Migrasome-mediated clearance of excess PLK4 defines a targetable vulnerability

Summary

Background

Centrosome amplification caused by Polo-like kinase 4 (PLK4) overexpression promotes tumour initiation, yet sustained PLK4 accumulation is detrimental to cancer cell viability. While proteasomal degradation limits PLK4 levels, whether alternative clearance mechanisms exist remains unknown.

Methods

We established inducible PLK4-overexpressing breast cancer cells and employed live-cell imaging, vesicular marker profiling, immunoprecipitation–mass spectrometry, and in vivo patient-derived xenograft models to investigate how cancer cells eliminate excess PLK4.

Findings

In this study, we identified a previously unrecognised mechanism in which excess PLK4 is rapidly expelled via migrasomes. The tetraspanin protein TSPAN6 directly binds PLK4 with high affinity and mediates its incorporation into migrasomes for extracellular release. In breast cancer samples, PLK4 expression decreased while TSPAN6 increased during progression. TSPAN6 knockdown blocked PLK4 expulsion, leading to multipolar spindle formation, apoptosis, and suppression of tumour growth and metastasis in vivo.

Interpretation

Migrasome-mediated clearance represents a non-proteasomal pathway maintaining centrosome homoeostasis in cancer cells. Targeting TSPAN6 to prevent PLK4 elimination selectively triggers mitotic catastrophe in PLK4-high tumours, highlighting a previously unrecognised therapeutic vulnerability.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC) (T2225006, T2488301, and 82272948 to ML), Beijing Municipal Natural Science Foundation (Key program Z220011 to ML, 5254051 to TW), NSFC (82371640 to BX, 82403693 to PW), and the “Clinic + X” program of Peking University (to PW).

Research in context.

Evidence before this study

Centrosome amplification caused by overexpression of Polo-like kinase 4 (PLK4) promotes tumour initiation but is lethal to cells if sustained. Most studies have focused on proteasome-dependent degradation as the only mechanism controlling PLK4 abundance. Although multiple E3 ligases were reported to target PLK4 for degradation, no study had identified alternative or extracellular routes for its removal. The role of migrasomes in regulating intracellular protein homoeostasis had also not been examined.

Added value of this study

We identified a previously unrecognised mechanism by which excess PLK4 is cleared from cancer cells through migrasomes. The tetraspanin TSPAN6 binds PLK4 with high affinity and mediates its selective incorporation into migrasomes for extracellular elimination. Inhibition of TSPAN6 blocks PLK4 expulsion, leading to mitotic failure, apoptosis, and suppression of tumour growth and metastasis in vivo. This work support migrasome-mediated PLK4 clearance as a pathway contributing to centrosome homoeostasis.

Implications of all the available evidence

Our findings reveal that cancer cells use migrasomes as an efficient non-proteasomal system to remove excess PLK4 and maintain viability. Targeting TSPAN6 disrupts this process and selectively kills PLK4-high tumours. This mechanism expands our understanding of intracellular protein clearance and highlights TSPAN6 as a potential therapeutic target for cancers characterised by centrosome amplification.

Introduction

Centrioles function as the core of centrosomes that organise the microtubule cytoskeleton in animal cells and drive the formation of the bipolar mitotic spindle.¹ As the most structurally divergent member of the polo family, the serine/threonine-protein kinase PLK4 is highly conserved across species and controls the duplication of centrioles in cells.^2,3 Thus, the cellular abundance of PLK4 is strictly controlled since excess PLK4 leads to centriole amplification, chromosomal instability (CIN), and aneuploidy.4, 5, 6 One major mechanism against excess PLK4 involves the ubiquitin-proteasome system, where the SCF (Skp, Cullin, F-box containing complex) ubiquitin ligase complex or E3 ubiquitin ligase Mind bomb targets PLK4 for proteasomal degradation, preventing its cellular accumulation.^7,8 Additionally, PLK4 autoregulates its own stability through autophosphorylation, promoting its interaction with components of the degradation machinery.^9,10 A recent study revealed that PLK4 abundance is translationally suppressed through conserved upstream open reading frames (uORFs) in the 5′ UTR of the mRNA. A portion of primordial germ cells in homozygous Plk4 uORF knockout mice contained extra centrioles.¹¹ Each of these mechanisms highlights the importance of maintaining appropriate PLK4 abundance in cells. However, the extent to which cells rely on these regulatory pathways, and whether more efficient protein degradation mechanisms exist, remain unknown.

It is well recognised that centrosome amplification induced by excess PLK4 can lead to CIN by causing errors in chromosome segregation during cell division, thereby serving as a driving force behind tumourigenesis.^12,13 On the other hand, for a tumour cell that has already become cancerous, the ongoing CIN intensively challenges the cell itself owing to the frequent generation of unviable karyotypes.¹⁴ Therefore, to ensure its survival and continuous proliferation, a mature tumour cell must evolve specific mechanisms to counter the continuously increasing trend of CIN. Disruption of TP53 has been suggested as a crucial mechanism that facilitates the tolerance of CIN, in both cultured cells and animal models.15, 16, 17, 18 Besides, scientists discovered that some tumour cells can cluster extra centrosomes via specific spindle-associated proteins, thereby forming bipolar spindles and completing mitosis.^19,20 Nevertheless, whether cancer cells possess special regulatory mechanisms for the PLK4 protein per se, allowing their further survival and proliferation, is still far from clear.

In this work, we observed a considerable number of PLK4 puncta surrounding host cells when culture dishes were precoated with fibronectin to retain more complete extracellular materials on the dish bottoms, and the number of puncta was positively correlated with the overexpression level of PLK4 in host cells. Using various vesicular markers, we determined that the extracellular PLK4 puncta represented migrasomes that were capable of rapidly eliminating excess PLK4 from host cells. Through immunoprecipitation-mediated mass spectrometry analysis of PLK4-overexpressing cells, we identified that TSPAN6, one of the most potent tetraspanins for enhancing migrasome formation, had a high affinity for PLK4, mediating its incorporation into migrasomes and subsequent exclusion. More interestingly, as tumour progression advanced, the expression patterns of PLK4 and TSPAN6 exhibited decreasing and increasing trends, respectively, in people with tumours. In in vivo models, knockdown of TSPAN6 blocked the expulsion of excess PLK4 by tumour cells into the extracellular microenvironment, leading to multipolar spindle formation and apoptosis. Taken together, this study not only uncovers a previously unreported mechanism for efficient clearance of excess PLK4 but also provides an alternative anti-tumour strategy targeting migrasome-mediated PLK4 elimination.

Methods

Ethic

Mice experiments were conducted in compliance with the guidelines of the Ethics Committee of the Peking University Health Science Center (SA2022289). Tumour samples for patient-derived xenograft (PDX) models were collected with the approval from Medical Science Research Ethics Committee of Peking University Third Hospital (S2022301) and written informed consent was obtained from all patients for the use of their samples.

Chemicals and antibodies

Unless otherwise specified, all chemicals were purchased from Sigma–Aldrich. Fibronectin (Cat# PHE0023) was purchased from Gibco and Alexa Fluor™ 555 WGA (Cat# W32464) from Thermo Fisher Scientific. DAPI (Cat# C1005) and Hoechst 33342 (C1027) were obtained from Beyotime. Primary antibodies were obtained as follows: anti-PLK4 (Cat# MABC544, RRID: AB_2893410) and anti-Centrin (Cat# 04–1624, RRID: AB_10563501) from Millipore; anti-LAMP1 (Cat# ab25630, RRID: AB_470708), LC3B (Cat# ab192890, RRID: AB_2827794), Proteasome 20S alpha + beta (Cat# ab22673, RRID: AB_2268907), and LAMP2A (Cat# ab18528, RRID: AB_775981) from Abcam; anti-ERp72 (Cat# 66365-1-Ig, RRID: AB_2881745), GM130 (Cat# 11308-1-AP, RRID:AB_2115327), LMAN2 (Cat# 11496-1-AP, RRID: AB_3085375), TMED10 (Cat# 15199-1-AP, RRID: AB_2204321), ITGB1 (Cat# 12594-1-AP, RRID: AB_2130085), and TSPAN6 (Cat# 12293-1-AP, RRID: AB_2213446) from Proteintech; anti-cleaved caspase-3 (Cat# 9664, RRID: AB_2070042), PDI (Cat# 2446, RRID: AB_2298935), and CYCS (Cat# 4272, RRID: AB_2090454) from Cell Signalling Technology; anti-HSC70 (Cat# NB120-2788, RRID: AB_2120309) and Ki67 (Cat# NB500-170, RRID: AB_10001977) from Novus; anti-PIGK (Cat# PA5-28337, RRID: AB_2545813), EOGT (Cat# MA5-53414, RRID: AB_3247885), TSPAN4 (Cat# PA5-69344, RRID: AB_2688603) and Calnexin (Cat# PA5-34754, RRID: AB_2552106) from Thermo Fisher Scientific. Secondary antibodies included Alexa Fluor™ 633 goat anti-mouse IgG (Cat# A-21126, RRID: AB_2535768), Alexa Fluor™ 555 goat anti-rabbit IgG (Cat# A-21429, RRID: AB_2535850), HRP-conjugated goat anti-mouse (Cat# 32430, RRID: AB_1185566), and goat anti-rabbit (Cat# 31466, RRID: AB_10960844) antibodies (all from Thermo Fisher Scientific).

Plasmid construct

PLK4-EGFP was expressed using a Tet-On system based on pLVX-TetO3G (rtTA-hygromycin) and pLVX-Tre3G-MCS vectors. Full-length PLK4 and EGFP coding sequences were cloned into the pLVX-Tre3G-MCS vector. For bacterial expression, coding sequences were subcloned into pGEX-4T-1 for GST-tagged fusion protein production. TSPAN6 knockdown was performed using single shRNA cloned into the pLKO.1 vector. The primer sequences were provided in Table S1.

Cell culture

HEK-293T, MDA-MB-231, HeLa, and A549 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM). MCF-7 cells were cultured in DMEM supplemented with 10 μg/mL insulin, and T47D cells were cultured in RPMI-1640 medium supplemented with 0.2 U/mL insulin. All media were supplemented with 10% fetal bovine serum (FBS), at 37 °C in a humidified 5% CO₂ incubator. For directional migration, cell culture medium was replaced with fresh medium containing 0.1 μg/mL doxycycline and 10 ng/mL EGF. All cell lines were authenticated by short tandem repeat (STR) profiling and routinely tested for mycoplasma contamination.

Western blot

Cells were lysed using RIPA buffer for total protein extraction. Protein samples were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and subsequently transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were then blocked in Tris buffered saline with Tween 20 (TBST) containing 5% skim milk for 2 h, followed by incubation with the indicated primary antibodies (1:500–1:1000 dilution) overnight at 4 °C. After three washes in TBST, membranes were then incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:1000 dilution) for 1 h at 37 °C. After three additional TBST washes, protein bands were detected using an enhanced chemiluminescence detection system (Amersham Biosciences).

Immunofluorescence

Cells were fixed with 4% paraformaldehyde in PBS for 30 min and permeabilized with 0.5% Triton-X-100 at room temperature for 25 min. After blocking in PBS containing 1% bovine serum albumin (BSA) for 1 h, cells were incubated with the indicated primary antibodies (1:100–1:500 dilution) at 4 °C overnight. Following three washes with PBS supplemented with 0.1% Tween 20 and 0.01% Triton-X 100, cells were incubated with appropriate fluorescent secondary antibodies for 1 h at room temperature. Following several additional washes, nuclei were counterstained with DAPI for 15 min. Samples were eventually mounted onto glass slides and imaged under a confocal laser scanning microscope (Zeiss LSM 880) with a 63×/1.40 oil immersion objective.

Cell-cycle and proliferation analysis

To evaluate DNA synthesis and cell-cycle distribution, ^PLK4−OEcells were cultured with or without doxycycline (1 μg/mL) for 24 h. DNA synthesis was detected using the BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 647 (Beyotime Biotechnology, China) according to the manufacturer's instructions. Total DNA was counterstained with 50 μg/mL propidium iodide (PI) in the presence of 100 μg/mL RNase A for 30 min at 37 °C. Stained samples were analysed using a NovoCyte flow cytometer (Agilent Technologies, USA), and the data were processed with NovoExpress software (ACEA Biosciences).

Isolation of migrasomes from cultured cells

Migrasomes were isolated by differential centrifugation to obtain crude migrasomes, followed by iodixanol–sucrose (OptiPrep; #LYSISO1, Sigma–Aldrich) density-gradient centrifugation for high-purity preparation.^21,22 Briefly, cells cultured in 15-cm dishes were trypsinized, and the resulting cell suspension containing detached migrasomes was gently transferred to 50-mL tubes. All subsequent steps were performed at 4 °C. The suspension was centrifuged twice at 600×g for 10 min to remove cells and large debris. The supernatant was then centrifuged at 2000×g for 20 min to further clear residual cell bodies and debris. Crude migrasomes were pelleted by centrifugation at 18,000×g for 30 min and then resuspended in a total volume of 800 μL buffer (400 μL extraction buffer mixed with 400 μL 10% OptiPrep) and loaded onto a discontinuous OptiPrep step gradient. The gradient consisted of 50%, 40%, 35%, 30%, 25%, 20%, 15%, 10%, and 5% OptiPrep (500 μL each), followed by the sample layer (800 μL; 5% OptiPrep). Gradients were centrifuged at 150,000×g for 4 h at 4 °C. Fractions were collected gently from top to bottom (500 μL per fraction). Fractions 4–6 were each diluted with 500 μL PBS and centrifuged at 18,000×g for 30 min at 4 °C. Pellets were washed once with PBS and centrifuged again at 18,000×g for 30 min to recover migrasomes. Purified migrasomes were used immediately for downstream western blotting.

Live-cell imaging

Prior to live imaging, ^PLK4−OEcells induced with doxycycline for 8 h were seeded into 35 mm glass-bottom dishes pre-coated with 10 μg/mL fibronectin in PBS for 1 h. Following an additional 12h of doxycycline induction, cells were stained with 1 μg/mL WGA and 10 μg/mL DAPI. For image acquisition, the TOKAI HIT live cell workstation was pre-equilibrated to 37 °C in a humidified chamber containing 5% CO₂. Time-lapse movies were acquired using an Andor Dragonfly 200 spinning disk confocal microscope equipped with a 60×/1.42 oil immersion objective and Andor Sona sCMOS cameras. Fluorescence channels included 405 nm (DAPI), 488 nm (PLK4-EGFP), and 561 nm (WGA). Images were captured every 5 min over a 300-min period for each field of view. Acquired images were processed and analysed using Imaris software.

3D reconstruction

Z-stack confocal images were acquired at 0.25 μm intervals using a Zeiss LSM 880 microscope equipped with a 63×/1.40 oil objective. Images were processed in Imaris software (Bitplane), where PLK4-EGFP and migrasome-associated proteins were reconstructed using 3D surface rendering. Migrasome volume was quantified using Imaris Volume-Specific Measurements.

Co-immunoprecipitation (Co-IP)

Cells were lysed in ice-cold IP lysis buffer for 15 min, centrifuged at 13,000 rpm for 10 min at 4 °C, and incubated with anti-PLK4 antibody for 4 h at 4 °C with rotation. Protein A/G-Sepharose beads were added and incubated overnight at 4 °C. Beads were washed three times, resuspended in 2× SDS loading buffer, boiled, and analysed by SDS–PAGE followed by immunoblotting.

Liquid chromatography-tandem mass spectrometry

Protein eluted from immunoprecipitation were separated on SDS-PAGE and visualised by Coomassie G-250, followed by destaining with water. Each gel lane was manually cut into six fractions. Proteins within each fraction were subjected to in-gel digestion, and the peptides were dried using a vacuum concentrator.

The extracted peptides were resuspended in 25 μL of Solvent A (0.1% formic acid in water) and loaded onto a Thermo Ultimate 3000 nano-UPLC system coupled with a Thermo Fusion LUMOS mass spectrometer. The chromatographic separation was performed using a 120-min gradient elution at a flow rate 0.300 μL/min. The analytical column was an Acclaim PepMap RSLC column (75 μm ID, 250 mm length, C18) coupled with a trap column (Acclaim PepMap™ 100, 75 μm ID, 20 mm length, C18). Mobile phase A was 0.1% formic acid in water, and mobile phase B was 100% acetonitrile containing 0.1% formic acid. The gradient program was as follows: 3% B for 5 min; 3%–22% for 75 min; 22%–35% for 12 min; 35%–90% for 11 min; 90% for 6 min; 90%–3% for 1 min; 3% for 10 min.

The Fusion LUMOS mass spectrometer was operated in data-dependent acquisition (DDA) mode using Xcalibur 4.1.50 software. Full MS scans were acquired in the Orbitrap (375–1500 m/z, 60,000 resolution), followed by MS/MS scans. The most intense ions selected under top-speed mode were isolated in Quadrupole with a 1.6 m/z window and activated by high energy collision dissociation (HCD) with a normalised collision energy of 30%, then detected in the Orbitrap at resolution of 15,000. The maximum injection time was set to dynamic mode, and the AGC target was set to standard mode.

Protein identification

Raw mass spectrometry data were processed using Proteome Discoverer software (Thermo Fisher Scientific, version 2.4). The resulting spectra were searched against the Homo sapiens protein database from UniProt (https://www.uniprot.org). Parameters were set as follows. Precursor mass tolerance: 20 ppm; fragment mass tolerance: 0.05 Da; variable modifications: oxidation of methionine, carbamidomethylation of cysteine, and acetylation of protein N-terminal; a maximum of two missed cleavage sites were allowed. Only proteins identified with at least two unique peptides and a false discovery rate (FDR) of less than 1% were retained for further analysis.

Bioinformatics analysis

For quality control of the PLK4-interacting proteome, hierarchical clustering and t-distributed stochastic neighbour embedding (t-SNE) were performed using the R packages stats and Rtsne, respectively, to assess the similarity across samples. Proteins with log2 (fold change) ≥ 1 and p value < 0.05 were considered as significantly enriched in PLK4-immunoprecipitated ^PLK4−OEcells compared to WT controls. A volcano plot was generated using R package ggplot2 to visualise the significantly enriched proteins. Gene Ontology (GO) functional enrichment analysis was performed using the WEB-based gene set analysis toolkit (WebGestalt, http://www.webgestalt.org/). The heatmap of proteins involved in specific GO terms was created using the R package pheatmap.

Transcriptome data of tumour samples with paired normal tissues across all available cancer types were obtained from TCGA database. Expression levels of PLK4 and associated centrosome genes were extracted, along with corresponding clinicopathological staging information of TCGA patients. Survival analysis was conducted using R (version 4.3.3) with the survival, survminer, and ggplot2 packages. To assess the prognostic relevance of PLK4 at different disease stages, patients with breast cancer from TCGA database were first stratified into N0 and N3 groups according to their lymph node metastasis status. Within each group, patients were further classified into high and low expression subgroups based on the median mRNA level of PLK4. Overall survival was compared using Kaplan–Meier analysis with log-rank tests.

The single-cell transcriptomic dataset GSE176078, deposited in the Gene Expression Omnibus (GEO) database, and comprising 130,246 cells from 26 primary human breast tumours, was analysed using R packages including SingleR, garnett, monocle3, Matrix, and org.Hs.eg.db to characterise the expression patterns of PLK4 and associated centrosome genes and to construct their pseudotime trajectories.

Protein expression data of PLK4 and TSPAN6 in patients with breast cancer were obtained from the CPTAC database. Pearson correlation analysis was performed to evaluate the association between PLK4 and TSPAN6 protein levels, with the correlation coefficient (r) and corresponding p value calculated accordingly. Changes in kinase activity from early-to late-stage breast tumour samples were assessed using Kinase-Substrate Enrichment Analysis (KSEA) based on global phosphoproteomic data from CPTAC clinical breast tumours.

Recombinant protein expression and protein purification

GST-tagged proteins were expressed in Escherichia coli BL21 (DE3) transformed with pGEX-4T-1 constructs. Cultures were induced with 0.5 mM IPTG at OD₆₀₀ of 0.6–0.8 and grown at 16–20 °C for 12–18 h. Cells were harvested and lysed in buffer containing PMSF and DTT by sonication (4 s on, 6 s off, 40% amplitude, 30 min, 30 °C). After centrifugation, supernatants were incubated with Glutathione Sepharose 4B beads for 4 h or overnight at 4 °C. Proteins were eluted and dialysed against PBS with sequential buffer changes (1 h, 1 h, 2 h, 2 h), and stored at 4 °C.

Microscale thermophoresis

MST assays were performed as previously described.²³ Purified GST-tagged TSPAN6, ITGB1, LMAN2 and TMED10 proteins were individually labelled using the RED-NHS protein labelling kit (NanoTemper, Germany) following the manufacturer's instructions. Each labelled protein at a constant concentration (10–50 nM) was incubated with twofold serial dilutions of purified GST-PLK4 protein in an MST-optimised buffer containing 50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl₂, and 0.05% Tween-20. Equal volumes of each binding reaction were mixed by pipetting and incubated for 15 min at room temperature. Samples were enclosed in standard-treated or premium-coated glass capillaries and loaded into the Monolith NT.115 instrument (NanoTemper, Germany). Measurement settings were as follows: fluorescence before 5 s, MST on 30 s, fluorescence after 5 s, and delay 25 s. For all the measurements, fluorescence intensity between 200 and 1000 counts were recorded. The measurement was performed at 20% and 40% MST power. F_norm = F1/F0 (F_norm: normalised fluorescence; F1: fluorescence after thermodiffusion; F0: initial fluorescence or fluorescence after T-jump). Binding curves were fitted using the Kd model derived from the law of mass action, assuming a 1:1 interaction between the target and ligand. During curve fitting, the target concentration was fixed, while all other parameters were freely fitted. Dissociation constants (Kd) values are reported with the fitting confidence (±), corresponding to the 68% confidence interval (±1σ) derived from nonlinear regression implemented in the NanoTemper analysis tool.

Autophagic flux assessment

Autophagic flux was analysed by immunofluorescence, immunoblotting, and flow cytometry. For immunofluorescence, cells on coverslips were treated with or without bafilomycin A1 (100 nM, 4 h), fixed, permeabilized, and stained with anti-LC3B (1:200 dilution) followed by Alexa Fluor-conjugated secondary antibodies (1:500 dilution). Nuclei were counterstained with DAPI. Confocal images were acquired using a Zeiss LSM 880 microscopy, and LC3 puncta were quantified in ImageJ.

For immunoblotting, cell lysates (RIPA buffer) were analysed for LC3 (1:1000 dilution) and β-actin (1:5000 dilution). LC3-II accumulation upon bafilomycin A1 treatment was used as a flux indicator.

Flow cytometric analysis was performed using the LC3 detection kit (Abcam ab224639). Treated cells were fixed, permeabilized, and stained with anti-LC3 and secondary antibodies. Data were acquired on a BD LSRFortessa and analysed using FlowJo. Mean fluorescence intensity (MFI) changes upon bafilomycin A1 treatment were used to quantify autophagic activity.

Isolation and culture of primary tumour cells

The isolation and culture of primary tumour cells were performed according to our previously described protocol.²⁴ Fresh tumour tissues were obtained from patients with breast cancer undergoing surgical tumour resection. Tumour tissues were first mechanistically cut into small fragments (<1 mm³) and subjected to enzymatic dissociation using 1× collagenase/hyaluronidase buffer (STEMCELL Technologies) at 37 °C with continuous shaking for 6 h. Then the digested suspension was then further treated with 0.25% trypsin, followed by incubation in a solution of dispase (5 units/ml) and DNase I (0.05 mg/mL; STEMCELL Technologies) to enhance cell dissociation for 5–10 min. Single-cell suspensions were obtained by passing the digested mixture through a 40-μm filter. The resulting suspension was seeded at a density of 0.5–1 × 10⁵/well onto 6-well plates coated with collagen I and incubated for 36–72 h. The culture medium composes of DMEM: F12 supplemented with 5% FBS, penicillin/streptomycin (1%), gentamycin (0.2%), EGF (10 ng/mL), adenine (20 μg/mL), cholera toxin (10 ng/mL), HEPES (15 mM), insulin (5 μg/mL), hydrocortisone (0.32 μg/mL) and ROCK inhibitor (5 μM). After achieving stable adherence, cells were either further expanded in culture or cryopreserved in a freezing solution consisting of FBS/DMSO (9:1) with 5 μM ROCK inhibitor and stored in liquid nitrogen.

shRNA lentivirus generation and shRNA knockdown

To knockdown TSPAN6 in ^PLK4−OEcells, lentiviral particles were generated by co-transfecting HEK293T cells with the pLKO.1 plasmid comprising shRNA and the packaging plasmids (psPAX2 and pMD2.G), using Lipofectamine 3000 according to the manufacturer's instructions. Six hours post-transfection, the medium was replaced with fresh growth medium, and the cells were cultured for an additional 48 h. The viral supernatant was subsequently collected, centrifuged at 3000×g for 5 min to remove cellular debris, and filtered through a 0.45-μm membrane. For viral concentration, the supernatant was subjected to ultrafiltration using a Centricon Plus-20 centrifugal filter at 4000×g, aliquoted, and stored at −80 °C until use.

For lentiviral transduction, ^PLK4−OEcells were seeded at a density of 1 × 10⁵ cells per well into 6-well plates and cultured at 37 °C in a humidified atmosphere with 5% CO₂ until reaching 30–40% confluence. Concentrated lentiviral supernatant was added at a multiplicity of infection (MOI) of 20 and incubated for 16 h.

PDX and cell-derived xenograft (CDX) mouse models

The PDX model was generated as described in our previous study.²⁴ Fresh tumour fragments (5–8 mm³) obtained from patients with breast cancer were subcutaneously implanted into the flanks of NOD-SCID mice aged 5–6 weeks. Tumour progression was monitored weekly using calliper-based measurements. Upon reaching a volume of 1000–1500 mm³, mice were euthanised, and the tumours were excised under sterile conditions. The excised tumours were sectioned into small pieces (4–8 mm³) and serially transplanted into the flanks of age-matched BALB/c Nude mice to establish the second generation of the PDX model. For the CDX model, 1 × 10⁶ MDA-MB-231 cells were suspended in 100 μL of a 1:1 mixture of Matrigel and PBS, and orthotopically injected into the fourth mammary fat pad of 4-week-old female BALB/c nude mice. To assess the functional relevance of TSPAN6 in vivo, 10 μL of lentiviral suspension containing 4 × 10⁷ viral particles encoding either TSPAN6-targeting shRNA or a non-targeting scramble control were intratumourally injected into established tumours. Tumour growth was observed over a period of 40–60 days. Mice were euthanised when tumours reached the ethically approved endpoint of ∼2500 mm³. Tumour volume was calculated using the formula: volume = (shortest diameter)² × (longest diameter) × 0.5. All mice were raised in the Animal Centre of Peking University Health Science Centre in an SPF environment.

Cell proliferation assay by xCELLigence system

Cell proliferation was assessed using the xCELLigence RTCA-DP system (Acea Bioscience).²⁵ Briefly, 20 μL complete culture medium was added to each well of E−16-well plates to obtain baseline impedance readings. Meanwhile, cells were digested by trypsinization, resuspended in complete medium, and seeded at a density of 5000–10,000 cells per well in 80 μL complete medium. Real-time monitoring was conducted for 60 h, with cell index values recorded at 60-min intervals. Data were collected and analysed with RTCA software (version 2.0).

Colony formation assay

To assess the cell proliferative capacity, ^PLK4−OEcells with or without TSPAN6 knockdown were digested using trypsin, resuspended in complete culture medium, and counted. A total of 500 cells were then seeded into each well of 6-well plates. The culture medium was replenished every 2 d over a 14-d incubation period. At the endpoint, cells were washed twice with PBS and fixed in 100% methanol for 30 min at room temperature. Finally, fixed colonies were stained with 0.05% crystal violet for 30 min, rinsed with PBS, and quantified.

Caspase-3 activity assay

To assess the impact of TSPAN6 knockdown on apoptosis, shTSPAN6 or control ^PLK4−OEcells were seeded into 4-well glass-bottom chamber slides at a density of 1 × 10⁵/well. After 12 h, cells were gently washed twice with complete culture medium. Subsequently, cells were incubated in the dark with GreenNuc™ Caspase-3 Substrate (1:200 dilution, Beyotime, #C1168) for 30 min to label positive apoptosis signal, followed by Hoechst 33342 (1:100 dilution) staining for 20 min to visualise nuclei. After two additional washes with complete medium to remove residual dyes, images were captured using a confocal laser scanning microscope (Zeiss LSM 880) with a 40× objective. Imaging acquisition parameters were kept consistent across all groups to ensure comparability.

Histologic staining

Tissue samples underwent fixation in 10% neutral-buffered formalin for 24 h before being dehydrated through a graded ethanol series (75%, 85%, 90%, and 100%), cleared in xylene, and embedded in paraffin. Embedded tissues were sectioned at a thickness of 5 μm using a microtome, and the sections were mounted on poly-l-lysine-coated glass slides for hematoxylin and eosin (HE) staining and immunohistochemistry (IHC). For IHC, sections were incubated with primary antibodies against Ki67 (1:1000 dilution) and cleaved caspase-3 (1:500 dilution). Images of immunostaining were captured using an Olympus BX51 microscope equipped with a DP73 CCD camera for subsequent analysis.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism (version 10.0, GraphPad Software, USA). Unless otherwise specified, all experiments were performed in triplicate using biological independent samples. Data were presented as means ± standard deviations (SD). Two-tailed unpaired Student's t-test were used for two group comparisons. For analyses involving multiple time points across groups, two-way analysis of variance (two-way ANOVA) was used to assess the group effect. p value < 0.05 was regarded as statistically significant. Statistical analysis details were indicated in the corresponding figures or figure legends.

Data availability

All raw data and search results for mass spectrometry of PLK4-immunoprecipitated proteome have been deposited in ProteomeXchange via iProX (www.iprox.org) with the identifiers No. PXD063698 (for ProteomeXchange) and No. IPX0011840000 (for iProX).

Role of funders

The funders of the study had no role in the study design, data collection, data analyses, interpretation, writing of report, or decision to submit it for publication. No authors were paid to write this article by a pharmaceutical company or other agency. No authors were precluded from accessing data in the study, and they accept responsibility to submit for publication.

Results

Excess PLK4 is excluded by migrasomes

To explore how cells cope with excess PLK4, we established an inducible cell line derived from the common breast cancer cell line MDA-MB-231, which overexpresses PLK4-EGFP following doxycycline treatment (Fig. 1A). Following a 24-h incubation with doxycycline, we observed an obvious overexpression of PLK4-EGFP in the cells (^PLK4−OEcells), and these cancer cells remained viable without overt impairment of proliferation (Fig. 1B, Fig. S1A, B). By contrast, enforcing sustained high-level PLK4 overexpression through repeated inductions (two additional inductions at 12-h intervals; three total) rendered the cells nonviable and triggered apoptosis (Fig. S1C, D). Accordingly, we used the single-induction regimen in subsequent experiments to preserve cancer cell viability. Most of ^PLK4−OEcells showed an increased number of centrioles labelled by the marker of Centrin (Fig. 1C). Besides the centrosome area, it is worth noting that excess PLK4 was observed along protrusions at the leading edge of cell membranes, as well as in sporadic dots outside host cells (Fig. 1D). To investigate the unexpected signal of PLK4, we precoated culture dishes with fibronectin to enable the retention of more complete extracellular materials on the dish bottoms. Interestingly, we found a considerable number of PLK4 dots surrounding host cells (Fig. 1D), and the number of dots was positively correlated with the overexpression level of PLK4 in host cells (Fig. 1E). To complement the imaging-based correlation, migrasomes were isolated from conditioned media under a Dox titration series and analysed by western blotting. PLK4-EGFP levels in migrasome fractions increased with intracellular PLK4 induction, with migrasome enrichment markers (PIGK/EOGT/TSPAN4) and calnexin depletion confirming fraction identity (Fig. S1E, F).²² To assess the generalisability of extracellular PLK4 dots, we likewise induced PLK4 overexpression in HeLa, A549, MCF-7 and T47D cells and cultured them on fibronectin-precoated dishes. Consistently, we detected extracellular PLK4 puncta at the cell periphery in all cell types (Fig. S1G, H). To explore how the extracellular PLK4 was formed, we measured the diameter of these dots and compared them with different types of extracellular vesicles, including exosomes, microvesicles, and migrasomes.²⁶ As seen in Fig. 1F, the diameter of extracellular PLK4 dots ranges from 0.5 to 2.5 μm, which aligns more closely with the size of migrasomes rather than exosomes or general microvesicles. To clarify the identity of PLK4-containing vesicles, we stained ^PLK4−OEcells with different types of vesicle markers, including those for migrasomes, the endoplasmic reticulum (ER), Golgi apparatus, mitochondria, lysosomes, and autophagosomes. The co-localisation of PLK4 with wheat germ agglutinin (WGA), a well-established migrasome marker,²⁷ was observed both inside and outside host cells during the early (0–3 h) and the subsequent stage (3–6 h) following PLK4 overexpression, confirming that the PLK4-containing vesicles were indeed migrasomes (Fig. 1G–I). As an orthogonal validation, tetraspanin 4 (TSPAN4), a canonical migrasome-enriched tetraspanin, also co-localised with WGA⁺ structures and extracellular PLK4 puncta, with most WGA⁺ migrasomes being TSPAN4⁺PLK4⁺ across replicates (Fig. S1I, J). In contrast, PLK4 signals were not detected in the endoplasmic reticulum (ER), Golgi apparatus, or mitochondria at either the early or later stages of overexpression. Although PLK4 was not detectable in lysosomes or autophagosomes during the initial phase of overexpression, the protein was found to co-localise with lysosomes and autophagosomes between 3 and 6 h post-overexpression (Fig. 1G–I). These results reveal that excess cellular PLK4 is expelled via migrasomes, and the portion remaining within the cell may be degraded by the autophagy-lysosome pathway.

Fig. 1.

Migrasomes exclude excess PLK4 from cells. (A) Schematic representation of PLK4-EGFP overexpression in MDA-MB-231 cells (^PLK4−OEcells) using a Tet-On inducible system. PLK4-EGFP was stably overexpressed 24 h after the addition of doxycycline. (B) The protein level of PLK4-EGFP in ^PLK4−OEcells was analysed by Western blot, with β-actin used as the loading control. (C) Fluorescent imaging showing the co-localisation of PLK4 with the centriole marker Centrin. PLK4 overexpression led to the formation of extra centrioles in ^PLK4−OEcells. Scale bar, 5 μm for each main image and 1 μm for the zoom in section (left). Statistical analysis of centriole foci numbers in WT and ^PLK4−OEcells. A total of 30 cells from three independent experiments were analysed for each group. Data are presented as mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant (right). (D) Fluorescent imaging of ^PLK4−OEcells cultured on dish bottoms coated with or without fibronectin (FN). A larger number of extracellular PLK4 dots were detected in the FN group. Scale bar, 10 μm for each main image and 2 μm for the zoom in section (left). The number of extracellular PLK4 dots surrounding individual cells was quantified in the –FN and +FN groups. A total of 30 cells from three independent experiments were included for each group. Data are presented as mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant (right). (E) Correlation analysis between the number of extracellular PLK4 dots and the intracellular fluorescence intensity of PLK4. A total of 30 cells from three independent experiments were included. r, Pearson correlation coefficient. (F) Diameter ranges of extracellular PLK4 dots, exosomes, microvesicles, and migrasomes. A total of 100 PLK4 dots were measured across three independent experiments. (G) Fluorescent staining of migrasome, lysosome, autophagosome, mitochondria, ER, and Golgi apparatus markers in ^PLK4−OEcells. Scale bar, 10 μm for each main image and 2 μm for the zoom in section. Enlarged views of the boxed areas are shown. (H) Intensity profiles of PLK4 and the indicated markers along the lines shown in (G). PLK4 exhibited strong colocalisation with migrasomes and partial colocalisation with lysosomes and autophagosomes. A total of 30 lines from three independent experiments were analysed for each group. (I) Quantification of the average number of PLK4-containing migrasomes, lysosomes, autophagosomes, mitochondria, ER, and Golgi apparatus per cell across different time intervals.

Dynamics of migrasome-mediated elimination of excess PLK4

To investigate the exclusion dynamics of PLK4-containing migrasomes (migrasomes^PLK4), ^PLK4−OEcells were monitored by live-cell imaging in culture dishes coated with fibronectin. Cells overexpressing either EGFP or GAPDH-EGFP were used as controls. During the entire monitoring period, only rare or small numbers of migrasomes were observed around host cells, although migrasome formation may continue beyond the imaging period. In contrast, upon the overexpression of PLK4, an intensive signal of WGA was enriched in the cortical area along the cell edge, implying a predisposition for migrasome formation (Fig. 2A, Movies S1–S3). Specifically, we observed a fair number of migrasomes formed and excluded by ^PLK4−OEcells during the early stage of PLK4 overexpression. As time progressed, an increasing number of migrasomes^PLK4 emerged and were released around the host cell (Fig. 2A and B). The number of migrasomes^PLK4 reached a plateau at 180–240 min post-PLK4 overexpression, and the exclusion process lasted for approximately 300 min (Fig. 2A and B, Movies S1–S3). Intriguingly, the area of the cells adhering to the bottom significantly shrank during the expulsion of migrasomes^PLK4 (Fig. 2A and C). This phenomenon corresponds with the migrasome formation since migrasomes carry away parts of the cell membrane, leading to a reduction in the cell's adherence area.²⁸ When examining the distribution of extracellular PLK4, we found that all of the excluded protein was packaged within migrasomes, with little protein detected along retraction fibres (RFs) (Fig. 2D, Fig. S2A, B). By performing 3D reconstruction based on a series of high-resolution scans of migrasomes using the Imaris program, we verified that extracellular PLK4 did be encapsulated by these WGA-labelled vesicles (Fig. 2E, Fig. S2C). Since migrasomes form on RFs, which are more actively extruded from the plasma membrane at the rear of the cell during forward movement, we examined the distribution of these migrasomes^PLK4 in migrating cells. In ^PLK4−OEcells that remained relatively immobile, migrasomes were evenly distributed around the cellular periphery. However, in ^PLK4−OEcells stimulated to move directionally by epidermal growth factor (EGF), migrasomes predominantly accumulated at the rear of the cells during a transient period of 55–65 min (Fig. 2F and G). This movement resulted in the more rapid formation of RFs and migrasomes at the rear of the migrating cells, with nearly all rear-positioned migrasomes containing PLK4 (Fig. 2H and I). Furthermore, to exclude potential non-specific effects induced by EGF, we focused on spontaneously migrating cells (without any chemotactic stimuli) and found that these cells also released a substantial number of migrasomes^PLK4 during migration (Fig. 2J). Taken together, these findings suggest that the migrasome-mediated exclusion of PLK4 is a rapid and efficient way for cells to clear excess PLK4.

Fig. 2.

Dynamics of migrasome-mediated removal of excess PLK4. (A) Live-cell imaging of MDA-MB-231 cells overexpressing EGFP, GAPDH-EGFP, or PLK4-EGFP, labelled with the migrasome marker WGA. ^PLK4−OEcells expelled a substantial number of migrasomes. Asterisks indicate migrasomes containing PLK4. Representative extracellular regions surrounding the host cells were magnified. Scale bar, 10 μm for each main image and 2 μm for the zoom in section. (B) Quantification of migrasomes released per cell from time-lapse images in (A). For the Empty-EGFP and GAPDH-EGFP groups, migrasomes were scored as WGA⁺ structures (total migrasomes). For the PLK4-EGFP group, migrasomes were scored as PLK4-loaded migrasomes, operationally defined as dual-positive WGA⁺/PLK4⁺ structures. A total of 30 cells from three independent experiments were included for each group. Data are presented as mean ± SD. (C) Changes in the two-dimensional area of cells overexpressing EGFP, GAPDH-EGFP, or PLK4-EGFP. A total of 30 cells from three independent experiments were included for each group. (D) Distribution of extracellular PLK4 surrounding the host cell. Representative regions containing RFs or migrasomes were magnified. Scale bar, 5 μm for each main image and 1 μm for the zoom in section. (E) 3D reconstruction using the Imaris program showing extracellular PLK4 encapsulated within migrasomes. Scale bar, 1 μm. (F) Distribution of migrasomes^PLK4 in immobile cells and migrating cells with the stimulation of EGF addition. Regions around the front, back, left, and right of the host cell were magnified. Scale bar, 10 μm for each main image and 1 μm for the zoom in section. (G) A lollipop plot illustrating the primary distribution direction of migrasomes from host cells with or without EGF stimulation. A total of 20 cells were included in the analysis for each group. (H) Statistical analysis of the migrasome number with or without EGF stimulation during time-lapse imaging. A total of 30 cells from three independent experiments were included for each group. (I) The proportion of rear-positioned migrasomes^PLK4 among all migrasomes of the cell. A total of 30 cells from three independent experiments were included for each group. Data are presented as mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (J) Time-lapse imaging focussing on the bottom area where a ^PLK4−OEcell previously resided before its spontaneously migration. A large number of migrasomes^PLK4 were observed released during their migration. Scale bar, 10 μm for each main image and 5 μm for the zoom in section.

TSPAN6 mediates the formation of migrasomes^PLK4

Next, we sought to understand how migrasomes^PLK4 are formed. Wild-type (WT) MDA-MB-231 cells and MDA-MB-231 ^PLK4−OEcells were lysed, followed by immunoprecipitation with a PLK4 antibody (Fig. 3A). The immunoprecipitated proteins were then subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. A total of 203 proteins that were highly enriched in ^PLK4−OEcells were potentially associated with PLK4 (Table S2, Fig. S3A–C). Among these, representative centrosome-related proteins, including cytoskeleton-associated protein 5 (CKAP5), nuclear mitotic apparatus protein 1 (NUMA1), dynein cytoplasmic 1 heavy chain 1 (DYNC1H1), DExH-box helicase 9 (DHX9), and kinesin family member 5B (KIF5B), were identified (Fig. 3B). In gene ontology (GO) enrichment analysis, the 203 proteins were significantly enriched not only in “cytoskeleton organisation,” “protein complex assembly,” and “mitotic cell cycle transition,” which are well-known biological processes regulated by PLK4,^29,30 but also in “protein targeting to membrane,” “endomembrane organisation,” and “actin filament organisation,” which are closely associated with migrasomes (Fig. 3C).^22,31 To identify which protein(s) facilitate the incorporation of PLK4 into migrasomes, we intersected the set of 203 potential PLK4-interacting proteins with the previously published in vitro purified migrasome proteome.²² This analysis narrowed down four protein candidates: integrin subunit beta 1 (ITGB1), tetraspanin 6 (TSPAN6), transmembrane P24 trafficking protein 10 (TMED10), and vesicular integral-membrane protein 36 (LMAN2) (Fig. 3D). Next, we examined the interaction between the four proteins and PLK4 by performing co-immunoprecipitation experiments using cell lysates from WT MDA-MB-231 cells. We found that PLK4 interacted with TSPAN6 and ITGB1, but not with LMAN2 or TMED10 (Fig. 3E). To confirm the direct binding of TSPAN6 and ITGB1 to PLK4, we expressed and purified these proteins in vitro with GST tags (Fig. 3F) and measured their binding affinities using microscale thermophoresis (MST), a sensitive technique for analysing biomolecular interactions.³² The binding curves revealed that TSPAN6 exhibited significant binding affinity for PLK4 (Kd = 3.9 μM), ITGB1 displayed low binding affinity (Kd = 22.8 μM), while LMAN2 and TMED10 showed almost no binding ability to PLK4 (Fig. 3G, Fig. S3D, E). Given that TSPAN6 showed the highest binding affinity for PLK4, we performed reverse co-immunoprecipitation using an anti-TSPAN6 antibody, which further confirmed the association between endogenous TSPAN6 and PLK4 (Fig. S3F). To gain structural insights, we modelled a putative TSPAN6–PLK4 complex and mapped the predicted interface (Fig. S3G).

Fig. 3.

TSPAN6 interacts with PLK4 for the formation of migrasomes^PLK4. (A) Coomassie Blue staining of proteins immunoprecipitated with an anti-PLK4 antibody from lysates of WT and ^PLK4−OEcells. (B) A total of 203 proteins were highly enriched in ^PLK4−OEcells that were potentially associated with PLK4. (C) GO analysis for the 203 proteins. X-axis represents the number of proteins enriched in each GO term. Colour schema from dark red to light red indicates the significance from high to low. (D) Intersection of the set of 203 proteins with published data on the in vitro purified migrasome proteome to identify protein candidate(s) that directly interact with PLK4. (E) Immunoprecipitation using a PLK4 antibody from cell lysates of WT MDA-MB-231 cells, followed by the analysis of TSPAN6, ITGB1, LMAN2, and TMED10. Whole-cell lysates were used as the input. (F) PLK4 and the four candidate proteins were expressed with a GST tag in Escherichia coli or using the Bac-to-Bac baculovirus expression system. The purified proteins were stained with Coomassie blue. (G) In vitro binding assay between PLK4 and its potentially interaction proteins (TSPAN6, ITGB1) analysed by MST. Inset, thermophoretic movement of fluorescently labelled proteins. Fnorm = F₁/F₀ (Fnorm: normalised fluorescence; F₁: fluorescence after thermodiffusion; F₀: initial fluorescence or fluorescence after T-jump). Kd, dissociation constant. (H) Knockdown efficiency of TSPAN6 in ^PLK4−OEcells analysed by Western blot, with β-actin used as the loading control. Quantification is presented as mean ± SD from three independent biological replicates. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (I) The protein level of PLK4-EGFP in control and TSPAN6-KD cells analysed by Western blot in cell lysates and in isolated migrasomes. β-actin served as the loading control for cell lysates. Migrasomes were isolated from equal volumes of conditioned medium collected from the same number of cells. Quantification is presented as mean ± SD from three independent biological replicates. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (J) The effect of TSPAN6-KD on the formation of RFs and migrasomes in ^PLK4−OEcells detected by immunofluorescence staining. Boxed areas are enlarged. Scale bar, 10 μm for each main image and 2 μm for the zoom in section. (K, L) The number of migrasomes and the percentage of migrasomes^PLK4 relative to total migrasomes per ^PLK4−OEcell with or without TSPAN6-KD. A total of 30 cells derived from three independent experiments were analysed for each group. Data are presented as the mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (M) Immunofluorescence staining of cell adhesion areas of ^PLK4−OEcells under EGF stimulation with or without TSPAN6-KD. Scale bar, 10 μm. (N, O) The area covered by migrasomes released by individual ^PLK4−OEcells under EGF stimulation with or without TSPAN6-KD, and the proportion of migrasomes^PLK4 among all migrasomes of these cells. A total of 30 cells from three independent experiments were included for each group. Data are presented as mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant.

Considering that TSPAN6 is one of the most potent tetraspanins for enhancing migrasome formation when overexpressed,³³ and given its high binding affinity for PLK4, we hypothesised that TSPAN6 mediates the incorporation of PLK4 into migrasomes. Thus, we knocked down TSPAN6 in ^PLK4−OEcells (Fig. 3H) and assessed the cellular level of PLK4 and migrasome formation. Following TSPAN6 knockdown, cellular PLK4 levels increased by approximately 50% compared to the control group (Fig. 3I), and this increase of PLK4 was also observed in additional PLK4-OE cell lines (^PLK4−OEHeLa, ^PLK4−OEA549, ^PLK4−OEMCF-7 and ^PLK4−OET47D) upon TSPAN6 knockdown (Fig. S4A). Consistent with impaired PLK4 export, PLK4 abundance in isolated migrasome fractions was markedly reduced upon TSPAN6 knockdown (Fig. 3I, Fig. S4A). Although TSPAN6 knockdown did not affect the formation of RFs, it significantly reduced migrasome formation around host cells, including ^PLK4−OEcells (^PLK4−OEMDA-MB-231) and all other PLK4-OE models tested (Fig. 3J–L, Fig. S4B–E). Upon EGF stimulation, although ^PLK4−OEcells (as well as other PLK4-OE cell lines) with TSPAN6 knockdown were capable of generating some migrasomes, these migrasomes contained little to no PLK4 protein (Fig. 3M–O, Fig. S4F–I). These data suggest that TSPAN6 has a high affinity for PLK4, mediating its incorporation into migrasomes, as summarised in the schematic model (Fig. S5).

Expression patterns of PLK4 and TSPAN6 in individuals with tumour

Given the fact that elevated PLK4 expression is associated with the induction of apoptosis,^4,20,34,35 tumour cells per se would bear potential apoptosis risk after they complete malignant transformation, and thus have to suppress the excess PLK4 for their continuous proliferation and evolution. Because our PLK4 overexpression model was established in MDA-MB-231 cells and that breast cancer remains the most common cancer worldwide, we performed an in-depth investigation of PLK4 expression patterns in patients with breast cancer. Specifically, we examined expression patterns of PLK4 in breast invasive carcinoma (BRCA) from The Cancer Genome Atlas (TCGA) database and normal breast tissue samples from Genotype-Tissue Expression database (GTEx). In this cross-sectional cohort, PLK4 mRNA levels were higher in stage II tumours than in stage I tumours, whereas stage III and stage IV tumours displayed lower PLK4 expression relative to stage II (Fig. 4A). These results indicate that PLK4 expression is not monotonically increased with stage in this cohort, but instead displays a stage-dependent distribution. Notably, the stage II–III difference should be interpreted as a stage-stratified between-group association in this cross-sectional cohort, rather than a temporal progression with stage. Consistent with this, key clinicopathological features were broadly comparable between stage II and stage III cases, and PLK4 expression differed across receptor status and race groups (Fig. S6A, B). Since PLK4 is the driving factor for centrosome replication, we further analysed the expression levels of several representative centrosome genes, including STIL Centriolar Assembly Protein (STIL), SAS-6 Centriolar Assembly Protein (SASS6), Centrosomal Protein 152 (CEP152), Centrosomal Protein 57 (CEP57), Centriolin (CNTRL), OFD1 Centriole And Centriolar Satellite Protein (OFD1), and Pericentriolar Material 1 (PCM1). Similar to PLK4, these centrosome genes showed stage-dependent differences in the TCGA-BRCA cohort, with lower expression observed in later-stage groups compared with stage II (Fig. 4B). To exclude potential batch effects arising from the inclusion of different sample batches in the TCGA database, we downloaded single-cell transcriptome sequencing data from a recent study comprising 130,246 cells from 26 patients with breast cancer³⁶ and analysed the expression trends of PLK4 and the aforementioned centrosome genes. Consistently, the expression of these genes exhibited a significant decline at stage III, with PLK4, STIL, CEP152, and CNTRL showing particularly pronounced reductions (Fig. 4C, Fig. S6C, D). Importantly, PLK4 expression was detected only in a subset of tumour cells. This PLK4-positive fraction was enriched for G2/M-phase cells and co-expressed canonical proliferation/mitotic regulators, DNA replication/repair and checkpoint genes, and mitotic spindle/kinetochore components (Fig. S6E–G). Moreover, within the PLK4-positive compartment, the distribution of cell-cycle phases differed across clinical stages (Fig. S6H), suggesting stage-associated shifts in proliferative states among PLK4-positive tumour cells. Next, we investigated the association between PLK4 expression and survival in patients with breast cancer. In early-stage breast cancer (N0, no lymph node metastasis), PLK4 expression was not correlated with survival as well as N1 (1–5 lymph node metastases) and N2 (5–10 lymph node metastases) groups (Fig. 4D, Fig. S6I). However, in advanced-stage breast cancer (N3, >10 lymph node metastases), patients with lower PLK4 expression exhibited significantly shorter survival, implying that suppression of PLK4 expression benefits the evolution of cancer cells (Fig. 4D). We further investigated the prognostic interplay between PLK4 expression and hormone receptor status (Fig. S6J). In the PLK4 low-expression subgroup, both ER-positive and PR-positive status were associated with significantly worse overall survival. Specifically, within this subgroup, ER-negative status was associated with poorer survival than ER-positive status during the first 10 years, though this disadvantage reversed after 10 years. A similar temporal pattern was observed for PR status. In contrast, within the PLK4 high-expression subgroup, neither ER nor PR status showed a significant association with patient survival. HER2 status had no significant impact on survival, regardless of PLK4 expression levels. These results reveal a significant interaction between PLK4 expression and hormone receptor status, where hormone receptor positivity is associated with better early but worse late survival only in the context of low PLK4 expression. Moreover, when assessing the activities of all kinases by measuring the phosphorylation levels of their substrates,³⁷ we found that in patients with advanced breast cancer, among all kinases with significantly altered activities (both increased and decreased), PLK4 ranked among the top-2 exhibiting the most pronounced reduction in activity (Fig. 4E). Interestingly, by analysing breast cancer samples from CPTAC, the NCI's largest public repository of proteomic tumour databases, PLK4 protein expression was observed to be significantly negatively correlated with its clearance factor, TSPAN6 (Fig. 4F). To validate the above expression of PLK4 and TSPAN6, we harvested primary tumour cells from PYMT mice at different time points during their tumour development (Fig. 4G). It is noteworthy that early-stage tumour cells exhibited increased PLK4 expression, whereas late-stage tumour cells showed suppressed expression. In contrast, the expression pattern of TSPAN6 was roughly opposite to that of PLK4 (Fig. 4H and I). Collectively, these findings suggest that during the later stages of tumour progression, PLK4 expression is suppressed by tumour cells, while the protein levels of TSPAN6 tend to increase.

Fig. 4.

Differential expression patterns of PLK4 and TSPAN6 in patients with breast cancer. (A) PLK4 mRNA expression in breast invasive carcinoma (BRCA) at different pathological stages analysed from TCGA database. Stages 1–4 are determined based on pathological staging by American Joint Committee on Cancer (AJCC). The box plots show the median, interquartile range (IQR), and whiskers. Kruskal–Wallis test followed by Dunn's post hoc test was used for multiple comparisons, with p < 0.05 considered statistically significant. (B) Stage-specific expression of PLK4 and its related centrosome genes in breast cancer analysed from TCGA database. The box plots show the median, interquartile range (IQR), and whiskers. Kruskal–Wallis test followed by Dunn's post hoc test was used for multiple comparisons, with p < 0.05 considered statistically significant. (C) Single-cell expression dynamics of PLK4 and its related centrosome genes in tumour cells during breast cancer progression. Stage II exhibits the highest number of tumour cells with positive expression of PLK4 and its related centrosome genes, while this number significantly decreases in Stage III. (D) Correlation analysis of PLK4 expression levels with patient survival in early stage (N0, 496 patients) and late stage (N3, 74 patients) breast cancer. Low PLK4 expression was associated with worse survival in N3 patients (Hazard Ratio (HR) = 0.24, 95% confidence interval: 0.06–0.94, p = 0.0272). (E) Fifty kinases with the most significant alterations in enzymatic activity in late-stage compared to early-stage tumour samples from patients with breast cancer analysed from the CPTAC database. Kinase activities were computationally estimated from phosphoproteomic data using kinase-substrate enrichment analysis. (F) Correlation analysis between PLK4 and TSPAN6 protein expression levels in patients with breast cancer from the CPTAC database. A significant negative correlation was observed (Pearson's r = −0.415, p = 0.000214). (G) Schematic representation of mammary tumour development across early, mid, and late stages in MMTV-PyMT mouse model. (H, I) Protein levels of PLK4 (H) and TSPAN6 (I) in mammary tumours at different stages of progression. β-actin was used as the loading control. Band intensities were quantified using ImageJ software, and relative protein abundance was normalised to β-actin. Data are presented as the mean ± SD from three independent repeats. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant.

Targeting TSPAN6 suppresses tumour growth and metastasis

Given the strict regulation of cellular PLK4 and the role of TSPAN6 in the clearance of excess PLK4, we investigated whether targeting TSPAN6 could induce tumour cell apoptosis. ^PLK4−OEcells that were treated with scramble RNA or shTSPAN6 RNA were monitored by the xCELLigence RTCA system. As shown in Fig. 5A, downregulation of TSPAN6 significantly suppressed cell proliferation over a long-term supervision of 60 h, which is further supported by colony formation assay using crystal violet staining (Fig. 5B). Consistently, a large amount of apoptotic signals was detected in shTSPAN6 cells (Fig. 5C and D). To confirm the anti-tumour effect of targeting TSPAN6, we collected primary tumour cells from patients with triple-negative breast cancer (TNBC) (Table S3) and stratified them by PLK4 protein expression into PLK4^high and PLK4^low populations. The majority of cases (15/18) were classified as PLK4^high (Fig. 5E). After implantation of pooled PLK4^high or pooled PLK4^low primary tumour cells into mice, we observed tumour growth in the mice over a period of 60 days, during which all the tumour cells progressed into clearly defined solid tumour masses. Overall, PLK4^high tumours developed slightly faster compared to PLK4^low tumours (Fig. 5F). Additionally, almost all the tumours exhibited active proliferation and a low level of apoptosis, as indicated by Ki67 and cleaved caspase-3 staining, respectively (Fig. 5G and H). Notably, when examining the intercellular spaces of tumour cells in frozen sections of the tumour mass using immunofluorescence, we identified a considerable number of extracellular particles containing PLK4 signals that were co-localised with TSPAN6 in PLK4^high tumours but not in PLK4^low tumours, suggesting that PLK4^high tumour cells in the in vivo microenvironment still exclude PLK4 (Fig. 5I and J, Fig. S7A, B). Correspondingly, in the observed fields, all PLK4^high tumour cells undergoing mitosis consistently exhibited bipolar spindles (Fig. 5K). Importantly, following the initiation of TSPAN6 knockdown using TSPAN6-shRNA lentivirus on Day 20 post-inoculation, a significant reduction in tumour size was observed in PLK4^high tumour mice rather than PLK4^low tumour mice (Fig. 5F). Moreover, the number of PLK4-containing particles in the intercellular spaces of PLK4^high tumour cells decreased significantly (Fig. 5I and J). As a result, intracellular levels of PLK4 protein of PLK4^high tumours increased (Fig. 5L), and these tumour cells undergoing mitosis frequently exhibited multipolar spindles, reflecting an impaired ability to support subsequent proliferation (Fig. 5I and K). Consistently, the Ki67 signal was markedly diminished, while the apoptotic signal showed a sharp increase within PLK4^high tumours instead of PLK4^low tumours (Fig. 5G and H). In response to concerns about potential off-target effects, we further performed a rescue experiment in the PLK4^high PDX model using a single lentiviral vector co-expressing the TSPAN6-targeting shRNA and an shRNA-resistant TSPAN6 cDNA (PLK4^high RESCUE). Intratumoural delivery at day 20 post-implantation markedly restored tumour growth compared with shTSPAN6 alone, accompanied by increased Ki67 staining and reduced cleaved caspase-3 staining (Fig. S7C, D). Importantly, the elevated intracellular PLK4 accumulation induced by TSPAN6 knockdown was attenuated in the PLK4^high RESCUE group (Fig. S7E), consistent with restoration of TSPAN6-dependent PLK4 handling in vivo. These results demonstrate that inhibition of TSPAN6 effectively kills PLK4^high tumours. Given that the lethality of most breast cancers arises from the metastasis of cancer cells, we sought to investigate whether TSPAN6 knockdown could inhibit this process. Forty-five days after primary PLK4^high tumour formation in PDX-mice, which exhibit a relatively stable metastasis rate, the mice were sacrificed for metastasis analysis. Among 30 PDX-mice with primary tumours, 13 showed evidence of distant metastasis in the lung and/or liver (Fig. 5M, N). However, when TSPAN6 was knocked down starting on Day 20 post-inoculation, lung and/or liver metastasis was detected in only 3 out of 30 PDX-mice (Fig. 5M, N). We quantified the number of metastatic nodules by examining serial sections. In the control group, 4–16 lung metastatic nodules and 3–14 liver metastatic nodules were found in distant organs 45 days after primary tumour formation (Fig. 5O, P). In contrast, the numbers of lung and liver metastatic nodules were significantly reduced following TSPAN6 knockdown (Fig. 5O, P).

Fig. 5.

Targeting TSPAN6 suppresses tumour growth and metastasis. (A) Real-time cell proliferation assay of ^PLK4−OEcells treated with scramble or shTSPAN6 RNA. Data are presented as mean values ± SD. (B) Cystal violet staining of shTSPAN6 or control ^PLK4−OEcells after 14 days of culture. Colony number of each group was quantified on the right. Data are presented as the mean ± SD from three independent repeats. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (C) Fluorescence staining for active apoptosis in shTSPAN6 or control ^PLK4−OEcells detected by GreenNuc Caspase-3 Assay Kit. The green fluorescence in nuclei indicate active apoptosis. Scale bar, 20 μm. (D) Western blot analysis of cleaved caspase3 in ^PLK4−OEcells treated with scramble or shTSPAN6 RNA. β-actin was used as the loading control. (E) Western blot analysis of PLK4 protein levels in primary tumour cells derived from 18 patients with triple-negative breast cancer (TNBC). β-Actin served as the loading control (left). Relative PLK4 protein levels were quantified as the ratio of sample to NC (sample/NC). Cases with sample/NC > 1 were classified as PLK4^high (red dots), whereas cases with sample/NC < 1 were classified as PLK4^low (blue dots) (right). (F) Growth curve of tumours in PDX-models treated with scramble or shTSPAN6 RNA. Tumour volumes are presented as mean values ± SD (n = 30 mice per group). p values were calculated by two-way ANOVA. p < 0.05 was considered statistically significant. (G, H) Immunohistochemical staining of Ki67 and cleaved-caspase3 in tumours from PDX-mice treated with scramble or shTSPAN6 RNA. Scale bar, 100 μm. (I) Immunofluorescent staining for PLK4, TSPAN6, and α-tubulin in frozen tumour sections from PLK4^high tumour mice treated with scramble or shTSPAN6 RNA. White arrowheads indicate PLK4 particles colocalized with TSPAN6 in the extracellular space. Yellow arrowheads indicate bipolar spindles in the control group and multipolar spindles in the shTSPAN6 RNA-treated group. Scale bar, 20 μm for each main image and 2 μm for the zoom in section. Regions marked by the arrowheads are shown at higher magnification in the right three columns. (J) Quantification of PLK4-containing particles in the extracellular space of frozen tumour sections from PLK4^high tumour mice treated with scramble or shTSPAN6 RNA. 20 fields of view from tumour sections of three mice per group were analysed. Data are presented as mean ± SD. p values were calculated using unpaired two-tailed Student's t-tests. p < 0.05 was considered statistically significant. (K) Proportion of bipolar spindles in mitotic tumour cells in frozen tumour sections from PLK4^high tumour mice treated with scramble or shTSPAN6 RNA. (L) Western blot analysis of intracellular PLK4 protein levels in PLK4^high tumours derived from PDX-mice treated with scramble or shTSPAN6 RNA. β-Actin was used as the loading control. Quantification is presented as mean ± SD from three independent biological replicates. p values were calculated using unpaired two-tailed Student's t-tests, with p < 0.05 considered statistically significant. (M) Liver and lung tissues were harvested from PDX-mice with or without detectable metastatic lesions. Arrowheads indicate representative metastatic nodules within the organs. (N) Metastatic incidence in the liver and lung tissues of PDX-mice treated with scramble or shTSPAN6 RNA. The number of mice analysed in each group is indicated above the corresponding bar. (O) Metastatic regions in liver and lung tissues visualised by H&E staining. Representative metastatic foci are indicated by arrowheads, with corresponding higher-magnification images shown on the right. Scale bar, 100 μm for each main image and 20 μm for the zoom in section. (P) Quantification of metastatic nodules in the lungs and livers of PDX-mice treated with scramble or shTSPAN6 RNA. Each dot represents an individual mouse bearing metastatic nodules in the indicated organ.

To further validate the in vivo relevance of TSPAN6 in an independent model, we established an MDA-MB-231 cell line–derived xenograft (CDX) and compared three groups: control, shTSPAN6, and an all-in-one rescue group co-expressing shTSPAN6 with an shRNA-resistant TSPAN6 cDNA (231 RESCUE). Consistent with the PDX results, intratumoural shTSPAN6 delivery at day 20 significantly suppressed tumour growth, whereas the rescue restored tumour growth (Fig. S7F). IHC analyses showed decreased Ki67 and increased cleaved caspase-3 signals in shTSPAN6 tumours, which were reversed by rescue (Fig. S7G). Moreover, shTSPAN6 increased intracellular PLK4 accumulation in tumour tissues, and this effect was attenuated in the rescue group (Fig. S7H). In metastasis analyses, shTSPAN6 markedly reduced lung and liver dissemination, while rescue partially restored metastatic burden, as assessed by gross inspection, histopathology, and metastatic nodule quantification (Fig. S7I–L). Collectively, these results indicate that targeting TSPAN6 to block PLK4 exclusion is an effective strategy for suppressing TNBC growth and metastasis.

Discussion

As a critical regulator of centriole duplication, PLK4's protein level is known to be regulated by the ubiquitin-proteasome system via a conserved phosphodegron and autophosphorylation-dependent degradation mechanism.7, 8, 9, 10 However, despite being a relatively efficient protein degradation pathway, the ubiquitin-proteasome system is still limited by the rate of ubiquitination of target proteins and its inability to efficiently degrade protein aggregates.^38,39 This raises the question of whether cells possess alternative mechanisms to regulate PLK4 levels, given the strict stoichiometric control required for this protein. Here, we inducibly overexpressed PLK4 in the breast cancer cell line MDA-MB-231 and observed that, during the early stages of overexpression, the cells formed internal vesicles containing PLK4 that were subsequently expelled as migrasomes. Mechanistically, the migrasomal protein TSPAN6, which has a high affinity for PLK4, interacts with PLK4 and facilitates the formation of migrasomes^PLK4. Knockdown of TSPAN6 resulted in the failure to form migrasomes^PLK4 in cancer cells, leading to apoptosis. Consequently, targeting TSPAN6 effectively inhibited tumour growth and metastasis in in vivo models.

Our study indicates that migrasome-mediated expulsion provides a pathway for PLK4 clearance distinct from proteasomal degradation. This swift and effective mechanism for regulating the abundance of key intracellular proteins is more closely aligned with the cell's survival strategies and evolutionary advantages. Interestingly, in our live-cell imaging, we observed that PLK4, shortly after being inducibly overexpressed, became surrounded by the migrasome marker WGA, even when the absolute abundance of PLK4 was not yet high. Although this surrounding did not form a closed vesicle within the cell, it seems to suggest that for proteins like PLK4, whose abundance is tightly regulated by the cell, the migrasome system targets and monitors them from the very onset of expression, ensuring their orderly clearance from the cell when necessary. A previous study has reported that migrasomes clear damaged mitochondria from the cell,⁴⁰ Compared to mitophagy, migrasome-mediated mitochondrial clearance is likely an auxiliary or complementary mechanism.^41,42 However, migrasome-mediated export may represent a prominent route for limiting PLK4 accumulation under the conditions tested, alongside previously reported PLK4 regulatory mechanisms. Our findings uncover an advanced strategy by which cells control protein levels and suggest that migrasomes may have greater potential in clearing intracellular proteins than in organelle clearance. Nevertheless, whether PLK4 kinase activity affects its recognition by TSPAN6 and recruitment into migrasomes remains unclear and warrants further investigation.

Beyond tumour-intrinsic protein homoeostasis, our findings may intersect with breast cancer progression through the tumour microenvironment (TME). Breast cancer lesions and metastatic niches exhibit substantial TME heterogeneity, with dynamic remodelling of immune and stromal ecosystems revealed by recent single-cell and spatial studies.43, 44, 45 In parallel, extracellular vesicle-mediated communication, particularly signals conveyed by cancer-associated fibroblast–derived vesicles, has been shown to promote invasion, metastatic niche formation, and immune modulation in breast cancer.⁴⁶ In this context, migrasomes^PLK4 expelled into the extracellular space may represent an additional class of tumour-derived extracellular structures with potential TME-relevant consequences. While the present study focuses on the cell-autonomous benefit of PLK4 extrusion, it will be important to determine in future work whether migrasomes^PLK4 are sensed or processed by stromal/immune cells, and whether they influence local inflammatory programs, phagocytic clearance, or metastatic niche conditioning.

Given that centrosome amplification is a hallmark of solid tumour cells, recent years have seen continuous efforts to explore the inhibition of centrosome amplification as a potential anticancer strategy. However, cancer cells lacking centrosomes due to PLK4 inhibition remain capable of forming bipolar spindles and can proliferate indefinitely.⁴⁷ Although a subsequent study has identified a correlation between the ubiquitin ligase TRIM37 and PLK4 through a genome-wide screen and has demonstrated that for specific tumours with amplification of the TRIM37 locus, PLK4 inhibition offers a novel approach for selectively inducing mitotic failure,⁴⁸ targeting PLK4 still cannot be a mainstream anticancer strategy. In fact, if centrosome amplification serves as a driver for tumour initiation by inducing high levels of genomic instability, then in a fully developed cancer cell, such instability may instead become a new burden and obstacle to its subsequent survival. Suppressing the inherent tendency for high PLK4 expression, for example, via PLK4 extrusion, may be a necessary choice for mature cancer cells to maintain stable survival. In our in vivo models, we detected extracellular PLK4 protein co-localised with TSPAN6 in tumour tissue sections. Inhibiting migrasome formation effectively blocked PLK4 extrusion, resulting in prevention of tumour cell proliferation and suppression of metastasis. Thus, our study not only uncovers a unique mechanism by which cells manage excess PLK4, but also provides a new strategy for killing tumours with initial feature of centrosome amplification.

Limitations of the study

Although we identify TSPAN6 as a key determinant of PLK4 loading into migrasomes, the molecular basis of substrate recognition remains incompletely defined. In particular, it is still unclear whether PLK4 kinase activity, autophosphorylation, or ubiquitination status influences its affinity for TSPAN6, recruitment to retraction fibres, or packaging into migrasomes. In addition, we did not directly examine how migrasomes^PLK4 engage the tumour microenvironment. Given the cellular diversity and dynamic remodelling of immune and stromal compartments in breast cancer, it will be important to determine whether PLK4-positive migrasomes are sensed, internalised, or processed by stromal or immune cells, and whether they modulate inflammatory programs, phagocytic clearance, or metastatic niche conditioning.

Contributors

ML conceived and designed the experiments; JM, YG, YJ, QW, MB, TW, YX, NZ, and PW performed the experiments; JM and BX operated mass spectrum equipment; JM and YG analysed bioinformatics data; ML, GW, YW, and BX wrote the manuscript. JM, YG, and ML had accessed and verified the underlying data. All authors read and approved the final version of the manuscript.

Data sharing statement

The data and materials supporting the conclusions of this article are included in the Supplementary Material files and also available from the corresponding authors on request.

Declaration of interests

The authors declare that they have no conflicts of interest.