Parkin Regulates Mitosis and Genomic Stability Through Cdc20/Cdh1

INTRODUCTION

Loss of function of the Parkin protein leads to death of dopaminergic neurons and causes Autosomal Recessive Juvenile Parkinsonism (AR-JP) (Kitada et al., 1998; Lucking et al., 2000). Parkin as a RING finger containing protein is capable of promoting mono- and polyubiquitination of target proteins (Moore et al., 2008; Olzmann et al., 2007; Walden and Martinez-Torres, 2012). The neuroprotective role of Parkin is linked to its role in mitophagy and removal of toxic substrates (Winklhofer, 2014). Parkin has also been identified as a candidate tumor suppressor in a wide variety of human cancers (Cesari et al., 2003; Fujiwara et al., 2008; Picchio et al., 2004; Veeriah et al., 2010; Yeo et al., 2012). However, how Parkin functions as a tumor suppressor remains unclear. At the cellular level, loss of Parkin has been associated with formation of micronuclei and multipolar spindles, implying a requirement for Parkin in proper chromosome segregation (Veeriah et al., 2010). Mechanistically, Cyclin E was proposed as a Parkin substrate contributing to mitotic defects (Veeriah et al., 2010). However, another group suggested that Cyclin E is not a Parkin substrate (Yeo et al., 2012). Therefore, how Parkin regulates mitosis remains unclear.

APC/C (anaphase-promoting complex/ cyclosome) is important for cell cycle regulation and is active from mitosis to the end of G1 phase (Castro et al., 2005; Nakayama and Nakayama, 2006; Peters, 2006). APC/C activity is controlled by two adaptor proteins, Cdh1 and Cdc20, which serve as coactivators for APC/C by binding to substrates, including Cyclin A, Cyclin B (for mitotic cyclins), Aurora A, Aurora B, Plk1, Nek2A (for mitotic kinases), Securin, Sgo1 (for chromosome segregation), Survivin (for mitotic checkpoint), Geminin, Cdc6 (DNA replication), Skp2 (F-box for S phase), and Ets2 and FoxM1 (for transcription) (Acquaviva et al., 2004; Castro et al., 2005). Dysregulation of this pathway could result in mitotic catastrophe, chromosome instability, or uncontrolled cell proliferation that will lead to cancer (Castro et al., 2005; Nakayama and Nakayama, 2006). Many Cdh1 or Cdc20 substrates, including Plk1, Aurora A, Aurora B, Cyclin B1, and Securin are highly expressed in many types of lung tumors (Kim et al., 2011; Penas et al., 2011). However, the mechanism underlying the overexpression of these mitotic regulators remains unclear as very few mutations were found in APC/C subunits and Cdc20 itself is found to be overexpressed in cancers (Penas et al., 2011). This suggests that key mitotic regulators might be regulated by other mechanisms. In addition, several studies suggest that both Cdc20 and Cdh1 have functions independent of APC/C (Clarke et al., 2003; Goh et al., 2000; Gourguechon et al., 2013; Pfleger et al., 2001; Thornton and Toczyski, 2003).

In this study, we found that Parkin regulates mitotic progression and its deficiency results in multiple mitotic defects including chromosome misalignment, chromosome lagging, chromosome bridge formation, prometaphase-like arrest, anaphase and cytokinesis failure. Parkin knockout (KO) Mouse Embryonic Fibroblasts (MEFs) showed increased levels of mitotic proteins such as Plk1, Aurora A, Aurora B, Cyclin B1, Cdc20, and UbcH10. More interestingly, Parkin forms a complex with Cdc20 or Cdh1. Parkin-Cdc20/Cdh1 complex is a distinct complex from APC/C-Cdc20/Cdh1 complex. Depleting APC11 or Parkin alone delayed Cyclin B1 degradation and mitotic exit, but did not recapitulate Cdc20 depletion. However, codepletion of APC11 and Parkin had the same effect as Cdc20 depletion. Parkin-deficient tumor cells have increased expression of mitotic regulators, such as Plk1 and Aurora A, making these cells sensitive to Plk1 or Aurora kinase inhibitors. Therefore, our studies reveal a critical function of Parkin in maintaining genomic stability and a targeted approaches to treat tumors with Parkin deficiency.

RESULTS

Parkin Regulates Mitosis

To understand the role of Parkin in mitosis, we monitored mitotic chromosome movement using time-lapse microscopy in Parkin-depleted U2OS cells (Figure 1A; Figure S1A) and Parkin knockout (KO) mouse embryonic fibroblasts (MEFs; Figures S1B-S1F). This analysis revealed a broad spectrum of mitotic defects including chromosome misalignment, chromosome lagging, chromosome bridge formation, prometaphase-like arrest, anaphase and cytokinesis failure (Figure 1A; Figure S1F). In addition, progression from nuclear envelope breakdown (NEBD) to anaphase onset was significantly delayed in Parkin KO MEFs compared to wild type (WT) MEFs (Figures S1B and S1H), a defect that was reversed by exogenous expression of WT Parkin (Figures S1G and S1H). These results demonstrate that Parkin deficiency results in multiple mitotic defects.

Figure 1

Parkin Regulates Mitosis

(A) Time-lapse analysis of mitotic U2OS cells transfected with Control or Parkin siRNA. 50 cells were counted in each experiment. Top: Quantification of abnormal mitotic cells. *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus Control siRNA by one-way ANOVA. Bottom left: Parkin and β–actin expression were shown; Bottom right: Representative images of cells with indicated misaligned chromosome, lagging chromosome, and Chromosome bridge were shown. Scale bar, 10 μm.

(B) Cells were synchronized at the G1/S transition by double-thymidine block, and then released into a drug-free medium. Cell were harvested at indicated times and analyzed by immunobloting. p27 ^{serves as a G}₀-G₁ phase marker; Cyclin E, early S phase; Skp2, G₁-S; p-H3, mitosis.

(C) Subcellular localization of Parkin during each stage of the cell cycle. U2OS cells were stained with antibodies against Parkin (Red) and Plk1 (Green) and DNA (Blue) stained with DAPI. White arrows, centrosome; Pink arrowheads, midzone, midbody, or midring from anaphase to cytokinesis. Scale bar represents 20 μm.

(D) Immunoblot analysis of mitotic factors in primary Parkin WT and KO MEFs (Passage 5).

(E) Immunoblot analysis of mitotic factors in primary Parkin WT and KO MEFs after releasing from serum starvation (for 72 h) and nocodazole arrest (for 18 h).

See also Figure S1 and S2.

Next, we examined Parkin levels at different stages of the cell cycle. Cells arrested at the G1/S boundary by double thymidine block (DTB) showed high Parkin levels. Upon release, Parkin levels decreased as cells progressed through S phase, and then peaked from G2 until early G1, without corresponding changes in mRNA levels (Figure 1B; Figure S1I). Furthermore, we found that Parkin localized to centrosomes, midzone, and midbody in various cells types, including U2OS cells (Figure 1C; Figures S1J and S1K) and IMR-90 lung fibroblasts (PDL=33) (data not shown). These results suggest that Parkin might have a direct role in mitotic regulation. To examine how Parkin might regulate mitosis, we examined the expression of key mitotic regulators. Immunoblot analysis of asynchronous or mitotic lysates from Parkin WT and KO MEFs showed increased levels of Plk1, Aurora A, Aurora B, Cyclin B1, Cdc20, and UbcH10 (Figures 1D and E). Other key mitotic regulators, such as Mad1, Mad2, Bub1, BubR1 and Bub3 were not affected. Cyclin E, whose upregulation has been linked to genomic instability in Parkin-deficient cells (Veeriah et al., 2010), was also present at normal levels. Furthermore, Parkin-depleted cells showed aberrant localization and expression of Plk1, Cyclin B1, and Aurora B as examined by immunofluorescence (IF) and immunoblot (Figures S2A and S2B; data not shown), respectively. Mitotic defects and up-regulation of Plk1 and Cyclin B1 in Parkin-depleted cells were reversed by expressing WT Parkin but not C431S, which abolishes Parkin's E3 ligase activity (Figures S2C and S2D) (Iguchi et al., 2013; Riley et al., 2013). These results suggested that Parkin regulates mitosis by controlling the levels of particular mitotic regulators through its E3 ligase activity. PINK1 knockdown did not affect Plk1 and Cyclin B1 levels, suggesting that Parkin's role in mitotic regulation is PINK1-independent (Figure S2E), and thus distinct from Parkin's established role in mitophagy.

Parkin-Mediated Ubiquitination Is a Mitotic Regulator

We hypothesized that Parkin directly regulates the levels of mitotic regulators, such as Plk1 and Aurora B, through its E3 ligase activity (Shimura et al., 2000). Consistent with this, we found that endogenous Parkin interacts with Plk1, Cyclin B1, Aurora A, Aurora B, and Nek2A (Figure 2A). Furthermore, overexpression of WT Parkin but not the C431S mutant markedly decreased levels of these mitotic regulators, which could be prevented by MG132 pre-treatment (Figure 2B), supporting the idea that Parkin regulates the abundance of these mitotic regulators through the proteasome pathway. Immunoblot analysis of tissue lysates from Parkin WT and KO mice revealed that Plk1, Aurora B and Cyclin B1 protein levels are elevated in tissues lacking Parkin (Figure 2C). Importantly, overexpression of Parkin in cells increased the polyubiquitination of Plk1, Aurora B, Cyclin B1, Aurora A, Securin, Aurora B and Nek2A, but not c-Myc and Cyclin E, whose expression are not regulated by Parkin (Figure 2D). Furthermore, the C431S mutation abolished Parkin's E3 ligase activity toward its substrates. Early studies suggest that Parkin mediates K48- or K63-linked polyubiquitylation in brain (Moore et al., 2008; Olzmann et al., 2007; Youle and Narendra, 2011). Interestingly, Parkin mostly mediated K11–linked polyubiquitin-chains in Plk1 ubiquitination (Figure S2F). Collectively, these data indicate that Parkin regulates the levels of a subset of mitotic proteins through the ubiquitin-proteasome pathway.

Figure 2

Parkin-Mediated Regulates the Levels and Ubiquitination of Mitotic Regulators

(A) HEK 293T cells were synchronized by nocodazole for 18 h, and mitotic and ansynchronized cells were collected for immunoprecipitation (IP)-immunoblot analysis with control IgG, anti-Plk1, Cyclin B1 and Parkin antibodies.

(B) HEK 293T cells were transfected with the indicated plasmids, and then treated with MG132 or left untreated. Cell lysates were blotted with the indicated antibodies.

(C) The lung, liver, kidney, spleen tissues of Parkin WT and KO mice (n = 3 mice/genotype) were lysed, and cell lysates were blotted with the indicated antibodies.

(D) HEK 293T cells were transfected with the indicated constructs and arrested in mitosis with nocodazole for 18 h (Left). Cells were synchronized at the G1/S transition by double-thymidine block, and then released into a new medium (Right). Cells were then treated with MG132. Ubiquitinated proteins were pull down under denaturing conditions by Ni-NTA agarose and analyzed by immunoblot. c-Myc and Cyclin E were shown as negative controls.

See also Figure S3.

In experiments designed to identify the E2 ubiquitin ligase for Parkin, we did not observe an interaction between Parkin and UbcH10, the E2 for APC/C in mitosis (data not shown) (Castro et al., 2005; Peters, 2006). Instead, UbcH7 (also called Ube2L3), the E2 for Parkin in cellular processes other than mitosis (Shimura et al., 2000; Wenzel et al., 2011), was significantly elevated and interacted with Parkin in mitosis (Figures S3A and S3B) and accumulated at various mitotic structures, including centrosomes, midzone, and midbody, just like Parkin (Figure S3C). Importantly, UbcH7 depletion caused mitotic defects similar to Parkin depletion (Figures S3D and S3E), further supporting the idea that UbcH7 acts as an E2 ubiquitin ligase for Parkin in mitosis.

Parkin-Cdc20/Cdh1 Acts as a Mitotic-Regulating Complex

Parkin regulates mitotic factors, which are also regulated by APC/C, raising the possibility that Parkin interacts with APC/C or its subunits. We next examined the interaction between Parkin and the APC/C subunits (Figure 3A). Endogenous Parkin co-immunoprecipitated with Cdc20 and Cdh1 from mitotic cell lysates, but not with APC/C components APC11 and APC2. Furthermore, we found that recombinant Parkin interacted with Cdc20 and Cdh1 under cell-free conditions, suggesting that Parkin directly interacts with Cdc20/Cdh1 (Figure 3B). Use of synchronized cell lysates indicated that Parkin first interacts with Cdc20 and then switches to Cdh1 after cells exit mitosis (Figure 3C). Taken together, these data suggest that Parkin forms a complex with Cdc20 or Cdh1 that does not include the APC/C.

Figure 3

Parkin-Cdc20/Cdh1 Complex Is A Mitotic Regulator during the Cell Cycle

(A) HEK 293T cells were synchronized by nocodazole and treated with MG132. Cell lysates were then subjected to IP and immunoblot as indicated.

(B) Purified Cdc20 or Cdh1 were incubated with GST or GST-Parkin coupled to GSH-Sepharose. Proteins retained on Sepharose were then blotted with the indicated antibodies.

(C) HEK 293T cells transfected with Flag-tagged WT Parkin were synchronized by nocodazole treatment. Cells were released and subjected to IP and immunoblot with the indicated antibodies.

(D and E) Cells were transfected with the indicated constructs and treated as in Figure 3A. Cells were subjected to IP and immunoblot with the indicated antibodies. APC11 (D) and APC2 (D and E) were shown as negative controls.

(F) in vitro ubiquitination of Cyclin B1, Securin and Nek2A by Parkin and Cdc20/Cdh1. Purified bacteria-produced His-Cyclin B1, Securin and Nek2A protein was incubated with different components as indicated for 90 min at 30°C. Samples were analyzed by immunobloting with ubiquitin antibody.

See also Figure S3 and Table S1.

Cdc20 and Cdh1 act as substrate-recognition subunits of APC/C (Castro et al., 2005; Peters, 2006), which led to the idea that Parkin might also target specific mitotic substrates through Cdc20 and Cdh1. Knockdown of Cdc20 or Cdh1 resulted in decreased binding of Parkin to various mitotic substrates, including Cyclin B1 and Aurora B (Figure 3D and data not shown). In contrast, knockdown of APC11 did not affect these interactions and Parkin's interaction with Cdc20/Cdh1 (Figure 3E). Moreover, we found Cdc20- and Cdh1-specific degron sequences (D-box and KEN box motifs) (Castro et al., 2005; Nakayama and Nakayama, 2006) in a series of established Parkin substrates, including Ataxin 2 and 3, Synaptotagmin XI, RanBP2, β-catenin, PCDP2-1, α and β tubulin, LIM kinase, PLC-γ1, MFN1 and 2, Mitochondrial Rho GTPase isoform 1, Septin 4 and 5, and Drp1 (Table S1) (Walden and Martinez-Torres, 2012), the latter of which has been shown to require Cdh1 for ubiquitination (Horn et al., 2011; Wang et al., 2011). To further confirm the role of Cdc20 and Cdh1 in Parkin-mediated ubiquitination, we performed in vitro ubiquitination assays. Parkin induced ubiquitination of Cyclin B1, Securin and Nek2A; however, their ubiquitination were abolished in the absent of Cdc20/Cdh1, Ube1 (E1), UbcH7 (E2) or Parkin (Figure 3F; Figure S3F). Furthermore the D-box / KEN-box mutants of these substrates were not polyubiquitinated by Parkin. These findings further strengthen the notion that Parkin-Cdc20 and -Cdh1 complexes act independently of APC/C-Cdc20 and -Cdh1 in regulating the abundance of key mitotic regulators.

UbcH7-Parkin-Cdc20 and -Cdh1 Complexes Regulate Mitosis Independently of UbcH10-APC/C-Cdc20 and -Cdh1 Complexes

Next we examined the functional interaction between Parkin and APC/C. Inactivation of APC/C by APC11 knockdown resulted in chromosome missegregation defects and upregulation of Plk1 (Figures 4A and 4B). We found that ectopic expression of Parkin in APC11-deficient cells reversed these mitotic abnormalities (Figures 4A and 4B). In addition, Parkin overexpression restored Plk1 levels and rescued mitotic errors induced by UbcH10 (APC/C E2) knockdown, but had no effect on UbcH7 (Parkin's E2)-induced mitotic defects (Figure 4C). These studies suggest that the UbcH7-Parkin-Cdc20 and -Cdh1 complexes regulate mitosis independently of UbcH10-APC/C-Cdc20 and -Cdh1 complexes. Although Parkin and APC/C show many similarities in mitosis, there are some differences in their localization. As shown in Supplemental Figures S4A and S4B, Parkin is localized in the centrosome or midbody like Cdc20, while APC3 is localized in the kinetochores, or the midring in mitosis. Furthermore, we found that UbcH7-Parkin-Cdc20 has target proteins such as α and β tubulin that are not regulated by APC/C (Figure S4C).

Figure 4

UbcH7-Parkin-Cdc20 and -Cdh1 Complexes Regulate Mitosis Independently of UbcH10-APC/C-Cdc20 and -Cdh1 Complexes

(A-C) Live-cell imaging analysis of chromosome segregation defects in U2OS cells infected with the indicated constructs and synchronized by nocodazole treatment. Cells were fixed and stained with DAPI. Representative images of cells with indicated missegregation events were shown. Scale bar, 20 μm (A). Analysis of numerical chromosome segregation errors. 100 cells were counted in each experiment. *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus Control shRNA by one-way ANOVA (B-C, Left). Immunoblot analysis with indicated antibodies (B-C, Right).

(D) Cells were infected with the indicated constructs, synchronized by nocodazole, and released. Cyclin B1 expression was then examined by immunoblot analysis (Top). FACS analysis for cell cycle profile (Bottom).

(E) Fluorescence quantification of Cyclin B1-GFP by time-lapse imaging in mitotic H2B-mRFP-expressing U2OS cells infected with the indicated shRNAs. Cells were plotted against time before and after prometaphase (shake off). *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus control shRNA by two-way ANOVA.

(F) Representative images of cells as indicated. Note. The frames of live cell imaging were recorded by shake-off for mitotic cells. Trypsin-EDTA was the treatment for Interphase cells as the control. Scale bar, 20 μm.

(G) After infection with the indicated shRNAs, cells were synchronized at the mitosis transition (prometaphase) by nocodazole treatment for 18 h. After harvesting the mitotic cells by shake off, cells were re-cultured and dividing cells were examined at the indicated time points. The data represent the average of three experiments and 100 cells were monitored in each experiment. Scale bar, 20 μm.

See also Figure S4.

Since Parkin and APC/C share the same coactivator Cdc20, one prediction is that mitotic defects caused by depletion of APC/C or Parkin alone would be less severe than those caused by depletion of Cdc20 (Huang et al., 2009). To test this idea, we studied whether Parkin affects Cdc20-mediated degradation of Cyclin B1 at the metaphase-to-anaphase transition. Depleting APC11 or Parkin alone delayed Cyclin B1 degradation and mitotic exit, but did not recapitulate Cdc20 depletion (Figures 4D, 4E and 4G; Figures S4C-S4E). However, codepletion of APC11 and Parkin phenocopied Cdc20 depletion (Figures 4D, 4E, 4F and 4G; Figures S4D and S4E).

Parkin Is Phosphorylated and Activated by Plk1 upon Mitotic Entry

We next sought to identify mitosis-specific regulation of Parkin. Phosphorylation is a common posttranslational modification and has been shown to involve protein stability and activity. We therefore scanned Parkin using GPS2.12, a tool for prediction of kinase-specific phosphorylation sites (Xue et al., 2008), which identified Ser 378 as a potential phosphorylation site by Plk1. We found that Parkin was phosphorylated at Ser 378 in mitosis (Figures 5A and 5B). Treatment of carbonyl cyanide m-chlorophenylhydrazone (CCCP), a mitochondrial-uncoupling reagent that activates Parkin during mitophagy (Iguchi et al., 2013; Riley et al., 2013), did not affect Ser 378 phosphorylation (Figure 5B). Ser 378 is predicted to be a Plk1 phosphorylation site (http://gps.biocuckoo.org/) and its surrounding residues fit with a consensus Plk1 phosphorylation site (D/ExS/TΦ, Φ: hydrophobic residues). In addition, Ser 378 is highly conserved among vertebrates (Figure 5C), suggesting that the phosphorylation of this site may have an evolutionarily conserved role in regulating Parkin activity. To test whether Plk1 regulates Parkin S378 phosphorylation, we treated cells with BI 2536, a Plk1 inhibitor, or infected cells with Plk1 shRNA. We found that Plk1 inhibition or deficiency blocked Parkin phosphorylation at Ser378 (Figures 5D and 5E). Conversely, overexpression of Plk1 or constitutively active Plk1 (T210D) (van de Weerdt et al., 2005), but not inactive Plk1 (T210A), increased Parkin phosphorylation (Figure 5F). Furthermore, Plk1 was able to phosphorylate recombinant WT Parkin but not S378A (Figure 5G). Interestingly, Ser 378 localizes within the IBR domain. In previous studies, it is established that the IBR domain assists the recruitment of proteins involved in the ubiquitination pathway (Chung et al., 2001; Zhang et al., 2000). Structurally, the IBR domain helps a close arrangement of the RING1 and RING2 domains, which facilitates protein interactions and subsequent ubiquitination (Beasley et al., 2007) In addition, the region involves in maintaining conformational flexibility, and it can affect Parkin's activity and stability (Trempe et al., 2013). We also found that the IBR domain is involved in Parkin's interaction with Cdh1. As shown in Figure S5A, Cdh1 and Plk1 could interact with the C terminal region of Parkin containing the RING1-IBR or IBR-RING2 domain. The RING2 domain alone, but not the RING1 domain, could interact with Cdh1. There results suggest that the IBR and RING2 domain could interact with Cdh1.

Figure 5

Parkin Is Phosphorylated by Plk1 at Ser378 and Activated during Mitosis

(A) Cells were synchronized at the G1/S transition by double-thymidine block, and cells were released. Cell were harvested at indicated times and analyzed by immunobloting.

(B) Cells were incubated in the absence or presence of nocodazole or CCCP and subjected to immunoblot analysis with the indicated antibodies.

(C) Comparison of the sequences surrounding S378 of Parkin orthologues. Ser378 was shown in red.

(D) Nocodazole-arrested mitotic cells were incubated in the absence or presence of the Plk1 inhibitor (BI 2536) and subjected to IP and immunoblot. Immunoprecipitates were incubated with or without λ phosphatase (PPase), and were analyzed by immunobloting with pS378 antibody.

(E) Cells were infected with the indicated constructs, synchronized by nocodazole, and released. Parkin phosphorylation at pS378 was then examined by immunoblot analysis.

(F) Cells were transfected with indicated plasmids, and Parkin phosphorylation at pS378 was examined.

(G) In vitro kinase assay of Parkin by Plk1. Parkin phosphorylation was visualized by pS378 Parkin antibody.

(H) Cells were transfected with the indicated plasmids, and then treated with nocodazole. Cell lysates were then blotted with the indicated antibodies.

(I) in vitro ubiquitination of Cyclin B1, Securin and Nek2A by WT Parkin and mutants (S65A, S65D, S378A and S378D). Purified bacteria-produced His-Cyclin B1, Securin and Nek2A protein was incubated with different components as indicated for 90 min at 30°C. Samples were analyzed by immunobloting with anti-Cyclin B1, Securin and Nek2A antibody.

(J and K) Cells were treated with chemical (J) or transfected with indicated constructs (K). Cells were then collected for IP-immunoblot analysis in the absence or presence of nocodazole.

See also Figure S5.

Previous studies suggest that Parkin activity is regulated by PINK1-mediated phosphorylation during mitophagy (Iguchi et al., 2013; Kane et al., 2014; Kondapalli et al., 2012), we therefore hypothesize that Parkin phosphorylation by Plk1 is also important for its function in mitosis. We found that mutation of S378 (S378A) abolished Parkin's effect toward Aurora A, Aurora B and Cyclin B1 (Figure 5H and data not shown). Mutating other phosphorylation sites mediated by Casein kinase- 1, protein kinase A, and protein kinase C did not affect Parkin's function (Yamamoto et al., 2005). Furthermore, we found that Parkin-mediated polyubiquitination of its mitotic substrates is abolished by the S378A mutation, while had no effect on CCCP-induced Tom20 ubiquitination (Figure 5I and S5B). Conversely, the S378D mutation, which mimics S378 phosphorylation, dramatically enhanced Parkin E3 ligase activity. On the other hand, mutating PINK1 phosphorylation site of Parkin (S65A) (Iguchi et al., 2013; Kane et al., 2014; Kondapalli et al., 2012), although abolished CCCP-induced Tom20 ubiquitination (Geisler et al., 2014), retained basal E3 ligase activity toward its mitotic substrates comparable to WT Parkin (Figure 5I, Figure S5B). The S65D mutant slightly increased Parkin E3 ligase activity toward mitotic substrates. However, it was not comparable to the dramatic increase caused by the S378D mutation.These results suggest that Plk1-mediatd phosphorylation of Parkin at S378 is another mode of Parkin activation and is important for its function in mitosis.

To further explore how Plk1-mediated phosphorylation affects Parkin function, we treated cells with BI 2536 and found that Plk1 inhibition resulted in decreased binding of Parkin to Cdc20 (Figure 5J). Furthermore, mutation of Ser 378 (S378A) abolished its interaction with Cdc20 during mitosis (Figure 5K). Therefore, S378 phosphorylation is important for Parkin's interaction with Cdc20.

Parkin Misregulation Is A Driving Event in Tumorigenesis

Cdh1 or Cdc20 substrates such as Plk1, Aurora A, Aurora B, Cyclin B1, and Securin are highly expressed in many types of tumors (Kim et al., 2011; Penas et al., 2011). However, very few mutations were found in APC/C subunits (Penas et al., 2011). On the other hand, Parkin was found to be mutated in several human cancers. Since Parkin has been identified as a candidate tumor suppressor and our results demonstrate Parkin's role in regulating mitosis, we hypothesized that Parkin has tumor suppressor function as a mitotic regulator. To further test this hypothesis, we examined the expression of Parkin substrates in cells expressing WT Parkin or cancer-derived Parkin mutants (Figure 6A). We chose three tumor-associated Parkin mutations (C360S, S378G and W453L) in cBioPortal (http://www.cbioportal.org/) for Cancer Genomics. C360 is located at the IBR Zinc region of Parkin, which is an important region to interact with Cdh1. Interestingly, S378, which is identified as a phosphorylation site by Plk1 in our study, is also mutated in cancers. The W453L mutation is found in both Parkinson's disease and cancer. We found that these cancer-derived mutations abolished Parkin E3 ligase activity and blocked the degradation of mitotic regulators, such as Cyclin B1 and Aurora B. We next determined Parkin expression level by immunohistochemical staining in 400 human lung specimens (normal and cancer) spotted on a tissue microarray (TMA; Figures S5C-S5E). Parkin expression is lower in NSCLC samples compared to lung normal next to its cancer, but not Plk1 (Figure S5C). Furthermore, we identified a negative correlation between Parkin and Plk1 expression (Figures S5C and S5F). Parkin KO MEFs showed more aneuploidy and polyploidy (Figures S6A and S6B). Furthermore, we found that WT MEFs became senescent when cultured in vitro, while Parkin KO MEFs readily escaped senescence and became transformed (Figures S6C-S6E). Parkin KO MEFs also became tumorigenic in vivo (Figure S6F). These results suggest that Parkin misregulation is a driving event in tumorigenesis.

Figure 6

Parkin Is A Key Mitotic Regulator Functioning as a Tumor Suppressor

(A) Cells were transfected with the indicated plasmids, and mitotic cells were analyzed by immunoblot for the indicated proteins.

(B) Schematic of the experiments.

(C) A549 cells stably transfected with doxycycline-inducible constructs encoding WT Parkin or mutant Parkin (S378A and S378D) were treated with doxycycline and subjected to IP and immunoblot as indicated (Top), in vivo ubiquitination (Bottom).

(D) Athymic nude mice were injected subcutaneously with A549 cells stably-transfected with vector or doxycycline-inducible Parkin constructs (WT, S378A and S378D). Two days after injection, doxycycline was administered in drinking water. Tumor growth was measured at the indicated times after injection. n=5 for each group. The image shows a representative mouse injected with the indicated cells (Top). Tumor volumes (mm^{) were measured at the indicated times after injection (Bottom). *,} p < 0.05, **, p < 0.01 and ***, p < 0.001 by two-way ANOVA.

(E) Cells were infected with the indicated constructs, were collected for FACS analysis.

See also Figure S6.

Parkin is a Key Mitotic Regulator Functioning as a Tumor Suppressor

We next examined whether the loss of Parkin contributes to the development of human tumors. As shown in Figure S5F, the expression of mitotic factors regulated by Parkin was much higher in all seven types of human lung cancer cell lines, while Parkin expression was low or lost in these lines in comparison to three lung normal cell lines. We prepared doxycycline-inducible expression vectors to express Parkin and Parkin mutant forms (S378A or S378D) in Parkin-low cells to study the role of Parkin in tumorigenesis (Figures 6B-6E). Induction of Parkin expression in A549 cells (Figure 6C) and other three lung cancer cells (Figure S6G) with doxycycline resulted in decreased Cyclin B1 levels without affecting Cyclin E levels (Figure 6C; Figure S6G and data not shown). In addition, Cyclin B1 became polyubiquitinated upon Parkin induction (Figure 6C). Induction of Parkin expression with doxycycline inhibited tumor growth (Figure 6D). Importantly, we found the S378D, but not the S378A mutant, had tumor suppressive function (Figure 6D). Furthermore we found that Parkin-depleted cells showed G2/M accumulation, indicating a mitotic defect (Figure 6E). These effects were rescued by reconstitution of WT Parkin and S378D mutant form but not S378A mutants. Similar results were obtained using Parkin KO MEFs (Figures S6H-S6JC). These results suggest that tumorigenicity was suppressed by Parkin expression.

We reasoned that the misregulation of mitotic regulators in Parkin-deficient cells might provide a valuable therapeutic target. As Plk1 is overexpressed in Parkin-deficient cells, we tested Plk1 inhibitor, BI 2536. We found that Parkin KO MEFs are more sensitive to BI 2536 than WT MEFs and BI 2536 inhibited transformation of Parkin KO MEFs (Figure 7A). Furthermore, we found that it is abolished by knockdown of Plk1 and BI 2536 treatment that Parkin depletion induced escaped senescence and transformation (Figure S7A). Transformed and down-regulation of senescence events in Parkin KO MEFs were reversed by expressing WT Parkin but not mutants C431S or S378A (Figure S7B). Similar results were obtained using Aurora A inhibitor, VX 680, or another Plk1 inhibitor, ON01910 in seven types of lung cancer cell lines (Figures S7C and S7D; data not shown). All of Parkin-deficient lung cancer cell lines but not lung normal fibroblast, WI-38 and IMR 90 cells, were significantly sensitive to Plk1 or Aurora A inhibition by BI 2536 or VX 680 (Figures S7E and S7F; data not shown). In addition, we observed excellent tumor inhibition with BI 2536 in vivo for tumors with Parkin-deficiency using xenograft models (Figure 7B).

Figure 7

Parkin deficient cells are more sensitive to Plk1 inhibitor

(A) Parkin WT or KO MEF cells were treated with increasing concentrations of BI 2536 for 3 days, fixed, and stained by 0.2% Crystal violet (Left). Results represent the means (±S.E.) of three experiments performed in triplicate. *, p < 0.05 and ***, p < 0.001 versus Parkin WT MEFs by one-way ANOVA (Right).

(B) Nude mice bearing Parkin WT (Left side) or KO (Right side) MEFs were treated i.v. for four cycles with either the vehicle control (indicated by closed black circles or closed red squares) or BI 2536 at a dose of 20 mg/kg twice weekly, n=10 per group. Mean transformed MEFs volumes for Parkin KO are shown. ***, p < 0.001 and ****, p < 0.0001 versus Parkin WT MEFs by two-way ANOVA.

(C) Schematic model.

See also Figure S7.

Our data presented here best fit a model in which ordered progression through mitosis is governed by two distinct E3 ligases, APC/C and Parkin, targeting mostly a common set of substrates for destruction through the shared use of Cdc20 and Cdh1 (Figure 7C). Our findings indicate that Parkin-deficiency results in overexpression of key mitotic regulators, aneuploidy, escaping from senescence, and cell transformation. Treating tumors with Parkin-deficiency with mitotic kinase inhibitors might be an effective therapeutic strategy.

Parkin Regulates Mitosis

To understand the role of Parkin in mitosis, we monitored mitotic chromosome movement using time-lapse microscopy in Parkin-depleted U2OS cells (Figure 1A; Figure S1A) and Parkin knockout (KO) mouse embryonic fibroblasts (MEFs; Figures S1B-S1F). This analysis revealed a broad spectrum of mitotic defects including chromosome misalignment, chromosome lagging, chromosome bridge formation, prometaphase-like arrest, anaphase and cytokinesis failure (Figure 1A; Figure S1F). In addition, progression from nuclear envelope breakdown (NEBD) to anaphase onset was significantly delayed in Parkin KO MEFs compared to wild type (WT) MEFs (Figures S1B and S1H), a defect that was reversed by exogenous expression of WT Parkin (Figures S1G and S1H). These results demonstrate that Parkin deficiency results in multiple mitotic defects.

Figure 1

Parkin Regulates Mitosis

(A) Time-lapse analysis of mitotic U2OS cells transfected with Control or Parkin siRNA. 50 cells were counted in each experiment. Top: Quantification of abnormal mitotic cells. *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus Control siRNA by one-way ANOVA. Bottom left: Parkin and β–actin expression were shown; Bottom right: Representative images of cells with indicated misaligned chromosome, lagging chromosome, and Chromosome bridge were shown. Scale bar, 10 μm.

(B) Cells were synchronized at the G1/S transition by double-thymidine block, and then released into a drug-free medium. Cell were harvested at indicated times and analyzed by immunobloting. p27 ^{serves as a G}₀-G₁ phase marker; Cyclin E, early S phase; Skp2, G₁-S; p-H3, mitosis.

(C) Subcellular localization of Parkin during each stage of the cell cycle. U2OS cells were stained with antibodies against Parkin (Red) and Plk1 (Green) and DNA (Blue) stained with DAPI. White arrows, centrosome; Pink arrowheads, midzone, midbody, or midring from anaphase to cytokinesis. Scale bar represents 20 μm.

(D) Immunoblot analysis of mitotic factors in primary Parkin WT and KO MEFs (Passage 5).

(E) Immunoblot analysis of mitotic factors in primary Parkin WT and KO MEFs after releasing from serum starvation (for 72 h) and nocodazole arrest (for 18 h).

See also Figure S1 and S2.

Next, we examined Parkin levels at different stages of the cell cycle. Cells arrested at the G1/S boundary by double thymidine block (DTB) showed high Parkin levels. Upon release, Parkin levels decreased as cells progressed through S phase, and then peaked from G2 until early G1, without corresponding changes in mRNA levels (Figure 1B; Figure S1I). Furthermore, we found that Parkin localized to centrosomes, midzone, and midbody in various cells types, including U2OS cells (Figure 1C; Figures S1J and S1K) and IMR-90 lung fibroblasts (PDL=33) (data not shown). These results suggest that Parkin might have a direct role in mitotic regulation. To examine how Parkin might regulate mitosis, we examined the expression of key mitotic regulators. Immunoblot analysis of asynchronous or mitotic lysates from Parkin WT and KO MEFs showed increased levels of Plk1, Aurora A, Aurora B, Cyclin B1, Cdc20, and UbcH10 (Figures 1D and E). Other key mitotic regulators, such as Mad1, Mad2, Bub1, BubR1 and Bub3 were not affected. Cyclin E, whose upregulation has been linked to genomic instability in Parkin-deficient cells (Veeriah et al., 2010), was also present at normal levels. Furthermore, Parkin-depleted cells showed aberrant localization and expression of Plk1, Cyclin B1, and Aurora B as examined by immunofluorescence (IF) and immunoblot (Figures S2A and S2B; data not shown), respectively. Mitotic defects and up-regulation of Plk1 and Cyclin B1 in Parkin-depleted cells were reversed by expressing WT Parkin but not C431S, which abolishes Parkin's E3 ligase activity (Figures S2C and S2D) (Iguchi et al., 2013; Riley et al., 2013). These results suggested that Parkin regulates mitosis by controlling the levels of particular mitotic regulators through its E3 ligase activity. PINK1 knockdown did not affect Plk1 and Cyclin B1 levels, suggesting that Parkin's role in mitotic regulation is PINK1-independent (Figure S2E), and thus distinct from Parkin's established role in mitophagy.

Parkin-Mediated Ubiquitination Is a Mitotic Regulator

We hypothesized that Parkin directly regulates the levels of mitotic regulators, such as Plk1 and Aurora B, through its E3 ligase activity (Shimura et al., 2000). Consistent with this, we found that endogenous Parkin interacts with Plk1, Cyclin B1, Aurora A, Aurora B, and Nek2A (Figure 2A). Furthermore, overexpression of WT Parkin but not the C431S mutant markedly decreased levels of these mitotic regulators, which could be prevented by MG132 pre-treatment (Figure 2B), supporting the idea that Parkin regulates the abundance of these mitotic regulators through the proteasome pathway. Immunoblot analysis of tissue lysates from Parkin WT and KO mice revealed that Plk1, Aurora B and Cyclin B1 protein levels are elevated in tissues lacking Parkin (Figure 2C). Importantly, overexpression of Parkin in cells increased the polyubiquitination of Plk1, Aurora B, Cyclin B1, Aurora A, Securin, Aurora B and Nek2A, but not c-Myc and Cyclin E, whose expression are not regulated by Parkin (Figure 2D). Furthermore, the C431S mutation abolished Parkin's E3 ligase activity toward its substrates. Early studies suggest that Parkin mediates K48- or K63-linked polyubiquitylation in brain (Moore et al., 2008; Olzmann et al., 2007; Youle and Narendra, 2011). Interestingly, Parkin mostly mediated K11–linked polyubiquitin-chains in Plk1 ubiquitination (Figure S2F). Collectively, these data indicate that Parkin regulates the levels of a subset of mitotic proteins through the ubiquitin-proteasome pathway.

Figure 2

Parkin-Mediated Regulates the Levels and Ubiquitination of Mitotic Regulators

(A) HEK 293T cells were synchronized by nocodazole for 18 h, and mitotic and ansynchronized cells were collected for immunoprecipitation (IP)-immunoblot analysis with control IgG, anti-Plk1, Cyclin B1 and Parkin antibodies.

(B) HEK 293T cells were transfected with the indicated plasmids, and then treated with MG132 or left untreated. Cell lysates were blotted with the indicated antibodies.

(C) The lung, liver, kidney, spleen tissues of Parkin WT and KO mice (n = 3 mice/genotype) were lysed, and cell lysates were blotted with the indicated antibodies.

(D) HEK 293T cells were transfected with the indicated constructs and arrested in mitosis with nocodazole for 18 h (Left). Cells were synchronized at the G1/S transition by double-thymidine block, and then released into a new medium (Right). Cells were then treated with MG132. Ubiquitinated proteins were pull down under denaturing conditions by Ni-NTA agarose and analyzed by immunoblot. c-Myc and Cyclin E were shown as negative controls.

See also Figure S3.

In experiments designed to identify the E2 ubiquitin ligase for Parkin, we did not observe an interaction between Parkin and UbcH10, the E2 for APC/C in mitosis (data not shown) (Castro et al., 2005; Peters, 2006). Instead, UbcH7 (also called Ube2L3), the E2 for Parkin in cellular processes other than mitosis (Shimura et al., 2000; Wenzel et al., 2011), was significantly elevated and interacted with Parkin in mitosis (Figures S3A and S3B) and accumulated at various mitotic structures, including centrosomes, midzone, and midbody, just like Parkin (Figure S3C). Importantly, UbcH7 depletion caused mitotic defects similar to Parkin depletion (Figures S3D and S3E), further supporting the idea that UbcH7 acts as an E2 ubiquitin ligase for Parkin in mitosis.

Parkin-Cdc20/Cdh1 Acts as a Mitotic-Regulating Complex

Parkin regulates mitotic factors, which are also regulated by APC/C, raising the possibility that Parkin interacts with APC/C or its subunits. We next examined the interaction between Parkin and the APC/C subunits (Figure 3A). Endogenous Parkin co-immunoprecipitated with Cdc20 and Cdh1 from mitotic cell lysates, but not with APC/C components APC11 and APC2. Furthermore, we found that recombinant Parkin interacted with Cdc20 and Cdh1 under cell-free conditions, suggesting that Parkin directly interacts with Cdc20/Cdh1 (Figure 3B). Use of synchronized cell lysates indicated that Parkin first interacts with Cdc20 and then switches to Cdh1 after cells exit mitosis (Figure 3C). Taken together, these data suggest that Parkin forms a complex with Cdc20 or Cdh1 that does not include the APC/C.

Figure 3

Parkin-Cdc20/Cdh1 Complex Is A Mitotic Regulator during the Cell Cycle

(A) HEK 293T cells were synchronized by nocodazole and treated with MG132. Cell lysates were then subjected to IP and immunoblot as indicated.

(B) Purified Cdc20 or Cdh1 were incubated with GST or GST-Parkin coupled to GSH-Sepharose. Proteins retained on Sepharose were then blotted with the indicated antibodies.

(C) HEK 293T cells transfected with Flag-tagged WT Parkin were synchronized by nocodazole treatment. Cells were released and subjected to IP and immunoblot with the indicated antibodies.

(D and E) Cells were transfected with the indicated constructs and treated as in Figure 3A. Cells were subjected to IP and immunoblot with the indicated antibodies. APC11 (D) and APC2 (D and E) were shown as negative controls.

(F) in vitro ubiquitination of Cyclin B1, Securin and Nek2A by Parkin and Cdc20/Cdh1. Purified bacteria-produced His-Cyclin B1, Securin and Nek2A protein was incubated with different components as indicated for 90 min at 30°C. Samples were analyzed by immunobloting with ubiquitin antibody.

See also Figure S3 and Table S1.

Cdc20 and Cdh1 act as substrate-recognition subunits of APC/C (Castro et al., 2005; Peters, 2006), which led to the idea that Parkin might also target specific mitotic substrates through Cdc20 and Cdh1. Knockdown of Cdc20 or Cdh1 resulted in decreased binding of Parkin to various mitotic substrates, including Cyclin B1 and Aurora B (Figure 3D and data not shown). In contrast, knockdown of APC11 did not affect these interactions and Parkin's interaction with Cdc20/Cdh1 (Figure 3E). Moreover, we found Cdc20- and Cdh1-specific degron sequences (D-box and KEN box motifs) (Castro et al., 2005; Nakayama and Nakayama, 2006) in a series of established Parkin substrates, including Ataxin 2 and 3, Synaptotagmin XI, RanBP2, β-catenin, PCDP2-1, α and β tubulin, LIM kinase, PLC-γ1, MFN1 and 2, Mitochondrial Rho GTPase isoform 1, Septin 4 and 5, and Drp1 (Table S1) (Walden and Martinez-Torres, 2012), the latter of which has been shown to require Cdh1 for ubiquitination (Horn et al., 2011; Wang et al., 2011). To further confirm the role of Cdc20 and Cdh1 in Parkin-mediated ubiquitination, we performed in vitro ubiquitination assays. Parkin induced ubiquitination of Cyclin B1, Securin and Nek2A; however, their ubiquitination were abolished in the absent of Cdc20/Cdh1, Ube1 (E1), UbcH7 (E2) or Parkin (Figure 3F; Figure S3F). Furthermore the D-box / KEN-box mutants of these substrates were not polyubiquitinated by Parkin. These findings further strengthen the notion that Parkin-Cdc20 and -Cdh1 complexes act independently of APC/C-Cdc20 and -Cdh1 in regulating the abundance of key mitotic regulators.

UbcH7-Parkin-Cdc20 and -Cdh1 Complexes Regulate Mitosis Independently of UbcH10-APC/C-Cdc20 and -Cdh1 Complexes

Next we examined the functional interaction between Parkin and APC/C. Inactivation of APC/C by APC11 knockdown resulted in chromosome missegregation defects and upregulation of Plk1 (Figures 4A and 4B). We found that ectopic expression of Parkin in APC11-deficient cells reversed these mitotic abnormalities (Figures 4A and 4B). In addition, Parkin overexpression restored Plk1 levels and rescued mitotic errors induced by UbcH10 (APC/C E2) knockdown, but had no effect on UbcH7 (Parkin's E2)-induced mitotic defects (Figure 4C). These studies suggest that the UbcH7-Parkin-Cdc20 and -Cdh1 complexes regulate mitosis independently of UbcH10-APC/C-Cdc20 and -Cdh1 complexes. Although Parkin and APC/C show many similarities in mitosis, there are some differences in their localization. As shown in Supplemental Figures S4A and S4B, Parkin is localized in the centrosome or midbody like Cdc20, while APC3 is localized in the kinetochores, or the midring in mitosis. Furthermore, we found that UbcH7-Parkin-Cdc20 has target proteins such as α and β tubulin that are not regulated by APC/C (Figure S4C).

Figure 4

UbcH7-Parkin-Cdc20 and -Cdh1 Complexes Regulate Mitosis Independently of UbcH10-APC/C-Cdc20 and -Cdh1 Complexes

(A-C) Live-cell imaging analysis of chromosome segregation defects in U2OS cells infected with the indicated constructs and synchronized by nocodazole treatment. Cells were fixed and stained with DAPI. Representative images of cells with indicated missegregation events were shown. Scale bar, 20 μm (A). Analysis of numerical chromosome segregation errors. 100 cells were counted in each experiment. *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus Control shRNA by one-way ANOVA (B-C, Left). Immunoblot analysis with indicated antibodies (B-C, Right).

(D) Cells were infected with the indicated constructs, synchronized by nocodazole, and released. Cyclin B1 expression was then examined by immunoblot analysis (Top). FACS analysis for cell cycle profile (Bottom).

(E) Fluorescence quantification of Cyclin B1-GFP by time-lapse imaging in mitotic H2B-mRFP-expressing U2OS cells infected with the indicated shRNAs. Cells were plotted against time before and after prometaphase (shake off). *, p < 0.05, **, p < 0.01 and ***, p < 0.001 versus control shRNA by two-way ANOVA.

(F) Representative images of cells as indicated. Note. The frames of live cell imaging were recorded by shake-off for mitotic cells. Trypsin-EDTA was the treatment for Interphase cells as the control. Scale bar, 20 μm.

(G) After infection with the indicated shRNAs, cells were synchronized at the mitosis transition (prometaphase) by nocodazole treatment for 18 h. After harvesting the mitotic cells by shake off, cells were re-cultured and dividing cells were examined at the indicated time points. The data represent the average of three experiments and 100 cells were monitored in each experiment. Scale bar, 20 μm.

See also Figure S4.

Since Parkin and APC/C share the same coactivator Cdc20, one prediction is that mitotic defects caused by depletion of APC/C or Parkin alone would be less severe than those caused by depletion of Cdc20 (Huang et al., 2009). To test this idea, we studied whether Parkin affects Cdc20-mediated degradation of Cyclin B1 at the metaphase-to-anaphase transition. Depleting APC11 or Parkin alone delayed Cyclin B1 degradation and mitotic exit, but did not recapitulate Cdc20 depletion (Figures 4D, 4E and 4G; Figures S4C-S4E). However, codepletion of APC11 and Parkin phenocopied Cdc20 depletion (Figures 4D, 4E, 4F and 4G; Figures S4D and S4E).

Parkin Is Phosphorylated and Activated by Plk1 upon Mitotic Entry

We next sought to identify mitosis-specific regulation of Parkin. Phosphorylation is a common posttranslational modification and has been shown to involve protein stability and activity. We therefore scanned Parkin using GPS2.12, a tool for prediction of kinase-specific phosphorylation sites (Xue et al., 2008), which identified Ser 378 as a potential phosphorylation site by Plk1. We found that Parkin was phosphorylated at Ser 378 in mitosis (Figures 5A and 5B). Treatment of carbonyl cyanide m-chlorophenylhydrazone (CCCP), a mitochondrial-uncoupling reagent that activates Parkin during mitophagy (Iguchi et al., 2013; Riley et al., 2013), did not affect Ser 378 phosphorylation (Figure 5B). Ser 378 is predicted to be a Plk1 phosphorylation site (http://gps.biocuckoo.org/) and its surrounding residues fit with a consensus Plk1 phosphorylation site (D/ExS/TΦ, Φ: hydrophobic residues). In addition, Ser 378 is highly conserved among vertebrates (Figure 5C), suggesting that the phosphorylation of this site may have an evolutionarily conserved role in regulating Parkin activity. To test whether Plk1 regulates Parkin S378 phosphorylation, we treated cells with BI 2536, a Plk1 inhibitor, or infected cells with Plk1 shRNA. We found that Plk1 inhibition or deficiency blocked Parkin phosphorylation at Ser378 (Figures 5D and 5E). Conversely, overexpression of Plk1 or constitutively active Plk1 (T210D) (van de Weerdt et al., 2005), but not inactive Plk1 (T210A), increased Parkin phosphorylation (Figure 5F). Furthermore, Plk1 was able to phosphorylate recombinant WT Parkin but not S378A (Figure 5G). Interestingly, Ser 378 localizes within the IBR domain. In previous studies, it is established that the IBR domain assists the recruitment of proteins involved in the ubiquitination pathway (Chung et al., 2001; Zhang et al., 2000). Structurally, the IBR domain helps a close arrangement of the RING1 and RING2 domains, which facilitates protein interactions and subsequent ubiquitination (Beasley et al., 2007) In addition, the region involves in maintaining conformational flexibility, and it can affect Parkin's activity and stability (Trempe et al., 2013). We also found that the IBR domain is involved in Parkin's interaction with Cdh1. As shown in Figure S5A, Cdh1 and Plk1 could interact with the C terminal region of Parkin containing the RING1-IBR or IBR-RING2 domain. The RING2 domain alone, but not the RING1 domain, could interact with Cdh1. There results suggest that the IBR and RING2 domain could interact with Cdh1.

Figure 5

Parkin Is Phosphorylated by Plk1 at Ser378 and Activated during Mitosis

(A) Cells were synchronized at the G1/S transition by double-thymidine block, and cells were released. Cell were harvested at indicated times and analyzed by immunobloting.

(B) Cells were incubated in the absence or presence of nocodazole or CCCP and subjected to immunoblot analysis with the indicated antibodies.

(C) Comparison of the sequences surrounding S378 of Parkin orthologues. Ser378 was shown in red.

(D) Nocodazole-arrested mitotic cells were incubated in the absence or presence of the Plk1 inhibitor (BI 2536) and subjected to IP and immunoblot. Immunoprecipitates were incubated with or without λ phosphatase (PPase), and were analyzed by immunobloting with pS378 antibody.

(E) Cells were infected with the indicated constructs, synchronized by nocodazole, and released. Parkin phosphorylation at pS378 was then examined by immunoblot analysis.

(F) Cells were transfected with indicated plasmids, and Parkin phosphorylation at pS378 was examined.

(G) In vitro kinase assay of Parkin by Plk1. Parkin phosphorylation was visualized by pS378 Parkin antibody.

(H) Cells were transfected with the indicated plasmids, and then treated with nocodazole. Cell lysates were then blotted with the indicated antibodies.

(I) in vitro ubiquitination of Cyclin B1, Securin and Nek2A by WT Parkin and mutants (S65A, S65D, S378A and S378D). Purified bacteria-produced His-Cyclin B1, Securin and Nek2A protein was incubated with different components as indicated for 90 min at 30°C. Samples were analyzed by immunobloting with anti-Cyclin B1, Securin and Nek2A antibody.

(J and K) Cells were treated with chemical (J) or transfected with indicated constructs (K). Cells were then collected for IP-immunoblot analysis in the absence or presence of nocodazole.

See also Figure S5.

Previous studies suggest that Parkin activity is regulated by PINK1-mediated phosphorylation during mitophagy (Iguchi et al., 2013; Kane et al., 2014; Kondapalli et al., 2012), we therefore hypothesize that Parkin phosphorylation by Plk1 is also important for its function in mitosis. We found that mutation of S378 (S378A) abolished Parkin's effect toward Aurora A, Aurora B and Cyclin B1 (Figure 5H and data not shown). Mutating other phosphorylation sites mediated by Casein kinase- 1, protein kinase A, and protein kinase C did not affect Parkin's function (Yamamoto et al., 2005). Furthermore, we found that Parkin-mediated polyubiquitination of its mitotic substrates is abolished by the S378A mutation, while had no effect on CCCP-induced Tom20 ubiquitination (Figure 5I and S5B). Conversely, the S378D mutation, which mimics S378 phosphorylation, dramatically enhanced Parkin E3 ligase activity. On the other hand, mutating PINK1 phosphorylation site of Parkin (S65A) (Iguchi et al., 2013; Kane et al., 2014; Kondapalli et al., 2012), although abolished CCCP-induced Tom20 ubiquitination (Geisler et al., 2014), retained basal E3 ligase activity toward its mitotic substrates comparable to WT Parkin (Figure 5I, Figure S5B). The S65D mutant slightly increased Parkin E3 ligase activity toward mitotic substrates. However, it was not comparable to the dramatic increase caused by the S378D mutation.These results suggest that Plk1-mediatd phosphorylation of Parkin at S378 is another mode of Parkin activation and is important for its function in mitosis.

To further explore how Plk1-mediated phosphorylation affects Parkin function, we treated cells with BI 2536 and found that Plk1 inhibition resulted in decreased binding of Parkin to Cdc20 (Figure 5J). Furthermore, mutation of Ser 378 (S378A) abolished its interaction with Cdc20 during mitosis (Figure 5K). Therefore, S378 phosphorylation is important for Parkin's interaction with Cdc20.

Parkin Misregulation Is A Driving Event in Tumorigenesis

Cdh1 or Cdc20 substrates such as Plk1, Aurora A, Aurora B, Cyclin B1, and Securin are highly expressed in many types of tumors (Kim et al., 2011; Penas et al., 2011). However, very few mutations were found in APC/C subunits (Penas et al., 2011). On the other hand, Parkin was found to be mutated in several human cancers. Since Parkin has been identified as a candidate tumor suppressor and our results demonstrate Parkin's role in regulating mitosis, we hypothesized that Parkin has tumor suppressor function as a mitotic regulator. To further test this hypothesis, we examined the expression of Parkin substrates in cells expressing WT Parkin or cancer-derived Parkin mutants (Figure 6A). We chose three tumor-associated Parkin mutations (C360S, S378G and W453L) in cBioPortal (http://www.cbioportal.org/) for Cancer Genomics. C360 is located at the IBR Zinc region of Parkin, which is an important region to interact with Cdh1. Interestingly, S378, which is identified as a phosphorylation site by Plk1 in our study, is also mutated in cancers. The W453L mutation is found in both Parkinson's disease and cancer. We found that these cancer-derived mutations abolished Parkin E3 ligase activity and blocked the degradation of mitotic regulators, such as Cyclin B1 and Aurora B. We next determined Parkin expression level by immunohistochemical staining in 400 human lung specimens (normal and cancer) spotted on a tissue microarray (TMA; Figures S5C-S5E). Parkin expression is lower in NSCLC samples compared to lung normal next to its cancer, but not Plk1 (Figure S5C). Furthermore, we identified a negative correlation between Parkin and Plk1 expression (Figures S5C and S5F). Parkin KO MEFs showed more aneuploidy and polyploidy (Figures S6A and S6B). Furthermore, we found that WT MEFs became senescent when cultured in vitro, while Parkin KO MEFs readily escaped senescence and became transformed (Figures S6C-S6E). Parkin KO MEFs also became tumorigenic in vivo (Figure S6F). These results suggest that Parkin misregulation is a driving event in tumorigenesis.

Figure 6

Parkin Is A Key Mitotic Regulator Functioning as a Tumor Suppressor

(A) Cells were transfected with the indicated plasmids, and mitotic cells were analyzed by immunoblot for the indicated proteins.

(B) Schematic of the experiments.

(C) A549 cells stably transfected with doxycycline-inducible constructs encoding WT Parkin or mutant Parkin (S378A and S378D) were treated with doxycycline and subjected to IP and immunoblot as indicated (Top), in vivo ubiquitination (Bottom).

(D) Athymic nude mice were injected subcutaneously with A549 cells stably-transfected with vector or doxycycline-inducible Parkin constructs (WT, S378A and S378D). Two days after injection, doxycycline was administered in drinking water. Tumor growth was measured at the indicated times after injection. n=5 for each group. The image shows a representative mouse injected with the indicated cells (Top). Tumor volumes (mm^{) were measured at the indicated times after injection (Bottom). *,} p < 0.05, **, p < 0.01 and ***, p < 0.001 by two-way ANOVA.

(E) Cells were infected with the indicated constructs, were collected for FACS analysis.

See also Figure S6.

Parkin is a Key Mitotic Regulator Functioning as a Tumor Suppressor

We next examined whether the loss of Parkin contributes to the development of human tumors. As shown in Figure S5F, the expression of mitotic factors regulated by Parkin was much higher in all seven types of human lung cancer cell lines, while Parkin expression was low or lost in these lines in comparison to three lung normal cell lines. We prepared doxycycline-inducible expression vectors to express Parkin and Parkin mutant forms (S378A or S378D) in Parkin-low cells to study the role of Parkin in tumorigenesis (Figures 6B-6E). Induction of Parkin expression in A549 cells (Figure 6C) and other three lung cancer cells (Figure S6G) with doxycycline resulted in decreased Cyclin B1 levels without affecting Cyclin E levels (Figure 6C; Figure S6G and data not shown). In addition, Cyclin B1 became polyubiquitinated upon Parkin induction (Figure 6C). Induction of Parkin expression with doxycycline inhibited tumor growth (Figure 6D). Importantly, we found the S378D, but not the S378A mutant, had tumor suppressive function (Figure 6D). Furthermore we found that Parkin-depleted cells showed G2/M accumulation, indicating a mitotic defect (Figure 6E). These effects were rescued by reconstitution of WT Parkin and S378D mutant form but not S378A mutants. Similar results were obtained using Parkin KO MEFs (Figures S6H-S6JC). These results suggest that tumorigenicity was suppressed by Parkin expression.

We reasoned that the misregulation of mitotic regulators in Parkin-deficient cells might provide a valuable therapeutic target. As Plk1 is overexpressed in Parkin-deficient cells, we tested Plk1 inhibitor, BI 2536. We found that Parkin KO MEFs are more sensitive to BI 2536 than WT MEFs and BI 2536 inhibited transformation of Parkin KO MEFs (Figure 7A). Furthermore, we found that it is abolished by knockdown of Plk1 and BI 2536 treatment that Parkin depletion induced escaped senescence and transformation (Figure S7A). Transformed and down-regulation of senescence events in Parkin KO MEFs were reversed by expressing WT Parkin but not mutants C431S or S378A (Figure S7B). Similar results were obtained using Aurora A inhibitor, VX 680, or another Plk1 inhibitor, ON01910 in seven types of lung cancer cell lines (Figures S7C and S7D; data not shown). All of Parkin-deficient lung cancer cell lines but not lung normal fibroblast, WI-38 and IMR 90 cells, were significantly sensitive to Plk1 or Aurora A inhibition by BI 2536 or VX 680 (Figures S7E and S7F; data not shown). In addition, we observed excellent tumor inhibition with BI 2536 in vivo for tumors with Parkin-deficiency using xenograft models (Figure 7B).

Figure 7

Parkin deficient cells are more sensitive to Plk1 inhibitor

(A) Parkin WT or KO MEF cells were treated with increasing concentrations of BI 2536 for 3 days, fixed, and stained by 0.2% Crystal violet (Left). Results represent the means (±S.E.) of three experiments performed in triplicate. *, p < 0.05 and ***, p < 0.001 versus Parkin WT MEFs by one-way ANOVA (Right).

(B) Nude mice bearing Parkin WT (Left side) or KO (Right side) MEFs were treated i.v. for four cycles with either the vehicle control (indicated by closed black circles or closed red squares) or BI 2536 at a dose of 20 mg/kg twice weekly, n=10 per group. Mean transformed MEFs volumes for Parkin KO are shown. ***, p < 0.001 and ****, p < 0.0001 versus Parkin WT MEFs by two-way ANOVA.

(C) Schematic model.

See also Figure S7.

Our data presented here best fit a model in which ordered progression through mitosis is governed by two distinct E3 ligases, APC/C and Parkin, targeting mostly a common set of substrates for destruction through the shared use of Cdc20 and Cdh1 (Figure 7C). Our findings indicate that Parkin-deficiency results in overexpression of key mitotic regulators, aneuploidy, escaping from senescence, and cell transformation. Treating tumors with Parkin-deficiency with mitotic kinase inhibitors might be an effective therapeutic strategy.

DISCUSSION

In this study, we have identified an unexpected function of Parkin in regulating mitotic progression through Cdc20/Cdh1. We found that Parkin interacts with Cdc20 and Cdh1, APC/C co-activators to mediate the degradation of several key mitotic regulators including Plk1, Nek2A, Cyclin B1, Securin, and Aurora A and B. Importantly, Parkin's interaction with Cdc20/Cdh1 is independent of APC/C. These studies identified a previously unknown ubiquitin ligase complex that plays an important role in mitosis.

The APC/C is a multi-subunit member of the RING finger family of E3 ubiquitin ligases for mitosis. During mitosis, APC/C mediates ubiquitination of many target substrates including Plk1, Aurora A and Aurora B, Cyclin A and Cyclin B, Survivin, Nek2A, and Securin through two adaptor proteins or coactivators, Cdc20 and Cdh1. Although APC/C is established to play important functions during mitosis, the role of Parkin in mitosis is just emerging. Our data suggest that Parkin functions in parallel with APC/C in regulating mitosis in our experimental systems. Previous studies suggested that loss of one of the APC/C subunit APC11, the catalytic subunit that recruits the E2 protein, showed delayed but not blocked mitosis (Izawa and Pines, 2011). Our study might explain the lack of total blockade of mitosis following APC11 knockdown. We propose that Parkin might compensate some of APC/C function. Consistent with this, we found that combination knockdown of Parkin and APC11 can block mitosis, just like cells depleted of Cdc20. The use of multiple E3 ligase complexes has been found in other cellular processes. For example, Skp2 and Fbw7 sharing the SCF complex to regulate the same target proteins, such as Myc and Cyclin E in Interphase (Nakayama and Nakayama, 2006). We propose that Parkin and APC/C may also function in a similar fashion.

Although Parkin and APC/C function together in cell lines we tested, germ line knockout (KO) mice of the core subunits of APC/C ubiquitin E3 ligase complex (Apc2 and APC10/Doc1) are embryonic lethal (Zhang et al., 2014), indicating that APC/C is essential for embryogenesis and development and that Parkin could not compensate for APC/C function in these cellular processes. On the other hand, Parkin KO mice do not show early lethality. Therefore, APC/C is a critical mitotic regulator; Parkin might collaborate with APC/C or compensate some APC/C function, at least in a subset of cell types. Although Parkin KO is not embryonic lethal, it does not mean that Parkin is not important. We observe chromosome instability in Parkin-depleted cells as well as cells Parkin KO mice. We also observed delay mitotic progression in these cells. These results suggest that even though Parkin is not essential for mitosis, it is important for optimal mitosis. Knockout mice of some other important regulators are also viable, such as Cks, Id-1, Sgo-1, Cdks (2, 3, 4, 6), and Skp2 (Berthet et al., 2003; Diril et al., 2012; Humbert et al., 2004; Nakayama et al., 2000; Watanabe and Kitajima, 2005; Yan et al., 1997).

Previous studies suggested that loss of Parkin activity up-regulates Cyclin E, thereby perturbing DNA replication and abnormal chromosome segregation (Veeriah et al., 2010). However, we failed to observe Cyclin E level changes in Parkin KO cells. Another study also failed to observe a regulation of Cyclin E by Parkin (Yeo et al., 2012). Our results suggest that Parkin has a more direct role in mitosis through its interaction with Cdc20/Cdh1 and regulation of key mitotic regulators through protein ubiquitination. Consistent with this, our results show that Parkin knockdown causes multiple abnormal mitotic phenotypes, including lagging chromosomes, chromosome misalignment, and prometaphase-like arrest.

In summary, we demonstrated that Parkin regulates mitosis regulators using APC/C co-activators, Cdc20 and Cdh1 and the Parkin-Cdc20/Cdh1 complex functions in parallel with the APC/C complex in mitotic regulation. Loss of Parkin causes increased levels of mitosis regulators that may contribute to various forms of chromosome instability including chromosome misalignment, chromosome lagging, centrosome amplification, aneuploidy, cytokinesis defects, and cause tumor formation. In addition, Parkin expression is reduced in several human malignancies including lung and brain cancers (Cesari et al., 2003; Picchio et al., 2004; Veeriah et al., 2010; Yeo et al., 2012). Finally, we found potential targeted therapies for cancers with low Parkin expression. Our results greatly increase current understanding of Parkin function in proliferating cells.

EXPERIMENTAL PROCEDURES

Mouse Strains and MEFs, Cells and Cell Lines and Reagents and Plasmids

Mouse strains and MEFs, cells and cell lines and reagents and plasmids are detailed in the Supplemental Experimental Procedures.

Time-Lapse Live Microscopy, Cell Synchronizations, FACS analysis, Gene Silencing by siRNAs and Lentiviral shRNAs, in vivo Kinase Assays and Reverse Transcription (RT)-PCR of cDNA

Time-lapse live microscopy, cell synchronization, FACS analysis, gene silencing by siRNAs and lentiviral shRNAs, in vivo kinase assays and RT-PCR of cDNA are detailed in the Supplemental Experimental Procedures.

Co-immunoprecipitation, Immunobloting, Antibodies, in vivo and in vitro Ubiquitination Assays, Immunofluorescence, Immunohistochemistry, Protein Purification and in vitro Binding Assay

For immunoprecipitation, extraction of proteins with a modified buffer from cultured cells was followed by immunoprecipitation and immunobloting with corresponding antibodies. For removing heavy chain, light-chain-specific anti-mouse and anti-rabbit IgG secondary antibodies were obtained from Jackson Immunoresearch. For in vivo ubiquitination, cells were lysed by Urea lysis buffer (8 M Urea, 0.1 M Na₂HPO₄, 0.1 M Tris / HCl (pH 8.0), 0.05 % Tween 20 and 0.01 M imidazole). After centrifugation, the supernatants were collected and incubated with 20 ml Ni-NTA agarose beads (Qiagen) for four hours at 4°C. The precipitates were washed three times with Urea wash buffer (8 M Urea, 0.1 M Na₂HPO₄, 0.1 M Tris / HCl (pH 8.0), 0.05 % Tween 20 and 0.02 M imidazole) and Native wash buffer (0.1 M Na₂HPO₄, 0.1 M Tris / HCl (pH 8.0), 0.05 % Tween 20 and 0.02 M imidazole), and were boiled with SDS loading buffer, and then subjected to SDS-PAGE followed by immunoblot analysis. Additional details for Co-immunoprecipitation, immunobloting, antibodies, in vivo and in vitro ubiquitination assays, Immunofluorescence, Immunohistochemistry, Protein purification and in vitro binding assay are provided in Supplemental Experimental Procedures.

Colony Formation or Foci assay, Senescence-Associated β-galactosidase (Gal) Staining, Chromosome Spreading and Centrosome Staining Assays, Doxycycline-Inducible

Parkin Tet-On A549 Cell Lines, Mouse Xenograft Tumor Model and Statistical Analysis

Colony formation or foci assay, SA β-gal Staining, chromosome spreading and centrosome staining assays, doxycycline-inducible Parkin Tet-On A549 cell lines, mouse xenograft tumor model and Statistical analysis are detailed in the Supplemental Experimental Procedures.

Supplementary Material

01

ACKNOWLEDGMENTS