LIMK1 as a Novel Kinase of β-Catenin Promotes Esophageal Cancer Metastasis by Cooperating With CDK5

Abstract

Metastasis is a major cause of cancer deaths, but the underlying molecular mechanisms remain largely unknown. Esophageal squamous cell carcinoma (ESCC) is a highly aggressive cancer with poor survival, yet the key kinases driving ESCC metastasis and their biological function have not been fully discovered. Here, a kinase-substrate map of metastatic ESCC is presented for the first time by conducting a phosphoproteomics analysis of 60 clinical specimens. By further consolidating data with CRISPR/Cas9 functional screening, LIM domain kinase 1 (LIMK1) is identified as a novel kinase of β-catenin. The in vitro and in vivo experiments demonstrated that LIMK1 cooperates with Cyclin-dependent kinase 5 (CDK5) to promote cancer metastasis in a phosphorylation-dependent manner. Mechanistically, LIMK1 and CDK5 synergistically phosphorylate β-catenin at S191, enhancing its phosphorylation and interaction with Nucleoporin 93, resulting in β-catenin nuclear translocation and activation of key pathways in cancer metastasis. High expression of LIMK1 and CDK5 is associated with poor prognosis of ESCC patients, and the clinical and functional significance of LIMK1/CDK5-Wnt/β-catenin axis is also verified in esophageal adenocarcinoma, gastric cancer, and lung cancer. Furthermore, the combination of LIMK1 and CDK5 inhibitors significantly suppresses metastasis in multiple models. This work highlights LIMK1 as a novel regulatory and targetable kinase of β-catenin, informing the treatment of advanced cancer.

Keywords: LIMK1; esophageal squamous cell carcinoma; kinase‐substrate map; metastasis; β‐catenin.

Figure 1

Activated Wnt signaling is a key pathway in cancer metastasis. A) Workflow of the experiment and the number of samples used for multi‐omics analysis. B) Heatmap showing the differentially expressed proteins among the four classes. The tiling bars above the heatmap show the distribution of different clinicopathological characteristics among the 20 metastatic ESCC patients. C) FC of proteins and phosphosites, and their correlations in LN and N. Pathways enriched with cancer‐related phosphoproteins. Red dots: phosphosites are greater than two‐fold changes in LN compared to N, and changes in phosphosites abundance are greater than changes in their corresponding protein abundance. D) Heatmap showing the activities of cancer hallmark‐related pathways across N, T, and LN groups. E) Boxplots showing the distribution of Wnt/β‐catenin signaling (upper) and EMT signaling (lower) among N, T, and LN groups. F) The phospho‐regulatory network links cancer‐related kinases (diamonds), phosphoproteins (hexagons), and proteins in Wnt signaling (circles). G) Phosphosites in Wnt/β‐catenin signaling were differentially expressed in N, T, and LN groups. N: normal tissues; T: tumor tissues; LN: lymph node metastatic tissues.

Figure 2

Integrated multi‐omics analysis with CRISPR/Cas9 screening identifies LIMK1 and CDK5 as key kinases responsible for ESCC metastasis. A) Diagram showing the approach used to systematically identify key kinases related to ESCC metastasis. B) Read counts of sgRNAs targeting LIMK1 or CDK5 in GeCKO‐transduced cells and input cells through in vitro (upper) and in vivo (lower) screening. C) LIMK1 and CDK5 promote ESCC cell invasion in a manner dependent on their phosphorylation activity. D) Western blot showing the effect of LIMK1, CDK5, and their mutant type on the expression levels of EMT markers in ESCC cells. E) RNA‐seq analysis of LIMK1 and CDK5 mRNA expression in N, T, and LN from metastatic ESCC cohort. F) Measurement of LIMK1 and CDK5 expression in N, T, and LN of ESCC by Western blot. G) proteomic analysis of CDK5 levels in metastatic ESCC cohort. H) Measurement of LIMK1 and CDK5 kinase activity of metastatic ESCC cohort by ssGSEA. N: normal tissues; T: tumor tissues; LN: lymph node metastatic tissues.

Figure 3

LIMK1 synergizes with CDK5 to promote ESCC metastasis. A, B) Transwell assays were performed to determine the invasive abilities of ESCC cells when treated with BMS‐5 and/or Dinaciclib (A) or subjected to knockdown via siRNA targeting LIMK1 and/or CDK5 (B). C) Transcriptome analysis of metastatic ESCC cohort and public datasets (GES53622, GES53624, and TCGA) demonstrates that the upregulation of LIMK1 and CDK5 affects multiple signaling pathways. D) RT‐qPCR analysis of SRC, ITGA3, CCNB1, and PCNA mRNA expression in ESCC cells overexpressing either LIMK1 or CDK5. E) Heatmap showing the CDK5 substrates with significantly altered phosphorylation levels in the metastatic ESCC cohort. F) The effect of BMS‐5 and/or Dinaciclib treatment on β‐catenin p‐Ser level was investigated by western blot. G) The nuclear and cytoplasmic distribution of β‐catenin was analyzed after treatment with BMS‐5 and/or Dinaciclib.

Figure 4

LIMK1 is a novel kinase responsible for the phosphorylation and nuclear translocation of β‐catenin. A) MEME software output showing the locations with the top‐scoring LIMK1 kinase recognition motifs. B) Correlation of phosphorylation of β‐catenin at S191 and LIMK1 expression in the metastatic ESCC cohort. C) Co‐IP assay confirming the exogenous interaction between LIMK1 and β‐catenin in ESCC cells. D) GST pull‐down assay showing a direct interaction between LIMK1 and β‐catenin. E–G) Co‐IP assay (E, F) and in vitro kinase assay (G) show that LIMK1 increases the phosphorylation of β‐catenin at the S191 site. H) MEME software output, showing the locations with the top‐scoring for CDK5 kinase recognition motif. I) Correlation of phosphorylation of β‐catenin at S191 and CDK5 expression in the metastatic ESCC cohort. J) CO‐immunoprecipitation assay was performed to confirm the exogenous interaction between CDK5 and β‐catenin in ESCC cells. K) GST pull‐down assay investigating the direct interactions between GST‐β‐catenin and CDK5‐His. L, M) Effect of the CDK5‐WT or CDK5‐K33T on the β‐catenin‐WT or β‐catenin‐S191 p‐Ser level in ESCC cells. N) Comparison of the phosphorylation level of ESCC cells overexpressing CDK5‐WT or CDK5‐K33T and β‐catenin‐WT or β‐catenin‐S191A in vitro. O) CO‐immunoprecipitation assay was performed to confirm the interaction between CDK5‐K33T and β‐catenin.

Figure 5

LIMK1 and CDK5 increase the binding of β‐catenin and NUP93 by phosphorylation to promote its nuclear translocation. A) Schematic diagram of the LIMK1 truncation mutants used in this study (upper). ESCC cells were transfected with LIMK1 mutants as indicated and collected for Co‐IP (lower). B) Schematic diagram of different β‐catenin truncation mutant constructs. C, D) β‐Catenin mutants were transfected into ESCC cells, and Co‐IP assays were performed to investigate the binding domains between β‐catenin and LIMK1 (C) or CDK5 (D). E, F) Predicted protein interactions based on structures of human LIMK1 (E) or CDK5 (F) and β‐catenin using ZDOCK software. LIMK1: pink, CDK5: carmine, β‐catenin: cyan. G) Subcellular fractionation was used to assess the nuclear and cytoplasmic distribution of β‐catenin in ESCC cells overexpressing LIMK1‐WT/LIMK1‐D460A or CDK5‐WT/CDK5‐K33T. H, I) Co‐IP was performed to evaluate the interactions between β‐catenin and NUP93 in the presence of LIMK1 or CDK5 in ESCC cells. J) Interaction between NUP93 and wild‐type or mutant β‐catenin was evaluated in ESCC cells.

Figure 6

LIMK1 and CDK5 are clinically and functionally important for tumor metastasis in multiple cancer types. A) Western blot was used to detect the expression of downstream proteins in the Wnt signaling pathway in ESCC cells overexpressing LIMK1/CDK5 and their mutant type. B) Boyden chamber invasion assay showed that knockdown of β‐catenin attenuated the effect of LIMK1 or CDK5 on ESCC cell invasion. C) Bioluminescence imaging and quantitative analysis showed that β‐catenin mediated the effect of LIMK1 or CDK5 on ESCC metastasis. D) Representative images and expression patterns of LIMK1 in 208 ESCC and 152 normal tissues are shown. E) Survival analysis for 207 ESCC patients stratified by tumor LIMK1 expression levels. F) Representative images and expression patterns of CDK5 in ESCC tissue microarray are shown. G) Survival analysis for 207 ESCC patients stratified by tumor CDK5 expression levels. H) Survival analysis for 207 ESCC patients stratified by tumor LIMK1 and CDK5 expression levels. I) Representative images and nuclear localization patterns of β‐catenin in 215 ESCC and 145 normal tissues are shown. J) Survival analysis for 214 ESCC patients stratified by tumor β‐catenin nuclear localization levels. K) Representative images of β‐catenin nuclear localization at different expression levels of LIMK1 and CDK5 in ESCC tissue microarray. L, M) Association between LIMK1 and/or CDK5 with β‐catenin nuclear localization in ESCC tissues.

Figure 7

The combination therapy targeting LIMK1 and CDK5 significantly inhibits ESCC metastasis. A) Transwell assays were performed to determine the invasive abilities of EAC, gastric cancer, and lung cancer cells when treated with BMS‐5 and/or Dinaciclib. B) The flow diagram shows the lung metastasis model, lymph node metastasis model in mice, and the treatment plan. C) Bioluminescence imaging and quantification of lung metastasis in mice intravenously injected with ESCC cells and treated with BMS‐5 and/or Dinaciclib. Hematoxylin‐eosin (H&E) staining shows lung metastasis as indicated. D) The lymph node metastasis model was established and the therapeutic effects of BMS‐5 and/or Dinaciclib were monitored by Bioluminescence imaging. E) Key indicators of drug toxicity were detected in mice treated with indicated inhibitor or DMSO.

Figure 8

Schematic model.