Plk4 Promotes Cancer Invasion and Metastasis through Arp2/3 Complex Regulation of the Actin Cytoskeleton

Polo like kinase 4 (Plk4) is a serine threonine kinase that localizes to the centriole throughout the cell cycle, and is essential for centriole duplication (1–4). Increased expression of Plk4 is described in a variety of common human epithelial malignancies, and is associated with inferior survival in several large cohorts of breast cancer patients (5–7). In particular, high levels of Plk4 mRNA are found in the triple-negative human breast cancers that are resistant to conventional systemic therapy, stimulating interest in Plk4 as a therapeutic target (8). Upregulation of Plk4 expression independently induces aneuploidy, loss of cell polarity, and hyperplasia in some nontransformed cell lines and proliferative tissues (9, 10). In the context of p53 dysfunction, increased Plk4 expression contributes to aneuploidy and tumorigenesis (10, 11). Yet, Plk4 also works as a haploinsufficient tumor suppressor in some contexts: Plk4^+/− mice are predisposed to form multipolar spindles and mitotic irregularities, aneuploidy, and ultimately hepatocellular carcinoma, whereas loss-of-heterozygosity at the Plk4 locus is seen in ∼50% of human hepatocellular carcinomas (12, 13). In sum, these results highlight the critical importance of tight regulation of Plk4 expression and activity, and the need to better understand its function(s).

Mechanistically, the oncogenic properties of overexpressed Plk4 have been attributed to the centriole amplification associated with its upregulation (9, 14, 15). Human cancers characteristically harbor supernumerary centrosomes (16–19), although the pathogenic role of this phenomenon has long been debated (20, 21). However, controlling centriole duplication may not be the only function of Plk4 relevant to tumorigenicity. In Plk4^+/− MEF lines that spontaneously acquire tumorigenicity over serial passaging, we reported a seemingly paradoxical association with changes in transcripts predictive of impaired cellular motility (22). Moreover, the Plk4^+/− MEFs displayed impaired migration, and siPlk4 suppressed spreading and invasion in Plk4^+/+ MEFs (22). Here we demonstrate that invasive and metastatic progression of human breast cancer line MDA-MB-231 in murine xenografts is dependent on Plk4 expression. An unbiased BioID screen for Plk4 interactors identified the actin-regulating complex Arp2/3 component Arp2 as a potential mediator of Plk4-induced motility. Further analysis revealed Polo-box 1-Polo-box 2–dependent binding of Plk4 to Arp2, phosphorylation of Arp2 at T237/T238 by Plk4, and dependence of Plk4-driven cell migration on Arp2.

Cells were grown at 37°C in DMEM (HeLa, MDA-MB-231, HEK293T), RPMI1640 (MDA-MB-435), or McCoy 5A medium (U2OS) supplemented with 10% FBS. MDA-MB-231 HTB-26 and HeLa cell lines were obtained from ATCC in 2013 and 2009, respectively. HEK293T and MDA-MB-435 cell lines were a kind gift from the Tony Pawson laboratory (obtained in 2010; Lunenfeld Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, Canada), and U2OS cells were a kind gift from the Laurence Pelletier laboratory (obtained in 2013; Lunenfeld Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, Canada). These cell lines were not further authenticated in our laboratory. Transient transfection was performed using Lipofectamine2000 or LipofectamineRNAiMAX (Invitrogen) according to manufacturer's instructions; a pool of four siRNAs and the individual constructs were utilized (Dharmacon; Supplementary Table S1).

Stable cell lines were generated as Flp-In HeLa or U2OS T-REx cell pools or clones. The jetPRIME DNA transfection reagent (Polyplus) was used to transfect cells grown in 6-well plates with 300 ng DNA and 3 μg pOG44 overnight, as per manufacturer's instructions. Transfected cells were selected using 200 μg/mL hygromycin (Multicell, 450141XL). HeLa, MDA-MB-231, or MDA-MB-435 cells expressing two Plk4 (SHCLNG-NM_014264; Sigma), luciferase or RFP (Sigma) shRNAs were generated through lentiviral infection. Lentiviruses were produced as described (23), and used to infect cells for 24 hours, followed by puromycin (1 mg/mL: HeLa and MDA-MB-231, 5 mg/mL: MDA-MB-435) mediated drug selection.

RNA was isolated using the RNeasy Mini Kit (Qiagen), treated with RNase-free DNase (Invitrogen), and reverse transcribed with SuperScript II Reverse Transcriptase (18064-014; Invitrogen) using Random Primers (48190-011; Invitrogen) according to manufacturer's instructions. Real-time RT-PCR was performed using SYBR Green PCR Master Mix (Applied Biosystems) on an ABI 7900HT apparatus. All quantifications were normalized to control endogenous GAPDH or RPII (Primers; Supplementary Table S2). Data generated by PCR software (SDS2.2.2; Applied Biosystems) were analyzed using the 2^−ΔΔC_t method (24).

HeLa cells transfected with siRNA for 48 hours or with FLAG-Plk4-wild-type (WT) or FLAG were serum-starved for 8 hours, then 1 × 10⁶ cells were plated on 1-μm pore size filters (PET membrane six-well insert, Falcon) in serum-free media, with DMEM + 20% FBS in the bottom chamber. Pseudopod isolation was performed as described previously (22)

Scratch wound migration, Golgi positioning, and RTCA transwell invasion assays were performed as described in ref. 22.

Cells were fixed, permeabilized, and blocked using 4% paraformaldehyde for 10 minutes, 0.5% Nonidet-P40 (Bioshop) for 20 minutes, and 10% FBS/PBS for 1 hour. Antibody incubations were in the blocking solution for 1 hour, and slides were mounted in Immuno-mount medium (Thermo Fisher Scientific). Immunofluorescence images were collected using the Olympia Deconvolution fluorescence microscope and softWoRx software (Applied Precision). Images were collected using a 100× or 60 × 1.4 NA oil objective (Olympus).

HeLa cells were seeded onto 6-well plates (150,000 cells/well) with a glass coverslip or 96-well plates (2,500 cells/well). At 3 hours, cells were fixed and immunostained, and 96-well plates were scanned and images acquired on the INCell Analyzer 6000, equipped with Nikon Plan Fluor 10×/NA 0.45 objective and 2,048 × 2,048 sCMOS camera. Automated cell size measurements were performed using a custom image analysis routine (Columbus2.3, PerkinElmer). Hoechst-stained nuclei were detected, followed by segmentation of cell borders based on F-actin signal, and determination of cell area.

All protocols were approved by the Toronto Center for Phenogenomics (TCP) Animal Care Committee. A total of 1.0 × 10⁶ MDA-MB-231 RFP or Plk4 shRNA cells were injected subcutaneously in the right flank of 5-week-old NOD SCID/J1303 mice (Jackson Laboratory). Tumor growth was monitored by palpation, size was measured with calipers, and volume calculated assuming an ellipsoid shape. Mice were sacrificed at 4- to 6-week or 7- to 10-week postinjection. Tumors were harvested and split for fixation in 10% buffered formalin phosphate (SF100-20; Thermo Fisher Scientific), or frozen in liquid nitrogen for RNA extraction. The right lung lobes, left liver lobe, and intraabdominal metastases were harvested and fixed for histology.

Fixed tissue was paraffin-embedded, sectioned (4 μm), deparaffinized in xylene, and rehydrated in graded ethanol. For tumors, serial coronal sectioning was performed at four levels, 200 μm apart. Samples were stained with hematoxylin and eosin (H&E) or the indicated antibodies (Supplementary Methods) according to standard protocols using the Veristain Gemini Automated Slide Stainer (Thermo Scientific) at the TCP Pathology Core, and examined with a Leica DMR upright microscope by a Royal College–certified pathologist in a blinded manner.

Images were taken at ×20 magnification by light microscopy, and the number and size of lung micrometastases determined using ImageJ. Human-vimentin–positive colonies with an area ≥500 μm² were considered metastases. The metastatic burden was expressed as the number and average size of metastasis present in two random sections in a defined area of 1.73 mm × 975 μm. For Ki67, two peripheral tumor sections were examined at ×40. Nuclear staining of Ki67 was considered positive and percent-positive nuclei quantified using ImageJ.

Tissue was placed in Ambion RNAlater-ICE solution overnight at −20°C, then disrupted and homogenized in RLT buffer (RNeasy Mini Kit, Qiagen) supplemented with β-mercaptoethanol using a rotor-stator homogenizer. Lysates were centrifuged for 10 minutes (4°C; 14,000 rpm) and the supernatant used for RNA purification using the RNeasy Mini Kit following manufacturer's protocol.

BioID purification and mass spectrometry data acquisition and analysis were performed as described in ref. 25. For details, see Supplementary Methods.

HEK293T cells transfected using PEI transfection reagent (Sigma) for 24 hours were lysed using TNTE lysis buffer (2 mmol/L Tris-HCl, pH 7.5, 120 mmol/L NaCl, 1% Triton X-100, 1 mmol/L EDTA), with protease inhibitor cocktail, 5 mmol/L NaF and 2 mmol/L NaOva. Beads were prewashed and blocked with 5% BSA. Extracts were centrifuged (14,000 rpm) for 10 minutes, and supernatants immunoprecipitated with anti-FLAG M2 affinity gel (Sigma; A2220) for 1.5 hours, or incubated with rabbit polyclonal anti-mCherry antibody (Abcam) for 1 hour followed by immunoprecipitation with ProteinG Sepharose beads (GE Healthcare) for 30 minutes. Beads were washed six times with lysis buffer supplemented with 500 mmol/L NaCl, boiled in Laemmli sample buffer, and analyzed using SDS-PAGE followed by immunoblotting.

Protein samples were separated by SDS-PAGE, transferred onto PVDF membranes, blocked with 5% milk-PBS-0.1%Tween, probed with primary antibodies at 4°C overnight, then HRP-linked secondary antibodies (GE Healthcare), and detected using SuperSignal West Femto Maximum Sensitivity Substrate (34095; Thermo Fisher Scientific). Band intensity was quantified using ImageJ. For phospho-Arp2 T237+T238 detection, cells were pretreated with 5.0 μmol/L pervanadate for 10 minutes, and following SDS-PAGE the PVDF membrane was blocked with, and primary antibody incubated in, 5% BSA-Tris buffered saline-0.1%Tween.

HeLa Plk4 and luciferase shRNA cells, and U2OS T-Rex YFP-Plk4 cells, were lysed in lysis buffer (25 mmol/L HEPES, 150 mmol/L NaCl, 1% NP40, 10% glycerol, 10 mmol/L MgCl₂, 1 mmol/L EDTA, 1 mmol/L orthovanadate, 1 mmol/L PMSF, 25 mmol/L NaF, protease inhibitor cocktail), and extracts incubated with 10 μL PAK-PBD beads for 45 minutes. After washing, beads were resuspended in Laemmli buffer and processed for immunoblotting.

HEK293T cells were transfected, lysed using TNTE lysis buffer with protease inhibitor cocktail, 5 mmol/L NaF and 2 mmol/L NaOva, extracts centrifuged (14,000 rpm) for 10 minutes, and supernatants immunoprecipitated with anti-FLAG M2 affinity beads. For FLAG-Plk4, the beads were washed three times with lysis buffer + 500 mmol/L NaCl and twice with kinase buffer (25 mmol/L Tris-HCl, pH 7.5, 25 mmol/L MgCl₂, 15 mmol/L sodium glycerolphosphate, 0.5 mmol/L NaOva, 2 mmol/L EDTA, 25 mmol/L NaF, 1 mmol/L DTT, 1.25 μg BSA). For FLAG-Arp2, the beads were washed three times with lysis buffer + 500 mmol/L NaCl, and protein eluted with 15 μg 3×FLAG-Peptide (APExBIO, A6001) with gentle mixing for 30 minutes. FLAG-Plk4 was then incubated with 3 μg Arp2 protein in 30 μL kinase buffer containing 100 μmol/L ATP with 10 μCi [γ-³³P]ATP. Kinase reactions were performed at 30°C for 30 minutes and terminated by adding Laemmli sample buffer. Proteins were separated by SDS-PAGE, stained with Colloidal Blue (Invitrogen), and dried using Bio-Rad gel dryer. Phosphorylation was visualized by autoradiography (Typhoon FLA9500, GE Healthcare).

CK-666 (SML0006, Sigma-Aldrich) and centrinone B (5690, TOCRIS Bioscience) were dissolved in DMSO and used at final concentrations of 50 μmol/L or 1,000 nmol/L, respectively.

Depletion of Plk4 to approximately 35% of control using shRNA markedly suppressed scratch wound healing by MDA-MB-231 breast cancer cells, without affecting viability or proliferation (Fig. 1A; Supplementary Fig. S1A and S1B). Similar impairment of directional migration was seen in HeLa cells, whether acutely or stably depleted of Plk4 (Supplementary Figs. S1C–S1H S2A, and S2B), and in MDA-MB-435 cells depleted of Plk4 (Supplementary Figs. S1I–S1K and S2C). The highly selective inhibitor of Plk4 kinase activity centrinone B (26) suppressed wound healing by MDA-MB-231 cells (Fig. 1B). Modest elevation of Plk4 levels using a stable inducible construct promoted migration of HeLa and U2OS cells (Supplementary Figs. S3A–S3D and S4A–S4D).

Invasion across a Matrigel-coated transwell membrane was impaired in Plk4-depleted MDA-MB-231 and HeLa cells (Fig. 1C; Supplementary Fig. S5A and S5B), and the impairment could be partially rescued by transfection of siPlk4-treated cells with nondegradable mutant Plk4 (Supplementary Fig. S5C). Plk4 overexpression via a tetracycline-inducible construct enhanced transwell migration in DLD-1 and HeLa cells (Fig. 1D; ref. 27). In a different system that measured invasiveness by quantifying protein in the protrusions of cells invading through a porous mesh, depletion and elevation of Plk4 levels similarly suppressed and stimulated, respectively, protrusion formation by HeLa cells (Fig. 1E).

We noted a striking alteration in the morphology of cultured HeLa cells stably depleted of Plk4, which appeared less fibroblastic and more clustered than control cells, consistent with a shift from a classic mesenchymal to a more epithelial phenotype (Fig. 2A). Indeed, Plk4-depleted cells gained expression of the epithelial marker E-cadherin, and lost expression of fibronectin and other mesenchymal markers (Fig. 2A).

Exposure of HeLa cells to increased Plk4 levels for 48 hours invoked a marked phenotypic change characterized by an exaggerated spindle morphology with extended protrusions (Fig. 2B). In an assay of spreading in response to the stimulus of replating, Plk4-transfected HeLa cells formed more protrusions and attained a greater cell area than did FLAG-transfected controls, whereas transfection with kinase-dead Plk4 had no effect on morphology or spreading (Fig. 2B and C), demonstrating the dependence of these phenotypes on kinase activity. Cells that were stably depleted of Plk4 were notably deficient in spreading, a defect that was partially rescued by transient transfection with FLAG-Plk4 (Fig. 2D); acute depletion by siRNA had a similar effect on spreading and could be rescued by nondegradable Plk4 mutants (Supplementary Fig. S5D and S5E). Similar suppression of cell spreading was observed in shPlk4-treated MDA-MB-231 cells (not shown).

We used a flank xenograft model in NOD SCID mice to investigate the effect of Plk4 on tumor progression in vivo. The effect of shPlk4 infection on Plk4 expression and invasion was confirmed to be stable over long-term culture of MDA-MB-231 human breast cancer cells (Supplementary Fig. S6a). In vivo tumor growth was mildly suppressed in Plk4-depleted versus control cells, beginning at approximately 5 weeks after subcutaneous injection (Fig. 3A). Plk4 expression in vivo was reduced to 25% to 45% of RFP control, and this reduction was durable up to at least 10 weeks, for each of the two shPlk4 constructs (Fig. 3B). Of note, the expression of Plks1-3 was comparable to shRFP controls (Supplementary Fig. S6b). Immunohistochemical staining for Ki67 revealed similar rates of cellular proliferation in Plk4-depleted versus RFP control tumors (Fig. 3C and D). H&E staining of the flank tumor nodules at 4 to 6 weeks after flank injection showed the presence of cancer invadopods reaching into the underlying skeletal muscle in RFP control but not shPlk4 tumors (Fig. 3E). This difference in invasive capacity became grossly apparent by 7 to 10 weeks, at which point RFP control cancers had penetrated into the peritoneal cavity and caused peritoneal carcinomatosis, whereas Plk4-depleted tumor nodules remained superficial to the abdominal wall musculature (Fig. 3F). Thus, although the effect of Plk4 on primary tumor growth appeared relatively minor, at least for the level of depletion achieved here, its effect on invasion into surrounding tissues was more profound.

We examined the lungs of mice with size-matched primary tumors (Supplementary Fig. S6C) at the time of sacrifice (7–8 weeks postinjection for RFP controls, 9–10 weeks for shPlk4), using both H&E and vimentin staining to detect, enumerate, and measure metastases. Lung metastases were significantly more numerous and larger in size in the controls than in the mice with Plk4-depleted tumors (Fig. 4A and B). Overall, at approximately 5 weeks postinjection, we noted evidence of locally aggressive behavior in 11 of 14 (79%) controls versus only 1 of 16 (6%) shPlk4 primary tumors (Fig. 4C). At 7 to 10 weeks, peritoneal invasion and lung metastases were seen in 100% of control mice, but only 7% and 29%, respectively, of Plk4-depleted tumors (Fig. 4C). As with cells growing in culture, Plk4-depleted tumors displayed a more “epithelial” gene expression pattern than did control tumors (Fig. 4D). Taken together with differences in motility gene expression observed between early passage Plk4^+/− and Plk4^+/+ MEFs (22), this is suggestive of transcriptional regulation downstream of Plk4 dosage.

Increased Rac1 activation at the leading edge is a hallmark of migrating cells, and forced expression of constitutively active Rac1 stimulates migration in several cell types. In many instances, Rac1 activation promotes cell spreading. Manipulation of Plk4 regulates Rac1 (Fig. 5A and ref. 14). The stimulatory effect of Plk4 on mammary gland invasiveness in vitro was previously attributed to activation of Rac1 by Plk4-induced centriole amplification (14). We therefore interrogated the role of Rac1 in mediating the enhancement of cancer cell migration by Plk4. As expected, transfection of U2OS cells with dominant negative Rac1T17N slowed migration (Fig. 5b). However, although Rac1T17N suppressed wound healing in the absence of tet-induced Plk4, stimulation of U2OS migration by Plk4 was not affected by the presence of Rac1T17N, suggesting mediation by an alternative mechanism.

To query the potential relationship between centriole amplification and altered motility, we assessed the effects of Plk4 overexpression or knockdown on centriole number in the cell lines used for the above functional studies (Supplementary Figs. S3C and S3D, S4C and S4D, S7A–S7H,; Supplementary Table S3). The direction of the change in centriole number generally correlated with the direction of change in motility, as illustrated for migration, but the magnitude of the changes was often disproportionate (Supplementary Figs. S3 and S4 and Supplementary Table S3). For instance, induction of Plk4 expression in U2OS cells using 0.01 μg/mL tetracycline had minimal impact on centriole number, but significantly increased wound healing, whereas higher levels of Plk4 were required to increase centriole number. There were similar findings with respect to centriole number and invasion (Fig. 1C; Supplementary Figs. S5A and S5B and S7A–S7F).

Directional migration toward a scratch wound is normally associated with the development of a polarized cell morphology, with the Golgi apparatus positioned on the side of the nucleus closest to the wound and actin filaments oriented toward the wound, as observed in siLuciferase-treated HeLa cells (Fig. 5C and D). Golgi and actin filaments failed to reposition normally in HeLa cells depleted of Plk4. Although polarity is a Cdc42-dependent phenotype (28), and although manipulation of Plk4 did indeed result in altered levels of active Cdc42 (Fig. 5E), dominant negative Cdc42T17N had no effect on directional migration in U2OS cells (Fig. 5F). Thus, although Plk4 regulates Rac1 and Cdc42 activation, it appeared likely that Plk4 regulates an additional component of the cell polarity and migration molecular machinery.

In an unbiased screen searching for candidate Plk4-interacting proteins by BioID in HEK293 cells, we identified several members of the Arp2/3 complex (Fig. 6A; Supplementary Fig. S8A; Supplementary Table S4), a seven-subunit conserved complex that binds existing actin filaments and initiates branching (reviewed in ref. 29). Dynamic actin polymerization and branching are critical for the coordinated formation of lamellopodia required for cell migration. We interrogated a panel of Arp2/3 complex proteins in reciprocal coimmunoprecipitation experiments with Plk4 (Supplementary Fig. S8B). Under stringent conditions, Plk4 associated specifically with Arp2, an interaction for which the Polo-box 1 and 2 domain was necessary and sufficient (Fig. 6B and C; Supplementary Fig. S8C). In HeLa cells migrating toward a scratch wound, RFP-Arp2 and FLAG-Plk4 localized similarly to the lamellopodial front, as did endogenous Arp2 and FLAG-Plk4 (Fig. 6D; Supplementary Fig. S8E). In contrast, the unrelated Rho GEF PLEKHG6 distributed widely in the cytoplasm, but did not concentrate with Plk4 at the leading edge (Supplementary Fig. S8E). Arp2 contains a consensus sequence for phosphorylation by Plk4, centered at T237/T238, which has recently been identified as an activation site (Supplementary Fig. S8D; refs. 30, 31). Using a phosphospecific antibody, we found that Arp2 phosphorylation at T237/T238 was significantly greater in HeLa cells transfected with wild-type Plk4, than cells transfected with kinase-dead Plk4 (Fig. 6E, top; Supplementary Fig. S9A), and was reduced in Plk4-depleted versus control cells (Fig. 6E, bottom), in keeping with the possible phosphorylation of Arp2 by Plk4 at this site in cellulo. An in vitro kinase assay showed dose-dependent phosphorylation of wild-type Arp2 in the presence of wild-type Plk4 (Fig. 6F), whereas T237/238A mutant Arp2 was not phosphorylated, suggesting that this is a bona fide Plk4 phosphorylation site (Fig. 6F).

As expected, depletion of Arp2 by siArp2 suppressed spreading of HeLa cells (pool of 4, and 3 of 4 individual constructs), as well as directional migration (Fig. 7A; Supplementary Fig. S9B and S9D). The Arp2 inhibitor CK-666 also impaired spreading and migration (Fig. 7A and B; Supplementary Fig. S9C). In contrast, depletion of Arp2 had no effect on centriole number, either in otherwise untreated cells or with Plk4-induced centriole amplification (Supplementary Fig. S9E), demonstrating separation of the two Plk4 functions. Arp2 did not localize to the leading edge in Plk4-depleted HeLa cells engaged in wound healing (Supplementary Fig. S9F). HeLa cell spreading induced by transfection with FLAG-Plk4 was blocked by depletion of Arp2, and the inhibition of spreading induced by Arp2 depletion was partially rescued by wild type but not T237/238A mutant Arp2 (Fig. 7C). Furthermore, the enhanced wound healing seen in U2OS cells stably expressing Plk4 was diminished by the Arp2 inhibitor CK-666 (Fig. 7D). Taken together, these results indicate that Arp2 binds to Plk4, is phosphorylated and activated in its presence, and this is required for stimulation of cancer cell motility by Plk4.

The results described here show that Plk4 promotes cancer cell migration and invasion, and cancer progression in vivo. Furthermore, we demonstrate that Plk4 interacts with the Arp2/3 complex, enhancing lamellipodia formation and cell migration. Plk4 is cell-cycle regulated at the transcriptional and posttranslational levels, the latter through dimerization, autophosphorylation, ubiquitination, and proteosomal degradation (3, 32–37). Plk4 transcripts are reportedly increased in several common human cancers (breast, colorectal, prostate, pancreatic; refs. 5–8, 38), which may simply reflect a higher mitotic index, but could also represent an unexplored change in Plk4 transcriptional regulation in cancer. In normal cells, Plk4 levels are highest in late M-phase (3, 12, 32, 36), perhaps playing a role in re-establishing polarity following cytokinesis, in keeping with the polarity defects we observed upon Plk4 depletion here, and in regenerating liver of Plk4 heterozygous mice after partial hepatectomy (12). Altered regulation of Plk4 in cancer cells appears to misappropriate this activity to promote migration and invasion, potentially creating a therapeutic window for anti-Plk4 therapy.

The appeal of Plk4 as a therapeutic target has been based not only on its increased expression in cancers that are resistant to conventional systemic therapies (7, 8, 39), but also on the notion that its overexpression contributes to pathologic centrosomal amplification (CA) that is relatively unique to cancer cells. In a search for evidence of a causative role for CA in cancer development and progression, others have shown that Plk4-induced CA is associated with acquisition of an invasive phenotype by benign cells (9, 14). We and others (14, 40) have demonstrated that Plk4 promotes Rac1 activation; however, this was insufficient for Plk4-induced motility in our system. Regulation of actin dynamics and cell motility occurs at multiple levels. Our unbiased screen using BioID revealed that Plk4 interacts with the Arp2/3 complex. Furthermore, Plk4 phosphorylates Arp2, and Arp2 activity is required for Plk4-induced spreading and directional migration. This suggests that Arp2 is a downstream effector of Plk4 in promoting cancer cell motility and invasion.

The Arp2/3 complex stimulates actin polymerization by generating nucleation sites on existing filaments that facilitate filamentous branching. The V domain of Wiskott-Aldrich syndrome family proteins (WASP, N-WASP, WAVE, and WASH) binds actin monomers, whereas the CA region binds the Arp2/3 complex to initiate a nucleation site. Phosphorylation on threonine and tyrosine residues T237, T238, and Y202 of the Arp2 subunit is necessary for activity of the Arp2/3 complex and lamellipodia formation (30). In the model proposed by Narayanan and colleagues, phosphorylation at these sites destabilizes the inactive state, allowing Arp2 to reorient itself into a state that is permissive for full activation by nucleation promoting factors (NPF; ref. 31). Recent evidence reveals that Nck-interacting kinase (NIK), a Ste20/MAP4K4 serine/threonine kinase, increases the activity of the Arp2/3 complex by directly binding and phosphorylating Arp2 at T237 and T238, leading to increased actin nucleating activity and promoting plasma membrane protrusions in response to EGF (41). Here we show that Plk4 phosphorylates Arp2, and this appears to be specific to the T237 and T238 residues, implying that Plk4′s effect on cell motility and invasion is mediated through increased actin nucleating activity of the Arp2/3 complex.

The usefulness of Plk4 inhibition as a therapeutic strategy has been questioned (42), not only in view of its role as a haploinsufficient tumor suppressor (13, 43). Although the first orally available potent Plk4 inhibitor to be tested in vivo, CFI-400945, suppressed the growth of murine xenografts, it was not clear to what extent this was due to inhibition of Plk4, as this agent also inhibits Aurora B (8). Furthermore, as pointed out by Holland and Cleveland, by interfering with the proteosomal degradation triggered by autophosphorylation, partial inhibition of Plk4 kinase activity may in fact lead to CA and its potential sequelae (37, 42). In this regard, it is interesting to note that CFI-400945 failed to deplete centrosomes. In contrast, the highly selective Plk4 inhibitor centrinone/centrinoneB developed by Wong and colleagues completely blocked centriole duplication, and fully depleted centrosomes within 4–5 cell cycles (26). Despite their acentriolar state, however, cancer cells continued to proliferate, albeit at a reduced rate. Notably, depletion of Plk4 to 25% to 30% of control levels had only a minor effect on tumor growth in NOD/SCID mice, but markedly reduced tumor penetration into the peritoneal cavity. Thus the utility of Plk4 inhibitors in cancer therapy may ultimately depend more on suppression of invasion and metastasis, as our data suggest. Centrinone is not yet available in an orally active form, but CFI-400945 is currently being studied in phase I trials in cancer patients.

Altered expression of one or more of the seven Arp2/3 complex subunits has been shown in several types of human epithelial malignancy, including gastric, lung, and colorectal cancer (44–48). In particular, high Arp2 expression correlates with aggressive behavior and poor prognosis (45–48). In keeping with this, silencing of individual Arp2/3 complex subunits, including Arp2, reduces migration in pancreatic and gastric cancer cell lines (46, 49). However, not all cell motility is dependent on Arp2/3 activity (50), as illustrated by the insensitivity of A2780 cancer cell migration to CK-666 (51). In considering the rational selection of patients for Plk4 inhibitor therapy, determination of Arp2 status may be useful.

No potential conflicts of interest were disclosed.

Conception and design: K. Kazazian, J.W. Dennis, A.-C. Gingras, C.J. Swallow

Development of methodology: K. Kazazian, O. Brashavitskaya, A.-C. Gingras, C.J. Swallow

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Kazazian, C. Go, H. Wu, O. Brashavitskaya, R. Xu, A.-C. Gingras

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Kazazian, C. Go, O. Brashavitskaya, R. Xu, A.-C. Gingras, C.J. Swallow

Writing, review, and/or revision of the manuscript: K. Kazazian, C. Go, J.W. Dennis, A.-C. Gingras, C.J. Swallow

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Kazazian, A.-C. Gingras

Study supervision: A.-C. Gingras, C.J. Swallow

Other (design, acquisition, and analysis of the proteomics data performed by her laboratory; supervised C. Go): A.-C. Gingras

We acknowledge Karina Pacholczyk, Mohammed Soliman, and Mikhail Bashkurov for technical assistance, Judy Pawling for manuscript review, and Laurence Pelletier and Sergio Grinstein for helpful discussions.

This work was supported by grants from the Cancer Research Society (C.J. Swallow), the Canadian Institutes of Health Research (CIHR; FDN 143301 to A.-C. Gingras), and by the Syd Cooper Program for the Prevention of Cancer Progression (C.J. Swallow). A.-C. Gingras is the Canada Research Chair in Functional Proteomics and the Lea Reichmann Chair in Cancer Proteomics. K. Kazazian is supported by a Johnson & Johnson Medical Products Surgeon Scientist Training Program Fellowship and Canadian Society of Surgical Oncology Research grant. C. Go is supported by a CIHR Banting studentship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.