Wnts activate at least two signaling pathways, the β-catenin-dependent and -independent pathways. Although the β-catenin-dependent pathway is known to contribute to G1–S transition, involvement of the β-catenin-independent pathway in cell cycle regulation remains unclear. Here, we show that Wnt5a signaling, which activates the β-catenin-independent pathway, is required for cytokinesis. Dishevelled 2 (Dvl2), a mediator of Wnt signaling pathways, was localized to the midbody during cytokinesis. Beside the localization of Dvl2, Fz2, a Wnt receptor, was detected in the midbody with the endosomal sorting complex required for transport III (ESCRT-III) subunit, CHMP4B. Depletion of Wnt5a, its receptors, and Dvl increased multinucleation. The phenotype observed in Wnt5a-depleted cells was rescued by the addition of purified Wnt5a but not Wnt3a, which is a ligand for the β-catenin-dependent pathway. Moreover, depletion of Wnt5a signaling caused loss of stabilized microtubules and mislocalization of CHMP4B at the midbody, which affected abscission. Inhibition of the stabilization of microtubules at the midbody led to the mislocalization of CHMP4B, while depletion of CHMP4B did not affect the stabilization of microtubules, suggesting that the correct localization of CHMP4B depends on microtubules. Fz2 was localized to the midbody in a Rab11-dependent manner, probably along stabilized microtubules. Fz2 formed a complex with CHMP4B upon Wnt5a stimulation and was required for proper localization of CHMP4B at the midbody, while CHMP4B was not necessary for the localization of Fz2. These results suggest that Wnt5a signaling positions ESCRT-III in the midbody properly for abscission by stabilizing midbody microtubules.
Wnt is a secreted ligand that regulates various cellular functions, including proliferation, differentiation, migration, and adhesion (Logan and Nusse, 2004). There are 19 Wnt members in the human and mouse genomes, and these Wnts activate at least two signaling pathways, the β-catenin-dependent and -independent pathways (Kikuchi et al., 2009; Sato et al., 2010). The β-catenin-dependent pathway contributes to G1–S transition by expressing cyclin D1 and c-Myc in a β-catenin/T cell factor (Tcf)-dependent manner, and abnormal accumulation of β-catenin causes tumorigenesis by disrupting cell cycle control (Kikuchi et al., 2009; MacDonald et al., 2009). In addition, it has been reported that low-density-lipoprotein receptor-related protein 6 (LRP6), which is a co-receptor for the β-catenin-dependent pathway, is efficiently phosphorylated by a mitotic cyclin Y-dependent protein kinase, Pftk, and that β-catenin-dependent transcription is enhanced in mitosis (Davidson et al., 2009). Expression of conductin (also referred to as Axil and Axin2), which is a target gene of the β-catenin-dependent pathway, was regulated in a cell-cycle-dependent manner, with its highest expression level occurring during mitosis, and overexpression of conductin caused chromosomal instability in colon cancer cell lines (Hadjihannas et al., 2006). Dishevelled (Dvl), a cytoplasmic mediator of Wnt signaling pathways, as well as LRP6 and a Wnt receptor Frizzled 2 (Fz2), was demonstrated to be involved in mitotic spindle orientation (Kikuchi et al., 2010). These results suggest that the β-catenin-dependent pathway not only promotes G1–S progression but also regulates mitotic progression. On the other hand, it remains unclear whether the β-catenin-independent pathway functions in cell cycle regulation.
From anaphase to telophase, an actomyosin-based contractile ring in the midzone drives the ingression of the cleavage furrow between the two reforming nuclei, resulting in the formation of the midbody, which is a narrow anti-parallel microtubule-enriched intercellular bridge connecting the two daughter cells, during cytokinesis (Hu et al., 2012; Steigemann and Gerlich, 2009). The intercellular bridge persists until abscission splits the daughter cells apart. An endosomal sorting complex required for transport (ESCRT) has been implicated in diverse cellular processes, such as endosomal membrane sorting and cytokinesis (Peel et al., 2011). ESCRTs and vascular protein sorting 4 (Vps4) are sequentially localized to the midbody to mediate fission events (Wollert et al., 2009). During cytokinesis, ESCRT-I protein tumor-susceptibility gene 101 (Tsg101) and apoptosis-linked gene 2-interacting protein X (ALIX) interact with a non-canonical coiled-coil region of centrosomal protein of 55 kDa (Cep55) (Carlton and Martin-Serrano, 2007). Cep55 is localized to the center of the midbody where it binds to the centralspindlin subunit mitotic kinesin-like protein-1 (Mklp1) (Fabbro et al., 2005). The ESCRT-III subunit charged multivesicular body protein (CHMP) binds to the microtubule-severing enzyme Spastin and accumulates on a region adjacent to Tsg101 to lead to complete abscission (Elia et al., 2011; Morita et al., 2007). Vps4 functions to dissociate ESCRT-III after membrane scission is completed (Hurley and Hanson, 2010). All of these ESCRTs and Vps4 are essential for cytokinesis, because their depletion leads to cytokinetic failure, which is assessed by an increase in multinucleated cells.
Progress in the understanding of the roles of ESCRTs in cytokinesis promoted us to analyze the relationship between Wnt signaling and ESCRTs during cytokinesis, because we previously found that Dvl localizes to the midbody (Kikuchi et al., 2010). The β-catenin-independent pathway was originally shown to regulate cytoskeletons, thereby coordinating cell motility and polarity (Veeman et al., 2003). Among 19 Wnt family members, Wnt5a is a representative ligand that activates this pathway. When Wnt5a binds to Fz and receptor tyrosine kinase-like orphan receptor 1/2 (Ror1/2), which function as co-receptors in the β-catenin-independent pathway, Dvl mediates Wnt5a signaling to regulate various cellular functions including adhesion, migration, invasion, and gene expression (Kikuchi et al., 2012). Extending our previous observation of the localization of Dvl2 to the midbody (Kikuchi et al., 2010), we found here that Fz2 is also localized to the midbody at telophase and that its localization is different from Dvl2 but similar to CHMP4B, an ESCRT-III subunit. Depletion of Wnt5a, Fz2, Ror2, or Dvl increased multinucleated cells and caused destabilization of midbody microtubules. These molecules were also required for the proper localization of CHMP4B at the midbody. In addition, Fz2 formed a complex with CHMP4B in a Wnt5a-dependent manner. These results suggest that Wnt5a-mediated β-catenin-independent pathway is involved in the positioning of ESCRT-III in the midbody properly, thereby leading to abscission.
Dvl2 is localized to the midbody at telophase
HA–Dvl2 and GFP–Dvl2 overexpressed in HeLaS3 and USOS cells, respectively, were concentrated to the center of the midbody during telophase (Kikuchi et al., 2010) (Fig. 1A; supplementary material Fig. S1A). Endogenous Dvl2 was also localized to the same region as ectopically expressed Dvl2 in HeLaS3 and a human breast carcinoma cell line, MDA-MB-435 cells (Fig. 1B; supplementary material Fig. S1A). The view of midbody morphology by electron microscopy reveals that anti-parallel microtubules are interdigitated and end at the center of the midbody, thereby creating the tight overlap region, where microtubule staining is blocked (Elad et al., 2011). This region is referred to as the stem body. Therefore, we used the term ‘stem body’ to indicate the central dark area of the midbody.
Molecules required for cytokinesis, such as motor proteins and microtubule associating proteins, are also accumulated at the midbody from telophase to abscission (Elad et al., 2011; Glotzer, 2009; Steigemann and Gerlich, 2009). Mklp1, a centralspindlin subunit, is present in the stem body, where it promotes midbody microtubule bundling (Sellitto and Kuriyama, 1988). In late telophase the localization of Mklp1 and GFP–Tsg101, an ESCRT-I subunit, became restricted to the stem body and they showed similar localization to GFP– or HA–Dvl2 (Fig. 1C). CHMP4B, an ESCRT-III subunit, accumulated on both sides of the stem body (Elia et al., 2011), and its localization was different from that of Dvl2 (Fig. 1C). GFP–Dvl2 at the intercellular bridge moved asymmetrically into one daughter cell after abscission (Fig. 1D) and was colocalized with mCherry-LC3 (supplementary material Fig. S1B). LC3 is known to be localized to the midbody during abscission and is required for lysosomal degradation of midbody ring after abscission (Pohl and Jentsch, 2009).
The accumulation of HA–Dvl2 to the stem body was reduced in Wnt5a-, Fz2-, or Ror2-depleted HeLaS3 cells but not in LRP6-depleted cells (Fig. 2A,C; supplementary material Fig. S2A), while the expression of HA–Dvl2 was still observed in the cytosol of dividing Wnt5a-depeleted cells as well as control cells (Fig. 2A). Endogenous Dvl2 localized to the midbody also decreased, whereas its total levels were not changed in Wnt5a-depleted cells as compared to control cells (Fig. 2B,D; supplementary material Fig. S2B). Therefore, it seems that a defect in the stem body localization of Dvl2 in Wnt5a- and its receptor-depleted cells does not reflect merely a failure of expression of Dvl2. The midbody localization of Dvl2 was restored by expression of mouse Wnt5a, which is resistant to siRNA for human Wnt5a, in Wnt5a-depleted cells (Fig. 2B,D; supplementary material Fig. S2C). These results suggest that Wnt5a signaling through Fz2 and Ror2 is required for the localization of Dvl2 to the stem body and is involved in cytokinesis.
Fz2 is localized to a different region of the midbody from Dvl2
Fz functions as a seven-transmembrane receptor for Wnt, and Fz2 has been shown to be involved in both the β-catenin-dependent and -independent pathways (Sato et al., 2010) (supplementary material Fig. S3A). While HA–Fz2 was localized to the cell surface membrane from prometaphase to anaphase, it accumulated at the midbody in early telophase (Fig. 3A; supplementary material Fig. S3B). At late telophase, FLAG– or HA–Fz2 was observed at both sides of the stem body clearly and showed the similar localization to CHMP4B (Fig. 3B; supplementary material Fig. S3B). However, FLAG–Fz2 showed localization different from Mklp1, GFP–Tsg101, and GFP–Dvl2 (Fig. 3B). FLAG–Fz2 was detected at the midbody when cells were stained with anti-FLAG-antibody before fixation and permeabilization (supplementary material Fig. S4A,B), indicating that the localization of Fz2 was not as a result of cytoplasmic aggregation by its overexpression but that Fz2 was present on the cell surface. Under the same staining conditions, FLAG–Dvl2 was not detected on the cell surface (supplementary material Fig. S4C). Receptor-tyrosine kinase-like orphan receptor 2 (Ror2), which is known as a receptor for Wnt5a (Green et al., 2008), was not clearly accumulated on the midbody but distributed throughout the cell cortex, including the intercellular bridge region (supplementary material Fig. S5).
Structured illumination microscopy (SIM) images into three dimensions (3D) revealed that HA–Dvl2 made web-like structures, extending into the interior of the stem body and that HA–Fz2 formed discontinuous ring structures along both sides of the stem body (Fig. 3C). Characterization of the organization of ESCRTs by SIM showed that Cep55, which associates with microtubules, is observed as a solid-disk like structure in the midbody and that CHMP4B forms distinctive ring-like structures which associate with the cell surface membrane (Elia et al., 2011). The similarity in localization and structure between Fz2 and CHMP4B suggests their functional relationship.
Failure of Wnt5a signaling increases multinucleated cells
There are three Dvl homologs in mammals, Dvl1, Dvl2, and Dvl3 (Wharton, 2003) (hereafter referred to collectively as Dvls). To clarify the roles of Dvls in cytokinesis, the number of multinucleated cells, which are a hallmark of cytokinesis defects, was counted in Dvls-depleted cells. Because Dvls were required for prometaphase and metaphase progression (Kikuchi et al., 2010), only the multinucleated cells that have the same size of nuclei were counted to exclude ones induced by lagging chromosomes. Depletion of either Dvl1, Dvl2, or Dvl3 alone by two kinds of siRNA for each Dvl in HeLaS3 cells increased multinucleated cells (Fig. 4A; supplementary material Fig. S6A), and expression of GFP–mouse Dvl2 rescued the phenotype of Dvl2-depleted cells (Fig. 4A; supplementary material Fig. S6B). The extent of the phenotype induced by depletion of all Dvl1, Dvl2, and Dvl3 was similar to or stronger than that by the depletion of each Dvl (Fig. 4A). Therefore, HeLaS3 cells in which all Dvls were depleted were used in the following studies.
It was examined whether Wnts and their receptors are involved in cytokinesis. Depletion of Fz2 or Ror2 increased multinucleated cells but that of LRP6 did not (Fig. 4B). Depletion of Wnt5a in HeLaS3 cells also increased multinucleated cells, and purified Wnt5a rescued the phenotype in a dose-dependent manner, but purified Wnt3a, which is a representative ligand for the β-catenin-dependent pathway, did not (Fig. 4C), suggesting that Wnt5a signaling is involved in cytokinesis specifically. Depletion of Wnt5a or Dvls in mouse embryonic fibroblasts (MEFs) also increased multinucleated cells (Fig. 4D; supplementary material Fig. S6C), indicating that Wnt5a signaling is required for cytokinesis in non-cancer cells.
Wnt5a and Dvl stabilize microtubules at the midbody
Because Wnt5a and Dvl are involved in the stability and polarized organization of microtubules (Ciani et al., 2004; Matsumoto et al., 2010; Schlessinger et al., 2007), it was examined whether they have a role in microtubule organization at the midbody. In control cells, thick stabilized microtubule bundles that are heavily detyrosinated, which was detected by anti-Glu-tubulin antibody, were observed in the intercellular bridge (Fig. 5A,B). On the other hand, thin microtubules were observed and the staining of Glu-tubulin was reduced in Wnt5a- or Dvls-depleted cells (Fig. 5A,B). EB1 tends to be concentrated to the microtubule plus-ends and its loss induces short microtubules in interphase (Morrison et al., 1998; Tirnauer and Bierer, 2000). In Wnt5a- or Dvls-depleted cells, EB1 was still observed in each extending microtubule at the interphase, but its staining at the cell periphery was reduced (Matsumoto et al., 2010) (supplementary material Fig. S7). During cytokinesis EB1 was concentrated at the midbody in control cells as described (Elad et al., 2011). In contrast, EB1 staining at the midbody was decreased in Wnt5a- or Dvls-depleted cells (Fig. 5A,C). These results suggest that the number of overlapping microtubules at the stem body is reduced in Wnt5a- or Dvls-depleted cells, resulting in decreased assembly of the midbody microtubules. There were more than 90% of cells with an intercellular bridge at shorter than 20 µm in control cells, while more than 50% of cells had the intercellular bridge at longer than 20 µm by depletion of Wnt5a or Dvls (Fig. 5D). Therefore, the length of the intercellular bridge was elongated in Wnt5a- or Dvls-depleted cells, suggesting that they failed to separate into two daughter cells because of the abscission defect.
Time-lapse imaging of GFP–α-tubulin-expressing HeLaS3 cells showed that the midbody appears to be formed normally both in control and Wnt5a-depleted cells. However, in Wnt5a-depleted cells midbody microtubule bundles became thinner in time, the intercellular bridge remained connected prior to completion of cytokinesis, and the length of microtubules was elongated consistently with the results from fixed cell assays (supplementary material Movies 1, 2). In addition, the duration time from telophase to abscission was delayed (Fig. 5E). Taken together, Wnt5a and Dvl could be required for the stabilization of midbody microtubules.
To examine whether microtubule stabilization is required for the localization of Dvl, EB1 was depleted to destabilize microtubules. The staining of microtubules and Glu-tubulin was indeed decreased in EB1-depleted cells (Fig. 5B). Consistent with the results, ∼36% of cells had the intercellular bridge at longer than 20 µm by depletion of EB1 (Fig. 5D). In addition, the localization of HA–Dvl2 to the stem body was reduced in EB1-depleted cells (Fig. 5F). Therefore, the localization of Dvl and the stabilization of microtubules at the midbody might be mutually dependent.
Wnt5a signaling regulates the proper localization of ESCRT-III at the midbody
Midbody microtubules are assembled at the stem body, and some centrosomal proteins are translocated to the midbody on microtubules by kinesin family proteins to control abscission (Sagona et al., 2010). Because the loss of Wnt5a signaling perturbs the stabilization of midbody microtubules, it was examined whether Wnt5a signaling is required for the localization of cytokinesis regulating machineries, such as centralspindlin and ESCRTs. Depletion of Wnt5a or Dvls in HeLaS3 cells did not affect the localization of Mklp1 and GFP–Tsg101 to the stem body (Fig. 6A,B). Although CHMP4B–GFP was still located to the midbody in Wnt5a- or Dvls-depleted cells, the correct localization of CHMP4B–GFP on both sides of the stem body was disrupted (Fig. 7A,B). The same results were observed in Fz2- or Ror2-depleted but not in LRP6-depleted cells (Fig. 7B). Consistently, endogenous CHMP4B was localized to both sides of the stem body in control cells as well and its correct localization was impaired in Wnt5a- or Dvls-depleted cells (Fig. 7C,D).
SIM images showed that endogenous CHMP4B is observed as double ring structures at the stem body in control cells (Fig. 7E). In contrast, CHMP4B was mislocalized and observed as dot-like structures in the stem body in Wnt5a- or Dvls-depleted cells (Fig. 7E).
As shown in Fig. 5A, Wnt5a and Dvl were required for the stabilization of midbody microtubules. To examine whether the stabilization of microtubules is necessary for the correct localization of CHMP4B, cells were treated with siRNA for EB1. CHMP4B was mislocalized in EB1-depleted cells (Fig. 7D), suggesting that the correct localization of CHMP4B depends on the stabilization of midbody microtubules. In addition, a biochemical study showed that HA–Fz2 forms a complex with CHMP4B–GFP and that Wnt5a enhances complex formation (Fig. 7F). Therefore, Fz2 may regulate the correct localization of ESCRT-III to both sides of the stem body via Wnt5a signaling.
In order to define the mechanism by which Wnt5a signaling regulates the proper localization of ESCRT-III at the midbody through microtubule stabilization, the trafficking of Fz2 to the midbody was examined as Fz2 interacted with CHMP4B. The localization of FLAG–Fz2 to both sides of the stem body was not observed in Wnt5a- or Dvl-depleted cells, while depletion of CHMP4B did not affect the localization of FLAG–Fz2 (Fig. 8A,B; supplementary material Fig. S8A). Thick microtubule bundles stained with Glu-tubulin were observed in CHMP4B-depleted cells as well as control cells (supplementary material Fig. S8B), suggesting that the stabilization of microtubules does not require CHMP4B. It is known that vesicular internalization is re-activated in late mitosis for membrane proteins to be trafficked to the midbody (Schweitzer et al., 2005). The Rab11-mediated recycling pathway, which depends on microtubules, is shown to be involved in the midbody accumulation of endosomes (Wilson et al., 2005). The localization of FLAG–Fz2 to the midbody was indeed decreased in Rab11- but not Rab5-depleted cells (Fig. 8A,B; supplementary material Fig. S8A). The absence of FLAG–Fz2 signal did not simply reflect lack of its expression in Wnt5a-, Dvls-, or Rab11-depleted cells, because FLAG–Fz2 was observed in the cytoplasm of dividing cells (see Fig. 8A). Therefore, Fz2 might be trafficked to the midbody in a Rab11-dependent manner presumably along stabilized microtubules. All of these results suggest that Wnt5a signaling determines the localization of Fz2 to both sides of the stem body and controls the proper localization of ESCRT-III.
Emerging evidence suggests interesting roles for Wnt signaling in mitotic progression. Cycling-Y binds to and mediates the phosphorylation of LRP6, resulting in activation of the β-catenin-dependent pathway that is required for G2/M transition (Davidson et al., 2009). We also reported that Fz2, LRP6, and Dvl2 are required for mitotic spindle orientation, which is an event in early mitotic phase (Kikuchi et al., 2010). The present study showed a novel function of Wnt5a signaling, which activates the β-catenin-independent pathway, in cytokinetic abscission by positioning ESCRT-III at the midbody through the stabilization of midbody microtubules. Our findings provide the first evidence that the Wnt5a-mediated β-catenin-independent pathway functions in later mitotic phases.
Microtubule-dependent localization of Fz2 and Dvl2 to the midbody is regulated by Wnt5a
Overexpressed and endogenous Dvl2 were localized to the stem body at telophase. In addition, Fz2 accumulated on the stem body at early telophase and relocated to both sides of the stem body at late telophase. We further observed the localization of Dvl and Fz2 by using SIM microscopy and demonstrated that they show unique structures in the midbody. Dvl was not merely accumulated to the interior of the midbody but made web-like structures. This structure might contribute to make open interphase to bind to other proteins that are involved in abscission. As previously reported, Dvl oligomerizes via its DIX domain (Kishida et al., 1999; Schwarz-Romond et al., 2007), and therefore it is intriguing to speculate that oligomerization might be required for the structure formation and efficient stabilization of midbody microtubules. SIM image also revealed that Fz2 forms discontinuous ring structures at late telophase. This structure strikingly resembles with CHMP4B at the midbody (Elia et al., 2011), and we demonstrated that the similarity in the localization of Fz2 and CHMP4B is related with their roles in cytokinesis (see below).
The localization of Dvl to the stem body was dependent on microtubules, because depletion of EB1 interfered with it. The localization of Fz2 to the stem body might be also dependent on microtubules, because it required Rab11, which is involved in microtubule-mediated trafficking of endosomal proteins to the midbody (see below). It has been demonstrated that Dvl is recruited to Fz2 on cell surface membranes in response to Wnt3a to activate the β-catenin-dependent pathway in interphase cells (Bilic et al., 2007). Because disruption of microtubules did not affect Wnt3a signaling (Sakane et al., 2010), it is unlikely that the binding of Dvl and Fz2 depends on microtubules in the activation of the β-catenin-dependent pathway. On the other hand, our data showed that Wnt5a, Fz2, and Ror2, but not LRP6, are required for the accumulation of Dvl2 to the stem body, suggesting that Wnt5a signaling positions Dvl to the midbody. Therefore, the trafficking of Dvl and Fz2, which depends on microtubules, might be specific in the Wnt5a-mediated β-catenin-independent pathway during cytokinesis.
Evidence has accumulated that Dvl stabilizes microtubules. Neurons from Dvl1 knockout mice developed a dendritic arbor of lesser complexity than neurons from wild-type mice (Rosso et al., 2005) and did not undergo growth cone remodeling to form synapses in response to Wnt probably through a failure of microtubule stabilization (Purro et al., 2008). Moreover, it was reported that Dvl inhibits GSK3β locally, resulting in changes in the phosphorylation levels of GSK3β targets, such as the microtubule-associated protein 1B, thereby regulating the stabilization of microtubules (Ciani et al., 2004) and that the binding of Dvl to adenomatous polyposis coli gene product (APC) increases the number of microtubules at the cell cortex (Matsumoto et al., 2010). Consistent with these results, our results suggest that Wnt5a signaling through Fz2, Ror2, and Dvl may be involved in the stabilization and bundling of midbody microtubules, resulting in the positioning of Fz2 and Dvl to the midbody. Alternatively, Wnt5a signaling may regulate the localization of some microtubule-interacting proteins, including Cep55, protein regulator of cytokinesis 1 (Prc1), kinesin family member 4 (Kif4), and Mklp2, rather than the stabilization of midbody microtubules itself, because these proteins contribute to the stabilization of midbody microtubules in the process of the midbody assembly by accumulating at the midbody and relocating into different parts of the midbody. Taken together, Wnt5a signaling to position Fz2 and Dvl at the midbody and the stabilization of midbody microtubules might couple functionally to complete membrane scission.
Similar spatial localization of Fz2 and CHMP4B at the midbody
The ESCRT-III protein CHMP4B was shown to move dynamically during cytokinesis (Elia et al., 2011). At the initial stage of cytokinesis Cep55, Tsg101, and CHMP4B are assembled at the midbody. Then, CHMP4B relocates outward to the constriction zone by polymerizing into a spiral, leading to the separation of two cells. Our results showed that centralspindlin and ESCRTs were observed in the midbody in Wnt5a- or Dvl-depleted cells where the stabilization of microtubules is decreased. One possible scenario is that their localization is not solely dependent on microtubules or that thin midbody microtubules might be sufficient to assemble them at the stem body, because midbody microtubules still formed intercellular bridges in Wnt5a- or Dvl-depleted cells. However, depletion of Wnt5a or Dvls and destabilization of microtubules impaired the proper localization of CHMP4B to both sides of the stem body at late telophase, thereby causing the failure of cytokinetic abscission. Fz2 showed the similar localization to CHMP4B and both proteins form a complex in a Wnt5a-dependent manner, suggesting that Fz2 directly contributes to the proper localization of CHMP4B. Consistently, depletion of Fz2 impaired the proper localization of CHMP4B, but CHMP4B was not required for the localization of Fz2.
Fz2 was located to the cell surface membranes from prometaphase to anaphase, but it disappeared from there and accumulated on the midbody at early telophase. In late telophase, Fz2 as well as CHMP4B relocated to both sides of the stem body as well as CHMP4B. Vesicular internalization has been shown to be re-activated from anaphase for membrane proteins to be trafficked to the midbody (Schweitzer et al., 2005). Rab11-containing endosomes have been shown to accumulate in the midbody and are required for cytokinesis (Wilson et al., 2005). The membrane trafficking by Rab11 in cytokinesis is dependent on microtubules through kinesin (Lin et al., 2002; Montagnac et al., 2008). Our results showed that depletion of Rab11 interferes with the midbody localization of Fz2. Therefore, it is intriguing to speculate that Fz2 is internalized in early telophase and accumulates on the midbody in a microtubule-dependent manner probably through Rab11.
It is important for abscission that CHMP4B concentrates on both sides of the stem body with appropriate timing. Wnt5a signaling through Fz2, Ror2, and Dvl might be required for the functional localization of ESCRT-III to both sides of the stem body, and especially Fz2 might have an important role in this event. As a novel function of Wnt5a-mediated β-catenin-independent signaling, it stabilizes midbody microtubules, thereby controlling cytokinesis.
Possible mechanism by which Wnt5a signaling regulates cytokinesis
Taken together, one possible model of Wnt5a-regulated cytokinesis is as follows (supplementary material Fig. S9). At early telophase Fz2 is accumulated to the stem body in a Rab11-dependent manner and Wnt5a signaling controls the accumulation of Dvl to the stem body. Accumulated Fz2 and Dvl2 at the midbody enhance the stabilization of midbody microtubules, which contributes to their efficient localization to the midbody. At late telophase Fz2 relocates to both sides of the stem body and Wnt5a signal strengthens the binding of Fz2 and CHMP4B, thereby positioning CHMP4B to both sides of the stem body to lead to complete abscission.
A recent study showed that a secreted extracellular matrix protein, hemicentin, regulates cytokinesis (Xu and Vogel, 2011), giving an example of the notion that a secreted ligand and its receptor-mediated signaling is involved in cytokinesis. Therefore, the present data may give another example of the involvement of secreted ligand-activated signaling in cytokinesis.
Materials and Methods
Materials and chemicals
pCS2/FLAG-Dvl2 and mCherry-LC3 were kindly provided by R. Habas (UMDNJ Robert Wood Johnson Medical School, Piscataway, NJ, USA) and T. Yoshimori (Osaka University, Osaka, Japan), respectively. pEGFP-N1/CHMP4B and pBjMyc/Tsg101 were from M. Maki (Nagoya University, Nagoya, Japan). pEGFP-C1/α–tubulin was from K. Kaibuchi (Nagoya University, Nagoya, Japan). pCS2/FLAG-Fz2 and pEGFPC1-Dvl2, pCGN/Dvl2 and pPGK-neo-Wnt5a were constructed as described (Kishida et al., 2007; Sato et al., 2010). Standard recombinant DNA techniques were used to construct pEGFP-C1/Tsg101 and pCGN/Fz2. GFP-Dvl2 and FLAG-Fz2 cDNAs were cloned into pLvSIN to construct lentiviral vectors (Takara Bio Inc., Shiga, Japan). The lentiviruses were produced in X293T cells by using Lenti-X™ Lentiviral Expression Systems (Takara Bio Inc., Shiga, Japan) according to manufacturer's instructions.
Both Wnt3a and Wnt5a were purified to homogeneity as described previously (Kishida et al., 2004; Komekado et al., 2007; Kurayoshi et al., 2007). All of the primary antibodies used in this study are listed in supplementary material Table S1. Secondary antibodies coupled to horseradish peroxidase (HRP) were purchased from Jackson ImmunoResearch Laboratories, Bar Harbor, ME, USA and secondary antibodies used for immunofluorescence were from Invitrogen, Grand Island, NY, USA.
Cell culture and transfection
HeLaS3, U2OS, X293T, MEF, and MDA-MB-435 were maintained in DMEM supplemented with 10% FBS and penicillin-streptomycin. HEK293T cells were maintained in DMEM/Ham’s F12 supplemented with 10% FBS and penicillin-streptomycin. U2OS stably expressing GFP–Dvl2 or HeLaS3 stably expressing mouse Wnt5a were generated by transfecting pEGFPC1/Dvl2 or pPGK-neo-Wnt5a, respectively, and the cells were selected and maintained in the same medium with parental cells containing 400 µg/ml G418. To generate HeLaS3 cells stably expressing GFP–Dvl2 or FLAG–Fz2, 50,000 parental cells/well in 12-well plates were treated with lentiviruses and 10 µg/ml polybrene, centrifuged 1000 rpm for 1 h and incubated for a further 24 h. Then, the cells were selected and maintained in the same medium with parental cells containing 400 µg/ml G418. MEFs were prepared from E13.5 embryos using a 3T3 protocol (Todaro and Green, 1963) and were maintained in DMEM supplemented with 10% FBS. To express proteins transiently, plasmids were transfected in cells using Lipofectamine LTX (Invitrogen, Grand Island, NY, USA) for HeLaS3 cells and Lipofectamine 2000 for HEK293T cells.
Treatment of cells with siRNA
In analyses with siRNAs for randomized control, Wnt5a, Fz2, LRP6, Ror2, Dvl1, Dvl2, Dvl3, EB1, Rab5, Rab11a, and Rab11b in HeLaS3 cells, the following target sequences were used. Randomized control, 5′-CAGTCGCGTTTGCGACTGG-3′; Wnt5a-1, 5′-GTTCAGATGTCAGAAGTAT-3′; Wnt5a-2, 5′-CTGTGGATAACACCTCTGT-3′; Fz2, 5′-CGGTCTACATGATCAAATA-3′; LRP6, 5′-ATTGCCCATCCTGATGGTA-3′ and 5′-CCAAAGTCCAAGCTCGAAT-3′; Ror2, 5′-GCAACCTTTCCAACTACAA-3′; Dvl1-1, 5′-GGAGGAGATCTTTGATGAC-3′; Dvl2-1, 5′-GGAAGAAATTTCAGATGAC-3′; Dvl3-1, 5′-GGAGGAGATCTCGGATGAC-3′; Dvl1-2, 5′-CCAAGATTATCTACCACAT-3′; Dvl2-2, 5′-GCTGGTGTCCTCAGATAAT-3′; Dvl3-2, 5′-CCAGCTATAAGTTCTTCTT-3′; EB1, 5′-GTGAAATTCCAAGCTAAGC-3′; CHMP4B, 5′-GGACATCGATAAAGTTGAT-3′; Rab5, 5′-AAGGCCGACCTAGCAAATA-3′; Rab11a, 5′-GCGATATCGAGCTATAACA-3′; Rab11b, 5′-GCACCTGACCTATGAGAAC-3′. The target sequences of siRNA for Dvl1, Dvl2, and Dvl3 in MEF cells were 5′-GGAGGAGATCTTCGATGAC-3′, 5′-GGAAGAGATCTCCGATGAC-3′, 5′-GGAAGAGATCTCGGACGAC-3′, respectively.
HeLaS3 and MEF cells were transfected with a mixture of siRNAs against genes of interest at 20 nM each using RNAiMAX (Invitrogen, Grand Island, NY, USA) and the cells were used for experiments at 72 h post-transfection. When necessary, after 48 h siRNA transfection, cells were transfected with plasmid cDNA and then the cells were analyzed a further 24 h after transfection.
Cell staining and image analysis
For cell staining and image analysis, cells were seeded on coverslips and fixed with 100% methanol for 20 min at −20°C. The cells were then washed three times with PBS, blocked for 20 min with 1% BSA in phosphate-buffered saline including 0.05% Tween-20 (PBST), and further incubated with primary antibodies overnight at 4°C. After the cells were washed three times with PBS, they were incubated with fluorescent secondary antibodies for 1 h at RT. Coverslips were washed extensively with PBS and then mounted in 50% glycerol in PBS. Processing and measurements were carried out using an LSM510 system (Carl Zeiss Microscopy Co., Ltd, Jena, Germany) and ImageJ software (NIH, Bethesda, MD) except for time-lapse images. The same magnification of scale bars was used in a series of panels such as Fig. 1A.
For time-lapse imaging, an inverted microscope (IX81; Olympus, Tokyo, Japan) was used. In Fig. 1D, supplementary material Fig. S1B, and supplementary material Movies 1, 2, the images were captured every 3 or 5 min for 12 h and treated with MetaMorph software (MDS Analytic Technologies, Tokyo, Japan). In Fig. 3C, thin (0.1 µm) Z-stacks of high-resolution images were collected in three rotations for each midbody using an ELYRA S1 microscope, and then images were reconstructed using ZEN software (Carl Zeiss Microscopy Co., Ltd, Jena, Germany). All measurements were performed on reconstructed images of single z-sections in ZEN.
In Fig. 5B,C, integrated fluorescence intensity of Glu-tubulin (detyrosinated α-tubulin) or EB1 of intercellular bridge area and that from perinuclear cytoplasmic region (1×1 µm square) adjacent to intercellular bridge in Wnt5a- Dvls- or EB1-depleted HeLaS3 cells were measured by ImageJ software and the ratio of the former to the latter was calculated and shown as arbitrary unit.
Counting of multinucleated cells
Cells were washed with PBS, and then fixed for 10 min with cold methanol at −20°C before staining with an anti-β-tubulin antibody and PI. Individual cells were carefully determined by taking Z-stack images from 0.5 µm-thick sections of area. All processing and measurements were carried out using an LSM510 system with Axiovision. In each experiment multinucleated cells among more than 150 cells, where microtubules and nuclei were visible clearly, were counted.
Immunoblotting and immunoprecipitation
Methods for immunoblotting and immunoprecipitation were described previously (Hino et al., 2005). Briefly, to immunoprecipitate proteins, cells were washed once with PBS and lyzed in lysis buffer (20 mM Tris-HCl at pH 7.5, 150 mM NaCl, 2 mM EGTA, 0.5% NP40 with protease inhibitors (1 mM phenylmethylsulfonylfluoride, 1 µg/ml leupeptin, and 1 µg/ml aprotinin) for 10 min on ice. After centrifugation, the supernatant was collected, and incubated with appropriate antibodies and 30 µl of protein-G–Sepharose (50% slurry). The mixtures were placed on a rotary mixer for 1.5 h at 4°C. The beads were then washed four times with lysis buffer and finally suspended in Laemmli’s sample buffer.
Quantitative real-time PCR
Total RNA was isolated from HeLaS3 cells treated with siRNA for Fz2 for 72 h. RNA sample (2 µg) was subjected to reverse transcription using murine leukemia virus reverse transcriptase (PE Applied Biosystems) in a total volume of 20 µl. Quantitative reverse transcription-PCR (RT-PCR) was performed using a LightCycler (Roche Molecular Biochemicals). Aliquots (2.5 µl) of the reverse transcription products were amplified in a reaction mixture (10 µl) containing LightCycler FastStart DNA Master SYBR Green I, 1 µM primer, and 2 mM MgCl2. Forward and reverse primers were as follows: human GAPDH, 5′-CCTGTTCGACAGTCAGCCG-3′ and 5′-CGACCAAATCCGTTGACTCC-3′; human Fz2, 5′-GAGCGTGATTGTGCTG-3′ and 5′-GCTCTGGGTAGCGGAA-3′.
Experiments were performed at least three times and the results were expressed as means ± s.e.m. Statistical analysis was performed using StatView-J 5.0 software (SAS Institute Inc.). Differences between the data were tested for statistical significance using t-tests. P-values less than 0.05 were considered statistically significant.
We are grateful to Drs R. Habas, M. Maki, and T. Yoshimori for donating cDNAs. We also thank Dr A. Sato for lentiviral production and the Center for Medical Research and Education in Osaka University for imaging using an ELYRA S1 microscope.
This work was supported by Grants-in-Aid for Scientific Research (A) (2009, 2010 and 2011) [grant number 212490170 to A.K.], for Scientific Research on Priority Areas (2011) [grant number 23112004 to A.K.], and for Young Scientists (B) (2010, 2011) [grant number 22700881 to K.K.] from the Ministry of Education, Science, and Culture of Japan, and by The Nagase Foundation (2011) [to A.K.].