs, but disappears between embryonic stages 9–12 (Figure 1F), that is, around midblastula transition when zygotic gene expression starts (Kimelman et al, 1987). This implies that NabKin is absent from non-embryonic somatic cells. We, therefore, asked if somatic cells contain an alternative, sequence-related actin-bundling kinesin.

The closest paralogue of NabKin is KIF14 (Figure 2A)—a kinesin that is essential for cytokinesis of human somatic cells (Carleton et al, 2006; Gruneberg et al, 2006) and that also features an unusual NTE (Figure 2B). Strikingly, we found by sedimentation assays and electron microscopy that the Xenopus and human KIF14–NTEs also bind and bundle F-actin (Figures 2C and D). Moreover, when transfected into PtK2 cells, both KIF14–NTEs localised to cytoplasmic stress fibres, the predominant sites of bundled F-actin in interphase cells (Figure 3A).

Figure 3.

Importin β inhibits the NabKin–NTE interaction with F-actin, RanGTP counteracts the effect. (A) PtK2 cells expressing the NTE of NabKin (XlNabKin^aa1-268), HsKIF14 (HsKIF14^aa1-345) or XtKIF14 (XtKIF14^aa1-341) fused to the N terminus of EGFP were fixed, stained for F-actin and DNA, and analysed by CLSM. Line plots represent pixel values along the dashed line. Note the correlation between NTE (green) and F-actin signals (red). The colocalisation might even be underestimated, because the assay neglects the avidity contribution from a potential coiled-coil-mediated dimerisation of the kinesins. Scale bar=10 μm. (B) Immobilised NabKin–NTE was incubated with diluted Xenopus oocyte extract in the absence or presence of 2 μM RanQ69L·GTP (a GTPase-deficient mutant; Bischoff et al, 1994). Following a washing step, bound proteins were eluted with buffer containing 5 μM RanQ69L·GTP to increase specificity of elution, and were analysed by SDS–PAGE and Coomassie staining (see also Supplementary Figure S4B). (C) Import of NabKin–NTE fused to EGFP into the nuclei of digitonin-permeabilised HeLa cells was performed for 10 min at RT. Reactions were stopped by fixation and examined by CLSM. Addition of Importin β facilitates nuclear import of NabKin–NTE–EGFP. Scale bar=10 μm. (D) NabKin–NTE–EGFP (2 μM) was assayed for actin-bundling activity (see also Figure 2C). Where indicated, 2 μM Importin-β was also present. The image was cut at black lines and rearranged for clarity. (E) Equal amounts of phalloidin beads were incubated in an undiluted high-speed egg extract and allowed to bind F-actin complexes in the absence or presence of 20 μM RanQ69L·GTP. Bound fractions were immunoblotted with indicated antibodies. RanGTP stimulated binding of NabKin, but not of Capping protein alpha.

Surprisingly, we could not detect any obvious sequence similarity between the actin-bundling NTEs of Xenopus NabKin and KIF14, even though their motor domains are 64%, and the remaining C-terminal domains are 46% sequence-identical. In an interspecies comparison, the sequence of the KIF14–NTE is also far less conserved than the other modules of the kinesin, yet the NTE is present in all known metazoan KIF14 family members. We assume that these NTEs represent disordered domains that evolve rapidly, because there is no evolutionary pressure to conserve a globular fold.

The NabKin–actin interaction is controlled by Importin-β and RanGTP

In contrast to the KIF14–NTEs, the EGFP-fused NabKin–NTE did not localise to cytoplasmic stress fibres, but to the nucleus (Figure 3A). This points to a nuclear import signal within the NTE and indicates that cytoplasmic actin binding is overruled by nuclear import, possibly because an importin masks the F-actin interaction site of NabKin.

In search for such importin, we incubated an oocyte extract with the immobilised NabKin–NTE and identified by mass spectrometry Importin-β as a prime candidate (Figure 3B). Indeed, Importin-β bound the NTE under cytoplasmic conditions and became disengaged by RanGTP, which marks the nucleus as the destination compartment of import (Görlich et al, 1996). Consequently, Importin-β conferred efficient import of NabKin–NTE–EGFP into nuclei of digitonin-permeabilised cells (Figure 3C).

Next, we tested if Importin-β influences NabKin’s interaction with actin. We found that Importin-β strongly inhibits NabKin’s actin-bundling activity (Figure 3D). To test if RanGTP counteracts this inhibitory effect, we used phalloidin beads and retrieved endogenous NabKin–actin complexes from Xenopus egg extract. Addition of RanGTP strongly promoted the NabKin–actin interaction (Figure 3E)—presumably by disengaging Importin-β from the NTE (Figure 3B).

NabKin interacts with meiotic actin- and microtubule-based structures

Despite the presence of a kinesin motor domain, NabKin is unlikely to interact with microtubules during the long meiotic prophase arrest, because NabKin localises to the tubulin-depleted nuclei (Figure 1D; Gard, 1999; Bohnsack et al, 2006). Contacts between NabKin and microtubules could, however, occur after breakdown of the nuclear envelope, when nuclear and cytoplasmic contents intermix. To test this, we investigated the localisation of NabKin during the maturation of oocytes into fertilisable eggs.

Within the first minutes after NEBD, NabKin colocalised with the transient microtubule array (Figure 4B and C) and then with the MI spindle itself (Figure 4D, see Supplementary Figure S7 for a summary of NabKin’s localisation during maturation). Higher-magnification analysis of MI spindles revealed that NabKin colocalised with individual spindle microtubules (Figure 4E).

Figure 4.

NabKin interacts with meiotic microtubule structures and cortical F-actin after NEBD. Oocytes were fixed at various stages of maturation, stained as detailed in Materials and Methods, and were analysed by CLSM as whole mounts of hemisections along the animal–vegetal axis to follow NabKin localisation during this process. Cartoons explain the analysed structures (see also Supplementary Figure S7). Scale bars A–D=50 μm. (A) Prophase I-arrested stage VI oocyte. (B) Formation of the disc-shaped transient microtubule array directly following NEBD. (C) Transient microtubule array moving towards cortex. (D) MI spindle formation at cortex. (E) High magnification of another MI spindle. Note that NabKin colocalises with the spindle and specifically also with individual microtubules extending from the spindle (indicated by arrows). Scale bar=5 μm.

Besides the localisation at spindle microtubules, a fraction of NabKin localised to the F-actin-rich cortical cap (Figure 4B–D). This dual localisation was maintained also following the spindle attachment to the oocyte cortex (Figure 5A); the cortex-to-spindle signal ratio, however, increased at this stage. Upon anaphase onset, we observed bright NabKin signals at the F-actin patch forming at the cortex underlying the spindle attachment site, and later at the nascent first PB emerging from this patch (Figure 5B and C).

Figure 5.

NabKin localisation during PB extrusion. CLSM analysis of first and second PB extrusion. Oocytes and eggs were fixed and stained as detailed in the Materials and Methods section. Matured oocytes (A–C) were analysed as whole mounts of hemisections along the animal–vegetal axis. Arrested (D) or activated (E, F) eggs were analysed as whole mounts. Scale bars=10 μm. (A) MI spindle attached to the oocyte cortex. High-sensitivity settings allowed detection of NabKin at the spindle and the cortex. The signal of NabKin at the spindle is usually weaker than the cortical signal. (B) Cortical F-actin cap located directly above the MI spindle during anaphase. (C) CLSM analysis of a nascent first PB. A maximum intensity projection of sections spanning 11.7 μm in depth is shown. (D) Egg arrested in metaphase of meiosis II. High-sensitivity settings allowed detection of NabKin at the MII spindle and the already extruded first PB. (E) A Ca²⁺-ionophore-activated egg extruding the second PB, stained for NabKin, F-actin and DNA. A maximum intensity projection of sections spanning 34.6 μm in depth is shown. Note that the NabKin signal is far stronger at the nascent PB II compared to the first PB I. (F) Another activated egg extruding the second PB, stained also for a marker of the actomyosin ring (myosin light chain 2). Note that NabKin colocalises with F-actin at the ‘roof’ of the PB as well as with the actomyosin ring structure. A maximum intensity projection of sections spanning 3.8 μm in depth is shown.

Following meiosis I, the remaining chromosomes attach to a MII spindle, which then arrests in metaphase II. At this stage, NabKin localised to the first PB and the MII spindle (Figure 5D). We then relieved the metaphase II arrest by Ca²⁺-ionophore treatment (which mimics fertilisation) and observed that NabKin also localised to the second PB, where it gave a far stronger signal than at the previously emitted first PB (Figure 5E). This indicates that NabKin marks the zone of active cell division (Figure 5E). Closer inspection revealed a particularly strong NabKin signal at the ‘roof’ of the emerging PB (Figure 5F), which is characterised by highly dynamic F-actin filaments (Zhang et al, 2008). A second NabKin population localised to the contractile F-actin ring (Figure 5F), which drives cell division, and which is marked by myosin II and more stable F-actin filaments (Zhang et al, 2008).

KIF14 localises to mitotic spindle and contractile ring of human cultured somatic cells

The somatic paralogue of NabKin is KIF14, which plays an essential role in somatic cytokinesis (Carleton et al, 2006; Gruneberg et al, 2006). During anaphase, KIF14 had previously been detected at the central spindle (Carleton et al, 2006; Gruneberg et al, 2006), a structure formed by overlapping, antiparallel non-kinetochore microtubules that eventually becomes part of the midbody. The midbody itself represents the last connection between the dividing cells, and KIF14 had been detected in this structure as well (Carleton et al, 2006; Gruneberg et al, 2006), possibly as a consequence of a persistent binding to the overlapping central microtubules.

Because of the presence of an actin-bundling NTE in KIF14, we wondered whether KIF14 also localises to the F-actin-rich cortex of the ingressing cleavage furrow, which may have been missed in previous studies. We expressed EGFP–KIF14 in HeLa cells and performed confocal 3D time-lapse microscopy. Consistent with previous studies (Carleton et al, 2006; Gruneberg et al, 2006), we observed KIF14 at the central spindle during anaphase and at the midbody during telophase (Figure 6A). In addition, we also detected a substantial fraction of EGFP–KIF14 at the cell cortex, colocalising with the contractile F-actin ring at the ingressing cleavage furrow (Figure 6A and B). Hence, we conclude that the capability to bind both actin and the microtubule cytoskeleton is conserved in the somatic paralogue of NabKin, KIF14.

Figure 6.

KIF14 localisation during mitotic cell division in live HeLa cells. (A) Confocal time-lapse micrographs of live HeLa cells expressing mRFP-α-tubulin and KIF14-mEGFP. Arrowheads indicate cortical pool of KIF14, arrows indicate central spindle. (B) Confocal time-lapse micrographs of live HeLa cells expressing LifeAct-mCherry, which we used to visualise F-actin (Riedl et al, 2008) and KIF14-mEGFP. Arrowheads indicate cortical pool of KIF14, arrows indicate central spindle. Time indicated (s) is relative to first image. Scale bars=10 μm.

The NabKin–actin interaction is crucial for PB extrusion

It was striking to see actin-bundling kinesins colocalising with critical actin and microtubule structures during cell division. We therefore wondered whether an interaction of NabKin with actin would be required for meiosis and, if so, at which stage of meiosis this would become relevant. To selectively disrupt the NabKin–actin interaction, we raised antibodies against the NTE, affinity-purified them and prepared Fab fragments (Supplementary Figure S6A). This yielded a specific, small-sized and monovalent inhibitor that selectively blocked the binding of NabKin to F-actin (Figure 7A). This control also confirmed that the NTE is crucial for the kinesin–actin interaction and argues against additional actin-binding modules within the NabKin molecule.

Figure 7.

Interaction of NabKin with F-actin is crucial for PB extrusion. (A) Equal amounts of phalloidin beads were incubated in undiluted high-speed extract from Xenopus eggs and allowed to bind F-actin complexes in the absence or presence of 50 μg/ml anti-NabKin–NTE Fab fragment. 20 μM RanQ69L·GTP was added to the extract to release inhibitory Importin β from the NabKin–NTE (see also Figure 3E). Bound fractions were immunoblotted with indicated antibodies. Addition of anti-NabKin–NTE Fab fragment dramatically reduced the amount of NabKin bound to the matrix, but not the amount of bound Capping protein alpha. (B) CLSM analysis of PB I extrusion. Oocytes received nuclear injections with 50 ng of indicated Fab fragments. Oocytes were fixed 3 h post NEBD to allow sufficient time for completion of meiosis I and arrest in metaphase of meiosis II (Gard et al, 1995; Castro et al, 2003). Only oocytes displaying the spindle at the cortex were included in the analysis, as this location marks the site of the PB extrusion. Note that these anti-NabKin–NTE Fab fragment-injected oocytes do not show an extruded PB. Instead, two spindles or an enlarged spindle with two metaphase chromosome plates (indicated by red arrows in lowest panel) were found in these depicted oocytes. Scale bar=10 μm. See Supplementary Figure S6D for additional analyses. (C) Statistical analysis of successful PB extrusion. Uninjected oocytes (59% extruded PB I; n=32; actual oocyte numbers are shown in the graph bars) and anti-GFP-Fab fragment-injected oocytes (53%; n=30) showed no significant difference (NS; P=0.63; χ²-test). The injected Fab fragment against the NabKin–NTE (8%; n=49) impaired PB extrusion highly significantly (***P<0.0001; χ²-test) as compared to uninjected or control injected cells.

Oocytes were microinjected with these anti-NabKin–NTE Fab fragments directly into the nucleus. They were then activated with progesterone and fixed at various time points. We could not detect prominent effects of NabKin inhibition on early meiotic maturation. Oocytes re-entered meiosis, underwent NEBD and formed the MI spindle. Sixty minutes post NEBD, this spindle was perpendicularly attached to the cortex in 71% of the injected oocytes (n=34, Supplementary Figure S6B), which is in line with the typical efficiency of in vitro maturation of non-injected oocytes (Gard, 1992).

Next, we addressed progression through the first asymmetric cell division and now observed a dramatic effect. The anti-NabKin–NTE Fab fragment not only displaced NabKin from the oocyte’s cortex (Figure 7B, Supplementary Figure S6C and D), but also reduced the number of successfully extruded PBs from >50% in uninjected or anti-GFP-injected oocytes down to 8% (Figure 7C; statistically significant at P<0.0001; χ²-test). We assume that these 8% actually represent cases where the microinjected Fab fragment provided only incomplete NabKin inhibition or where microinjection failed altogether. This would agree with the general experience that only ∼90% of attempted nuclear injections are successful (Guille, 1999) and with the fact that the few PBs observed in anti-NabKin–NTE injected oocytes all showed a bright cortical NabKin signal. Together, this suggests that the oocyte cannot extrude the PB without NabKin and that the NabKin–actin interaction is required for efficient cell division in meiosis I.

In some anti-NabKin–NTE Fab fragment-injected oocytes, we preserved two coexisting metaphase II spindles (Figure 7B, upper anti-NabKin–NTE-injected panel). Such phenotype had previously been observed after failed extrusion of one chromosome set into a PB during meiosis I (Gard et al, 1995; Elbaz et al, 2010). In most cases, however, the two metaphase II spindles had already merged to form a single enlarged spindle, which either contained a single metaphase plane of aligned chromosomes, or sometimes two metaphase planes (Figure 7B, lower anti-NabKin–NTE-injected panel).

As NabKin also localises to the site of the second PB extrusion (Figure 5E and F), it is tempting to assume that inhibition of NabKin will also block cytokinesis of meiosis II. However, already a single failure of PB exclusion has dramatic consequences, namely a polyploid egg.

Discussion

The nuclei of stage VI X. laevis oocytes are 100 000 times larger in volume than typical somatic cell nuclei. They are filled by a liquid protein solution with chromosomes occupying only a tiny volume fraction—a setting that makes these nuclei very fragile in the absence of their intranuclear actin skeleton. With their actin scaffold, they are, however, remarkably stable (Bohnsack et al, 2006). The structural organisation of this skeleton and whether it is built from canonical F-actin filaments have been debated for a long time (Gall, 2006; Jockusch et al, 2006). Our new observation that oocyte nuclei contain several prominent F-actin binders (Figure 2C and Table I) now strongly supports the notion that this structure indeed relies on classical F-actin. The Arp2/3 complex was undetectable in oocyte nuclei, suggesting that nuclear filament nucleation is Arp-independent. Instead, we detected Inverted formin-2, a potent nucleator of unbranched filaments (Chhabra and Higgs, 2006), which may promote F-actin assembly inside nuclei. Our data further indicate that the nuclear actin filaments are bundled by NabKin and Supervillin (Pestonjamasp et al, 1997), and are cross-linked by Filamin-A (Wang et al, 1975) into a scaffold that appears suitable for stabilising the giant oocyte nuclei.

An unbiased visualisation of the native intranuclear actin structure is, however, a formidable challenge. Any fixation of the giant oocytes might alter the structure, and the opaqueness of the oocyte precludes any live cell imaging inside nuclei. Isolated nuclei are highly transparent, but aqueous isolation buffers might perturb the actin assemblies (Paine et al, 1992; Gall, 2006). We therefore isolated nuclei in oil (which avoids any contact with an exogenous buffer) and stained them by microinjection of the actin marker LifeAct-GFP (Riedl et al, 2008). This confirmed our biochemically derived conclusions and revealed a clearly filamentous, interconnected actin network inside these native oocyte nuclei (Supplementary Figure 9).

Besides maintaining the structural integrity of oocyte nuclei, the nuclear F-actin network probably fulfils additional interesting functions, such as the immobilisation of the >1000 nucleoli of the oocyte (Brangwynne et al, 2011), epigenetic reprogramming (Miyamoto et al, 2011) and, upon NEBD, the congression of the widely scattered chromosomes towards the MI spindle (Lénárt et al, 2005; Mori et al, 2011). Our interactome data might therefore comprise factors involved in these functions as well. As our approach specifically selected for F-actin binders, it is not surprising that interactors of non-filamentous actin forms, such as actin’s nucleotide exchange factor profilin (Carlsson et al, 1977) or the transcription factor Mal (Vartiainen et al, 2007), remained missing from our interaction map (Table I).

NabKin is a particularly interesting F-actin interactor, because it is a novel actin-binding kinesin, which may interconnect actin filaments with microtubules. So far, only two metazoan motor proteins with similar integrative potential have been described: CHO1, an alternatively spliced variant of the MKLP1 kinesin (Kuriyama et al, 2002) containing an actin-binding module that appears, however, dispensable for the kinesin’s function in cytokinesis (Matuliene and Kuriyama, 2004). The second example is Myo10, a microtubule-binding myosin (Weber et al, 2004) that acts mainly in spindle assembly and nuclear positioning (Weber et al, 2004; Woolner et al, 2008).

NabKin shares numerous properties with its somatic paralogue KIF14. Both are essential for cytokinesis, KIF14 during somatic cell division and NabKin during meiotic PB extrusion. Both bind endogenous microtubule structures, in particular the meiotic/mitotic spindle. NabKin and KIF14 contain an actin-binding NTE and localise to endogenous actin structures, in particular the contractile actomyosin ring. The contractile ring is in close proximity to spindle microtubules (Green et al, 2011; Fededa and Gerlich, 2012), and it is an attractive hypothesis that NabKin and KIF14 interconnect the two cytoskeletal systems during cytokinesis. Indeed, our data suggest that the NabKin–actin interaction is critical for its fundamental role in cell division.

Microtubule-dependent ATPase activity was so far demonstrated for human KIF14 (Carleton et al, 2006), which strongly argues for KIF14 being a functional motor. The perfect conservation of the relevant sequence motifs (Supplementary Figure 3) along with the clearly evident microtubule binding (Figure 4B–E) suggests that NabKin’s motor domain is functional as well. It is, however, still unclear when the KIF14/NabKin motors become active and how the motor activity is regulated. Likewise, it remains to be shown whether actin filaments are moved against microtubules, which forces are generated and why a lack of these contributions causes cytokinesis to fail.

The functional specialisation between a somatic KIF14 and the female meiotic NabKin appears specific to the amphibian lineage. This predicts that KIF14 of other metazoans functions not only during mitotic cell division, but may also play a role during meiotic PB extrusion. The consideration begs interesting questions, namely whether, and how, KIF14 from non-amphibian species can switch between a meiotic and a somatic mode of regulation.

NabKin’s specialisation to the highly asymmetric cell division of the giant oocyte is already evident. NabKin is recruited to meiotic structures that have no mitotic equivalent; for example, to the actin-based roof of the PB (Figure 5F). Here it is possible that NabKin somehow contributes to the dynamic anchoring of the spindle inside the nascent PB. Another example is the transient microtubule array (Figure 4B and C), which cooperates with actin to collect chromosomes that are too widely scattered to be collected by spindle microtubules alone (Gard et al, 1995; Lénárt et al, 2005; Mori et al, 2011). Possibly, NabKin also participates in this process and NabKin deficiency could lead to an increased rate of chromosome loss. In this context, it is also interesting that we identified several DNA-binding proteins in our actin-bound fraction (such as Msh2, MCM3/5 and RuvB-like 1) that are now good candidates for connecting chromosomes to a chromosome-collecting actin network.

NabKin localisation is controlled by nucleo-cytoplasmic transport, similar to several meiotic MAPs (Carazo-Salas et al, 1999; Kalab et al, 1999; Gruss et al, 2001): In interphase, importins keep these MAPs inside nuclei and thus separated from cytoplasmic microtubules. Following NEBD, the importins form inhibitory complexes with the MAPs. RanGTP, originating from chromatin-bound RanGEF, relieves this inhibition locally. This mechanism favours spindle formation in the vicinity of the meiotic chromosomes. Likewise, RanGTP unlocks the Importin-β-mediated inhibition of the actin-bundling NabKin–NTE. A meiotic RanGTP gradient might therefore also guide NabKin to the cortical F-actin structure near the meiotic spindle.

Without confinement by an intact nuclear envelope, a RanGTP gradient will be blurred by diffusion, limiting the spatial resolution of the gradient to a few micrometres (Görlich et al, 2003), which is similar in dimension to an entire mitotic somatic cell. This implies that such gradient will be more ‘informative’ in a large oocyte and could explain as to why NabKin, the oocyte-specific form of the F-actin-bundling kinesin, became responsive to RanGTP.

We expect that NabKin is not the only importin-inhibited and RanGTP-stimulated actin interactor relevant for female meiosis. Additional players should not only bind actin but also importins, and thus show a nuclear localisation prior to NEBD. Our list of nuclear F-actin binders (Table I) therefore includes promising candidates for additional Ran-controlled factors that may aid communication between meiotic chromosomes and cortical F-actin structures.

Methods

Purification of phalloidin

Phalloidin was isolated from Amanita phalloides by a modified version of a published protocol (Enjalbert et al, 1992). Mushrooms were homogenised and extracted in 2 vol. 50% MeOH. Extracts were cleared by centrifugation, 2 vol. acetone were added and the mixture was kept for 1 h at −20 °C. The solution was cleared by filtration, and MeOH and acetone were removed from the extract by vacuum rotary evaporation. Lipids were extracted with 1 vol. of diethyl ether and aqueous phase was collected. Residual ether was removed from the aqueous phase by vacuum rotary evaporation; the extract was filtered and toxins were separated by reversed-phase chromatography on a C18 GraceVydac column. Phalloidin-containing fractions were identified by mass spectrometry, were snap-frozen and then lyophilised.

Phalloidin-based affinity matrix

See Supplementary Figure 1 for an overview of the reaction steps. Ten milligram of lyophilised phalloidin were dissolved in 100 μl anhydrous pyridine (reaction a). A 16-fold molar excess of 4-Toluenesulfonyl chloride (Fluka) was dissolved in 50 μl CHCl₃ and added slowly on ice to monotosylate the unique primary hydroxyl group (the δ-hydroxyl group of the dihydroxyleucine) of phalloidin (Wieland et al, 1983). The reaction was rotated under argon for 30 min at 4 °C and was stopped by the addition of 6 vol. of diethyl ether. The precipitated monotosylated phalloidin was purified by reversed-phase chromatography (Diphenyl, GraceVydac) and its identity confirmed by mass spectrometry.

For introduction of a readily reactable functional group (reaction b), 6 mg monotosylated phalloidin was dissolved in 1.2 ml MeOH and slowly added to a 2000-fold molar excess of 1,8-dithiol-PEG₃ (3,6-dioxa-1,8-octanedithiol; Sigma) in MeOH in the presence of triethylamine (Roth; equimolar to 3,6-dioxa-1,8-octanedithiol). Triethylamine deprotonates the thiol groups and aids solubility of 3,6-dioxa-1,8-octanedithiol. The resulting phalloidin derivative containing a highly flexible and hydrophilic spacer with a terminal SH group, thiol-PEG₃-phalloidin, was extracted from the reaction by addition of 1 vol. 100 mM ammonium acetate (pH 5, readjust after extraction), which caused phase separation between surplus 3,6-dioxa-1,8-octanedithiol and toxin-containing aqueous phase.

Thiol-PEG₃-phalloidin was purified by reversed-phase chromatography (C18, GraceVydac) and its identity confirmed by mass spectrometry. It was then lyophilised and reacted with an ∼1.2 M excess of maleimido-PEG_∼70-trimethoxysilane (NANOGS Inc.) in DMF:DMSO:MeOH 1:3.5:1 for 40 min under argon (reaction c). Unreacted maleimide groups were quenched with an excess of methoxy-PEG₆-thiol (2,5,8,11,14,17,20-Heptaoxadocosane-22-thiol; Polypure). In the control reaction, maleimido-PEG_∼70-trimethoxysilane was reacted only with methoxy-PEG₆-thiol. Subsequently, the silane reactions were mixed with 1 μm SiMAG-Silanol superparamagnetic beads (Chemicell) in toluene: EtOH 9:1 (reaction d). Silane deposition was started by the addition of 1% v/v triethylamine and was allowed to proceed for 36 h at room temperature (RT) under argon. Matrices were washed in toluene and coated with PEG_6-9-silane ([3-methoxy(polyethyleneoxy)propyl]trimethoxysilane; 90%; SIM6492.7; Gelest) for 36 h in toluene with 1% v/v trimethylamine (reaction e). Matrices were washed thoroughly in toluene and subsequently in EtOH, and stored in EtOH at 4 °C.

F-actin affinity chromatography

Oocytes were manually microdissected in ‘5:1/HEPES buffer’ (10 mM HEPES; pH 7.5, 83 mM KCl, 17 mM NaCl) as previously described (Liu and Liu, 2006). To prepare compartment-specific extracts, 300 washed nuclei and the respective cytoplasmic fractions were diluted 1:4 in ‘5:1/HEPES buffer’ (10 mM HEPES; pH 7.5, 83 mM KCl, 17 mM NaCl) and homogenised. Latrunculin A was added to all samples to a final concentration of 0.2 μM. F-actin was allowed to depolymerise for 1 h at 4 °C, and aggregates, membrane fragments, as well as yolk particles, were removed from the extract by centrifugation (30 000 g; 10 min; 4 °C). Equilibrated phalloidin or control beads were added and actin polymerisation was induced by addition of 2 mM Mg²⁺, an energy-regeneration system (comprising 1 mM ATP and 20 mM creatine phosphate) and 0.01% Triton X-100. Samples were rotated for 2 h at RT. The matrix was then washed with polymerisation buffer and eluted with SDS sample buffer. Eluates were separated by SDS–PAGE (NuPAGE, Invitrogen) and stained by colloidal Coomassie.

Differential centrifugation assay

The assay was performed as previously described (Dixon et al, 2008) with the following modifications. All proteins were cleared by centrifugation (130 000 g; 5 min; RT) prior to the experiment. G-actin (10 μM) was mixed with candidate protein (3 μM, unless stated otherwise) in a total volume of 50 μl F-buffer (100 mM KCl, 2 mM MgCl₂, 0.5 mM ATP, 10 mM Tris; pH 7.5) and allowed to polymerise for 40 min at RT. The samples were then subjected to ‘low-speed’ centrifugation (13 000 g; 15 min; 22 °C) and supernatants were subsequently subjected to ‘high-speed’ centrifugation (130 000 g; 15 min; 22 °C). Pellets were washed carefully with F-buffer and resolved in SDS sample buffer. ‘High-speed’ supernatants were acetone precipitated (5 vol.; −80 °C; 1 h), resolved in the same volume SDS sample buffer as the other pellets, and identical volumes were analysed by SDS–PAGE and Coomassie staining.

Protein identification by mass spectrometry

SDS–PAGE-separated protein samples were processed as described (Shevchenko et al, 1996), with modifications introduced according to Schmidt and Urlaub (2009). Peptides were analysed on LTQ-Orbitrap XL (Thermo Fisher Scientific) under standard conditions, and MS and MS/MS spectra were searched against the NCBI non-redundant database using MASCOT as the search engine. Results were analysed directly or annotated with the Scaffold 3.0 software.

Recombinant protein expression

Recombinant proteins were expressed at 18 °C for 12–16 h in E. coli, usually with an N-terminal oligo histidine tag followed by a SUMO protease cleavage site, and purified by Ni(II) chelate chromatography. Proteins were loaded in 50 mM Tris/HCl pH 7.5, 500 mM NaCl, 20 mM imidazole and 1 mM DTT, and eluted with 0.3 M imidazole pH 7.5. Exchange to F-actin buffer (10 mM Tris/HCl pH 7.5, 100 mM KCl, 2 mM MgCl₂) was on Sephadex G25 columns. Maps of the DNA constructs used and more detailed protocols for protein expression are available on request.

Antibodies

Polyclonal antibodies against Capping protein alpha (gi|147900955, residues 1–279), Drebrin (gi|147899740, residues 1–600), Inverted formin-2 (gi|148237492, residues 1–1099), NabKin (IMAGE:7010529, residues 1–268) and Supervillin (gi|148230951, residues 1–450) were raised in rabbits, and affinity-purified against immobilised antigens. Commercially available antibodies were used against dsDNA (HYB331-01, Santa Cruz), Filamin-A (ab11074, Abcam), α-tubulin (DM1A, Sigma), myosin light chain 2 (AT3B2, Novus Biologicals) and Arp2 (ab47654, Abcam).

Immunoblot analysis

Samples were separated by SDS–PAGE and electroblotted onto nitrocellulose membranes. Membranes were blocked in 5% milk in PBS and incubated with primary antibody at 1 μg/ml in blocking solution. IRDye secondary antibodies (LI-COR Biosciences) were used at 1:10 000 dilution of commercial stock and visualised with an Odyssey infrared imaging system (LI-COR Biosciences).

Negative-stain electron microscopy

F-actin was mixed with the candidate protein as described above and incubated for 30 min. A drop of the protein solution was applied for 30 s to a 400-mesh electron microscope cooper grid coated with a carbon film and stained 2 × 20 s on drops of 1% aqueous uranyl acetate without rinsing. Samples were analysed on a Hitachi H-7600 transmission electron microscope at 80 kV, equipped with a MegaView 3 CCD camera (Soft Imaging Solutions).

Import assay

Permeabilised cells were prepared as previously described (Jäkel and Görlich, 1998). HeLa cells were permeabilised on ice with 25 μg/ml Digitonin (high purity; Calbiochem) in transport buffer (20 mM HEPES pH 7.5, 5 mM MgAc, 110 mM KAc, 1 mM EGTA, 250 mM Sucrose). Nuclear import of 2 μM NabKin–NTE–EGFP was performed for 10 min at RT with an energy-regeneration system and a Ran system. One incubation additionally received 2 μM Importin-β. The reactions were stopped by formaldehyde fixation and examined by confocal laser-scanning microscopy (CLSM).

Importin-binding assay

NabKin–NTE (zz-tagged) was immobilised to IgG sepharose, incubated with Xenopus oocyte extract (diluted 1:4 in 10 mM Tris; pH 7.5, 50 mM NaCl) in the presence or absence of 2 μM RanQ69L·GTP, washed and eluted with 5 μM RanQ69L·GTP.

Transfection and fluorescence analysis of cultured cells

Cells were transfected using Lipofectamine (Invitrogen) according to manufacturer’s instructions and fixed with 3.7% formaldehyde in PBS after 36 h. Cells were stained for DNA (Hoechst) and F-actin (phalloidin–TRITC), and examined by confocal microscopy using a Leica SP5 equipped with a HCX PL APO × 63, 1.4 NA oil. Line plots were generated using ImageJ.

Immunofluorescence on Xenopus oocytes and eggs

Oocyte immunofluorescences visualised whole-mounted hemisections along the animal–vegetal axis of the oocyte (except Figure 2E). For this, we adapted a published protocol (Becker and Gard, 2006). Tubulin and F-actin structures require mutually exclusive fixation protocols for optimal preservation (Gard, 1999). Tubulin was fixed by 100% MeOH overnight at RT (Becker and Gard, 2006), F-actin was fixed by ‘FAMeOH’ (3.5% (w/v) formaldehyde, 7.5% MeOH, PBS; pH 7.4) for 2 h at RT.

Oocytes were matured in 10 μg/ml progesterone at RT and fixed between 0–60 min post NEBD for observation of MI spindle formation, between 90–120 min post NEBD for cortical actin cap visualisation and between 180–210 min post NEBD for PB extrusion. Samples were then transferred to TBSN (10 mM Tris pH 7.4, 155 mM NaCl, 0.1% NP-40), hemisected, blocked by 2% BSA in TBSN for 1 h and incubated with primary antibodies for 24 h. A dsDNA reactive monoclonal antibody (HYB331-01, Santa Cruz) was used to achieve maximum signal intensity for DNA, especially for the analysis of PB extrusion. Anti-NabKin antibodies were used at 10 μg/ml, anti-dsDNA and anti-tubulin (DM1A, Sigma) antibodies at 1:100 dilution from the commercial stock. F-actin was visualised using phalloidin–TRITC at 1 μM final concentration. Samples were washed for 24 h, incubated with Alexa Fluor dye-conjugated secondary antibodies (Invitrogen) at 1:200 dilution from commercial stock for 24 h and again washed for 24 h. For simultaneous detection of tubulin and DNA with mouse monoclonal antibodies, we stained first for DNA as detailed above, blocked the secondary antibodies with mouse normal serum, and stained with a directly FITC-labelled tubulin antibody (DM1A; Sigma) at a 1:100 dilution. Samples were then dehydrated in MeOH (for tubulin visualisation) or acetone (for F-actin visualisation). Acetone dehydration was not as effective as MeOH, but allowed the preservation of phalloidin–TRITC staining. Samples were cleared with Murray’s solution, mounted as previously described (Becker and Gard, 2006) and subsequently examined by CLSM using a Leica SP5 equipped with a HC PL APO × 20, 0.7 NA Glyc and an HCX PL APO × 63, 1.3 NA Glyc objective (both Leica). For analysing Xenopus eggs, jelly coats were removed with cysteine and cells were fixed directly or 5–40 min after activation by addition of 0.16 μg/ml Ca²⁺-ionophore A23187 (Blow and Laskey, 1986).

Eggs were then treated essentially as described for the oocytes above, except that the hemisectioning was omitted and DNA was visualised by Hoechst 33342. A Gaussian blur filter (ImageJ, Sigma Radius 1.00) has been applied to the Z-stacks, shown as maximum intensity projections in Figures 5D and E, to reduce shot noise.

For the cryostat sections, oocytes were embedded in TissueTek (Sakura) on a thin strip of aluminium foil and plunge frozen in isopentane cooled to −150 °C. Samples were mounted, cryostat-sectioned and fixed on coverslip in 3.7% formaldehyde in PBS.

Live cell imaging

Confocal live imaging was performed on a commercial Zeiss LSM 780 microscope using a × 40, 1.4 NA Oil Plan-Apochromat objective. The microscopes was equipped with an EMBL incubation chamber (European Molecular Biology Laboratory), providing a humidified atmosphere at 37 °C with 5% CO₂.

Fab fragment generation

Fab fragments were generated using the Fab Preparation Kit (Pierce; Rockford, USA), according to manufacturer's instructions, and precipitated by 80% saturated ammonium sulphate. Pellets were resuspended in PBS and residual ammonium sulphate was removed using Zeba Desalting Columns (Pierce) according to manufacturer's instructions.

Preparation of Xenopus egg extract

Xenopus egg extract was prepared as previously described (Blow and Laskey, 1986).

Assembly of the NabKin sequence

The mass spectrometric analysis of the NabKin band gave (in relation to its Coomassie staining intensity) unexpectedly few matching peptides, most of which were initially assigned to KIF14. This indicated that the candidate protein was indeed related to, but not identical with KIF14, and that its sequence was not contained in NCBI non-redundant database. We therefore searched the X. tropicalis genome for the missing KIF14 paralogue. By that we could identify a genomic NabKin sequence that gave a large number of additional matching peptides. The position of the 26 introns was reconstructed with the help of X. laevis and X. tropicalis EST clones, the homology to the KIF14 gene and the mass spectrometry data set. The X. laevis NabKin cDNA sequence corresponding to amino acids 1–705 (IMAGE:7010529/IMAGp998K0714717Q, SourceBioScience) and 871–1541 (IMAGE: 4040452/IMAGp998E139296Q, SourceBioScience) was then determined by direct sequencing of indicated EST clones. To fill the gap (residues 706–870), we sequenced gene-specific cDNA (Superscript III, Invitrogen; according to manufacturer’s instructions) from total oocyte RNA (peqGOLD RNApure, Peqlab; according to manufacturer’s instructions). The X. laevis NabKin sequence is deposited at GenBank: NabKin KC342235.

Identification and annotation of the kinesin class-3 family proteins

The kinesin genes have been identified by TBLASTN and PSI-BLAST searches against the sequenced eukaryotic genomes, which have been obtained via lists available from the diArk database (Hammesfahr et al, 2011). All hits were manually analysed at the genomic DNA level together with the multiple sequence alignment of all kinesin proteins to reveal homologous regions. If available for a given species, EST data from the NCBI EST database has been analysed to help in the annotation process. All sequence-related data (names, corresponding species, GenBank ID’s, alternative names, corresponding publications, domain predictions and sequences) and references to genome sequencing centres are available through CyMoBase ( http://www.cymobase.org; Odronitz and Kollmar, 2006). The alignment has been generated manually and curated as a structure-guided multiple sequence alignment (e.g., alignment gaps have only been introduced in known loop regions).

Building of phylogenetic trees

The phylogenetic tree shown in Figure 2A is simplified; the detailed analysis, including the extensive tree, is available on request. In brief, trees were generated based on the conserved kinesin motor domains (corresponding to amino acids 5–352 of HsKIF1A) using three different methods: (1) An unrooted phylogenetic tree was generated using the neighbour-joining and the bootstrap-resampling (1000 replicates) method as implemented in ClustalW (Chenna et al, 2003). (2) ProtTest was used to determine the most appropriate of the available 112 possible amino acid substitution models (Darriba et al, 2011). The tree topology was calculated with the BioNJ algorithm, and both the branch lengths and the model of protein evolution were optimised simultaneously. The LG+I+G+F model as determined by the AIC and AICc criteria was employed in a maximum likelihood analysis of the kinesin-3 data set using the bootstrap-resampling method (1000 replicates) as implemented in RAxML (Stamatakis et al, 2008). (3) Posterior probabilities were generated using MrBayes v3.1.2 (Ronquist and Huelsenbeck, 2003). Two independent runs with 4 000 000 generations, four chains and a random starting tree were computed using the mixed amino acid option. From the 1000th generation, MrBayes used the WAG model (Whelan and Goldman, 2001). Trees were sampled every 1000th generation and the first 100 000 of the trees were discarded as ‘burn-in’ before generating a consensus tree. Sequence comparisons of NabKin and KIF14 domains were done using EMBOSS Matcher alignment.

Domain and motif prediction

Protein domains were predicted using the SMART (Letunic et al, 2011) and Pfam (Finn et al, 2010) web server. Coiled-coil regions were predicted using the Coils software (Lupas et al, 1991).

Statistical analysis

Prism 5 software (GraphPad Software, Inc.) was used for performing the χ²-test.