ODF2 maintains centrosome cohesion by restricting β-catenin accumulation Free

The centrosome functions as microtubule-organizing center (MTOC) in interphase and constitutes the spindle poles in mitosis. Centrosomes are unique organelles consisting of a pair of centrioles and its associated pericentriolar material (PCM), and are duplicated once every cell cycle (Kellogg et al., 1994; Doxsey, 2001; Bornens, 2002). However, as centriole duplication is semi-conservative, centrosomes comprise structurally and functionally diverse centrioles. The older or mother centriole is characterized by the presence of distal and subdistal appendages, and initiates microtubule (MT) polymerization and anchoring. Moreover, it is the mother centriole that is transformed into a basal body to initiate the formation of a primary cilium, an essential sensory organelle present on nearly every cell of the vertebrate body (Dawe et al., 2007; Gerdes et al., 2009; Veland et al., 2009; Hoyer-Fender, 2010, 2013). A signature protein of the mature centriole and the basal body is ODF2 (also known as cenexin), first identified as the major protein component of the sperm cytoskeletal outer dense fibers (ODFs) (Brohmann et al., 1997; Shao et al., 1997; Turner et al., 1997; Lange and Gull, 1995; Nakagawa et al., 2001; Hoyer-Fender et al., 2003; Schweizer and Hoyer-Fender, 2009). ODF2 is essential for cilia formation and, beyond that, is critically important for embryonic development, as ODF2 deficiency in mice leads to pre-implantation lethality (Ishikawa et al., 2005; Anderson and Stearns, 2009; Salmon et al., 2006). Alternative splicing of Odf2 creates functionally diverse isoforms illustrated by the N-terminal cenexin insertion present in so-called cenexin isoforms but not in the abundant testicular ODF2 form (Hüber and Hoyer-Fender, 2007). This insertion in conjunction with the coiled-coil region is required for centrosomal targeting and for the formation of subdistal appendages and basal feet (Hüber et al., 2008; Tateishi et al., 2013). Furthermore, a C-terminal extension of ∼150 amino acids, which specifies human and rat cenexin isoforms, is important for centrosomal targeting and ciliogenesis as well as for the recruitment of Polo-like kinase 1 (Plk1) to the centrosome (Soung et al., 2006, 2009; Rivkin et al., 2008; Chang et al., 2013; Tateishi et al., 2013). The binding site for the polo-box domain of Plk1 is generated by Cdk1-mediated phosphorylation of S796 in the C-terminal extension of human isoform denoted cenexin 1 (Lee et al., 1998; Barr et al., 2004; Elia et al., 2003; Soung et al., 2006, 2009). At the onset of mitosis, activated Plk1 promotes centrosome separation and the formation of a bipolar spindle that is fundamental to ensure correct partitioning of duplicated chromosomes.

Although centrosomes are duplicated synchronously with DNA replication, they are tightly connected and function as a single entity until the onset of mitosis. Parental centrioles are tethered by a proteinaceous linker until the G2/M transition when the linker is severed by the action of kinases. At the onset of mitosis, cyclin B2–Cdk1 activates Plk1, which is recruited to the centrosome by ODF2. Activated Plk1 in turn phosphorylates Mst2, which counteracts the phosphatase activity of the Nek2–PP1γ complex. Subsequently, the activated Nek2 kinase phosphorylates the pivotal tethering proteins C-Nap1 (also known as Cep250) and rootletin, leading to their dissociation and severing of the proteinaceous linker to allow for centrosome separation (Fry et al., 1998; Mayor et al., 2000, 2002; Faragher and Fry, 2003; Bahe et al., 2005; Mardin et al., 2011; Nam and van Deursen, 2014; Hardy et al., 2014). Likewise important for centrosome separation and establishment of a bipolar spindle is β-catenin, which is in accordance with the observation that Wnt signaling can induce centrosome splitting, eventually causing the centrosome aberrations observed in cancer cells (Kaplan et al., 2004; Hadjihannas et al., 2010). Logically, several Wnt pathway components have been localized at the centrosome (Fumoto et al., 2009; Itoh et al., 2009; Kim et al., 2009; Hadjihannas et al., 2010; Mbom et al., 2013). β-catenin, specifically, colocalizes with rootletin and is also phosphorylated by Nek2. However, increased Nek2 activity results in rootletin-independent binding of (stabilized) β-catenin to centrosomal sites, and this is essential for centrosome separation (Bahmanyar et al., 2008). Nek2, although phosphorylating the same regulatory sites in the N-terminus of β-catenin as GSK3β, seems to inhibit binding of the E3 ligase β-TrCP, thus preventing β-catenin ubiquitylation and degradation. Nek2-mediated phosphorylation of β-catenin therefore stabilizes β-catenin at the centrosome (Mbom et al., 2014). The negative regulator of β-catenin, conductin (also known as Axin2), binds to C-Nap1. Conductin promotes β-catenin phosphorylation to stimulate its proteasomal degradation. Consequently, knockdown of conductin reduces the phosphorylation of β-catenin at the centrosome, and knockdown of C-Nap1 or rootletin abolishes β-catenin phosphorylation. However, since loss of Nek2 neither affected centrosomal linker dissolution nor cell cycle progression, an alternative pathway for linker dissolution must operate (Fletcher et al., 2004; Mardin and Schiebel, 2012). This model proposes that centrosome cohesion is promoted by phosphorylation-mediated β-catenin degradation whereas unphosphorylated, and therefore stabilized, β-catenin promotes centrosome splitting (Hadjihannas et al., 2010). Separated centrosomes afterwards form the two poles of the bipolar spindle.

Since formation of a bipolar spindle is crucial for correct chromosome segregation, the association of centrosomal abnormalities with genome instability and tumorigenesis is hardly surprising (Salisbury et al., 1999; Wang et al., 2004). ODF2 is a marker protein of the mature centrosome and essential for cell viability. Since aberrant expression of Odf2 is frequently observed in cancer cells, it has been acknowledged as both a cancer and testis antigen (Whitehurst, 2014). Previous data, obtained in human cancer cells, have hinted that ODF2 is involved in bipolar spindle formation and normal mitotic progression (Soung et al., 2006, 2009). We used here the mouse fibroblast cell line NIH3T3 to determine the involvement of ODF2 in centrosome cohesion. We show that depletion of ODF2 induced centrosome splitting. Depletion of ODF2 is accompanied by an increase in centrosomal β-catenin. These data are corroborated by reporter gene assays that show that overexpression of ODF2 inhibits canonical Wnt signaling. Our results thus demonstrate that ODF2 reduces β-catenin stability, thus maintaining centrosome cohesion and preventing premature centrosome splitting.

To determine the importance of ODF2 for centrosome cohesion, we investigated the effect of ODF2 depletion on centrosome and spindle pole splitting. To this end, NIH3T3 cells were transfected either with Odf2 siRNA or a non-targeting control siRNA and treated with the Eg5 (also known as KIF11) inhibitor VS-83 (depletion of ODF2 by siRNA is shown in Fig. S1). The mitotic kinesin Eg5 is essential for spindle pole separation leading to monoaster formation when inhibited (Kapitein et al., 2005). Spindle poles were detected by γ-tubulin staining, and their appearance was considered either monopolar and bipolar as shown in Fig. 1A. In control siRNA-treated cells more than 70% of cells showed monopolar spindles 48 h after transfection (n=1077). However, depletion of ODF2 decreased the proportion of cells with monopolar spindles, with a concurrent increase in the proportion of cells with separated spindle poles up to ∼35% (n=1113), corresponding to a 1.3-fold change, which is statistically significant (P<0.05, Student's t-test type 2) (Fig. 1B). Decreased centrosome cohesion might in turn provoke chromosome segregation defects as indicated by the formation of tripolar spindles. When ODF2 was depleted by siRNA transfection, we found an increase of the number of tripolar spindles from 15% (n=32 from a total of 215 cells) in control cells to 23% (n=23 from a total of 100 cells) in siRNA-treated cells.

We next asked whether decreased spindle pole cohesion is already reflected by a relaxation of centrosome cohesion in the G2 phase. NIH3T3 cells were transfected either with Odf2 siRNA or a non-target control siRNA and the centrosomes immunologically detected by staining for either γ-tubulin or pericentrin. The distances of the two spots, which characterize the G2 centrosome, were measured using the software program Amira (Amira 5.3.2; Stalling et al., 2005). We found a statistically significant increase of the distance between the two γ-tubulin spots (Odf2 siRNA n=69, control siRNA n=104) as well as between the two pericentrin spots (Odf2 siRNA n=43, control siRNA n= 64) when ODF2 is depleted, indicative of a relaxation of centrosome cohesion (Fig. 1C,D).

These data were corroborated by experiments where ODF2 depletion was mediated by the short hairpin vector sh3 (Fig. S2). To this end, NIH3T3 cells were transfected either with the control vector K07, the short hairpin vector sh3, or sh3 and human cenexin (hCenexin) vectors for rescue. Simultaneous transfection of a plasmid encoding human histone H2A fused to Emerald (H2A–Em) served to identify transfected cells. At 48 h post transfection, cells were incubated with the Eg5 inhibitor VS-83 for 2 h and finally processed for immunodecoration of acetylated tubulin. The numbers of monopolar, bipolar and multipolar spindles were counted in cells transfected with K07 (n=687), sh3 (n=588), and sh3 plus hCenexin (n=293) that were H2A–Em positive (Fig. 1E,F). sh3-mediated depletion of ODF2 significantly promoted the formation of bipolar and multipolar spindles, whereas hCenexin expression rescued the phenotype (P<0.05 and P<0.01, respectively) (Fig. 1F). We additionally observed an increase in the distance of centrosomes in monopolar spindle poles when ODF2 is depleted, from 1.4290 µm in K07-transfected control cells (n=54) to 1.623 µm in ODF2-depleted cells (n=51) that was rescued by hCenexin expression to 1.408 µm (n=42; P<0.05) (Fig. 1G). Our data indicate that relaxation of centrosome cohesion that occurs upon ODF2 depletion eventually provokes centrosome amplification, causing the formation of multipolar spindle poles.

ODF2 is a marker protein of the mature centriole. The Wnt pathway component β-catenin localizes to the centrosome and its stabilization has been reported to impair centrosome cohesion, thus provoking premature centrosome splitting (Lim et al., 2009; Mbom et al., 2013). We therefore queried whether the centrosome separation as observed upon ODF2 depletion is reflected by a shift in the amount of centrosomal β-catenin. Centrosomal localization of ODF2 and β-catenin is shown in Fig. 2. Furthermore, there is an obvious increase of the amount of β-catenin at the centrosome when ODF2 is reduced. Depletion of ODF2 was mediated by siRNA transfection in NIH3T3 cells and monitored 48 h later. Expression of Centrin2–GFP was used as a centrosomal marker. Successful depletion of ODF2 by Odf2 siRNA was verified by anti-ODF2 staining (Fig. 2E–H). In control siRNA-transfected cells, ODF2 is clearly visible at the centrosome, which is highlighted by Centrin2–GFP colocalization (Fig. 2A–D). Odf2 siRNA transfection caused an obvious decrease of ODF2. Depletion of ODF2 by means of Odf2 siRNA transfection, additionally, affected centrosomal β-catenin (Fig. 2I–P). Compared to control siRNA-transfected cells, the amount of centrosomal β-catenin is increased when ODF2 is depleted (Fig. 2M–P compared to I–L), which was additionally confirmed by quantification of centrosomal β-catenin (Fig. 2Q). siRNA-mediated depletion of ODF2 resulted in an ∼2-fold increase of β-catenin in interphase centrosomes (P<0.01; Student's t-test; n=14 for control siRNA-treated cells and n=27 for Odf2 siRNA-treated cells). To further substantiate these data, we depleted ODF2 via expression of a short hairpin construct (sh3). Cells were co-transfected with either the non-target control plasmid (K07) and an expression vector encoding Centrin2 fused to Cherry (Centrin2-Cherry), or sh3 and Centrin2-Cherry (sh3/Centrin2-Cherry). To rescue sh3-mediated ODF2 knockdown, a human cenexin expression plasmid (hCenexin) was transfected concurrently with sh3/Centrin2-Cherry plasmids. At 48 h post transfection, cells were processed for immunocytology and the centrosomal area, indicated by Centrin2–Cherry labelling, quantified (Fig. 3A–C). Expression of the short hairpin sequence targeting ODF2 (sh3) resulted in a significant decrease of ODF2 protein at the centrosome compared to that seen upon the expression of the non-target control K07 (Student's t-test; P<0.001; K07, n=30; sh3, n=30) (Fig. 3D). In contrast, the centrosomal amount of β-catenin is increased when ODF2 is depleted (P<0.001; K07, n=26; sh3, n=28). Concurrent expression of human cenexin counteracted the effects of sh3-mediated ODF2 depletion, resulting in a significant decrease in the amount centrosomal β-catenin compared to sh3-transfected cells (P<0.001; sh3+hCenexin: n=25) but an amount that is similar to the amount of centrosomal β-catenin in K07-transfected cells (P>0.1) (Fig. 3D). Our results, therefore, demonstrate that ODF2 suppresses accumulation of β-catenin at the centrosome.

To further corroborate our previous results indicating a negative effect of ODF2 on β-catenin accumulation or stability, we took advantage of the canonical Wnt reporter gene assay. The assay was performed using the reporter gene OT, which expresses firefly Luciferase under the control of three copies of the wild-type Tcf-4-binding sequence (Fig. 4). Co-transfection of expression plasmid encoding β-catenin activated the reporter resulting in a ∼2.5-fold increase compared to the control transfected only with the reporter OT and the control plasmid phRL. Transfection of the Odf2 expression plasmid 13.8NC-GFP (OT; phRL; ODF2/13.8NC-GFP) did not significantly alter the transcriptional activity of OT beyond the background level (OT; phRL). However, ODF2/13.8NC significantly inhibited the β-catenin-mediated transcriptional upregulation of the reporter vector (P<0.001) [OT; phRL; β-catenin; ODF2 (13.8NC-GFP) compared to OT; phRL; β-catenin]. siRNA-mediated knockdown of ODF2 did not significantly alter the transcriptional activity of OT (OT; phRL; β-catenin; Odf2 siRNA compared to OT; phRL; β-catenin) most likely owing to the general low amount of endogenous ODF2 (Fig. 4). Moreover, inhibition of canonical Wnt reporter gene activity was also found upon expression of ODF2/13.8NCΔGFP, in which the GFP tag is omitted, demonstrating that the GFP tag is not responsible for canonical Wnt reporter gene inhibition (data not shown).

Our studies so far have demonstrated that overexpression of ODF2 represses β-catenin-mediated reporter gene activity and inhibits the accumulation of β-catenin at the centrosome. To further substantiate whether ODF2 affected β-catenin stability, the relative amount of β-catenin was quantified on western blots. HEK293 cells were transfected with either β-catenin or ODF2 (13.8NC-GFP) expression plasmids or both, and the quantity of β-catenin (at ∼100 kDa) relative to α-tubulin quantity was assessed (Fig. 5). Overexpression of ODF2 (13.8NC-GFP) caused a reduction of the relative amount of β-catenin both of the endogenous as well as of the overexpressed β-catenin (Fig. 5A). Calculation of the relative amount of β-catenin in three independent biological replicas revealed an overall reduction to ∼40–60% compared to the control (Fig. 5B). The reduction in the relative amount of β-catenin is statistically significant [P<0.05 for endogenous β-catenin and P<0.01 in cells overexpressing both β-catenin and ODF2 (13.8NC-GFP)].

Collectively, our data support the view that ODF2 promotes β-catenin degradation. Accordingly, depletion of ODF2 caused an increase in centrosomal β-catenin. We thus asked whether both proteins are present in the same complex. Cells transfected with ODF2 constructs were lysed, and the endogenous β-catenin was immunoprecipitated. Western blotting revealed that not only ODF2 (13.8NC, without GFP tag) but also its C-terminally (NC2-GFP) and N-terminally (N2C-GFP) truncated versions co-precipitated with β-catenin (Fig. 6A, co-IP). Neither β-catenin nor ODF2 (here fused to GFP, 13.8NC-GFP) were precipitated using an unrelated, control IgG (Fig. 6A, co-IP, +control IgG) whereas both proteins are present in the lysate (input). Furthermore, neither β-catenin nor ODF2 bound unspecifically to the beads (Fig. 6A, co-IP, −). We, additionally, excluded the possibility that binding to β-catenin could be mediated by the GFP fusion tag by immunoprecipitation of β-catenin from lysate from cells expressing GFP (Fig. S3). The co-IP data are corroborated by pulldown assays in which bacterially expressed ODF2 fused to 6×His (13.8NC-6×His) was incubated with HEK293 cell lysates and affinity purified on Ni-NTA agarose beads. As shown in Fig. 6B, β-catenin was trapped in the presence of ODF2–His but not in its absence (control).

To further elucidate whether ODF2 is a component of the β-catenin destruction complex, we co-transfected HEK293 cells with both the ODF2 construct 13.8NC-GFP and Axin1-Myc, and immunoprecipitated the ODF2–GFP fusion protein. Immunoblotting revealed expression of ODF2–GFP and Axin1–Myc (Fig. 6C, input; Axin1–Myc ∼130 kDa), and co-precipitation of both proteins [Fig. 6C, +ODF2 (13.8NC-GFP), Axin1–Myc]. Neither unspecific binding to beads, by omitting the precipitating antibody (Fig. 6C, −control), nor unspecific binding to control IgG was found (Fig. 6C, +control IgG) although both ODF2 (13.8NC without the GFP-tag) and the endogenous Axin1 are present in the lysate (input control, without Axin1-Myc transfection). Collectively, our results support the view that ODF2 coexists in a complex with β-catenin and Axin1.

ODF2 depletion caused an increase in centrosomal β-catenin and consequently centrosome splitting. We thus questioned whether centrosome splitting at the onset of mitosis is accompanied by a decrease in ODF2. Cycling NIH3T3 cells were immunostained for γ-tubulin and ODF2, and the level of centrosomal ODF2 quantified with ImageJ. Interphase centrosomes were identified either through the presence of a single γ-tubulin spot per cell (G1 centrosome) or the presence of twin γ-tubulin spots (G2 centrosome). Mitotic centrosomes were identified through the presence of distant γ-tubulin spots in conjunction with chromosome condensation. We found a significant decrease of centrosomal ODF2 on the mitotic centrosome (n=27) compared to the interphase centrosome (n=53) (P<0.001, Student's t-test) indicating that centrosome splitting at the onset of mitosis occurs at the same time as a decrease in ODF2 (Fig. 7).

ODF2 is a component of the insoluble centrosomal scaffold, and beyond that specifically associates with the subdistal appendages of the mother centriole (Lange and Gull, 1995; Nakagawa et al., 2001; Tateishi et al., 2013). Moreover, since appendage formation is crucially dependent on ODF2, they are lacking in ODF2-deficient F9 cells (Ishikawa et al., 2005). In epithelial cells, appendages on the mother centriole persist throughout the cell cycle, including mitosis, whereas they are formed on the immature former daughter centriole not earlier than through the second half of mitosis. Throughout G2, when the recruitment of appendage proteins has already started, and the first half of mitosis, appendages are not visible on the immature centriole (Vorobjev and Chentsov, 1982; Kobayashi and Dynlacht, 2011; Lee and Rhee, 2015). In G0/G1 phase, ODF2 preferentially associates with the mother centriole, but becomes detectable in both mother and daughter centrioles towards the G1/S transition, just before centriole duplication (Nakagawa et al., 2001). High-resolution imaging in ciliated cells has revealed a ring-like arrangement of ODF2 at the distal part of the basal body, indicating recruitment of ODF2 to additional sites during the centriole to basal body transition (Herawati et al., 2016). Moreover, these results illustrate the promising power of new imaging tools. Since for most centrosomal proteins, including the Wnt pathway components, only a rather imprecise centrosomal location has been reported, super-resolution microscopy techniques are important to clarify the exact location of centrosomal proteins during cell cycle progression.

Centrosomes are unique cellular organelles in cycling cells that are duplicated once per cell cycle. However, duplicated centrosomes stick together to function as a single entity until onset of mitosis when the proteinaceous linker is dissolved to allow bipolar spindle formation. The serine/threonine kinase Plk1 plays a critical role in centrosome separation and maturation by mediating the recruitment of PCM components. Plk1 itself is recruited to the centrosome by the ODF2 isoform known as cenexin, which is a marker protein of the mature centriole. Cenexin provides the docking site for the Polo-box domain when phosphorylated by CDK1. Activated Plk1 then leads to activation of downstream kinases to phosphorylate linker proteins eventually resulting in dissolution of the proteinaceous linker and centrosome disjunction (Soung et al., 2006; Soung et al., 2009; Mardin et al., 2011; Nam and van Deursen, 2014; Joukov et al., 2014). The ultimate kinase responsible for linker protein phosphorylation is Nek2, which, in addition, phosphorylates β-catenin, resulting in its stabilization (Bahmanyar et al., 2008; Mbom et al., 2014).

β-catenin, as well as components of the β-catenin degradation complex, are located at the centrosome, and stabilization of β-catenin promotes centrosome splitting (Fumoto et al., 2009; Itoh et al., 2009; Kim et al., 2009; Hadjihannas et al., 2010; Mbom et al., 2013). Moreover, besides its well-known effect on β-catenin stabilization, aberrant activation of Wnt signaling also caused centrosome splitting that is eventually followed by centrosome aberrations, a hallmark of cancer cells (Kaplan et al., 2004; Hadjihannas et al., 2010). The significance of β-catenin accumulation for centrosome disjunction inspired a new model in which β-catenin has been attributed a central role. In this model, centrosome cohesion is promoted by phosphorylation-mediated β-catenin degradation, whereas stabilized β-catenin promotes centrosome splitting (Hadjihannas et al., 2010). However, the Nek2 kinase, which appeared responsible for both, linker dissolution and β-catenin stabilization, seems to be dispensable for centrosome disjunction and cell cycle progression (Fletcher et al., 2004; Mardin and Schiebel, 2012). Accordingly, another pathway for linker dissolution, in addition to the Nek2 pathway, must operate, which credits the prominent role β-catenin plays.

Previously, the HeLa human cancer cell line was used to examine the relevance of the Plk1-recruiting protein ODF2. The results, obtained by using a stable knockdown cell line, indicated that ODF2 is involved in bipolar spindle formation and normal mitotic progression. However, up to now, the effect appeared to be unique for the HeLa cancer cell line since the data could not be reproduced in the ODF2-deficient F9 cell line (Ishikawa et al., 2005; Soung et al., 2006, 2009).

Here, by undertaking a transient depletion of ODF2 in the NIH3T3 mouse fibroblast cell line, we show that ODF2 is required to prevent premature β-catenin accumulation at the centrosome and centrosome disjunction. Quantitative immunoblotting and reporter gene assays using the canonical Wnt signaling pathway indicated that ODF2 affected the abundance of β-catenin. However, since ODF2 is a coiled-coil protein without any known enzymatic activity, its negative effect on β-catenin accumulation is most probably mainly indirect. Given that ODF2 is a microtubule-associated protein, one likely explanation is that it might prevent β-catenin accumulation by either stabilizing the β-catenin destruction complex or promoting β-catenin degradation (Donkor et al., 2004) (Fig. 8). In fact, co-precipitation of ODF2 together with β-catenin and Axin1 corroborates this view. Taken together, our data suggest that ODF2 controls the amount of β-catenin at the centrosome. Decreasing the level of ODF2 by means of both siRNA and shRNA caused an increase in the amount of centrosomal β-catenin and centrosome disjunction. License of centrosome disjunction at the onset of mitosis could, therefore, be granted by a decrease in the amount of endogenous ODF2 that, in turn, would enable accumulation of β-catenin. This view is corroborated by the observation that the amount of ODF2 fluctuates within the cell cycle, revealing a significant higher amount of ODF2 in interphase centrosomes than in mitotic centrosomes.

Our data indicate that ODF2 plays a central role in centrosome cohesion besides acting solely as a docking site for the recruitment of Plk1 at mitosis. In interphase centrosomes, ODF2 keeps centrosomal β-catenin low to maintain centrosome cohesion. At the onset of mitosis, accumulation of β-catenin is supported by a decrease of ODF2, thus acting in line with the activation of the Nek2 pathway to finally promote centrosome splitting (Fig. 8). However, given that ODF2 serves as a docking site for the recruitment of the Plk1 kinase to activate Nek2, a decrease of ODF2 is likely to also counteract Nek2 activation. Nevertheless, supposing that ODF2 plays an important role in preventing β-catenin accumulation at the interphase centrosome, in the G2/M centrosome, β-catenin accumulation could be maintained even in the absence of activated Nek2 by just decreasing the amount of ODF2. Thus, depletion of ODF2 would not only abrogate recruitment of Plk1 and subsequent activation of the Nek2 pathway to promote linker dissolution and β-catenin stabilization but, additionally, also abolish restraint β-catenin accumulation (Fig. 8).

Owing to the nature of centriole duplication, the two spindle poles of the bipolar spindle are not identical but comprise either the old mother centriole or the newly formed mature centriole, finally resulting in asymmetric centrosome inheritance. Stem cells, stereotypically, inherit the centrosome that nucleates more-stable microtubules, and that is, in most cases, the old centrosome. ODF2 not only defines the old centrosome but is also responsible for microtubule stability. Depletion of ODF2 thus disturbs asymmetric cell division, resulting in a depletion of the stem cell pool during neurogenesis (Wang et al., 2009; Tylkowski et al., 2014; Gasic et al., 2015; Hung et al., 2016). The preimplantation lethality observed in ODF2-deficient mice may therefore be caused by perturbing replenishing of the stem cell pool (Salmon et al., 2006). The establishment and the orientation of a bipolar mitotic spindle, as well as the anchoring of MTs at the centrosome, are also controlled by β-catenin. Likewise, β-catenin is of utmost importance for sustaining a stem cell pool (Huang et al., 2007; Chilov et al., 2011). However, beyond experimental evidence, a clear picture of its molecular function is still missing (Bryja et al., 2017). Although we have shown that ODF2 counteracts β-catenin accumulation at the centrosome, future work has to elucidate whether and how both proteins converge in MT formation and spindle positioning.

To sum up, our results corroborate the data obtained by Soung et al. (2006, 2009) using the HeLa cancer cell line suggesting that ODF2 is involved in bipolar spindle formation. Furthermore, we demonstrate that ODF2 restricts β-catenin accumulation at the centrosome, which, in turn, prevents aberrant centrosome splitting and spindle formation. Since experimental depletion of ODF2 causes centrosome splitting and fosters the formation of tripolar spindles, aberrant expression of ODF2 launches a link to chromosomal instability (CIN), which is a hallmark of tumors (de Cárcer and Malumbres, 2014).

NIH3T3 or HEK293 cells (ATCC; regularly authenticated and tested for contamination) were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS), 1000 U/ml penicillin, 1000 µg/ml streptomycin and 20 mM L-glutamine at 37°C and 5% CO₂. NIH3T3 cells were transfected using Transfectin (BioRad) or Lipofectamin 2000 (ThermoFisher Scientific Biosciences); HEK293 cells were transfected with XtremeGene HP (Roche). Odf2 siRNA (stealth siRNA ODF2MSS207236; Life Technologies) and control siRNA (siGenome Non-targeting siRNA #1; ThermoFisher Scientific Biosciences) were transfected with silentFect (BioRad). At 24 h before siRNA transfection cells, were initially transfected with an expression plasmid encoding Centrin2–GFP to highlight centrosomes. Short hairpin constructs sh3 (specifically targeting the sequence GAACTCCTCCAGGAGATAC of mouse ODF2; Tylkowski et al., 2014) or K07 (Origene), which functions as a control since homology to any known mRNA is missing, and the expression plasmids encoding human cenexin (hCenexin) (Soung et al., 2006), human centrin 2 fused to Cherry (Centrin2-Cherry), and histone H2A fused to Emerald (Addgene #54110) were transfected with Lipofectamin 2000 (ThermoFisher Scientific Biosciences). Eg5-inhibitor VS-83 was used at a final concentration of 5 µM for 2 h (Millipore-Calbiochem). For immunocytology, NIH3T3 cells grown on coverslips were fixed in methanol, pre-incubated in 0.3% Triton X-100 in PBS for 10 min, washed in PBS and then incubated in 1% bovine serum albumin, 0.5% Tween-20 in PBS (PBT) for blocking unspecific binding sites. Incubation with the antibodies was performed at 37°C for 1 h. Primary antibodies used were against β-catenin (ab79089, Abcam), ODF2 (ESAP15572, Antibodies online; Brohmann et al., 1997), γ-tubulin (T6557; Sigma-Aldrich) and pericentrin (ab4448, Abcam), all diluted 1:100. Secondary antibodies were goat anti-mouse-IgG antibody conjugated to Alexa Fluor 488 (Molecular Probes, Eugene), goat anti-rabbit IgG antibody conjugated to MFP-488 (Mobitec, #MFP-A1008), goat anti-mouse-IgG antibody conjugated to Alexa Fluor 555 (Invitrogen) and goat anti-rabbit IgG antibody conjugated to Alexa Fluor MFP590 (Molecular Probes, Eugene). For detection of acetylated tubulin [anti-acetylated α-tubulin (clone 6-11B-1), sc-23950 Santa Cruz Biotechnology] cells were fixed in 3.7% paraformaldehyde in PBS and processed as described above. DNA was counterstained with DAPI. Images were taken by confocal microscopy (LSM 510, Zeiss), and processed using Adobe Photoshop 5.0. ImageJ was used for quantification of the centrosomal area. Amira 5.3.2 or ImageJ were used for measuring centrosomal distances (Stalling et al., 2005). For statistical analyses, a one-tailed, type 2 Student's t-test was used.

Reporter gene assays were performed in NIH3T3 or HEK293 cells. Cells were seeded at a density of 1.5×10⁵ or 1×10⁶ cells, respectively, per well of a 12-well plate. After 24 h Odf2 siRNA (stealth siRNA ODF2MSS207236; Life Technologies; 12 nmol per well) was transfected, when required. Transfection of reporter gene constructs (1 µg/well) together with expression plasmids (each 100 ng/well) was performed 24–48 h before luciferase assays. As reporter genes for the canonical Wnt assay, we used plasmids OT (containing three copies of the wild-type Tcf-4 binding sequence) and OF (containing the mutant Tcf-4 binding sequence) (Shih et al., 2000). For activation of Wnt-reporter genes we used a plasmid encoding β-catenin (Morin et al., 1997; Korinek et al., 1997). Reporter gene plasmids OT and OF, and the expression plasmid for β-catenin were kindly provided by Bert Vogelstein, The Howard Hughes Medical Institute & Sidney Kimmel Comprehensive Cancer Center, Baltimore, MD. Constructs were co-transfected with phRL-SV40 (Promega, Madison, USA), used as internal control. The Dual-Glo Luciferase Assay System (Promega, USA) was used for measuring firefly and Renilla luciferase activity using the Centro LB 960 luminometer (Berthold Technologies, Germany). Fold changes were calculated based on the relative luminescence (firefly luminescence/Renilla luminescence). Each experiment was performed in triplicate and repeated at least 3 times.

ODF2 isoform 13.8NC (Hüber et al., 2008) was subcloned into pET28a(+) using EcoRI and NotI to yield p13.8NC-6xHis. E. coli BL21, inoculated with p13.8NC-6His, was pre-cultured in 5 ml lysogeny broth (LB) medium at 37°C. Large-scale expression cultures were inoculated with pre-culture diluted 1:500 and expression was induced by auto induction in ZYM 5052 overnight at 37°C (Studier, 2005). The inclusion bodies of the 13.8NC-6xHis fusion protein were dissolved in wash buffer [50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.2 mM PMSF and 1 mM DTT] and purified by three repetitive sonication and centrifugation steps (5 s pulse at 50 W followed by centrifugation at 13,000 g for 30 min). The pellet was additionally washed and sonicated two times in wash buffer containing 1% Triton X-100 (v/v). The pellet was then dissolved in 100 mM Tris-HCl (pH 8.0), 50 mM glycine and 8.5 M urea. 5 mM reduced glutathione (GSSH) and 0.5 mM oxidized glutathione (GSSG) were added and stirred overnight at 4°C. Refolding of the protein was performed by dialyzing against refolding buffer (100 mM Tris-HCl, 400 mM L-arginine, 1 mM EDTA and 0.2 mM PMSF, pH 8.0) with decreasing urea concentrations (4 M, 2 M, 1 M and 0.5 M urea) over several days at 4°C. The penultimate dialyzing step was performed against refolding buffer without urea, and the final step was performed with 1×PBS (pH 7.6). Each dialyzing step was performed overnight at 4°C. The amount of 6×His-13.8NC fusion protein was quantified by comparing it to BSA standards via Coomassie Blue staining.

For in vitro interaction experiments, 6×His-13.8NC was added to washed Ni-NTA agarose (Qiagen GmbH, Hilden) [wash buffer at pH 7.6; 50 mM NaPi, 500 mM NaCl, 30 mM imidazole, 0.2 mM PMSF and protease inhibitors (Protease Inhibitor Cocktail Tablets, complete Mini, Roche, Mannheim, Germany)] and incubated for 2 h at 4°C with constant agitation. NIH3T3 cell lysates were added to the resin and incubated for at least 2 h at 4°C. The resin was then washed four times with wash buffer, followed by a final washing step performed overnight. Proteins were eluted with 50 mM NaPi, 500 mM NaCl, 500 mM imidazole, 0.2 mM PMSF. Finally, probes were analyzed by SDS-PAGE and western blotting (Laemmli, 1970; Towbin et al., 1979) using mouse anti-His tag (1:1000, #70796-3, Novagen) and rabbit anti-β-catenin (1:1000, ab79089, Abcam) antibodies. As a negative control, Ni-NTA agarose was incubated with NIH3T3 cell lysates without 6×His-13.8NC and processed as described.

NIH3T3 cells or HEK293 cells were transfected with expression plasmids using either Transfectin reagent (BioRad, Munich, Germany) for NIH3T3 cells or X-tremeGene HP (Roche Diagnostics) for HEK293 cells. Generation of ODF2 truncations (NC2-GFP and N2C-GFP) is described in Donkor et al. (2004). At 24 h post transfection, cells were trypsinized and washed two times in PBS. Cells were then lysed in lysis buffer [1×PBS containing 1% Nonidet P40, 100 µg/ml (w/v) PMSF and protease inhibitor mix (protease inhibitor cocktail tablets, complete Mini, Roche, Mannheim, Germany)] for 20 min on ice using 1 ml buffer per 10⁷ cells followed by passing it at least 10 times through a syringe with a 21-gauge needle. After centrifugation (20,000 g for 15 min at 4°C), the supernatant was incubated with 2 µg/ml of an anti β-catenin antibody for 2 h at room temperature. Not adding antibody and adding an unrelated IgG were performed as negative controls. Immobilized protein G–agarose (Thermo Fisher Scientific Biosciences) was washed three times with lysis buffer and added to the supernatant. Incubation was carried out overnight at 4°C. Afterwards, the resin was washed four times in lysis buffer. Proteins were finally extracted in SDS sample buffer by boiling for 5 min. Proteins were analyzed by SDS-PAGE and immunoblotting using rabbit anti-GFP (1:1000, ab26422; Abcam), rabbit anti-ODF2 (1:1000, Brohmann et al., 1997), and mouse anti β-catenin (1:1000, 7F7.2; Millipore) antibodies, respectively. For co-immunoprecipitation of ODF2 with Axin1, cells were co-transfected with ODF2 (construct 13.8NC-GFP) and Axin1-Myc (Addgene #21287; Zeng et al., 1997). Proteins were co-precipitated using anti GFP antibody. Omitting anti-GFP antibody or incubation with an unrelated IgG, respectively, served as controls. Detection was achieved by using rabbit anti-GFP (1:1000, ab26422, Abcam) and rabbit anti-Axin1 (1:1000, C76H11; Cell Signaling) antibodies, respectively.

For the kind gift of plasmids, we thank our colleagues Kyung S. Lee (National Institutes of Health, Bethesda), Jeffrey Salisbury (Mayo Clinic, Rochester), Roger Tsien (Howard Hughes Medical Institute, La Jolla), Bert Vogelstein (The Howard Hughes Medical Institute & Sidney Kimmel Comprehensive Cancer Center, Baltimore) and Terry Yamaguchi (National Institutes of Health, Frederick, MD).