Breast cancer gene 1 (BRCA1) contributes to the regulation of centrosome number. We previously identified receptor for activated C kinase 1 (RACK1) as a BRCA1-interacting partner. RACK1, a scaffold protein that interacts with multiple proteins through its seven WD40 domains, directly binds to BRCA1 and localizes to centrosomes. RACK1 knockdown suppresses centriole duplication, whereas RACK1 overexpression causes centriole overduplication in a subset of mammary gland-derived cells. In this study, we showed that RACK1 binds directly to polo-like kinase 1 (PLK1) and Aurora A, and promotes the Aurora A–PLK1 interaction. RACK1 knockdown decreased phosphorylated PLK1 (p-PLK1) levels and the centrosomal localization of Aurora A and p-PLK1 in S phase, whereas RACK1 overexpression increased p-PLK1 level and the centrosomal localization of Aurora A and p-PLK1 in interphase, resulting in an increase of cells with abnormal centriole disengagement. Overexpression of cancer-derived RACK1 variants failed to enhance the Aurora A–PLK1 interaction, PLK1 phosphorylation and the centrosomal localization of p-PLK1. These results suggest that RACK1 functions as a scaffold protein that promotes the activation of PLK1 by Aurora A in order to promote centriole duplication.
The centrosome is a major microtubule organizing center and facilitates proper bipolar spindle formation (Conduit et al., 2015). An abnormal number of centrosomes causes aberrant cell division (e.g. monopolar and multipolar division) and chromosome segregation errors, leading to carcinogenesis (Milunović-Jevtić et al., 2016; Cosenza et al., 2017; Levine et al., 2017). Therefore, the regulation of centrosome number is important for the maintenance of chromosomal stability and tumor suppression.
Centrosomes consist of a pair of centrioles, termed the mother centriole and the daughter centriole, and proteinaceous pericentriolar material surrounding the mother centriole (Conduit et al., 2015). The proper number of centrosomes is stably maintained by a regulatory network that ensures duplication only once per cell cycle (Conduit et al., 2015; Fujita et al., 2016; Nigg and Holland, 2018). At entry into mitosis, a cell has two centrosomes, each containing a pair of tightly engaged centrioles. During mitosis, centrioles lose their connection in a process termed disengagement, and the two centrioles are joined by a flexible linker. In late G1 phase of the following cell cycle, a new daughter centriole is generated on the side wall of each mother centriole. The daughter centrioles grow during S phase, resulting in the formation of two pairs of centrioles in two centrosomes in G2 phase. In late G2 phase, the linker is degraded and the two centrosomes separate to form bipolar spindles.
During centriole duplication, the newly formed daughter centrioles suppress the formation of additional daughter centrioles to prevent multiple rounds of duplication (Tsou and Stearns, 2006a; Nigg, 2007). Although the mechanism underlying this suppression is poorly understood, disengagement is thought to release the suppression (Tsou and Stearns, 2006a; Nigg, 2007; Shukla et al., 2015). Therefore, the timing of disengagement is important for the regulation of centriole duplication to maintain the proper number of centrosomes.
Centrosome amplification, a common condition in many cancers, including breast cancer (Chan, 2011), is associated with chromosomal instability and aggressive phenotypes (Denu et al., 2016; Godinho et al., 2014; Schneeweiss et al., 2003). Breast cancer gene 1 (BRCA1) mutations are responsible for hereditary breast and ovarian cancer syndrome, and loss or mutation of BRCA1 causes centrosome amplification (Starita et al., 2004; Ko et al., 2006). Certain BRCA1 variants found in familial breast cancers are associated with deficiencies in the regulation of centrosome number (Kais et al., 2012). BRCA1 germline mutation and negative expression of BRCA1 are significantly associated with centrosome amplification in breast cancer tissues (Shimomura et al., 2009; Watanabe et al., 2018).
BRCA1 forms multi-protein complexes with BRCA1-associated RING domain protein 1 (BARD1) and other proteins that function in a number of cellular processes, including DNA repair and centrosome regulation (Takaoka and Miki, 2018). In previous studies, we identified Obg-like ATPase 1 (OLA1) (Matsuzawa et al., 2014; Yoshino et al., 2018) and receptor for activated C kinase 1 (RACK1) (Yoshino et al., 2019) as components of the BRCA1–BARD1 complex. OLA1 localizes to centrosomes and directly binds to BRCA1, BARD1 and γ-tubulin. Knockdown or overexpression of OLA1 increases the fraction of cells with centrosome amplification caused by centriole overduplication in mammary tissue-derived cells (Matsuzawa et al., 2014; Yoshino et al., 2018). RACK1 also localizes to centrosomes and directly binds to BRCA1, OLA1 and γ-tubulin, and interacts with BARD1. Knockdown of RACK1 suppresses centriole duplication in some breast cancer cells. Conversely, overexpression of RACK1 increases the fraction of cells with centrosome amplification caused by centriole overduplication in mammary tissue-derived cells. Adequate RACK1 expression levels contribute to the proper localization of BRCA1 to centrosomes. Alterations in the expression levels or missense mutations of OLA1 and RACK1 are associated with various cancers, as reported in the COSMIC database (https://cancer.sanger.ac.uk/cosmic). These data suggest that the centrosomal BRCA1–BARD1 complex containing OLA1 and RACK1 is important for the regulation of centrosome number and tumor suppression. However, the underlying molecular mechanism remains unknown.
Centriole duplication is regulated by several mitotic kinases (Wang et al., 2014; Nigg and Holland, 2018). Polo-like kinase 1 (PLK1) is a critical regulator of centriole disengagement during mitosis (Tsou et al., 2009; Shukla et al., 2015), and its activity is tightly regulated spatially and temporally throughout the cell cycle. Aurora A kinase phosphorylates PLK1 at Thr210 under the control of CDK1 and Bora to activate PLK1 in late G2 phase and initiate mitotic entry (Macůrek et al., 2008; Schmucker and Sumara, 2014; Seki et al., 2008; Bruinsma et al., 2014). In centrosomes, PLK1 activation by Aurora A and the scaffold protein CEP192 is essential for centrosome maturation and bipolar spindle formation during G2 and mitotic phases (Joukov et al., 2014; Meng et al., 2015). In addition, PLK1 contributes to centrosome amplification induced by S phase or G2 arrest or DNA damage (Lončarek et al., 2010; Zou et al., 2014).
PLK4, another member of the PLK family, is essential for centriole duplication (Zitouni et al., 2014). PLK4 activation in late G1 phase triggers cartwheel formation, which is the initial step of centriole duplication (Kleylein-Sohn et al., 2007; Kim et al., 2013; Bettencourt-Dias et al., 2005). Although the function of these mitotic kinases in centriole duplication has been studied extensively, the mechanism by which the regulation of centriole duplication by these kinases is perturbed in cancer is not well understood.
In the present study, we investigated the effect of RACK1 overexpression on centriole overduplication in breast cancer cells. The results showed that RACK1 overexpression-induced centrosome amplification involved the activities of Aurora A and PLK1. RACK1 directly bound to Aurora A and PLK1, promoted their interaction, and was involved in the centrosomal localization of Aurora A and the activation of PLK1 at the centrosome. These results suggest that RACK1 functions as a scaffold involved in the regulation of the Aurora A/PLK1 signaling axis and centriole duplication.
RACK1 overexpression increases the fraction of cells with centrosome amplification during S and G2 phases
We previously reported that RACK1 overexpression increases the proportion of cells with centrosome amplification by inducing centriole overduplication (Yoshino et al., 2019). Cell cycle arrest at S or G2 phase causes centriole overduplication (Balczon et al., 1995; Lončarek et al., 2010). To examine the effect of RACK1 overexpression on cell cycle progression, MCF7 cells were transfected with empty vector (control) or HA-tagged RACK1 (HA–RACK1) expression vector, and analyzed by flow cytometry. No significant differences in cell cycle distribution were observed between HA–RACK1-overexpressing MCF7 cells (RACK1-OE cells) and control cells (Fig. 1A), as reported previously in RACK1-knockdown cells (Yoshino et al., 2019).
The effect of RACK1 overexpression on inducing centrosome amplification was investigated by measuring centrosome number in each cell cycle phase. MCF7 cells were transfected with control or HA–RACK1 vector and synchronized to the G1/S transition by a double thymidine block (Fig. 1B, Protocol 1). Then, cells were released into fresh medium and fixed for immunocytofluorescence with anti-HA and γ-tubulin antibodies after 0, 4, 8, and 12 h of incubation, which represented G1/S transition (G1/S), S, G2, and G1 phases, respectively (Fig. 1C). RACK1-OE cells were identified as HA-positive cells and subjected to analysis of centrosome number. Cells with amplified (more than two) centrosomes were counted, and their percentage was calculated. RACK1 overexpression significantly increased the proportion of cells with centrosome amplification in S and G2 cells, but not in G1 and G1/S cells (Fig. 1D,E). These data indicate that the centrosome amplification observed in RACK1-OE cells occurred during S and G2 phases.
To confirm that RACK1 overexpression induces centriole overduplication in MCF7 cells, as previously observed in Hs578T cells (Yoshino et al., 2019), we stained the centriole with anti-CP110 antibody, the mother centriole with anti-CEP152 antibody, and the daughter centriole with anti-SAS6 antibody in RACK1-OE MCF7 cells (Fig. S1A). RACK1 overexpression increased the fraction of cells with a CP110 foci/CEP152 foci ratio of >2 (Fig. S1B,D) and that of cells with extra SAS6 foci that did not pair with CEP152 foci (Fig. S1C,D). These results suggest that centriole overduplication is also induced in some RACK1-OE MCF7 cells.
PLK1, Aurora A and PLK4 activities contribute to centrosome amplification induced by RACK1 overexpression
To identify the signaling pathway involved in centrosome amplification upon RACK1 overexpression, RACK1-OE cells were treated with inhibitors of kinases important for centriole duplication, namely, MLN8054, BI6727 and centrinone B, which are inhibitors of Aurora A, PLK1, and PLK4, respectively (Wong et al., 2015; Manfredi et al., 2007; Rudolph et al., 2009). These inhibitors block cell cycle progression; treatment with MLN8054 or BI6727 causes G2/M arrest, and treatment with centrinone B causes G1 arrest. To minimize the effect on the cell cycle, cells were synchronized at G1/S by a double thymidine block and released into fresh medium containing inhibitors for 4 h (Fig. 2A, Protocol 2). As shown in Fig. 2B,C, treatment with BI6727 and centrinone B, but not MLN8054, decreased the proportion of cells with centrosome amplification in RACK1-OE cells. Because Aurora A is the upstream activator of PLK1, we speculated that Aurora A might contribute to centrosome amplification at an earlier cell cycle phase than that affected by PLK1. To verify this, MLN8054 treatment was started at 12 h before the release from a double thymidine block, and centrosome number was analyzed at 4 h after the release (Fig. 2A, Protocol 3). Under these conditions, MLN8054 treatment significantly decreased the fraction of cells with centrosome amplification in RACK1-OE cells (Fig. 2D,E). These results suggest that activation of Aurora A, followed by PLK1 and PLK4 activation contributes to centrosome amplification induced by RACK1 overexpression.
RACK1 binds to PLK1 and Aurora A and promotes their interaction
Overexpression of RACK1 increases the fraction of cells with centriole overduplication (Yoshino et al., 2019; Fig. S1). Activation of PLK4 initiates centriole duplication at centrioles licensed via disengagement (Kleylein-Sohn et al., 2007; Kim et al., 2013; Bettencourt-Dias et al., 2005). Disengagement requires activation of the PLK1 signaling pathway (Shukla et al., 2015; Tsou et al., 2009; Tsou and Stearns, 2006b). Therefore, we presumed that RACK1 overexpression activates PLK1 and Aurora A to induce centriole overduplication upstream of PLK4 activation. To explore how PLK1 and Aurora A contribute to centrosome amplification induced by RACK1 overexpression, the interaction of RACK1 with PLK1 and Aurora A was analyzed in immunoprecipitation experiments. Endogenous RACK1 associated with endogenous PLK1 and Aurora A (Fig. 3A,B). These associations were confirmed using exogenously expressed tagged proteins (Fig. S2A,B). To determine which domain of PLK1 interacts with RACK1, we performed immunoprecipitation between the kinase domain or polo-box domains of PLK1 and RACK1, which revealed that both the kinase domain and polo-box domains interacted with RACK1 (Fig. S2C,D).
RACK1 is composed of seven WD domains and functions as a scaffold protein that binds to numerous proteins, resulting in their activation (Ron et al., 2013). Because RACK1 interacted with both PLK1 and Aurora A, we determined whether RACK1 co-expression affects the interaction between PLK1 and Aurora A. The interaction between PLK1 and Aurora A was markedly strengthened by RACK1 co-overexpression in a dose-dependent manner (Fig. 3C; Fig. S2E). Consistent with this, knockdown of RACK1 decreased the interaction between PLK1 and Aurora A (Fig. S2F). Co-overexpression of RACK1 increased PLK1 phosphorylation at Thr210 (Fig. 3C). These results indicate that RACK1 contributes to the formation of the Aurora A–PLK1 complex and may activate PLK1.
To confirm the effect of RACK1 on promoting the Aurora A–PLK1 interaction, recombinant Aurora A, PLK1 and RACK1 proteins expressed in E. coli were purified for pull-down assays. RACK1 directly bound to Aurora A and both the kinase domain and polo-box domains of PLK1 (Fig. 3D,E; Fig. S2G). As shown in Fig. 3E, PLK1 weakly interacted with Aurora A in the absence of RACK1, whereas pre-binding of PLK1 to RACK1 markedly increased the Aurora A–PLK1 interaction. Reversing the reaction sequence (i.e. first Aurora A and then RACK1) did not increase the interaction of PLK1 with Aurora A. These data suggest that RACK1 functions as a scaffold protein to enhance the Aurora A–PLK1 interaction.
RACK1 binding to Aurora A and PLK1 was further investigated by identifying the domains of RACK1 responsible for binding to Aurora A and to the kinase and polo-box domains of PLK1 using recombinant maltose binding protein (MBP)-tagged RACK1 fragments. Each fragment contained the two indicated WD domains (Fig. S3A). Aurora A bound strongly to the WD23 and WD34 fragments of RACK1 and weakly to the WD67 fragment (Fig. S3B). The kinase domain of PLK1 bound to WD23 and weakly to the WD34 and WD67 fragments of RACK1 (Fig. S3C). The polo-box domains of PLK1 bound strongly to WD23 and weakly to WD12, WD34, and WD56 (Fig. S3D).
RACK1 is involved in the localization of Aurora A to the centrosome and PLK1 phosphorylation to promote centriole duplication in S phase
To examine the physiological role of RACK1 in PLK1 and Aurora A activities at the centrosome, we analyzed the effect of RACK1 knockdown on the localization of phosphorylated and total PLK1 (denoted p-PLK1 and t-PLK1, respectively) and Aurora A to centrosomes. Western blot analysis confirmed that RACK1 siRNA downregulated RACK1 without affecting Aurora A and t-PLK1 expression (Fig. S4A). RACK1 knockdown decreased the fraction of cells containing centrosomal p-PLK1 in S phase but not that in G1 or G1/S phase (Fig. 4A,B). RACK1 knockdown also decreased the fraction of cells with centrosomal p-PLK1 signals in G2 phase, although to a smaller extent than in S phase (Fig. S4B,C). The fluorescence intensity of p-PLK1 at centrosomes in S phase cells was also decreased significantly (Fig. S4D). The centrosomal localization of t-PLK1 was not affected by RACK1 knockdown, which was confirmed by examining the fluorescence intensity of t-PLK1 at centrosomes in S phase (Fig. 4C,D and Fig. S4E). Consistent with these results, immunoprecipitation followed by western blotting revealed a markedly decreased p-PLK1 level in RACK1-knockdown cells under asynchronized and S-phase-synchronized conditions (Fig. 4E). These results suggest that RACK1 is involved in the phosphorylation of PLK1 at the centrosome in S phase.
RACK1 knockdown also decreased the fraction of cells with centrosomal Aurora A signals in S phase but not in G1 or G1/S phase (Fig. 5A,B). We also observed that the fraction of S phase cells with high fluorescence intensity of centrosomal Aurora A was decreased by RACK1 knockdown, although the difference between RACK1-knockdown cells and the control cells was not statistically significant (Fig. S5A). To exclude off-target effects of the siRNA, the experiments in S phase were repeated using a different siRNA against RACK1, which yielded similar results (Fig. S5B–D). These results suggest that RACK1 is involved in the activation of PLK1 at the centrosome by promoting the localization of Aurora A to centrosome in S phase.
We previously reported that knockdown of RACK1 decreases the number of centrioles per cell in S and G2 phases (Yoshino et al., 2019). We hypothesized that the RACK1 knockdown-induced decrease in PLK1 activation in S phase was responsible for the reduction of centriole number. To determine whether decreased PLK1 activation reduces centriole number in S phase, MCF7 cells were treated with BI6727 according to Protocol 2 in Fig. 2A without transfection, and cells were fixed at 4 h after the release. Centrinone B was used as a control for suppressing centriole duplication. As shown in Fig. 5C, BI6727 treatment decreased the number of centrin foci per cell, as observed previously for RACK1 knockdown (Yoshino et al., 2019). These results suggest that the kinase activity of PLK1 in G1/S and S phase is involved in centriole duplication.
RACK1 overexpression increases the localization of phosphorylated PLK1 and Aurora A to centrosomes
To determine the effect of RACK1 overexpression on PLK1 and Aurora A at the centrosome, the localization of p-PLK1, t-PLK1 and Aurora A to centrosomes was analyzed in RACK1-OE cells. RACK1 overexpression increased the fraction of cells with centrosomal p-PLK1 signals (Fig. 6A,B), whereas it had no effect on the centrosomal localization of t-PLK1 (Fig. 6C,D). Consistent with these results, RACK1 overexpression increased the fluorescence intensity of p-PLK1 at centrosomes (Fig. S6A), but not that of t-PLK1 (Fig. S6B). The RACK1 overexpression-induced increase in the fraction of cells with centrosomal p-PLK1 signals was confirmed in another breast cancer cell line, T47D (Fig. S6C,D). Western blot analysis showed increased p-PLK1 levels in RACK1-overexpressing HEK-293T cells, whereas no significant changes of t-PLK1 expression were observed (Fig. 6E). Cell cycle synchronization revealed that the proportion of cells with centrosomal p-PLK1 signals was increased in G1, G1/S and S phases in RACK1-OE cells (Fig. 6F; Fig. S7A). The fraction of cells with a centrosomal Aurora A signal was also increased in G1, G1/S and S phases in RACK1-OE cells (Fig. 6G; Fig. S7B). RACK1 overexpression also significantly increased the fluorescence intensity of Aurora A at centrosomes in asynchronized cells (Fig. S7C).
RACK1 overexpression induces premature centriole disengagement through the kinase activities of PLK1 and Aurora A
The kinase activity of PLK1 is critical for centriole disengagement, and PLK1 overactivation causes premature centriole disengagement (Lončarek et al., 2010; Kong et al., 2014). We therefore examined centriole disengagement in RACK1-OE cells using C-Nap1 (also known as CEP250) as a marker of disengagement in G1, G1/S and S phases. C-Nap1 localizes to the basal part of the mother centriole and disengaged daughter centrioles (Mayor et al., 2000). Engaged centrioles have one C-Nap1 focus, whereas disengaged centrioles have two C-Nap1 foci (Tsou and Stearns, 2006b). Therefore, the presence of more than two C-Nap1 foci in G1/S and S phases indicates premature centriole disengagement (Fig. 7A).
To visualize centrioles and identify transfected cells, MCF7 cells were co-transfected with the GFP–centrin2 vector and HA–RACK1. Most control cells contained two C-Nap1 foci at G1, G1/S and S phases (Fig. 7B,C). However, the fraction of cells with more than two C-Nap1 foci increased in G1/S and S phases in RACK1-OE cells, indicating premature disengagement. In G1 phase, the fraction of cells with more than two C-Nap1 foci in RACK1-OE cells was small and comparable to that in control cells.
To test whether formation of excess C-Nap1 foci depended on Aurora A or PLK1 kinase activities, cells were treated with MLN8054 or BI6727 according to Protocol 4 in Fig. 7D, and the number of C-Nap1 foci was counted. Both MLN8054 and BI6727 significantly decreased the fraction of cells with more than two C-Nap1 foci in RACK1-OE cells (Fig. 7E,F). These results suggest that RACK1 overexpression causes centriole overduplication by promoting premature centriole disengagement in some cells, and this effect is mediated by abnormal activation of Aurora A and PLK1.
Missense variants of RACK1 fail to enhance Aurora A–PLK1 signaling
We previously reported that several missense variants of RACK1 fail to increase the proportion of cells with centrosome amplification when overexpressed (Fig. 8A; Table S1) (Yoshino et al., 2019). The centrosomal localization of these RACK1 variants was similar to that of wild-type RACK1 (Yoshino et al., 2019). We therefore examined the effect of these RACK1 variants on enhancing the interaction between Aurora A and PLK1. The RACK1 variants T94A, Y302F and T303M caused a significantly lower level of enhanced interaction between Aurora A and PLK1 compared with the effect of wild-type RACK1 in our repeated experiments (Fig. 8B; Fig. S8A). To examine whether these variants lose centrosome regulating activity through a defect in PLK1 phosphorylation, T94A and T303M variants were overexpressed in MCF7 cells, and p-PLK1 localization to centrosomes was analyzed. Wild-type RACK1 and the K280E variant were used as controls. The K280E variant decreases binding to BRCA1, localization of BRCA1 to the centrosome, and centrosome amplification induced by its overexpression in cells derived from normal mammary epithelium (Yoshino et al., 2019). As shown in Fig. 8C,D, the T94A and T303M variants did not increase the fraction of cells with centrosomal p-PLK1 signals, whereas wild-type RACK1 and the K280E variant significantly increased the fraction of cells with centrosomal p-PLK1 signals. Consistent with these results, western blot analysis confirmed that the T94A and T303M variants failed to increase p-PLK1 in HEK-293T cells (Fig. 8E).
We previously reported that adequate expression levels of RACK1 are important for normal centriole duplication in mammary tissue-derived cells; RACK1 knockdown suppresses centriole duplication, whereas its overexpression causes centriole overduplication in a subset of mammary gland-derived cells (Yoshino et al., 2019). In this study, we showed that RACK1 contributes to the regulation of centriole duplication by modulating the Aurora A/PLK1 signaling axis (Fig. 8F).
To investigate how RACK1 regulates centriole duplication, we analyzed centrosome amplification induced by RACK1 overexpression in cells synchronized by a double thymidine block. In another breast cancer cell line, Hs578T, centrosome amplification was observed in more than 30% of RACK1-OE cells at 36 h after transfection (Yoshino et al., 2019), and most cells died within 3 days after transfection. RACK1 overexpression induced cleavage of caspases and poly(ADP-ribose) polymerase (PARP) (markers of apoptosis) in Hs578T cells (Fig. S8B). Thus, it was difficult to perform cell cycle synchronization in RACK1-OE Hs578T cells. By contrast, in MCF7 cells, RACK1 overexpression caused centrosome amplification in 18% of cells at 72 h after transfection, and a small fraction of cells underwent cell death by apoptosis (Fig. S8C). RACK1 overexpression increased the fraction of cells with centriole overduplication in MCF7 cells as well as in Hs578T cells (Fig. S1). Therefore, we used MCF7 cells for cell cycle synchronization experiments in this study.
An increase in the fraction of cells with centrosome amplification was observed at 4 and 8 h after the release, indicating that centrosome overduplication in RACK1-OE cells occurs during S and G2 phases (Fig. 1D,E). In the following G1 phase at 12 h after the release, the increase in the fraction of cells with centrosome amplification was not observed (Fig. 1D,E). We presume that most cells with extra centrosomes died during mitosis, possibly by apoptosis (Fig. S8C). As shown in Fig. 2, the activity of both PLK1 and PLK4 was involved in centrosome amplification in RACK1-OE cells during G1/S and S phases, whereas Aurora A inhibition before G1/S phase was required to suppress centrosome amplification in RACK1-OE cells. These results are consistent with the role of Aurora A as an upstream activator of PLK1. Collectively, these results indicated that the Aurora A/PLK1 signaling axis and PLK4 activity contribute to centrosome amplification induced by RACK1 overexpression. Because centriole disengagement mediated by PLK1 is followed by the initiation of centriole duplication governed by PLK4, these results suggest that RACK1 overexpression leads to centrosome amplification associated with PLK1 overactivation mediated by Aurora A in some cells.
Aurora A requires a scaffold protein such as Bora, Furry or CEP192 to activate PLK1 (Bruinsma et al., 2014; Seki et al., 2008; Joukov et al., 2014; Macůrek et al., 2008; Ikeda et al., 2012). The present finding that RACK1 binds to Aurora A and PLK1 suggests that RACK1 also acts as a scaffold for the Aurora A–PLK1 interaction. This was supported by the finding that RACK1 enhances the interaction between Aurora A and PLK1 in a dose-dependent manner in cells (Fig. 3C; Fig. S2E). Moreover, RACK1 knockdown decreased the interaction between Aurora A and PLK1 (Fig. S2F). In addition, a pulldown assay showed that Aurora A bound to the PLK1–RACK1 complex more efficiently than to PLK1 alone (Fig. 3E). These data support the idea that RACK1 has a scaffolding role for the interaction between Aurora A and PLK1. RACK1 bound to both the kinase domain and polo-box domains of PLK1. Because Thr210 is located in the kinase domain, the interaction between RACK1 and the kinase domain of PLK1 may contribute to PLK1 phosphorylation, whereas the polo-box domains would contribute to interactions with PLK1 substrates (Zitouni et al., 2014). This suggests that RACK1 may be phosphorylated by PLK1; however, investigating this possibility exceeds the scope of this paper.
RACK1 knockdown decreased the centrosomal localization of p-PLK1 in S phase, but not that of t-PLK1 (Fig. 4), which was confirmed by measuring their fluorescence intensities at centrosomes (Fig. S4D,E). These results suggest that the physiological activation of PLK1 in S phase is supported by RACK1 (Fig. 8F). Overexpression of RACK1 increased the localization of p-PLK1, but not that of t-PLK1, to centrosomes (Fig. 6), indicating that RACK1 overexpression promoted PLK1 phosphorylation at the centrosome (Fig. 8F). The increase of centrosomal p-PLK1 was observed in G1, G1/S and S phases. Increased expression of RACK1 induced PLK1 overactivation in the wrong cell cycle phases, including G1/S and S phases, resulting in an increase in cells with premature disengagement (Fig. 7). The level of p-PLK1 in cell lysates was decreased upon RACK1 knockdown and increased upon RACK1 overexpression. This was consistent with the change in the centrosomal localization of p-PLK1.
Similar to what was seen for p-PLK1, knockdown of RACK1 decreased the centrosomal localization of Aurora A in S phase in some cells (Fig. 5). The localization of Aurora A to centrosomes increased in G1, G1/S and S phases in some RACK1-OE cells (Fig. 6G). This suggests that the recruitment of Aurora A to centrosomes by RACK1 contributes to the regulation of PLK1 activation at centrosomes (Fig. 8F). Although RACK1 overexpression altered the localization of Aurora A, the effect was smaller than that on the localization of p-PLK1 to centrosomes (Fig. 6F,G). These results suggest that, in addition to increasing the centrosomal localization of Aurora A, the effect of RACK1 on the Aurora A–PLK1 interaction contributes to PLK1 activation at the centrosome.
Analysis of the number of C-Nap1 foci indicated that premature disengagement occurred in G1/S and S phases more frequently in RACK1-OE cells than in control cells, and PLK1 and Aurora A activities were required for the increase in premature disengagement. The ratio of C-Nap1 to centrin foci is commonly used as a marker of disengagement (Tsou and Stearns, 2006b). However, because RACK1 overexpression induces centriole overduplication and increases centriole number, the C-Nap1 to centrin ratio was not suitable for estimating disengagement status in our experiments. As shown in Fig. 7A, the number of C-Nap1 foci was sufficient to indicate premature disengagement from G1/S to S phase. RACK1 overexpression increased the fraction of cells with elevated numbers of C-Nap1 foci, and this effect was antagonized by inhibition of PLK1 or Aurora A. These data suggest that RACK1 overexpression induces premature disengagement by promoting the overactivation of the Aurora A/PLK1 signaling axis in some cells (Fig. 8F).
Adequate spatio-temporal activation of PLK1 is regulated by several distinct mechanisms. At mitotic entry, Bora provides the scaffold for the interaction between Aurora A and PLK1 required for PLK1 activation (Macůrek et al., 2008; Seki et al., 2008; Bruinsma et al., 2014). After the entry into mitosis, PLK1 activation by Aurora A is amplified by the scaffold activity of CEP192 at centrosomes (Meng et al., 2015; Joukov et al., 2014). RACK1 is involved in the activation of PLK1 in S phase, thus acting earlier than Bora or CEP192. Inhibition of PLK1 during G1/S to S phase suppressed centriole duplication (Fig. 5F), similar to the effect of RACK1 knockdown (Yoshino et al., 2019), suggesting that PLK1 activity supported by RACK1 in S phase is required for centriole duplication. Activation of PLK1 promotes centriole elongation and centriole maturation (Lončarek et al., 2010; Kong et al., 2014). However, the localization of SAS6 to the centrosome is not affected by PLK1 depletion (Lončarek et al., 2010), suggesting that PLK1 is not essential for procentriole formation. By contrast, we showed that knockdown of RACK1 decreases the fraction of cells with centrosomal localization of SAS6 in G2 phase (Yoshino et al., 2019), indicating that RACK1 might be required for procentriole formation. We showed that RACK1 knockdown decreased the localization of p-PLK1 to centrosomes only in S and G2 phases, and not during the G1/S transition, the period during which procentriole formation is initiated (Fig. 4A,B; Fig. S4B–D). Therefore, RACK1 might be involved in procentriole formation in a manner that is independent of PLK1 activation. Future studies should investigate the involvement of RACK1 and PLK1 in centriole duplication.
We previously reported that missense variants of RACK1 affect its function in centrosomes. In this study, the T94A, Y302F and T303A variants failed to enhance the interaction between Aurora A and PLK1 (Fig. 8B). In addition, overexpression of the T94A and T303M variants did not cause the upregulation of p-PLK1 at the centrosome in contrast to that of wild-type RACK1 and the K280E variant (Fig. 8C,D). The T94 residue is located in WD3, which comprises RACK1 fragments that bind to Aurora A, and the kinase domain and polo-box domains of PLK1 (Fig. S3). The Y302 and T303 residues are located in WD7. The RACK1 fragment that includes WD7 showed weak binding to Aurora A and the kinase domain of PLK1. Thus, alteration of T94A, Y302F and T303 residues may affect the scaffolding activity of the RACK1 protein for the Aurora A–PLK1 interaction. These data suggest that mutations of RACK1 alter centrosome regulating activities in several distinct ways: proteins with T94A, Y302F and T303M mutations fail to regulate PLK1 phosphorylation by Aurora A, and that with the K280E mutation fails to properly localize BRCA1 to the centrosome. NetPhos 3.0 (http://www.cbs.dtu.dk/index.html) and NetworKIN (http://networkin.info/index.shtml/) predictions identified T94, Y302 and T303 as candidate residues for phosphorylation; therefore, the phosphorylation of these residues may affect the activity of RACK1 in enhancing the Aurora A–PLK1 interaction. T94A and T303M have been reported in colon cancer (COSMIC database; https://cancer.sanger.ac.uk/cosmic), suggesting that may contribute to carcinogenesis or induce a cancer phenotype by dysregulating Aurora A–PLK1 signaling. The effect of the T97A/T98A, L206F, K245C and Y246F variants on altering the centrosome regulatory activity of RACK1 remains unclear. In addition to further investigate these variants, future studies should examine the phosphorylation of these residues and the responsible kinases, and evaluate the association between RACK1 variants and carcinogenesis.
In conclusion, RACK1 was identified as a new scaffold protein for the Aurora A–PLK1 interaction that regulates centriole duplication by activating the Aurora A/PLK1 signaling axis. Overactivation of this signaling pathway by RACK1 overexpression may cause centrosome amplification.
MATERIALS AND METHODS
Cell lines and culture
MCF7 (ATCC HTB-22), T47D (ATCC HTB-133), and HEK-293T (ATCC CRL-3216) cells and Hs578T cells (ATCC HTB-126) were purchased from ATCC. MCF7 and HEK-293T cells were maintained in DMEM supplemented with 8% fetal bovine serum (FBS). T47D cells were maintained in RPMI-1640 supplemented with 10 µg/ml recombinant human insulin and 8% FBS. Hs578T cells were maintained in DMEM supplemented with 10 µg/ml recombinant human insulin and 8% FBS. All cells were incubated in an atmosphere containing 5% CO2. Cell line identities were verified using the GenomeLab Human STR Primer set (Beckman Coulter).
Preparation of MCF7-tet-shRACK1 cells
To produce lentiviral particles, pLKO-tet-shRACK1 was co-transfected with psPAX2 and pMD2.G (Addgene plasmids #12260 and #12259, respectively; deposited by Didier Trono) into HEK-293T cells. At 48 h after transfection, the supernatant was collected and filtered to remove cell debris. MCF7 cells were infected by lentivirus in the supernatant and infected cells were selected with puromycin for >7 days.
MLN8054 and BI6727 were purchased from AdooQ Bioscience. Centrinone B was purchased from Tocris Bioscience. These inhibitors were used at the concentrations of 1 μM for MLN8054, 100 nM for BI6727 and 500 nM for centrinone B.
Polyethylenimine MAX (Polysciences) was used for plasmid transfection. The TransIT-X2 Dynamic Delivery System (Mirus Bio) was used for siRNA transfection and siRNA and plasmid co-transfection.
The plasmids pCY4B-FLAG-RACK1, pCY4B-HA-RACK1, pCMV-Myc-RACK1 and pET-RACK1 were described previously (Yoshino et al., 2019). pCY4B-HA-RACK1 missense variants were generated by site-directed mutagenesis (Yoshino et al., 2019). To construct pCY4B-FLAG- and pCY4B-HA-PLK1, the coding sequences of PLK1 were amplified using primers containing XhoI and NotI sites using HeLa cDNA as the template. The amplified PCR products were subcloned into the XhoI/NotI sites of the pCY4B-FLAG and pCY4B-HA vectors (Okano et al., 2000). The construction of pCY4B-FLAG- and pCY4B-HA-Aurora A was as described for PLK1 plasmids. To construct pET-Aurora A, the coding sequence of Aurora A was subcloned from pCY4B-HA-Aurora A into the XhoI/NotI sites of the pET-44a(+) plasmid (Merck). To construct pLKO-tet-shRACK1-puro, annealed oligonucleotides (5′-CTAGCCAGGGATGAGACCAACTATGCTCGAGCATAGTTGGTCTCATCCCTGGTTTTTG-3′ and 5′-AATTCAAAAACCAGGGATGAGACCAACTATGCTCGAGCATAGTTGGTCTCATCCCTGG-3′) were subcloned into the NheI/EcoRI sites of EZ-Tet-pLKO-puro [Addgene plasmid #85966; deposited by Cindy Miranti (Frank et al., 2017)].
The sequences of the siRNAs designated as RACK1-1 and RACK1-2 were 5′-CAGGCUAUCUGAACACGGU-3′ (Silencer™ Select siRNA, Thermo Fisher Scientific) (Yoshino et al., 2019) and 5′-GAGGUUGUGGUGCUAGUUUCUCUdAdA-3′ (DsiRNA, Integrated DNA Technologies). The Silencer™ negative control siRNA template set (Thermo Fisher Scientific) was used as the negative control.
Cell cycle analysis
Cell cycle progression was analyzed by flow cytometry as described previously (Yoshino and Ishioka, 2015). In brief, cells were trypsinized, fixed with ethanol at 4°C and stored at −20°C until analysis. The fixed cells were suspended in phosphate-buffered saline (PBS) containing 0.1% Triton X-100, 200 µg/ml RNase A (Thermo Fisher Scientific) and 50 µg/ml propidium iodide, and analyzed on a Cytomics FC500 flow cytometer (Beckman Coulter).
Cell cycle synchronization
Cells were synchronized in G1/S transition using a double thymidine block protocol (Ma and Poon, 2011) with minor modifications. In brief, cells were incubated for 6 h after transfection and treated with 2 mM thymidine in complete medium for 24 h. Cells were washed with PBS, released in fresh medium without thymidine for 15 h, and treated with 2 mM thymidine in medium for 24 h. After the second thymidine block, cells were washed with PBS, released in fresh medium for the indicated duration, and fixed. In experiments using kinase inhibitors, each chemical was added as indicated in the protocols shown in the respective figures.
For immunostaining of centrosomes, cells were permeabilized in PBS containing 0.2% Triton X-100, 1 mM EGTA and 1 mM MgCl2 for 1 min, and fixed in methanol for 15 min at −80°C. After washing and blocking, the cells were incubated with primary antibodies in 1% BSA in PBS overnight. The following antibodies were used: anti-HA (3F10, 1:400, Roche), anti-γ-tubulin (GTU-88, 1:500, Sigma-Aldrich), anti-p-PLK1 (D5H7, 1:200, Cell Signaling Technology), polyclonal anti-t-PLK1 (A300-251A, 1:2000, Bethyl Laboratories), polyclonal anti-Aurora A (A300-071A, 1:3000, Bethyl Laboratories), anti-centrin (C7736, 1:1500, Sigma-Aldrich), polyclonal anti-CEP250/C-Nap1 (14498-1-AP, 1:2000, Proteintech), anti-CP110 (140-195-5, 1:200, Merck), anti-CEP152 (A302-479A, 1:2000, Bethyl Laboratories) and anti-SAS6 (91.390.21, 1:200, SantaCruz Biotechnology) antibodies. After incubation with primary antibodies, cells were washed with PBS containing 0.1% Tween-20 (PBS-T), and incubated with 1% BSA in PBS containing Alexa Fluor 488- and 568-conjugated secondary antibodies (Thermo Fisher Scientific) for 30 min. After washing with PBS-T, cells were mounted in mounting medium with DAPI (Vector Laboratories). Immunocytofluorescence images were acquired with the BZ-X710 fluorescent microscope system (Keyence) equipped with a CFI Plan Apo λ 100×H/1.45 NA oil immersion objective (Nikon) and monochrome cooled CCD camera (2/3 inches, 2.83 megapixels).
Cells were lysed in 1× SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% sucrose, and 5% 2-mercaptoethanol) supplemented with a protease inhibitor cocktail (Roche) and a phosphatase inhibitor cocktail (5 mM sodium fluoride, 200 µM sodium orthovanadate, 1 mM sodium molybdate, 2 mM sodium pyrophosphate, and 2 mM disodium β-glycerophosphate as final concentrations), yielding whole-cell lysates. SDS-PAGE and western blotting were performed as previously described (Yoshino and Ishioka, 2015). Primary antibodies were as follows: anti-FLAG (M2, 1:5000, Sigma-Aldrich), anti-HA (3F10, 1:3000, or 16B12, 1:3000, BioLegend, San Diego, CA, USA), polyclonal anti-t-PLK1 (A300-251A, 1:10,000), monoclonal anti-PLK1 (E-2, 1:1000, Santa Cruz Biotechnology), anti-p-PLK1 (D5H7, 1:3000), polyclonal anti-Aurora A (A300-071A, 1:10,000), anti-γ-actin (2F3, 1: 5000, Wako), anti-caspase 9 (#9502, 1:1000, Cell Signaling Technology), anti-caspase 3 (#9662, 1:1000, Cell Signaling Technology) and anti-PARP (#9532, 1:1000, Cell Signaling Technology) antibodies. Secondary antibodies were as follows: horseradish peroxidase (HRP)-tagged anti-mouse-IgG (GE Healthcare), HRP-tagged anti-rabbit-IgG (GE Healthcare) and HRP-tagged anti-mouse native form IgG (TrueBlot, Rockland). Signals were detected using an ECL substrate (ATTO) on a CCD imager (Image Quant LAS 4000 mini, GE Healthcare).
Cells were lysed in lysis buffer (10 mM HEPES pH 7.6, 250 mM NaCl, 0.5% NP-40, 5% glycerol, and 5 mM EDTA) supplemented with protease inhibitor and phosphatase inhibitor cocktails, and then cleared by centrifugation at 5000 g for 10 min. The lysate was rotated with anti-FLAG antibody (2H8, Trans Genic) or anti-p-PLK1 antibody (K50-483, BD Pharmingen) at 4°C overnight. The lysate was rotated for another hour with Protein G–Sepharose (GE Healthcare). After vigorous washing, proteins were eluted in 1× sample buffer and analyzed by western blotting.
His-tagged proteins were expressed in BL21-competent cells. The bacterial cell pellets were lysed with extraction buffer (50 mM phosphate buffer pH 8.0, 300 mM NaCl, and 0.1% NP-40). His-tagged proteins were purified from the lysate using Ni-NTA-agarose (Qiagen). MBP-tagged proteins were obtained using the pMAL Protein Fusion and Purification System (New England Biolabs) according to the manufacturer's instructions.
For the GST pulldown assay, GST-tagged proteins were expressed in BL21 cells, and bacterial cell pellets were lysed in sonication buffer (20 mM Tris-OAc pH 7.9, 120 mM potassium acetate, 10% glycerol, 1% Triton X-100, 1 mM EDTA and 1 mM DTT). After centrifugation at 12,000 g for 30 min, the supernatant was incubated with glutathione–Sepharose beads (GE Healthcare) at 4°C for 1 h. After vigorous washing, the beads were incubated with the purified His-tagged proteins in pulldown incubation buffer (10 mM HEPES pH 7.4, 100 mM KCl, 1 mM MgCl2, 0.1 mM CaCl2) at 4°C or at room temperature for 1–4 h. The beads were washed and analyzed by western blotting.
For the MBP pulldown assay, purified MBP-tagged proteins were incubated with amylose resin in column buffer (20 mM Tris-HCl pH 7.5, 200 mM NaCl, 1 mM EDTA, 1 mM DTT) at 4°C for 1 h. After washing the resin, the indicated prey proteins were added and incubated in pulldown incubation buffer at room temperature for 1 h. After vigorous washing, the beads were analyzed by western blotting.
Image processing and measurement of fluorescence intensity
ImageJ v1.49 (http://imagej.nih.gov/ij/) was used for image analysis. For analysis of fluorescence intensity at centrosomes, integrated signal intensity (ID) was measured in a fixed-size area around each centrosome. For analysis of p-PLK1 and t-Aurora A, the ID was used as the fluorescence intensity at centrosomes. For the analysis of t-PLK1, the adjusted fluorescent intensity was calculated by subtracting the ID of the background at cytoplasm from the ID at the centrosome.
Statistical analysis was performed using JMP 12 software (SAS Institute Inc). Graphs were constructed using Excel 2016 (Microsoft). Statistical comparisons between two different samples were made using a two-tailed Welch's test. Comparisons between more than two samples were made by ANOVA. When the result of ANOVA indicated a significant difference, post-hoc comparisons were made using a Dunnett's test to calculate P-values. The P-value was calculated using Wilcoxon rank sum test to compare intensity data. P<0.05 was considered significant.
We thank Dr Akira Yasui and Dr Ryutaro Shirakawa, Institute of Aging, Development, and Cancer, Tohoku University, for useful suggestions and discussion, and Satoko Kaneko for technical assistance.
Conceptualization: Y.Y.; Methodology: Y.Y., A.K.; Validation: Y.Y., A.K.; Formal analysis: Y.Y., A.K.; Investigation: Y.Y., A.K., H.Q., S.E., Z.F., K.S., R.K.; Data curation: Y.Y., A.K., N.C.; Writing - original draft: Y.Y.; Writing - review & editing: Y.Y., N.C.; Supervision: N.C.; Project administration: N.C.; Funding acquisition: Y.Y., N.C.
This study was supported by grants-in-aid from the Japan Society for the Promotion of Science (JSPS) KAKENHI grant numbers JP16K18409 (to Y.Y.), JP18K15233 (to Y.Y.), JP16H04690 (to N.C. and Y.Y.), JP19H03493 (to N.C. and Y.Y.), the Foundation for Promotion of Cancer Research in Japan, the Sasakawa Scientific Research Grant from The Japan Science Society (to Y.Y.), Friends of Leukemia Research Fund, Research Grant of the Princess Takamatsu Cancer Research Fund, and Research Program of the Smart-Aging Research Center, Tohoku University (to N.C.).
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.238931.reviewer-comments.pdf
The authors declare no competing or financial interests.