ABSTRACT
Upregulation of the developmental Wnt planar cell polarity (Wnt/PCP) pathway is observed in many cancers and is associated with cancer development. We have recently shown that PRICKLE1, a core Wnt/PCP pathway component, is a marker of poor prognosis in triple-negative breast cancer (TNBC). PRICKLE1 is phosphorylated by the serine/threonine kinase MINK1 and contributes to TNBC cell motility and invasiveness. However, the identity of the substrates of MINK1 and the role of MINK1 enzymatic activity in this process remain to be addressed. We used a phosphoproteomic strategy to identify MINK1 substrates, including LL5β (also known as PHLDB2). LL5β anchors microtubules at the cell cortex through its association with CLASP proteins to trigger focal adhesion disassembly. LL5β is phosphorylated by MINK1, promoting its interaction with CLASP proteins. Using a kinase inhibitor, we demonstrate that the enzymatic activity of MINK1 is involved in PRICKLE1–LL5β complex assembly and localization, as well as in cell migration. Analysis of gene expression data reveals that the concomitant upregulation of levels of mRNA encoding PRICKLE1 and LL5β, which are MINK1 substrates, is associated with poor metastasis-free survival in TNBC patients. Taken together, our results suggest that MINK1 may represent a potential target for treatment of TNBC.
INTRODUCTION
Triple-negative breast cancer (TNBC) represents 15–20% of breast cancers and is characterized by poor prognosis. The main features of TNBC are molecular heterogeneity, high cell proliferation and metastatic dissemination, and a limited access to specific therapies (Lyons, 2019). Transformation of mammary epithelial cells in cancer leads to the acquisition of specialized motile capacities that allow the most malignant cells to leave the primary tumor and invade the surrounding space and distant organs. For all cancers, including TNBC, such metastatic programs account for 90% of deaths (Hanahan and Weinberg, 2011). Hence, a better understanding of the molecular mechanisms triggering the migration steps of TNBC cells is essential in order to define novel therapeutic targets and approaches. Recently, we and others discovered that an evolutionarily conserved group of genes controlling the establishment and maintenance of a morphogenetic process regulated by Wnt planar cell polarity (Wnt/PCP) signaling is deregulated in solid tumors, including TNBC, and that this alteration is tightly linked to cancer progression and dissemination (Daulat and Borg, 2017; VanderVorst et al., 2019).
The Wnt/PCP pathway utilizes a set of Wnt (co-)receptors and adaptors that include the cytoplasmic scaffold protein PRICKLE1, orthologs of which play pivotal roles in embryonic development of Drosophila (Gubb and García-Bellido, 1982), zebrafish (Veeman et al., 2003) and Xenopus (Takeuchi et al., 2003). This evolutionarily conserved adaptor contains a PET domain at the N terminus followed by three LIM domains and a C-terminal farnesylation site (Jenny et al., 2003). Previously, we have shown that overexpression of PRICKLE1 is a marker of poor prognosis in TNBC (Daulat et al., 2016; 2019). We and others have demonstrated that, at the molecular level, PRICKLE1 regulates the subcellular localization of associated proteins such as VANGL2 (Daulat et al., 2012; Jenny et al., 2003), RICTOR (Daulat et al., 2016), ARHGAP21 and ARHGAP23 (Zhang et al., 2016) to coordinate oriented cellular migration. PRICKLE1 has signaling activity through its association with RICTOR, regulating phosphorylation of AKT kinases in TNBC cell migration (Daulat et al., 2016). During developmental processes in Xenopus, Prickle1 is asymmetrically localized in cells following phosphorylation by the serine/threonine kinase Mink1 (Daulat et al., 2012). MINK1 is a conserved Ste20-like serine/threonine kinase that belongs to the germinal center kinase family, and is involved in the migration of human fibrosarcoma and breast cancer cell lines (Hu et al., 2004). We have previously demonstrated that MINK1 is tightly associated with PRICKLE1 and that the MINK1–PRICKLE1 complex relocates at the cell periphery of TNBC cells upon PRICKLE1 phosphorylation by MINK1 (Daulat et al., 2012).
In the present study, we aimed to identify additional substrates of MINK1 in TNBC cells in order to better understand the signaling cascade regulated by MINK1 and its involvement in the prometastatic program. Using a combined strategy of protein purification and phosphoproteomics, we identified a set of potential MINK1 substrates that included LL5β (also known as PHLDB2), an adaptor protein that is associated with PRICKLE1 (Lim et al., 2016). LL5β is a pleckstrin homology domain protein that is partially located at the cell membrane (Takabayashi et al., 2010) and that anchors microtubules (MTs) through direct association with cytoplasmic linker-associated proteins (CLASPs), a family of MT plus-end-binding proteins (Lansbergen et al., 2006). This LL5β–CLASP interaction leads to the disassembly of focal adhesions and promotes cell migration (Stehbens et al., 2014). We uncovered a two-step phosphorylation cascade triggered by MINK1 that promotes the subcellular localization, association and function of PRICKLE1 and LL5β at the plasma membrane of TNBC cells. In addition, we show that the combined overexpression of mRNA encoding PRICKLE1 and LL5β represents an independent marker of poor prognosis in TNBC.
RESULTS
A screen for MINK1 substrates identifies LL5β as a putative candidate
Using the MDA-MB-231 TNBC cell line as a model, we have previously shown that downregulation of expression of the MINK1–PRICKLE1 complex inhibits TNBC progression (Daulat et al., 2016). MINK1 is thought to directly phosphorylate PRICKLE1 to promote its promigratory function (Daulat et al., 2012). To further explore the MINK1 mechanism of action and identify additional MINK1 substrates, we carried out a phosphoproteomic approach using the stable isotope labeling by amino acids in cell culture (SILAC) methodology followed by phosphopeptide enrichment and identification (Fig. 1A). We generated MDA-MB-231 cells stably transfected with control shRNA or shRNA targeting MINK1 and cultured them in media containing either heavy- or light-isotope labeled arginine following the SILAC procedure. We confirmed the downregulation of MINK1 expression by western blot analysis (Fig. 1B). After lysis and affinity purification, a total of 1323 phosphopeptides of differential abundance were identified by mass spectrometry analysis. Among them, we selected phosphopeptides having a log2 ratio difference below −1, accounting for a total of 44 peptides representing 37 proteins (Fig. 1C; Table S1). Among the phosphopeptides downregulated in the MINK1 shRNA condition, we identified a MINK1 phosphopeptide, validating our procedure. Previously, we have shown that PRICKLE1 is not only a direct partner of MINK1 but also a substrate of the protein kinase (Daulat et al., 2012). We cross-analyzed our phosphoproteomic dataset with the list of the PRICKLE1 complex-associated proteins identified in MDA-MB-231 cells that we published previously to uncover putative direct substrates of the MINK1 kinase (Daulat et al., 2019). This analysis identified five proteins as being both binders and potential substrates of MINK1 (Fig. 1D). Among them, the most prominent protein associated with the MINK1–PRICKLE1 complex was LL5β (PHLDB2). We verified by western blot analysis that total LL5β expression was not altered by MINK1 downregulation (Fig. 1B). We then expressed GFP or GFP–PRICKLE1 in MDA-MB-231 cells and confirmed by co-immunoprecipitation followed by western blot analysis that endogenous LL5β associates with GFP–PRICKLE1 (Fig. 1E), validating our mass spectrometry results (Fig. 1D). Next, we confirmed the existence of the endogenous complex by immunoprecipitation of endogenous PRICKLE1 followed by detection of LL5β and MINK1 by western blot analysis (Fig. 1F). To confirm the specificity of the association of PRICKLE1 with LL5β, we tested in parallel the association with LL5α (also known as PHLDB1), which shares 70% amino acid sequence identity with LL5β. We co-expressed either GFP–LL5α or GFP–LL5β with FLAG–PRICKLE1 in HEK 293T cells. After FLAG immunopurification, we found that LL5β, but not LL5α, was associated with PRICKLE1 (Fig. 1G). To characterize the role of MINK1 in the interaction between PRICKLE1 and LL5β, we used a PRICKLE1 mutant with a deletion of the second LIM domain (PRICKLE1 ΔLIM2), which is required for MINK1 association with PRICKLE1 (Daulat et al., 2012, 1). We observed that Venus–PRICKLE1 ΔLIM2 does not interact with LL5β, suggesting a role of MINK1 in this association. Moreover, the interaction between PRICKLE1 and LL5β was strengthened when using a PRICKLE1 mutant containing a T370D mutation, which mimics PRICKLE1 phosphorylation by MINK1 on threonine 370 (Daulat et al., 2012) (Fig. 1H). Conversely, expression of a PRICKLE1 mutant containing a T370A mutation led to a decrease in the interaction between PRICKLE1 and LL5β. Taken together, our results show that, among the MINK1–PRICKLE1 complex-associated proteins, LL5β represents a putative phosphorylated substrate of the serine/threonine kinase MINK1.
LL5β is associated with PRICKLE1 and is a putative substrate of the serine/threonine kinase MINK1. (A) Schematic of the phosphoproteomic approach used in our study. shRNA Ctl, control non-targeting shRNA. (B) Western blot analysis and detection of MINK1 and LL5β expression in lysates from MDA-MB-231 cells transfected with the indicated shRNAs. Tubulin is shown as a loading control. (C) Histogram reporting the proteins and phosphopeptides regulated in MDA-MB-231 cells upon MINK1 downregulation. The expanded histogram on the right shows phosphopeptides with abundance more than 2-fold lower in the MINK1 shRNA-treated sample than in the shRNA Ctl sample. (D) Venn diagram representation of the cross analysis of the putative MINK1 phosphorylation sites and previously identified PRICKLE1-associated proteins. (E) Co-immunoprecipitation (IP) assay followed by a western blot (WB) analysis to detect the interaction between GFP–PRICKLE1 and endogenous LL5β in MDA-MB-231 cells. (F) Endogenous co-immunoprecipitation of PRICKLE1 with LL5β and MINK1 in MDA-MB-231 cells following treatment with either siRNA targeting PRICKLE1 or non-targeting (NT) control siRNA. (G) Co-immunoprecipitation assay followed by a western blot analysis to determine the selectivity of association between LL5β and PRICKLE1 in HEK 293T cells. LL5β, but not LL5α, binds to PRICKLE1. (H) Co-immunoprecipitation experiment in HEK 293T cells showing that a PRICKLE1 mutant deleted of its LIM2 domain and unable to bind MINK1 does not associate with LL5β (lane 3). A PRICKLE1 mutant with a phosphomimic site (T370D) presents higher binding to LL5β (lane 4) than wild-type PRICKLE1 (lane 2), and a PRICKLE1 mutant unable to be phosphorylated (T370A) shows lower binding to LL5β (lane 5). All data are representative of at least two independent experiments. Lysate blots in E–H represent 5% of the total lysate.
LL5β is associated with PRICKLE1 and is a putative substrate of the serine/threonine kinase MINK1. (A) Schematic of the phosphoproteomic approach used in our study. shRNA Ctl, control non-targeting shRNA. (B) Western blot analysis and detection of MINK1 and LL5β expression in lysates from MDA-MB-231 cells transfected with the indicated shRNAs. Tubulin is shown as a loading control. (C) Histogram reporting the proteins and phosphopeptides regulated in MDA-MB-231 cells upon MINK1 downregulation. The expanded histogram on the right shows phosphopeptides with abundance more than 2-fold lower in the MINK1 shRNA-treated sample than in the shRNA Ctl sample. (D) Venn diagram representation of the cross analysis of the putative MINK1 phosphorylation sites and previously identified PRICKLE1-associated proteins. (E) Co-immunoprecipitation (IP) assay followed by a western blot (WB) analysis to detect the interaction between GFP–PRICKLE1 and endogenous LL5β in MDA-MB-231 cells. (F) Endogenous co-immunoprecipitation of PRICKLE1 with LL5β and MINK1 in MDA-MB-231 cells following treatment with either siRNA targeting PRICKLE1 or non-targeting (NT) control siRNA. (G) Co-immunoprecipitation assay followed by a western blot analysis to determine the selectivity of association between LL5β and PRICKLE1 in HEK 293T cells. LL5β, but not LL5α, binds to PRICKLE1. (H) Co-immunoprecipitation experiment in HEK 293T cells showing that a PRICKLE1 mutant deleted of its LIM2 domain and unable to bind MINK1 does not associate with LL5β (lane 3). A PRICKLE1 mutant with a phosphomimic site (T370D) presents higher binding to LL5β (lane 4) than wild-type PRICKLE1 (lane 2), and a PRICKLE1 mutant unable to be phosphorylated (T370A) shows lower binding to LL5β (lane 5). All data are representative of at least two independent experiments. Lysate blots in E–H represent 5% of the total lysate.
LL5β is a direct substrate of MINK1
To assess whether LL5β is a direct substrate of MINK1, we performed an in vitro kinase assay, mixing recombinant LL5β and MINK1 kinase, followed by mass spectrometry analysis. To do so, we generated recombinant GST–LL5β and GST–PRICKLE1 in Escherichia coli, and 2 µg of the resulting affinity-purified proteins were submitted to an in vitro kinase assay using the commercially available recombinant kinase domain of MINK1 in the presence or absence of ATP (Fig. 2A, upper panel). As we have previously characterized threonine 370 of PRICKLE1 as a phosphorylation site for MINK1 (Daulat et al., 2012), we used this posttranslational modification as a positive control in our experiment. Hence, we could verify MINK1 phosphorylation of PRICKLE1 at threonine 370 by western blotting using a homemade antibody directed against this phosphosite (Fig. 2A, lower panel). Data demonstrating the specificity of this antibody are shown in Fig. S1A,B. After in vitro phosphorylation by MINK1, recombinant PRICKLE1 and LL5β were analyzed by mass spectrometry to identify the phosphorylation sites. Label-free quantification (LFQ) was used, and the most prominent phosphorylated sites detected were ranked according to their relative LFQ intensity. Our analysis confirmed that PRICKLE1 is phosphorylated by MINK1 at threonine 370 as well as at other sites (Fig. 2B). Two phosphorylation sites, threonine 894 and threonine 217, were identified in LL5β (Fig. 2B), threonine 894 being one of the phosphosites detected in our SILAC screen (Fig. 1B). Alignment of the amino acid sequence encompassing threonine 894 showed that this sequence is conserved among vertebrate species (human, zebrafish and mouse; Fig. 1C). In addition, this sequence fits very well with the typical consensus sequence motif of the MINK1 family kinase activity (Wang et al., 2016). Of note, this phosphorylation site is located within the domain of LL5β required for CLASP binding (Fig. 2C) (Lansbergen et al., 2006). Taken together, our data show that MINK1 can directly phosphorylate LL5β at threonine 894.
LL5β is a direct substrate of MINK1. (A) Upper panel: silver staining of an acrylamide gel loaded with purified GST–PRICKLE1 and GST–LL5β subjected to an in vitro kinase assay using the kinase domain of MINK1. Lower panel: western blot (WB) analysis of the PRICKLE1 kinase assay using an anti-pT370-PRICKLE1 antibody. (B) Representation of phosphopeptide abundance for phosphorylation sites identified in GST–PRICKLE1 and GST–LL5β by mass spectrometry analysis following the MINK1 in vitro kinase assay. Data are presented as mean+s.d. (C) Schematic of LL5β domains and an amino acid (a.a.) sequence alignment of the peptide containing threonine 894 (asterisk) phosphorylated by MINK1. Orthologous LL5β sequences from human, zebrafish (DANRE) and mouse are shown. PH, pleckstrin homology. All data are representative of at least two independent experiments.
LL5β is a direct substrate of MINK1. (A) Upper panel: silver staining of an acrylamide gel loaded with purified GST–PRICKLE1 and GST–LL5β subjected to an in vitro kinase assay using the kinase domain of MINK1. Lower panel: western blot (WB) analysis of the PRICKLE1 kinase assay using an anti-pT370-PRICKLE1 antibody. (B) Representation of phosphopeptide abundance for phosphorylation sites identified in GST–PRICKLE1 and GST–LL5β by mass spectrometry analysis following the MINK1 in vitro kinase assay. Data are presented as mean+s.d. (C) Schematic of LL5β domains and an amino acid (a.a.) sequence alignment of the peptide containing threonine 894 (asterisk) phosphorylated by MINK1. Orthologous LL5β sequences from human, zebrafish (DANRE) and mouse are shown. PH, pleckstrin homology. All data are representative of at least two independent experiments.
Inhibition of MINK1 catalytic activity phenocopies loss of MINK1
The role of MINK1 in the promigratory process of TNBC cells has previously been demonstrated using siRNA- or shRNA-mediated knockdown experiments (Daulat et al., 2016). We decided to assess the role of the catalytic activity of MINK1 in cell migration using a commercially available protein kinase inhibitor. As MINK1 shares a high degree of homology with its paralogs TNIK and MAP4K4 in the region of the ATP-binding pocket (Read et al., 2019), we tested KY05009, a chemical compound previously described as a TNIK inhibitor (Kim et al., 2014). We set up an in vitro kinase assay using the recombinant active kinase domain of MINK1 and found that KY05009 inhibited MINK1 with an IC50 of 1.2 nM (Fig. 3A). We next immunoprecipitated GFP–PRICKLE1 expressed in MDA-MB-231 cells and detected PRICKLE1 phosphorylation by western blot analysis using a homemade antibody to detect PRICKLE1 phosphorylation at threonine 370 (anti-pT370-PRICKLE1; Fig. S1A,B). We observed that siRNA-mediated downregulation of MINK1 led to a decrease of PRICKLE1 phosphorylation (Fig. 3B). We next treated MDA-MB-231 cells with various concentrations of KY05009 and observed a dose-dependent inhibition of PRICKLE1 phosphorylation (Fig. 3C). We have previously shown that downregulation of MINK1 leads to a decrease of total AKT phosphorylation and activity (Daulat et al., 2016). Serum-starved or fetal calf serum (FCS)-treated MDA-MB-231 cells were treated with 1 µM KY05009 and analyzed for AKT phosphorylation. Stimulation with 5% FCS led to phosphorylation of AKT on serine 473. This phosphorylation was strongly decreased by KY05009 treatment (Fig. S2A). These data confirm that MINK1 phosphorylates PRICKLE1 in cells.
Characterization of KY05009, a MINK1 inhibitor. (A) In vitro kinase assay using a PRICKLE1 peptide encompassing the MINK1 phosphorylation site. Determination of the IC50 of KY05009. Data are presented as mean±s.e.m. of three experiments. (B) MDA-MB-231 cells stably expressing GFP–PRICKLE1 were subjected to treatment with siRNA targeting MINK1 or non-targeting (NT) control siRNA. Following GFP immunoprecipitation (IP), phosphorylation status of PRICKLE1 at threonine 370 was assessed by western blotting (WB) using anti-pT370-PRICKLE1 antibody. Tubulin is shown as a loading control. Lysate blots represent 5% of the total lysate. (C) MDA-MB-231 cells stably expressing GFP–PRICKLE1 were treated with the indicated concentrations of KY05009 before GFP immunoprecipitation and western blotting. Phosphorylation of PRICKLE1 at threonine 370 is dependent on MINK1 catalytic activity. The ratio of phosphorylated PRICKLE1 to GFP is indicated for the blots shown. (D) Anti-vinculin staining of MDA-MB-231 cells seeded on collagen-coated coverslips. Cells were treated with the indicated combinations of siRNA and KY05009 (1 μM) prior to staining. (E) Quantification of the area of the structures stained by vinculin as described in D. (F) Phalloidin staining of F-actin in MDA-MB-231 cells seeded on collagen-coated coverslips. Cells were treated with the indicated combinations of siRNA and KY05009 (1 μM) prior to staining. (G) Measurement of cell area for cells as described in F. (H) Western blot analysis to confirm expression of the indicated constructs in MDA-MB-231 cells. (I) Top panel: α-vinculin staining of MDA-MB-231 cells seeded on collagen-coated coverslips. Bottom panel: expression of the indicated GFP constructs. (J) Quantification of the area of the structures stained by vinculin in cells as described in I. Graphs in E,G and J represent mean±s.e.m. of the number of cells indicated on the graph (E,G) or n>25 cells (J). Data shown in B–J are from one experiment representative of three independent experiments. **P<0.01; ***P<0.001 (two-tailed unpaired Student's t-test). Scale bars: 20 μm.
Characterization of KY05009, a MINK1 inhibitor. (A) In vitro kinase assay using a PRICKLE1 peptide encompassing the MINK1 phosphorylation site. Determination of the IC50 of KY05009. Data are presented as mean±s.e.m. of three experiments. (B) MDA-MB-231 cells stably expressing GFP–PRICKLE1 were subjected to treatment with siRNA targeting MINK1 or non-targeting (NT) control siRNA. Following GFP immunoprecipitation (IP), phosphorylation status of PRICKLE1 at threonine 370 was assessed by western blotting (WB) using anti-pT370-PRICKLE1 antibody. Tubulin is shown as a loading control. Lysate blots represent 5% of the total lysate. (C) MDA-MB-231 cells stably expressing GFP–PRICKLE1 were treated with the indicated concentrations of KY05009 before GFP immunoprecipitation and western blotting. Phosphorylation of PRICKLE1 at threonine 370 is dependent on MINK1 catalytic activity. The ratio of phosphorylated PRICKLE1 to GFP is indicated for the blots shown. (D) Anti-vinculin staining of MDA-MB-231 cells seeded on collagen-coated coverslips. Cells were treated with the indicated combinations of siRNA and KY05009 (1 μM) prior to staining. (E) Quantification of the area of the structures stained by vinculin as described in D. (F) Phalloidin staining of F-actin in MDA-MB-231 cells seeded on collagen-coated coverslips. Cells were treated with the indicated combinations of siRNA and KY05009 (1 μM) prior to staining. (G) Measurement of cell area for cells as described in F. (H) Western blot analysis to confirm expression of the indicated constructs in MDA-MB-231 cells. (I) Top panel: α-vinculin staining of MDA-MB-231 cells seeded on collagen-coated coverslips. Bottom panel: expression of the indicated GFP constructs. (J) Quantification of the area of the structures stained by vinculin in cells as described in I. Graphs in E,G and J represent mean±s.e.m. of the number of cells indicated on the graph (E,G) or n>25 cells (J). Data shown in B–J are from one experiment representative of three independent experiments. **P<0.01; ***P<0.001 (two-tailed unpaired Student's t-test). Scale bars: 20 μm.
We have previously shown that downregulation of MINK1 in TNBC cells leads to increases in focal adhesion size and cell spreading, and to a decrease in cell motility (Daulat et al., 2016). We next treated MDA-MB-231 cells with KY05009 and in parallel with siRNA targeting MINK1, then stained the cells with anti-vinculin antibody to assess the size of focal adhesions by immunofluorescence. We observed that KY05009 treatment or downregulation of MINK1 expression increased the size of focal adhesions (Fig. 3D,E) and cell spreading (Fig. 3F,G). We did not observe a cumulative effect on the size of focal adhesions and cell spreading when cells were co-treated with KY05009 and siRNA targeting MINK1 (Fig. 3D–G). To further confirm the role of the catalytic activity of MINK1, we generated MDA-MB-231 cells expressing either wild-type MINK1, kinase-dead MINK1 (MINK1 K54R) or kinase domain-deleted MINK1 (MINK1 KinDel) (Fig. 3H). Expression of these mutant forms led to an increase of focal adhesion sizes (Fig. 3I, quantification in Fig. 3J), suggesting a dominant-negative role of MINK1, which is in line with previously published data using a HT1080 cell line (Hu et al., 2004). To demonstrate a role for MINK1, and not its paralogs, in cellular remodeling, we treated MDA-MB-231 cells with siRNAs targeting MINK1, MAP4K4 or TNIK, seeded them on collagen-coated coverslips and stained for actin (Fig. S2B,C). The efficiency of siRNA-mediated knockdown was validated using RT-qPCR (Fig. S2D). Only MINK1 downregulation led to a cellular phenotype similar to that observed following KY05009 treatment, with a strong reorganization of the cytoskeleton and increased cell spreading (Fig. S2B, quantification in Fig. S2C). Taken together, our data suggest that, in MDA-MB-231 cells, KY05009 behaves as a MINK1 inhibitor that phenocopies loss of MINK1 (Daulat et al., 2016).
MINK1 activity promotes the association of PRICKLE1 with LL5β and stabilizes PRICKLE1 subcellular localization
We next decided to explore the role of MINK1 catalytic activity in the formation and localization of the MINK1–PRICKLE1–LL5β complex. We expressed epitope-tagged versions of the three proteins in HEK 293T cells and assessed their interaction. We found that GFP–LL5β associated with FLAG–MINK1 (Fig. 4A, lane 4) and that co-expression of HA–PRICKLE1, a direct MINK1 interactor, strongly potentiated the association between the two proteins (Fig. 4A, lane 3). We next used MDA-MB-231 cells stably expressing GFP–PRICKLE1, which endogenously express LL5β. We performed an immunoprecipitation of GFP–PRICKLE1 using GFP antibodies and detected the presence of LL5β associated with PRICKLE1. This interaction was decreased by two independent siRNAs targeting MINK1 (Fig. 4B). To determine whether the catalytic activity of MINK1 is required for this association, we treated MDA-MB-231 cells with 1 µM KY05009. This treatment clearly decreased the association between PRICKLE1 and LL5β (Fig. 4C). As demonstrated in Fig. 1H, PRICKLE1 is bound to LL5β through its LIM2 domain, and this interaction is positively regulated by the phosphorylation of PRICKLE1 by MINK1 on threonine 370. We next examined, using immunofluorescence and confocal microscopy, the subcellular localization of endogenous PRICKLE1 in MDA-MB-231 cells using a homemade anti-PRICKLE1 antibody, the specificity of which was validated using two independent siRNAs targeting PRICKLE1 (Fig. S1C). We observed that, in MDA-MB-231 cells, PRICKLE1 was localized at the plasma membrane in an actin-rich subregion within the lamellipodia. Downregulation of MINK1 (Fig. 4D, quantification in Fig. 4E) or inhibition of its catalytic activity (Fig. 4F, quantification in Fig. 4G) led to a profound change in cellular morphology, as previously described (Daulat et al., 2016). In MINK1-deficient cells, we observed a drastic relocalization of PRICKLE1 from the plasma membrane to actin bundles. Similar observations were obtained using SUM149PT, another TNBC cell model (Fig. S3). Furthermore, we monitored LL5β subcellular localization upon MINK1 and PRICKLE1 downregulation by siRNAs or treatment with KY05009. We did not observe any influence of MINK1 or PRICKLE1 expression on LL5β membrane localization (Fig. 4H, quantification in Fig. 4I). We also monitored PRICKLE1 localization in cells with downregulated LL5β expression and observed a relocalization of PRICKLE1 to actin bundles (Fig. S4A, quantification in Fig. S4B). We then generated a mutant of LL5β, which we named LL5β (-ADA), where serine 892 and threonine 894 were mutated to alanine. We mutated both residues to exclude a possible alternative phosphorylation of serine 892 by MINK1. As in MINK1-deficient cells, expressing GFP–LL5β (-ADA) led to a relocalization of PRICKLE1 from the plasma membrane to actin bundles, compared to the localization observed in cells expressing wild-type LL5β (Fig. S4C, quantification in Fig. S4D). Finally, we did not notice any effect on CLASP2 localization when GFP–LL5β or GFP–LL5β (-ADA) were expressed in MDA-MB-231 cells (Fig. S4E, quantification in Fig. S4F). Taken together, our data suggest that MINK1 activity promotes and stabilizes PRICKLE1, but not LL5β, localization at the plasma membrane.
MINK1 catalytic activity promotes PRICKLE1 association with LL5β and localization at the plasma membrane. (A) Co-immunoprecipitation experiment showing the association between epitope-tagged MINK1, PRICKLE1 and LL5β in HEK 293T cells. IP, immunoprecipitation; WB, western blotting. (B) Co-immunoprecipitation in MDA-MB-231 cells of endogenous LL5β with GFP–PRICKLE1 in the presence of control non-targeting (NT) or MINK1 siRNAs (M10, M09). Tubulin is shown as a loading control. (C) Co-immunoprecipitation of endogenous LL5β with GFP–PRICKLE1 under KY05009 treatment (1 μM) or DMSO vehicle control. (D) PRICKLE1 and actin antibody staining of MDA-MB-231 cells treated with a siRNA targeting MINK1 or control NT siRNA. (E) Quantification of PRICKLE1 subcellular localization at the cell cortex (case A) or on actin fibers (case B) for cells as shown in D. n>200 cells per condition. (F) MDA-MB-231 cells treated with KY05009 or DMSO vehicle control and stained with PRICKLE1 and actin antibody. (G) Quantification of PRICKLE1 subcellular localization, performed as described in E, for cells as shown in F. (H) MDA-MB-231 cells treated with the indicated siRNAs and stained for tubulin and LL5β. (I) Quantification of LL5β plasma membrane localization for cells as shown in H. n>100 cells per condition. Images and blots are representative of three independent experiments. Lysate blots in A–C represent 5% of the total lysate. Nuclei in D,F and H were stained using DAPI. Scale bars: 20 µm.
MINK1 catalytic activity promotes PRICKLE1 association with LL5β and localization at the plasma membrane. (A) Co-immunoprecipitation experiment showing the association between epitope-tagged MINK1, PRICKLE1 and LL5β in HEK 293T cells. IP, immunoprecipitation; WB, western blotting. (B) Co-immunoprecipitation in MDA-MB-231 cells of endogenous LL5β with GFP–PRICKLE1 in the presence of control non-targeting (NT) or MINK1 siRNAs (M10, M09). Tubulin is shown as a loading control. (C) Co-immunoprecipitation of endogenous LL5β with GFP–PRICKLE1 under KY05009 treatment (1 μM) or DMSO vehicle control. (D) PRICKLE1 and actin antibody staining of MDA-MB-231 cells treated with a siRNA targeting MINK1 or control NT siRNA. (E) Quantification of PRICKLE1 subcellular localization at the cell cortex (case A) or on actin fibers (case B) for cells as shown in D. n>200 cells per condition. (F) MDA-MB-231 cells treated with KY05009 or DMSO vehicle control and stained with PRICKLE1 and actin antibody. (G) Quantification of PRICKLE1 subcellular localization, performed as described in E, for cells as shown in F. (H) MDA-MB-231 cells treated with the indicated siRNAs and stained for tubulin and LL5β. (I) Quantification of LL5β plasma membrane localization for cells as shown in H. n>100 cells per condition. Images and blots are representative of three independent experiments. Lysate blots in A–C represent 5% of the total lysate. Nuclei in D,F and H were stained using DAPI. Scale bars: 20 µm.
The LL5β–CLASP2 interaction is promoted by MINK1-dependent phosphorylation
Previous results have demonstrated that LL5β anchors MTs at the plasma membrane through its direct association with CLASPs, a family of MT plus-end-binding proteins (Lansbergen et al., 2006; Stehbens et al., 2014). CLASP2 is present in the list of PRICKLE1 complex-associated proteins purified from MDA-MB-231 cells (Daulat et al., 2019). Interestingly, threonine 894 of LL5β phosphorylated by MINK1 lies within the domain involved in binding to CLASP proteins (Fig. 2C) (Lansbergen et al., 2006). To determine whether LL5β and CLASP1 or CLASP2 are required for MDA-MB-231 cell migration, we selected specific siRNAs targeting these proteins (Fig. S5A–C) and demonstrated their importance in cell motility by conducting single-cell migration assays (Fig. S5D,E). These results confirmed data previously published by others in the same cellular context (Astro et al., 2014) and in gastric cancer cells (Lim et al., 2016). We co-immunoprecipitated GFP–LL5β with endogenous CLASP2 in MDA-MB-231 cells (Fig. 5A). However, downregulation of MINK1 strongly diminished this interaction, suggesting that the association between LL5β and CLASP2 is partially dependent on the MINK1–PRICKLE1 complex. We assessed the colocalization between LL5β and CLASP2 by immunofluorescence in MDA-MB-231 cells and confirmed that LL5β was colocalized with CLASP2 within the lamellipodia. Downregulation of MINK1 by siRNA treatment decreased colocalization of LL5β and CLASP2 (Fig. 5B) and led to fewer lamellipodia per cell (Fig. 5C). Furthermore, we quantified the colocalization of LL5β and CLASP2 in regions enriched in tubulin at the plasma membrane and observed a decrease of the Pearson correlation coefficient (PCC) in the absence of MINK1 (Fig. 5D). Thus, MINK1 promotes CLASP2–LL5β interaction and colocalization at the plasma membrane.
Phosphorylation of LL5β is required for its interaction with CLASP in MDA-MB-231 cells. (A) Co-immunoprecipitation of LL5β and CLASP2 is decreased upon loss of MINK1. MDA-MB-231 cells expressing GFP– LL5β were treated with MINK1 siRNA or control non-targeting (NT) siRNA before immunoprecipitation (IP) and analysis by western blotting (WB), as indicated. Tubulin is shown as a loading control. The ratio of CLASP2 to GFP is indicated for the blots shown. (B) Colocalization of LL5β and CLASP2 in MDA-MB-231 cells under treatment with the indicated siRNAs. White boxes mark regions shown in the magnified images below as examples of areas measured for colocalization analysis, with PCC values indicated. (C) Quantification of lamellipodia per cell for MDA-MB-231 cells treated with the indicated siRNAs. (D) PCC of colocalization between CLASP2 and LL5β for cells as described in B. (E) Co-immunoprecipitation of LL5β and CLASP2 is decreased for the LL5β (-ADA) mutant, which has mutation of the threonine 894 phosphosite. Lysates of MDA-MB-231 cells expressing GFP– LL5β or GFP– LL5β (-ADA) were used for immunoprecipitation and western blotting as indicated. The ratio of CLASP2 to GFP is indicated for the blots shown. (F) Influence of KY05009 treatment on the interaction between CLASP2 and the wild-type and mutant forms of LL5β in MDA-MB-231 cells. The ratio of CLASP2 to GFP is indicated for the blots shown. (G) Colocalization of LL5β and CLASP2 in MDA-MB-231 cells under treatment with 1 µM KY05009 or DMSO vehicle for 16 h. Magnified views beneath the main images show examples of areas measured for colocalization analysis, with PCC values indicated. (H) Quantification of lamellipodia per cell for MDA-MB-231 cells treated with DMSO or 1 µM KY05009. (I) PCC of colocalization between CLASP2 and LL5β for cells as described in G. Images and blots are representative of three independent experiments. Lysate blots in A,E and F represent 5% of the total lysate. Data in C,D,H,I are presented as mean±s.e.m. of n cells as indicated in the figure. Scale bars: 20 µm. Statistics were performed using the results of three independent experiments. *P<0.05; **P<0.01; ***P<0.001 (two-tailed unpaired Student's t-test).
Phosphorylation of LL5β is required for its interaction with CLASP in MDA-MB-231 cells. (A) Co-immunoprecipitation of LL5β and CLASP2 is decreased upon loss of MINK1. MDA-MB-231 cells expressing GFP– LL5β were treated with MINK1 siRNA or control non-targeting (NT) siRNA before immunoprecipitation (IP) and analysis by western blotting (WB), as indicated. Tubulin is shown as a loading control. The ratio of CLASP2 to GFP is indicated for the blots shown. (B) Colocalization of LL5β and CLASP2 in MDA-MB-231 cells under treatment with the indicated siRNAs. White boxes mark regions shown in the magnified images below as examples of areas measured for colocalization analysis, with PCC values indicated. (C) Quantification of lamellipodia per cell for MDA-MB-231 cells treated with the indicated siRNAs. (D) PCC of colocalization between CLASP2 and LL5β for cells as described in B. (E) Co-immunoprecipitation of LL5β and CLASP2 is decreased for the LL5β (-ADA) mutant, which has mutation of the threonine 894 phosphosite. Lysates of MDA-MB-231 cells expressing GFP– LL5β or GFP– LL5β (-ADA) were used for immunoprecipitation and western blotting as indicated. The ratio of CLASP2 to GFP is indicated for the blots shown. (F) Influence of KY05009 treatment on the interaction between CLASP2 and the wild-type and mutant forms of LL5β in MDA-MB-231 cells. The ratio of CLASP2 to GFP is indicated for the blots shown. (G) Colocalization of LL5β and CLASP2 in MDA-MB-231 cells under treatment with 1 µM KY05009 or DMSO vehicle for 16 h. Magnified views beneath the main images show examples of areas measured for colocalization analysis, with PCC values indicated. (H) Quantification of lamellipodia per cell for MDA-MB-231 cells treated with DMSO or 1 µM KY05009. (I) PCC of colocalization between CLASP2 and LL5β for cells as described in G. Images and blots are representative of three independent experiments. Lysate blots in A,E and F represent 5% of the total lysate. Data in C,D,H,I are presented as mean±s.e.m. of n cells as indicated in the figure. Scale bars: 20 µm. Statistics were performed using the results of three independent experiments. *P<0.05; **P<0.01; ***P<0.001 (two-tailed unpaired Student's t-test).
We next examined whether the association between LL5β and CLASP was regulated by MINK1 phosphorylation of threonine 894. We performed immunoprecipitation with an anti-GFP antibody in lysates from MDA-MB-231 cells stably expressing GFP–LL5β or GFP–LL5β (-ADA). Western blot analysis showed that the association between LL5β and CLASP2 was decreased by 2-fold with GFP–LL5β (-ADA) compared to the association with GFP–LL5β (Fig. 5E). To further evaluate whether the remaining interaction could be due to other MINK1-dependent phosphorylation events, we treated the cells with KY05009 prior to lysis and immunopurification. We observed that treatment of MDA-MB-231 cells expressing GFP–LL5β with KY05009 led to a decrease in GFP–LL5β–CLASP2 binding to the same range as that between GFP–LL5β (-ADA) and CLASP2 (Fig. 5F, left panel). However, KY05009 treatment did not further decrease the binding between GFP–LL5β (-ADA) and CLASP2, suggesting that no other phosphorylation event due to MINK1 activity is implicated in the LL5β–CLASP2 association (Fig. 5F, right panel). As with MINK1 siRNA treatment (Fig. 5B), KY05009 treatment led to a decrease in colocalization between LL5β and CLASP2 at the plasma membrane, and to a decrease in the number of lamellipodia per cell (Fig. 5G–I). Taken together, our data show that MINK1 phosphorylates LL5β within the CLASP2-binding site and that this event promotes the complete LL5β–CLASP2 association at the cell cortex.
Targeting MINK1 catalytic activity inhibits cell motility
The pharmacological inhibition of MINK1 by KY05009 demonstrates a role of the kinase activity of MINK1 in the assembly and localization of the PRICKLE1–LL5β–CLASP1/2 complex and on cell morphology. We next monitored single-cell motility and observed a decrease in the cumulative distance traveled by MDA-MB-231 cells expressing either MINK1 K54R or MINK1 KinDel (Fig. 6A, quantification in Fig. 6B). These data show a dominant-negative effect of MINK1 mutants with deficient kinase activity. A similar inhibition was obtained when cells were treated with KY05009 (Fig. 6C, quantification in Fig. 6D). No cumulative effect was observed when cells were simultaneously treated with KY05009 and siRNA targeting MINK1 (Fig. 6C, quantification in Fig. 6D). Furthermore, MDA-MB-231 cell migration was also decreased under KY05009 treatment in a wound healing assay (Fig. S6A, quantification in Fig. S6B). These results were well correlated with our previous results using siRNAs targeting MINK1 and PRICKLE1 in MDA-MB-231 cells (Daulat et al., 2016). We obtained similar results using SUM149PT, another TNBC cell line, treated with either siRNAs targeting MINK1 and PRICKLE1 (Fig. S6C) or KY05009 (Fig. S6D).
Inhibition of MINK1 catalytic activity inhibits cell motility. (A) Single-cell migration of MDA-MB-231 cells expressing the indicated MINK1 constructs. (B) Quantification of cumulative and Euclidean distances migrated for the experiment shown in A. (C) Single-cell migration of MDA-MB-231 cells that were treated with either MINK1 siRNA or non-targeting (NT) control siRNA, and either DMSO or 1 μM KY05009, as indicated. (D) Quantification of cumulative and Euclidean distances migrated for the experiment shown in C. (E) Single-cell migration of MDA-MB-231 cells expressing the indicated LL5β constructs. (F) Quantification of cumulative and Euclidean distances migrated for the experiment shown in E. Experiments were performed at least three times, and 15 cells were analyzed for each condition. A,C and E show individual cell tracks plotted in Excel. Mean±s.e.m. are represented in B,D and F. Statistical analysis was performed against the control conditions using a two-tailed unpaired Student's t-test. *P<0.05; **P<0.01; ***P<0.001; n.s., not significant.
Inhibition of MINK1 catalytic activity inhibits cell motility. (A) Single-cell migration of MDA-MB-231 cells expressing the indicated MINK1 constructs. (B) Quantification of cumulative and Euclidean distances migrated for the experiment shown in A. (C) Single-cell migration of MDA-MB-231 cells that were treated with either MINK1 siRNA or non-targeting (NT) control siRNA, and either DMSO or 1 μM KY05009, as indicated. (D) Quantification of cumulative and Euclidean distances migrated for the experiment shown in C. (E) Single-cell migration of MDA-MB-231 cells expressing the indicated LL5β constructs. (F) Quantification of cumulative and Euclidean distances migrated for the experiment shown in E. Experiments were performed at least three times, and 15 cells were analyzed for each condition. A,C and E show individual cell tracks plotted in Excel. Mean±s.e.m. are represented in B,D and F. Statistical analysis was performed against the control conditions using a two-tailed unpaired Student's t-test. *P<0.05; **P<0.01; ***P<0.001; n.s., not significant.
To further explore the role of MINK1-dependent LL5β phosphorylation in cell migration, MDA MB-231 cells expressing either GFP–LL5β or GFP–LL5β (-ADA) were seeded and cell motility was measured. We observed a strong decrease in cell motility of GFP–LL5β (-ADA)-expressing cells in comparison to that of GFP–LL5β-expressing cells (Fig. 6E, quantification in Fig. 6F). We noticed that GFP–LL5β expression led to a slight decrease in cell motility; however, no change of Euclidian distance was observed. In conclusion, our data demonstrate that the catalytic activity of MINK1 is required for TNBC cell migration.
Unfavorable prognostic value of combined PRICKLE1 and LL5β expression in TNBC
We have previously shown that upregulation of PRICKLE1 expression is associated with poor metastasis-free survival (MFS) in basal breast cancer (Daulat et al., 2016, 2019), a molecular subtype mainly composed of TNBC, as well as in TNBC (Daulat et al., 2019). In our present series of TNBC, which includes 471 informative patients, we confirmed that PRICKLE1 upregulation was associated with shorter MFS, with 72% 5-year MFS (95% CI, 64–81%) in the PRICKLE1-low group versus 56% (95% CI, 40–64%) in the PRICKLE1-high group (P=0.0027, log-rank test) (Fig. 7A). We next assessed the prognostic value of LL5β (referred to hereafter as PHLDB2) expression in this patient series. We observed that upregulation of PHLDB2 expression was associated with shorter MFS, with 5-year MFS of 70% (95% c.i., 64–77%) versus 47% (95% c.i., 38–58%) for the PHLDB2-low group and PHLDB2-high group, respectively (P=0.0001, log-rank test; Fig. 7B). Since both genes did not show correlation of their expression level (Fig. S7A), we tested an eventual complementary prognostic value. Interestingly, the patients with upregulation of expression of both genes (PRICKLE1-high and PHLDB2-high group, also referred to here as high/high) displayed 35% 5-year MFS (95% c.i., 25–49%), whereas patients with downregulation of expression of both genes displayed 74% 5-year MFS (95% c.i., 65–84%), and those with opposite deregulation of both genes (upregulated expression of one gene and downregulated expression of the other) displayed 67% 5-year MFS (95% c.i., 59–76%; P=1.16×10−6, log-rank test, four classes comparison) (Fig. 7C). Since these last three patient groups showed similar MFS, they were merged and thereafter designated as the ‘no PRICKLE1-high and PHLDB2-high’ group (also referred to here as no high/high). The 5-year MFS of this latter group was 69% (95% c.i., 64–76%) versus a 5-year MFS of 35% (95% c.i., 25–49%) for the PRICKLE1-high and PHLDB2-high group (P=7.15×10−8, log-rank test; Fig. 7D). Interestingly, such prognostic value of the combined expression of PRICKLE1 and PHLDB2 was not observed in the hormone receptor-positive HER2 (ERBB2)-negative (HR+/HER2−) and HER2-positive subtypes (Fig. S7B,C). The prognostic complementarity of expression of both genes in TNBC was confirmed using the likelihood ratio (LR) test. The two-gene combination provided more prognostic information than PRICKLE1 alone [change in LR (ΔLR-χ2)=14.45, P<2.2×10−16] and PHLDB2 alone (ΔLR-χ2=9.97, P<2.2×10−16) (Fig. S7D). Finally, in a multivariate analysis for MFS (Wald test; Fig. 7E), the two-gene combination maintained its prognostic value (P=1.90×10−2), whereas the prognostic values of pathological tumor size (pT) and pathological axillary lymph node status (pN) tended to remain significant (P<0.10). Taken together, these data suggested a permissive function of PRICKLE1 and PHLDB2 in the metastatic process of TNBC.
Prognostic value of PRICKLE1 and PHLDB2 expression in TNBC. (A) Kaplan–Meier MFS curves for TNBC patients grouped according to level of PRICKLE1 mRNA expression. The percentage 5-year MFS and number of patients for each group are shown on the graph, along with the total number of patients. (B) As in A, but with patients grouped according to level of PHLDB2 mRNA expression. (C) As in A, but with patients grouped according to both PRICKLE1 and PHLDB2 mRNA expression levels. The patients are separated into four classes (4K). Table shows number of patients and 5-year MFS for each class, as well as P-values from two tests comparing PHLDB2 classes within each PRICKLE1 class. (D) Similar to C, but with patients separated into two classes (2K; ‘high/high’ versus ‘no high/high’). Dashed lines in A–D indicate the PHLDB2-low class. (E) Uni- and multi-variate analyses of MFS in TNBC patients. HR, hazard ratio.
Prognostic value of PRICKLE1 and PHLDB2 expression in TNBC. (A) Kaplan–Meier MFS curves for TNBC patients grouped according to level of PRICKLE1 mRNA expression. The percentage 5-year MFS and number of patients for each group are shown on the graph, along with the total number of patients. (B) As in A, but with patients grouped according to level of PHLDB2 mRNA expression. (C) As in A, but with patients grouped according to both PRICKLE1 and PHLDB2 mRNA expression levels. The patients are separated into four classes (4K). Table shows number of patients and 5-year MFS for each class, as well as P-values from two tests comparing PHLDB2 classes within each PRICKLE1 class. (D) Similar to C, but with patients separated into two classes (2K; ‘high/high’ versus ‘no high/high’). Dashed lines in A–D indicate the PHLDB2-low class. (E) Uni- and multi-variate analyses of MFS in TNBC patients. HR, hazard ratio.
DISCUSSION
Wnt/PCP signaling is now recognized as a pathway that is not only important in embryonic development and adult life but also in diseases, especially cancer (Daulat and Borg, 2017). In the present study, we focused on TNBC, an aggressive subtype of breast cancer, and further studied MINK1, a Wnt/PCP prometastatic serine/threonine kinase with poorly described functions, particularly regarding the role of its kinase activity and the identity of its substrates. We characterized a two-step phosphorylation event triggered by MINK1 in TNBC. In our working model (Fig. 8), which is based on our previous findings (Daulat et al., 2012, 2016) and on the findings of the present study, MINK1 phosphorylates PRICKLE1, another Wnt/PCP molecule, at threonine 370 and, in doing so, promotes PRICKLE1 localization at the leading edge of migratory TNBC cells (Daulat et al., 2016). At this location, MINK1, in complex with PRICKLE1, associates with LL5β, which is then directly phosphorylated by MINK1 (Figs 1, 2). This second phosphorylation event enhances the association between CLASP2 and LL5β, which leads to an increase in cell motility (Fig. 6). It has been shown that CLASPs promote MT stabilization and tether MTs to focal adhesions. In doing so, CLASPs facilitate the local delivery of proteins for exocytosis and extracellular matrix degradation, leading to release of integrin–matrix connections and, in turn, promotion of integrin internalization and focal adhesion disassembly, as has been described in a previous study (Stehbens et al., 2014). We have shown that MINK1 and PRICKLE1, similarly to LL5β, are involved in the regulation of focal adhesion dynamics that promote the internalization of β1-integrin (Daulat et al., 2016). We propose that the MINK1–PRICKLE1–LL5β–CLASP2 protein complex is involved in the disassembly of focal adhesions and the promotion of cell migration (Fig. 8).
Working model. (A) MINK1 phosphorylates PRICKLE1, promoting PRICKLE1 subcellular localization within the lamellipodia of the cell and binding to LL5β. (B) MINK1 phosphorylates LL5β. (C) This second phosphorylation event allows the direct binding of CLASPs to LL5β and tethering of MTs to focal adhesions. (D) Altogether, this two-step phosphorylation process promotes disassembly of focal adhesions and promotes cellular migration. P, phosphorylation; VCL, vinculin. Figure created with BioRender.com.
Working model. (A) MINK1 phosphorylates PRICKLE1, promoting PRICKLE1 subcellular localization within the lamellipodia of the cell and binding to LL5β. (B) MINK1 phosphorylates LL5β. (C) This second phosphorylation event allows the direct binding of CLASPs to LL5β and tethering of MTs to focal adhesions. (D) Altogether, this two-step phosphorylation process promotes disassembly of focal adhesions and promotes cellular migration. P, phosphorylation; VCL, vinculin. Figure created with BioRender.com.
MINK1 has two paralogs (TNIK and MAP4K4), which share high degree of homology within their kinase domain, especially the ATP-binding pocket, rendering the generation of specific inhibitors for each family member difficult (Read et al., 2019). In this report, we show that treatment of TNBC cells with KY05009, an inhibitor of TNIK, phenocopies the loss of MINK1 or PRICKLE1 in terms of actin cytoskeleton reorganization, cell spreading and AKT kinase phosphorylation (Fig. 3). We used KY05009 to demonstrate the role of MINK1 activity in the regulation of assembly of the PRICKLE1–LL5β–CLASP2 complex and the implication of MINK1 catalytic activity in the promigratory process (Figs 4–6). Similarly to HT1080 cells, MDA-MB2-31 cells have a lower cell motility when the MINK1 (K54R) mutant is expressed and acts as dominant negative (Fig. 6A,B) (Hu et al., 2004). Taken together, our findings should stimulate the development of specific MINK1 inhibitors for cancer treatment. Of note, a recent study has shown that Mink1-knockout mice are viable (Li et al., 2019), suggesting that MINK1 inhibition might be associated with low side effects. Using a SILAC method and in vitro kinase assays, we defined threonine 894 of LL5β as a MINK1 phosphorylation site (Figs 1, 2). A whole-proteome database search for sequences matching a consensus motif based on the MINK1 phosphorylation sites of PRICKLE1 at threonine 370 and LL5β at threonine 894 (D[ST]LX[RK][RK], where X indicates any amino acid) identified ARHGAP21 and ARHGAP23 as putative MINK1 substrates. Interestingly, these two Rho-GTPase proteins have been shown to be associated with PRICKLE1 and to promote cellular migration through the regulation of Rho activity (Zhang et al., 2016). This result opens another avenue for further investigation of these potential MINK1 substrates. SMAD2 has previously been shown to be phosphorylated and subsequently inhibited by MINK1 in T helper cells (Fu et al., 2017). We did not find SMAD2 phosphopeptides in the list of MINK1 substrates in MDA-MB-231 cells (Fig. 1), suggesting cell context specificity of MINK1 substrates or dependency on the PRICKLE1-associated proteins that may not occur in lymphocytes. In conclusion, in MDA-MB-231 cells, PRICKLE1 acts as a scaffold protein that is able to bring substrates like LL5β close to MINK1 and, hence, provide specificity in cell signaling (Good et al., 2011).
Previously, it has been shown that LL5β and CLASPs form a complex with ELKS (also known as ERC1) at the cell cortex (Lansbergen et al., 2006). Our previous analysis of the PRICKLE1-associated proteins identified by mass spectrometry revealed the presence of LL5β and CLASP2 peptides, but not of ELKS peptides (Daulat et al., 2019). We hypothesize that our purification procedure from Daulat et al., 2019 was not appropriate to recover this interaction or that two independent CLASP–LL5β protein complexes co-exist, one containing ELKS and one containing MINK1–PRICKLE1. However, we confirmed that CLASP proteins and LL5β contribute to cell motility, as has already been demonstrated in previous studies (Astro et al., 2014; Lim et al., 2016).
Using a novel anti-PRICKLE1 antibody, we observed that PRICKLE1 is localized in cortical actin-containing structures of TNBC cells (Fig. 4). Loss or inhibition of MINK1 led to a strong reorganization of the actin cytoskeleton and delocalization of PRICKLE1 in actin bundles without modifying LL5β and CLASP2 membrane localization (Fig. 4D–I). This observation is in line with the fact that the plasma membrane recruitment of LL5β, a phosphatidylinositol (3,4,5)-trisphosphate-binding protein, is under the control of phosphoinositide 3-kinase activity (Lansbergen et al., 2006). However, CLASP2 and LL5β were found to be less colocalized in the remaining lamellipodia when MINK1 was downregulated or inhibited using KY05009 (Fig. 5B–D,G–I). Added to the fact that MINK1 inhibition impairs the association between CLASP2 and LL5β (Fig. 5A,E,F), this suggests that the activity of MINK1 plays a role in the regulation and localization of LL5β–CLASP2, an important membrane complex implicated in MT tethering and cell migration (Fig. 8).
We have found that MINK1 controls, through direct phosphorylation, the activity of PRICKLE1 and LL5β – two promigratory proteins that, when co-upregulated in TNBC, represent markers of very poor prognosis (Fig. 7; Fig. S7). Interestingly, high expression of both genes provides additional prognostic information above that provided by each gene alone, suggesting a likely synergic effect on the metastatic program. Moreover, the prognostic value is molecular subtype-dependent, as it is only observed in TNBC.
In conclusion, we describe a MINK1-dependent pathway involving members of the Wnt/PCP pathway. The dependency of MINK1 activity on Wnt/PCP-related receptors and extracellular signals remains uncertain for the moment; this issue should be addressed in the future. However, previous work has reported that TNIK phosphorylates transcription factor TCF family members, a key component of the Wnt pathway, and represents a pharmaceutical target in colon cancers having constitutively active TNIK- and TCF-dependent Wnt signaling (Mahmoudi et al., 2009). MINK1, TNIK and MAP4K4 have also been implicated in Hippo signaling, which is frequently deregulated in cancer (Meng et al., 2015). This suggests that targeting of MINK1 as well as other members of the MINK1 kinase family might have therapeutic value in TNBC and other cancers.
MATERIALS AND METHODS
Cell culture, reagents and antibodies
HEK 293T, MDA-MB-231 and SUM149PT cells were obtained from ATCC. Cells were grown in DMEM or DMEM/F12 (Gibco) containing 10% FCS (Invitrogen). Transfections were performed using polyethylenimine (Santa Cruz), Lipofectamine 2000 (Thermo Fisher Scientific) and Lipofectamine LTX (Thermo Fisher Scientific). siRNA was used with a reverse transfection protocol using Lipofectamine RNAimax (Thermo Fisher Scientific). Cells were tested for mycoplasma regularly. Antibodies targeting MINK1 were obtained from Bethyl (A302-192A; 1:1000 for western blotting). The following antibodies were obtained from Cell Signaling Technology: anti-AKT (C67E7; 1:2000 for western blotting), anti-pS473-AKT (9271; 1:2000 for western blotting). Anti-CLASP2 antibody was obtained from Absea Biotechnology Ltd (KT68; 1:200 for immunofluorescence). Anti-LL5β (BE-A304-582A; 1:1000 for western blotting) was obtained from Bethyl, anti-tubulin (T6074; 1:2000 for western blotting and immunofluorescence) and anti-actin (A1978; 1:1000 for immunofluorescence) antibodies were obtained from Merck. Anti-pT370-PRICKLE1 and anti-PRICKLE1 were produced and purified from rabbits injected with FPGLSGNADDpTLSR (where ‘pT’ indicates phosphorylated threonine) and QETPEDPEEWADHEDY peptides, respectively (Covalab commercial service). KY05009 was purchased from Merck. KY05009 treatments were for 16 h. The PRICKLE1 antibody used for western blot analysis and endogenous immunoprecipitation was obtained from Proteintech (22589; 1:1000 for western blotting). Anti-GFP (632569; 1:2000 for western blotting and immunofluorescence) was obtained from Takara. Anti-FLAG (F1804 product number; 1:2000 for western blot analysis) was obtained from Merck. Anti-HA (BLE901513 product number; 1:2000 for western blotting) was obtained from Biolegend. Anti-CLASP1 (050801A06 product number; 1:1000 for western blot analysis) was obtained from Absea Biotechnology Ltd. Anti-vinculin (MAB3574-C; 1:2000 for immunofluorescence). Anti-LL5β (HPA035147; 1:200 for immunuflorescence) was obtained from Merck. Phalloidin Alexa Fluor™ 647 was purchased from Thermo Fisher Scientific and used at 1:500. Slides were mounted with prolong (P36935) obtained from Life Technologies.
Affinity purification, immunoprecipitation and western blotting
At 48 h post-transfection, cells were lysed with TAP lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA, 10% glycerol, 0.1% NP40) and incubated at 4°C for 30 min to solubilize proteins. Affinity purification and immunoprecipitations were performed using GFP-Trap beads (made in-house) for 3 h at 4°C. After extensive washes with lysis buffer, proteins were eluted with 2× Laemmli sample buffer and heated at 95°C for 5 min in the presence of β-mercaptoethanol (Sigma). Whole-cell lysates or purified protein samples were resolved by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred onto Biotrace NT nitrocellulose transfer membranes (Cytiva). Western blotting was performed with antibodies as indicated in the figures, followed by chemiluminescence detection using appropriate HRP-conjugated secondary antibodies and ECL reagent (Cytiva). Western blot images were acquired using either a Syngene GBOX or Amersham ImageQuant 800 imaging system.
Endogenous immunoprecipitation was performed using 4 mg of MDA-MB-231 cell lysate in TAP lysis buffer. Lysates were incubated with 2 µg of antibody for 4 h at 4°C. Protein A agarose beads were used to immunoprecipitate the antibody–Prickle1 complex for 30 min at 4°C. Beads were extensively washed before western blot analysis. Flag immunoprecipitation was performed using 4 mg of HEK293T cell lysate in TAP lysis buffer. Lysates were incubated with 20 μl of FLAG M2 beads (A2220) obtained from Merck for 4 h at 4°C. Beads were extensively washed before western blot analysis
PRICKLE1 and PHLDB2 mRNA expression analysis in breast cancer samples
PRICKLE1 and PHLDB2 mRNA expression in breast cancer was analyzed in our own gene expression dataset (353 patients with invasive adenocarcinoma) coupled with publicly available datasets. The 17 public datasets comprising at least one probe set representing PRICKLE1 and PHLDB2 were collected from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database, the European Bioinformatics Institute (EBI) ArrayExpress database or the TCGA portal (Table S2). This resulted in a total of 5883 non-redundant pre-therapeutic samples of non-metastatic, non-inflammatory, primary, invasive breast cancers with clinicopathological annotations (Table S3). Data analysis required pre-analytic processing as previously described (Sabatier et al., 2011). To be comparable across datasets and to exclude bias from population heterogeneity, PRICKLE1 and PHLDB2 expression levels were standardized within each dataset using the luminal A population as reference. All steps were performed in R (http://www.cran.r-project.org/) using Bioconductor and associated packages.
PRICKLE1 and PHLDB2 upregulation and downregulation were defined using the median level as cut-off. The estrogen receptor (ER), progesterone receptor (PR) and HER2 statuses of samples were based on mRNA expression of ESR1, PGR and ERBB2 genes, respectively, and defined as discrete values (positive or negative) using a two-component Gaussian mixture distribution model (Lehmann et al., 2011). The molecular subtypes of tumors were defined as HR+/HER2− (ER- and/or PR-positive and HER2-negative), HER2+ (HER2-positive, regardless of ER and PR status), and TN (ER−, PR− and HER2-negative). Metastasis-free survival (MFS) was calculated from the date of diagnosis until the date of distant relapse. Follow-up was measured from the date of diagnosis to the date of last news for event-free patients. Survivals were calculated using the Kaplan–Meier method, and phalloidin Alexa Fluor™ 647 curves were compared using the log-rank test. Univariate and multivariate prognostic analyses for MFS were performed using Cox regression analysis (Wald test). The variables tested in the univariate analysis included patients' age (less than versus greater than or equal to 50 years), pathological tumor type (lobular versus ductal), size (pT2, pT3 versus pT1), grade (3 versus 1–2), axillary lymph node status (positive versus negative), delivery of adjuvant chemotherapy (yes versus no) and the PRICKLE1 and PHLDB2 expression status (high/high versus no-high/high). Multivariate analysis included the variables with a P value inferior to 5% in univariate analysis. The likelihood ratio (LR) tests were used to assess the prognostic information provided beyond that of each gene, assuming a χ2 distribution. Changes in the LR values (LR-Δχ2) quantified the relative amount of information of one Cox model compared with another. All statistical tests were two-sided at the 5% level of significance. Statistical analysis was done using the survival package (version 2.30) in the R software (version 2.15.2; http://www.cran.r-project.org/).
Samples of human origin and associated data were obtained from the IPC/CRCM Tumor Bank that operates under authorization # AC-2013-1905 granted by the French Ministry of Research. Prior to scientific use of samples and data, patients were appropriately informed and asked to express their consent in writing, in compliance with French and European regulations. The project was approved by the IPC Institutional Review Board and was conducted according to the principles expressed in the Declaration of Helsinki.
Migration assays
A total of 25,000 cells were seeded on collagen-coated 6-well plates. Cells were monitored using live-cell imaging using Metamorph 7.8.1 (Molecular Devices) and followed for 19 h. Cells were then tracked manually using ImageJ (NIH, Bethesda, MD) tracking plugin. The values for the assessment of cumulative distance and Eulidean distance were measured. Cumulative distance is the total distance traveled by cells, and Euclidean distance is the length between the initial and the final point measured (Siret et al., 2018). Wound healing assays were performed using an insert from IBIDI. A total of 70,000 cells were seeded and the next day the insert was removed. Cell migration was followed using live-cell imaging, and analysis was performed using an in-house generated script and Metamorph software.
In vitro kinase assay
The MINK1 assay was purchased from Promega (Charbonnieres-les-bains, France). Enzyme, substrate, ATP and inhibitors were diluted in kinase buffer as per the manufacturer's instructions. Kinase reactions were performed in a 384-well plate in a final volume of 5 µl. The reaction was initiated using 1 µl of inhibitor for each concentration (in 1% DMSO), 1 µl of enzyme and 3 µl of substrate/ATP mix (60 min, room temperature). Next, 5 μl of ADP-Glo™ reagent was used to stop the kinase reaction by ATP depletion (40 min, room temperature). Then, ADP formed by the kinase reaction was detected by adding 10 µl of Kinase Detection Reagent (30 min, room temperature). Luminescence was recorded using a PHERAstar plate reader. MINK1 kinase was used at optimized concentrations of 2 ng/well, ATP was used at 5 µM, DTT at 50 µM and the substrate of MINK1 [PRICKLE1 peptide encompassing MINK1 phosphorylation site was used FPGLSGNADDTLSR (Covalab)] was used at 0.25 ng/µl.
Phosphoproteomics
MDA-MB-231 cells were grown in DMEM supplemented with 0.073 mg/ml heavy or light arginine [cells were grown in the medium for 10 doubling times (Thermo Fisher Scientific)] and dialysed FBS (Thermo Fisher Scientific) for 15 days until reaching near-100% isotope labelling, as assessed by mass spectrometry analysis. Cells were treated with shRNA targeting MINK1 (light sample) or non-targeting shRNA (heavy sample). shRNAs were cloned into PLKO backbone (#10878 Addgene), and lentiviral particles were produced using PSPAX and VSV-G systems (#8455 and #8454 Addgene) in HEK 293T cells. The following shRNA sequence was used to target MINK1 (shRNA #03): 5′-AGCGGCTCAAGGTCATCTATG-3′. The 5′-AGGTTAAGTCGCCCTCGCTCG-3′ control shRNA sequence was 5′-AGGTTAAGTCGCCCTCGCTCG-3′. Cells were lysed in RIPA buffer (20 mM HEPES pH 8.0, 9 M urea, 1 mM sodium orthovanadate, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate). A total of 1–2×108 cells per condition were used (20 mg of protein). Cells were collected and sonicated at 15 W output with two bursts of 30 s each. The lysate was cleared by centrifugation at 20,000 g at 4°C for 15 min, and the supernatants were stored for mass spectrometry analysis. After protein quantification, lysates from each condition were mixed to 1:1 ratio and digested using trypsin. Phosphopeptides were enriched using TiO2 columns (Thermo Fisher Scientific) and analyzed by mass spectrometry. For full details of for digestion, enrichment, MS, database searching and phosphopeptide quantification methods, see Daulat et al., 2016.
Identification of phosphorylation sites using mass spectrometry
Identification of LL5β and PRICKLE1 phosphorylation sites was performed using in vitro kinase assay followed by liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis. Recombinant MINK kinase domain (Kinase Logistics) was incubated with recombinant GST–PRICKLE1 or GST–LL5β proteins purified from E. coli and immobilized on Glutathione beads. Phosphorylation reactions were performed in kinase buffer (25 mM HEPES pH 7.4, 25 mM β-glycerophosphate, 25 mM MgCl2, 0.1 mM Na3VO4, 0.5 mM DTT) supplemented with 20 µM ATP at 37°C for 1 h. Reactions were stopped by five successive washes with kinase buffer and the addition of 4× Laemmli sample buffer. Proteins were resolved by SDS–PAGE, extracted and digested with trypsin. The resulting peptide mixture was analyzed by LC-MS/MS. Peptide and protein identification was performed as described previously (Daulat et al., 2016), but the search parameters were modified to allow for the presence of one or more phosphate groups on serine, threonine or tyrosine residues. MS/MS spectra attributed to a phosphorylated peptide in PRICKLE1 or LL5β were manually inspected and validated to locate the position of the phospho-amino acid.
Confocal imaging
Cells were seeded on coverslips pre-treated with rat tail collagen (Roche). Cells were fixed using paraformaldehyde or ice-cold methanol followed by permeabilization using PBS containing Triton X-100 at 0.2%. Cells were treated with the primary antibodies indicated in the figures, followed by secondary antibody treatment. Samples were imaged on an LSM 880 confocal microscope (Zeiss) with a UV laser and 63× objective. Confocal images were analyzed using ImageJ software.
Image analysis
Lamellipodial structures were counted manually in each field of view and reported over the number of cells counted through the observation of DAPI staining of the nucleus. PRICKLE1 plasma membrane localization and cytosolic localization were manually assessed and reported after analyzing over 200 cells per condition for the quantifications shown in Fig. 4E and 4G. LL5β plasma membrane localization for over 100 cells per condition was manually assessed and is reported in Fig. 4I. Colocalization of CLASP2 and LL5β was measured on area selected within the plasma membrane and enriched in MT ends, as shown in Fig. 5B and Fig. 5G. Colocalization were measured using ImageJ software and PCCs were extracted. Measurements of cell and focal adhesion area were perfomed manually using ImageJ software.
DNA constructs and siRNAs
The following sequences of siRNAs were used: MINK1 (#09) 5′-GGAACAAACUGCGGGUGUA-3′, MINK1 (#10) 5′-GAAGUGGUCUAAGAAGUUC-3′, MAP4K4 (#06) 5′-CAAAAGGGCUCAAAGACUA-3′, MAP4K4 (#08) 5′-GAAAUACUCUCAUCACAGA-3′, TNIK (#03) 5′-UAAGCGAGCUCAAAGGUUA-3′, TNIK (#04) 5′-GAACAUACGGGCAAGUUUA-3′, LL5β (#03) 5′-GAACGGAAAAGCAGUAUUA-3′, LL5β (#17) 5′-GGAUCUACCUCAUAGCGUA-3′, LL5β (#09) 5′-GGGCAAUUCCAAACGAACA-3′, CLASP1 (#01) 5′-GCACAGACUUUAACACUAA-3′, CLASP1 (#02) 5′-GGACAGCUCUGGAUAACAA-3′, CLASP1 (#04) 5′-GCUGUUAGGUUAAUUAUUC-3′, CLASP1 (#17) 5′-CCGAAAGUAGUGUGCGUAA-3′, CLASP2 (#02) 5′-UCAGAACGCUCCUAUAGUU-3′, CLASP2 (#03) 5′-CAAGAUUGGUUGUUUGUAC-3′, CLASP2 (#03) 5′-GAGAUUAUGCCAGGUCUAA-3′, CLASP2 (#04) 5′-GAGAUUAUGCCAGGUCUAA-3′, CLASP2 (#05) 5′-CUAAUGAGAUUUACAGUUG-3′. The LL5α construct was obtained from Yuko Mimori-Kiyosue (RIKEN BDR, Kobe, Japan). LL5β cDNA was obtained from Joshua Sanes (Harvard, MA, USA). PRICKLE1 and MINK1 constructs were described previously (Daulat et al., 2012). Site-directed mutagenesis was performed using the Q5® site-directed mutagenesis kit protocol (NEB). GST–LL5β was cloned into pGEX plasmid vector (Merck).
RT-qPCR assays
Total RNA was isolated from MDA-MB-231 cells using an RNA isolation kit (RNeasy Mini Kit, Qiagen) according to the manufacturer's instructions. After DNase (Turbo DNA-free Kit, Promega) treatment, a standard PCR (GoTaq G2 Green Master Mix, Promega) was performed on RNA samples to check genomic DNA contamination. cDNA was synthesized from 1 µg of total RNA using a reverse transcriptase kit (SuperScript II Reverse Transcriptase, Invitrogen). Quantitative real-time RT-PCR was performed using a SYBR Green kit (Applied Biosystems) and CFX96 real-time PCR system (Bio-Rad) as follows: 95°C for 3 min, 40 cycles at 95°C for 15 s, 60°C for 30 s and 72°C for 30 s. Primers used were as follows: MINK1 forward primer, 5′-CTGATGTTGCTGGACCGAAG-3′; MINK1 reverse primer, 5′-AGAATCTTGTTCCGGAGCCA-3′; TNIK forward primer, 5′-AAGGTAACACGTTGAAAGAGGAG-3′; TNIK reverse primer, 5′-AGTCAGCAAGACATTTTGCCC-3′; MAP4K4 forward primer, 5′-GACTCCCCTGCAAAAAGTCTG-3′; MAP4K4 reverse primer, 5′-GTCCATAGGTGCCATTTCCAA -3′; ACTINB forward primer, 5′-CACCATTGGCAATGAGCGGTTC-3′; ACTINB reverse primer, 5′-AGGTCTTTGCGGATGTCCACGT-3′.
Acknowledgements
The authors wish to thank Dr Mukhtar Ahmad for the critical review of the manuscript, Emilie Beaudelet and Yves Toiron for technical assistance in mass spectrometry analysis, Sébastien Germain for assistance with the isolated cell trajectory analysis and Dr Rudra Kashyap for his help to generate recombinant PRICKLE1 and LL5β proteins.
Footnotes
Author contributions
Conceptualization: A.M.D., J.-P.B.; Methodology: A.M.D.; Validation: A.M.D.; Formal analysis: A.M.D.; Investigation: A.M.D., M.S.W., S.A., M.K., J.A.-B., P.F., F.B., L.C.; Data curation: P.F., F.B., L.C.; Writing - original draft: A.M.D., F.B., J.-P.B.; Writing - review & editing: A.M.D., J.-P.B.; Supervision: J.-P.B.; Project administration: A.M.D., J.-P.B.; Funding acquisition: J.-P.B.
Funding
This work was funded by Ligue Contre le Cancer (Label Ligue to J.-P.B. and F.B., and a fellowship to A.M.D.), Fondation de France (fellowship to A.M.D.), Fondation ARC pour la Recherche sur le Cancer (grant to J.-P.B.), and Alliance Nationale pour les Sciences de la Vie et de la Santé (AVIESAN) through the NANOTUMOR project. M.S.W. was a recipient of the Ciência sem Fronteiras PhD program from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Brazil. Marseille Proteomics (IBiSA) is supported by Institut Paoli-Calmettes (IPC) and Canceropôle PACA. J.-P.B. is a scholar of Institut Universitaire de France.
Peer review history
The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.259347.reviewer-comments.pdf.
References
Competing interests
The authors declare no competing or financial interests.