Article Text

PDF

Original article
DYRK1A modulates c-MET in pancreatic ductal adenocarcinoma to drive tumour growth
  1. Jeroni Luna1,2,
  2. Jacopo Boni2,3,
  3. Miriam Cuatrecasas1,4,5,
  4. Xavier Bofill-De Ros1,2,
  5. Estela Núñez-Manchón1,
  6. Meritxell Gironella1,4,
  7. Eva C Vaquero1,4,
  8. Maria L Arbones2,6,
  9. Susana de la Luna2,3,7,8,
  10. Cristina Fillat1,2,9
  1. 1Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain
  2. 2Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER), Barcelona, Spain
  3. 3Centre for Genomic Regulation (CRG), The Barcelona Institute for Science and Technology, Barcelona, Spain
  4. 4Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
  5. 5Departament de Patologia, Centre de Diagnòstic Biomèdic (CDB), Hospital Clínic de Barcelona i Banc de Tumors-Biobanc Clinic-IDIBAPS, Barcelona, Spain
  6. 6Institut de Biologia Molecular de Barcelona (IBMB), Barcelona, Spain
  7. 7Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain
  8. 8Universitat Pompeu Fabra (UPF), Barcelona, Spain
  9. 9Universitat de Barcelona (UB), Barcelona, Spain
  1. Correspondence to Dr Susana de la Luna, Centre for Genomic Regulation (CRG), Barcelona 08003, Spain; susana.luna{at}crg.eu and Dr Cristina Fillat, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona 08036, Spain; cfillat{at}clinic.ub.es

Abstract

Background and aims Pancreatic ductal adenocarcinoma (PDAC) is a very aggressive tumour with a poor prognosis using current treatments. Targeted therapies may offer a new avenue for more effective strategies. Dual-specificity tyrosine regulated kinase 1A (DYRK1A) is a pleiotropic kinase with contradictory roles in different tumours that is uncharacterised in PDAC. Here, we aimed to investigate the role of DYRK1A in pancreatic tumorigenesis.

Design We analysed DYRK1A expression in PDAC genetic mouse models and in patient samples. DYRK1A function was assessed with knockdown experiments in pancreatic tumour cell lines and in PDAC mouse models with genetic reduction of Dyrk1a dosage. Furthermore, we explored a mechanistic model for DYRK1A activity.

Results We showed that DYRK1A was highly expressed in PDAC, and that its protein level positively correlated with that of c-MET. Inhibition of DYRK1A reduced tumour progression by limiting tumour cell proliferation. DYRK1A stabilised the c-MET receptor through SPRY2, leading to prolonged activation of extracellular signal-regulated kinase signalling.

Conclusions These findings reveal that DYRK1A contributes to tumour growth in PDAC, at least through regulation of c-MET accumulation, suggesting that inhibition of DYRK1A could represent a novel therapeutic target for PDAC.

  • pancreatic cancer
  • carcinogenesis
  • cell growth
  • molecular mechanisms

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Dual-specificity tyrosine regulated kinase  1A (DYRK1A) has distinct functions in different tumours.

  • DYRK1B is a survival factor in pancreatic tumours.

  • DYRK1A positively regulates epidermal growth factor receptor (EGFR) accumulation in glioblastoma.

  • Overexpression of c-MET is a common alteration in pancreatic ductal adenocarcinoma (PDAC).

What are the new findings?

  • DYRK1A has a protumorigenic role in PDAC.

  • DYRK1A stabilises c-MET by inhibiting it being targeted for degradation.

  • Patient tumours with DYRK1Ahigh also showed high c-MET and EGFR expression levels. DYRK1A protein expression positively correlated with c-MET.

How might it impact on clinical practice in the foreseeable future?

  • Pharmacological inhibition of DYRK1A could represent a promising therapy for c-METhigh and EGFR-positive PDAC tumours.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) represents over 90% of all pancreatic malignancies. Although PDAC is a relatively rare tumour, it is the fourth leading cause of cancer deaths worldwide, with a median survival of 6 months and a 5-year survival rate of 8%.1 Recent estimates indicate that PDAC will become the second leading cause of cancer-related mortality by 2030.2 This dismal outlook is in part due to late diagnosis, at which point more than 70% of patients present with advanced disease. Recently, gemcitabine plus nab paclitaxel and FOLFIRINOX chemotherapies have shown a modest improvement in survival of patients with advanced disease.3 4

Pancreatic carcinogenesis is well defined as a multistage process accompanied by distinct morphological and histological changes, resulting from molecular alterations that are acquired during malignant transformation. Exome, whole-genome sequencing and copy-number variation analysis have revealed a complex mutational landscape.5 6 Recently, integrated genomic studies of PDAC samples identified 32 recurrently mutated genes that aggregate into 10 core molecular pathways.7

Protein kinases are key regulators of signal transduction pathways, and thereby their aberrant activation plays crucial roles in multiple cellular processes that lead to cancer progression.8 Indeed, protein kinases represent the most effective class of therapeutic targets in cancer.9 The dual-specificity tyrosine-regulated kinases (DYRKs) are a highly conserved family of protein kinases within the CMGC group of the eukaryote kinome. Based on phylogenetic analysis, the family is subdivided in two groups, known as class I and class II DYRKs, with the two closest paralogous DYRK1A and DYRK1B comprising class I in humans.10

Based on the activity of DYRK interactors and substrates, it is assumed that these kinases are pleiotropic proteins with widespread cellular functions, including the regulation of cell proliferation, survival and differentiation.10 Both DYRK1A and DYRK1B appear to control cell proliferation through phosphorylation of components of the cell cycle machinery.11 Indeed, changes in the expression of these two kinases have been observed in tumour samples. DYRK1B is normally expressed at low levels in most tissues but has increased expression levels in ovarian and pancreatic cancers.12 13 Reports on DYRK1A offer a less clear picture. Thus, while DYRK1A mRNA levels are found to be reduced in patients with acute myeloid leukaemia (AML), and DYRK1A overexpression inhibits proliferation of AML cell lines,14 the DYRK1A kinase behaves as a potent megakaryoblastic tumour promoter in a Down syndrome context, in which DYRK1A is in trisomy.15 Likewise, pharmacological inhibition of DYRK1A compromised survival of glioblastoma cells and reduced tumour burden in an intracranial glioblastoma model;16 however, other studies suggest that DYRK1A (and DYRK1B) cooperate to restrain tumorigenesis in the nervous system.17 Recently, it has been reported that administration of an inhibitor against DYRK1A in mice with head and neck squamous cell carcinoma xenograft tumours induced regression of tumour growth.18

Here we show that DYRK1A is upregulated in PDAC and provide evidence that DYRK1A favours tumour progression via stabilisation of the receptor tyrosine kinase (RTK) c-MET. Consistently, loss of Dyrk1a in pancreatic cancer mouse models is associated with tumour growth impairment and increased survival. Moreover, expression of DYRK1A significantly correlated with c-MET in patients with PDAC. Thus, targeting DYRK1A could represent a therapeutic strategy for PDAC.

Materials and methods

Clinical specimens

Samples from patients diagnosed with PDAC were provided by the Biobank Hospital Clinic of Barcelona IDIBAPS and were obtained from patients scheduled to undergo surgery. Written informed consent was obtained from all patients in accordance with the Declaration of Helsinki. 

Mouse models

For xenograft experiments, 2.5×106 PANC-1 cells transduced with shRNAs were suspended in 200 µL of serum-free Dulbecco’s modified Eagle medium (DMEM) and injected subcutaneously into the flanks of athymic nude-Foxn1nu mice (Envigo). Tumours were measured every 3 days, and tumour volume was calculated using the formula volume= π/6 × length × width2. Mice were sacrificed at 44 days after cell inoculation, and tumours were removed. Each tumour was cut in half and either fixed with 4% paraformaldehyde and embedded in paraffin or quickly frozen in liquid nitrogen for further analysis.

Ela-myc transgenic mice were kindly provided by E Sandgren (University of Wisconsin).19 K-RasLSL-G12D;p53LSL-R172H (Howard Hughes Medical Institute) and Pdx1-Cre (University of Cincinnati) strains, in a C57BL/6 background, were interbred to generate compound mutants K-RasLSL-G12D;p53LSL-R172H;Pdx1-Cre mice (named KPC mice), as previously described.20 Ela-myc:Dyrk1a+/- and Ela-myc:Dyrk1a+/+ mice were bred inhouse by crossing Ela-myc males (C57BL/6 background) with Dyrk1a+/- females (mixed 129/Sv and C57BL/6 background21). Transgenic mice and wild type (WT) littermates were used in this study. For harmine treatments, Ela-myc:Dyrk1a+/+ mice were treated with 30 mg/kg/day harmine hydrochloride (TCI Europe) or vehicle (saline) by intraperitoneal injection 5 days per week, starting at 6–7 weeks of age.

Cell lines

HEK-293T, U2OS and the human PDAC cell lines PANC-1, BxPC-3 and MIA PaCa-2 were obtained from the American Type Culture Collection. All cell lines were grown in DMEM supplemented with 10% fetal bovine serum and antibiotics (100 u/mL penicillin, 100 µg/mL streptomycin). All cell lines were routinely tested for mycoplasma by PCR. The calcium-phosphate method was used for transient transfection. Plasmids for expressing flag-tagged Sprouty-2 (SPRY2) proteins have been previously described.22 Information on lentiviral transduction, clonogenic assays and cumulative cell curves is included in supplementary methods.

mRNA quantification and gene expression analysis

mRNA quantification was performed by reverse transcriptase coupled to quantitative PCR and data are represented as 2^CT (see supplementary methods for description). Computational analysis is described in supplementary methods.

Protein analysis

A complete description for western blot analysis, immunohistochemistry and in vitro kinase assays is provided in supplementary methods.

Statistical analysis

All analyses were performed with GraphPad Prism V.6.0c (GraphPad Software). Samples were evaluated for normality using a Shapiro-Wilk test. Statistically significant differences in pairwise comparisons were defined with a two-tailed unpaired Student’s t-test for normal samples, and with a two-tailed unpaired Mann-Whitney test for samples not passing the normality test. Multiple comparisons were performed using one-way analysis of variance (ANOVA) or two-way ANOVA with Bonferroni’s post hoc test. Differences were considered statistically significant at p values less than 0.05. Correlation analyses were performed with a Spearman’s correlation test. Survival was analysed by the Kaplan-Meier method and evaluated with a log-rank (Mantel-Cox) test. Data in graphs are presented as the mean±SEM. Tukey’s method was used for plotting the whiskers and outliers in box plots, *p<0.05; **p<0.01; ***p<0.001.

Results

DYRK1A is overexpressed in PDAC

Recent evidence suggests that DYRK1A is deregulated in certain cancers, and that this could modulate tumour growth. To explore the function of DYRK1A in PDAC, we analysed DYRK1A expression by immunohistochemistry in primary human tumours and normal adjacent pancreas samples. In normal human pancreas, DYRK1A expression was restricted to islets of Langerhans (figure 1A). However, in tumour samples, strong DYRK1A staining was detected in PDAC and pancreatic intraepithelial neoplasia (PanIN) lesions (H-score, n=22: 195.7±10.09, mean±SEM), predominantly nuclear (figure 1A). In line with what has been described,13 high cytoplasmic DYRK1B expression was detected in PDAC and PanIN lesions (H-score, n=7: 254.6±21.58) (supplementary figure S1A). DYRK1A mRNA was also increased in PDAC samples as compared with normal adjacent tissue in two inhouse sample sets, and in an independent cohort studied from publicly available microarray data (GSE62452)23 (figure 1B,C). Differential expression of DYRK1B mRNA was only observed in one set of samples (supplementary figure S1B,C). These results indicate that DYRK1A shows increased expression in human PDAC, and that it is deregulated, at least at the transcriptional level.

Supplementary file 1

Figure 1

DYRK1A is highly expressed in human pancreatic tumours. (A) Representative images of DYRK1A immunohistochemical staining in normal pancreas, low and high grade pancreatic intraepithelial neoplasia (PanIN) and pancreatic ductal adenocarcinoma (PDAC) (n=22) (scale bar, 50 µm). The inset in the normal pancreas image corresponds to islets of Langerhans. Notice a predominant nuclear pattern of staining with DYRK1A. Upper panels, H&E; lower panels, DYRK1A staining. (B) qRT-PCR analysis of DYRK1A mRNA in PDAC samples and adjacent non-tumour tissue of two different sample sets normalised to UBL4A (set 1, n=11) or HPRT (set 2, n=14) (Mann-Whitney test). (C) DYRK1A mRNA levels by microarray analysis of PDAC samples and adjacent non-tumour tissue (GSE62452; n=69; Student’s t-test). (D) Box plots of DYRK1A expression in PDAC TCGA samples classified using the published mRNA signatures for the indicated PDAC subtypes.7 (progenitor-like, n=53; aberrantly differentiated endocrine exocrine (ADEX), n=38; squamous, n=31; immmunogenic, n=27; one-way analysis of variance (ANOVA)). *p<0.05; **p<0.01.

Published studies have reported gene expression subtypes of pancreatic cancer.7 24 25 No differences in DYRK1A expression were observed among subtypes when analysing the Cancer Genome Atlas (TCGA) PDAC cohort and the classification of two of these studies (supplementary figure S1D,E).However, increased expression of DYRK1A was found in the immunogenic subtype (figure 1D), using Bailey’s classification.7 No statistically significant differences were observed for DYRK1B expression among the different subtypes (data not shown). No differences in  DYRK1A expression were found among tumour grades from TCGA and GSE62452 cohorts (supplementary figure S1F).

Finally, although no statistically significant changes on the median survival were observed when comparing the 50% of patients with higher and lower DYRK1A levels using the TCGA cohort or the GSE62452 cohort (supplementary figure S1G), the 5-year survival percentage on the ‘low-DYRK1A’ group was 27% while it was 9.6% on the ‘high-DYRK1A’ group in the GSE62452 cohort.

DYRK1A downregulation reduces in vitro proliferation and in vivo tumourigenicity in PANC-1 xenografts

To determine the function of DYRK1A in PDAC, we studied the behaviour of the human pancreatic cancer cell line PANC-1, which expresses DYRK1A and DYRK1B (supplementary figure S2A), after reducing DYRK1A expression by lentiviral delivery of specific shRNAs (see supplementary figure S2B for knockdown efficiency). DYRK1A downregulation led to reduced cell proliferation and reduced number of colonies in clonogenic assays, with comparable results using three independent shRNAs (figure 2A,C). The reduction in cell proliferation was also induced by treatment with the DYRK inhibitor harmine (figure 2B), suggesting the requirement for DYRK kinase activity. Additionally, the antiproliferative effects were also observed in DYRK1B knockdown cells (supplementary figure S2C–E), in agreement with previous observations.13 Double knockdown of DYRK1A and DYRK1B proteins also showed reduced number of colonies, although we did not observe any additive effect (supplementary figure S2F). Finally, silencing of DYRK1A in BxPC-3 cells, a PDAC cell line lacking DYRK1B, led to similar results (supplementary figure S2A,G–I).

Figure 2

DYRK1A reduction impairs pancreatic tumour cell proliferation and migration. (A) Cell proliferation assay of shControl and DYRK1A knockdown PANC-1 cells. The relative cell numbers at day 0 were arbitrarily set to 1 (n=5; two-way analysis of variance (ANOVA), only comparisons at day 7 are shown). (B) Cell proliferation assay of PANC-1 cells treated with two different concentrations of harmine. The relative cell numbers at day 0 were arbitrarily set to 1 (n=2, in triplicate; two-way ANOVA, only comparisons at day 4 are shown). (C) Colony formation assay of shControl and DYRK1A knockdown PANC-1 cells. Images correspond to a representative experiment, and the graph shows the quantification of colony numbers (n=2–3 independent experiments, performed in triplicate; Mann-Whitney test). (D,E) Silencing of DYRK1A in PANC-1 cells significantly reduced its ability to migrate (D) and invade (E) in transwell assays (n=4; Mann-Whitney test). *p<0.05; **p<0.01; ***p<0.001.

A hallmark of tumorigenesis is its ability to promote cell invasion. To determine whether DYRK1A is required for migration in PANC-1 cells, we assessed the effect of silencing DYRK1A in transwell assays. The reduction in DYRK1A expression with two independent shRNAs greatly diminished the ability of PANC-1 cells to migrate (figure 2D) and to invade through a matrigel layer (figure 2E). The impairment was also observed in conditions of DYRK1B knockdown (supplementary figure S2J,K). Thus, PANC-1 invasiveness requires the presence of DYRK1A and/or DYRK1B.

To evaluate the impact of DYRK1A in tumour progression, PANC-1 cells expressing shCtrl or shDYRK1A were inoculated subcutaneously into the flanks of nude mice. Follow-up of tumour volume revealed a clear growth delay in tumours with DYRK1A knockdown (figure 3A). Consistently, tumour immunohistochemistry analysis of the Ki67 marker confirmed a reduced proliferation rate in DYRK1A knockdown tumours (figure 3B). These results indicate that reduced levels of DYRK1A impair proliferation, leading to slower tumour progression.

Figure 3

DYRK1A contributes to tumour proliferation. (A) Follow-up of tumour growth of PANC-1 and PANC-1-shDYRK1A xenografts (n=6 per group; two-way analysis of variance (ANOVA), only comparisons at day 44 are shown). DYRK1A protein levels are shown in supplementary figure S3. (B) Representative images of immunohistochemical staining of the proliferation marker Ki67 in the excised subcutaneous tumours (scale bar, 100 µm). Quantification of the percentage of Ki67 positive nuclei (n=5; Mann-Whitney test). *p<0.05; **p<0.01; ***p<0.001.

DYRK1A haploinsufficiency enhances survival of Ela-myc mice

To gain further insight to a role of DYRK1A in tumour initiation and progression, we first evaluated the expression of DYRK1A in two genetically modified mouse models of pancreatic carcinogenesis: the PDAC K-Ras driven model KPC and the Ela-myc transgenic mice, in which the c-Myc oncogene is expressed under the control of the elastase promoter.19 DYRK1A was highly expressed in the pancreata of the two mouse models, both at the protein and mRNA levels (figure 4A–D for Ela-myc and supplementary figure S4A,B for KPC). DYRK1B overexpression was also observed in these models (supplementary figure S4A,B, S5A–C). Next, we crossed Ela-myc transgenic mice with Dyrk1a+/–heterozygous mice (Dyrk1a–/– mice are embryonic lethal21) and checked that inactivation of one Dyrk1a allele resulted in reduced expression of the kinase in the Ela-myc context (supplementary figure S5F). Interestingly, a significant increase in survival was observed in mice with only one copy of Dyrk1a, with a median survival of 118 days for Ela-myc:Dyrk1a+/+ and of 194 days for the Ela-myc:Dyrk1a+/–genotype (figure 4E). All mice with only one copy of Dyrk1a survived longer than 6 months, at which time all Ela-myc:Dyrk1a+/+ mice had already died. Indeed, when considering long-term survivors (eg, longer than 120 days), only 30% of Ela-myc survived, whereas 81% of mice with the combined genotype survived (figure 4F). Improved survival was also observed in mice treated with the DYRK1A inhibitor harmine, with 60% of Ela-myc:Dyrk1a+/+-treated mice surviving longer than 120 days (figure 4E,F). Evaluation of Ki67 staining showed a clear reduction in staining in Ela-myc:Dyrk1a+/–mice (figure 4G), indicating that Dyrk1a haploinsufficiency decreased proliferation in the tumour tissue. Indeed, histopathological analysis of the pancreata at 11 weeks and 16 weeks from Ela-myc:Dyrk1a+/+ and Ela-myc:Dyrk1a+/–mice indicated reduced proliferation at every time point analysed (supplementary figure S5D,E). Taken together, the data show that DYRK1A inhibition limits proliferation of tumour lesions in Ela-myc mice, leading to a delay in tumour growth and enhanced survival, pointing to a role for DYRK1A as a mediator of pancreatic cancer progression.

Figure 4

DYRK1A deficiency enhances mouse survival and reduces tumour cell proliferation in Ela-myc mice. (A) Immunodetection of DYRK1A in the pancreas of wild type (WT) and Ela-myc mice (scale bar, 100 µm). (B,C) DYRK1A expression in pancreas of WT and Ela-myc mice detected by western blot (B) and quantified relative to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) levels (C) (n=6; Mann-Whitney test). (D) qRT-PCR analysis of Dyrk1a mRNA in Ela-myc (n=12) and WT mice (n=4; Mann-Whitney test). (E) Kaplan-Meier survival curves of mice with Ela-myc:Dyrk1a+/+ (n=12), Ela-myc:Dyrk1a+/– (n=11), or Ela-myc:Dyrk1a+/+ that had been treated daily with harmine (n=10). Statistical significance was only observed between Ela-myc:Dyrk1a+/+ and Ela-myc:Dyrk1a+/- (log-rank test). (F) Pie charts represent the percentage of animals of each group that survived longer than 4 months (light grey) (Fisher’s exact test: Ela-myc:Dyrk1a+/+ vs Ela-myc:Dyrk1a+/-, p=0.0361; Ela-myc:Dyrk1a+/+ vs Ela-myc:Dyrk1a+/++harmine, p=0.3913). (G) Representative images of immunohistochemical staining of the proliferation marker Ki67 in the excised pancreatic tumours of Ela-myc:Dyrk1a+/+ and Ela-myc:Dyrk1a+/-, and quantification of the positive nuclei for this marker (n=6 per genotype; Mann-Whitney test). **p<0.01.

DYRK1A regulates c-MET in PDAC

Inhibition of DYRK1A has been found to reduce epidermal growth factor receptor (EGFR)-dependent growth in glioblastomas.16 In PDAC, EGFR signalling is essential for K-ras–driven carcinogenesis.26 Furthermore, the hepatocyte growth factor (HGF)/c-MET signalling pathway is also known to be involved in this malignancy.27 Indeed, activation of EGFR and c-MET by their ligands triggers common downstream signalling pathways, and their crosstalk appears to be important for tumour development.28 On this basis, we sought to investigate an EGFR-mediated and/or c-MET-mediated effect in PDAC that is dependent on DYRK1A.

First, we asked whether EGFR and/or c-MET expression was affected in the Ela-myc transgenic mice with or without monoallelic Dyrk1a inactivation. Expression of EGFR in Ela-myc pancreas did not significantly differ from that in WT pancreas (supplementary figure S5F,G). In contrast, increased c-MET expression was detected in Ela-myc mice samples (figure 5A and supplementary figure S5F), which was very likely due to increased c-Met mRNA levels (figure 5B). Notably, whereas no changes in c-Met transcript levels were observed in the Ela-myc mice containing only one copy of Dyrk1a, c-MET protein levels were reduced in the Dyrk1a monoallelic genotype (figure 5A,B and supplementary figure S5F).

Figure 5

DYRK1A regulates c-MET and epidermal growth factor receptor (EGFR) post-transcriptionally. (A) Western blot quantification analysis of DYRK1A and c-MET protein levels in pancreas derived from wild type (WT), Ela-myc:Dyrk1a+/+ and Ela-myc:Dyrk1a+/– mice (n=4–6; one-way analysis of variance (ANOVA)). (B) qRT-PCR analysis of Dyrk1a and c-Met expression in pancreas derived from Ela-myc:Dyrk1a+/+ and Ela-myc:Dyrk1a+/– mice (n=4–9; one-way ANOVA; the increase of Dyrk1a mRNA in Ela-myc:Dyrk1a+/+ vs WT mice was statistically significant in pairwise comparisons using a Mann-Whitney test, p=0.0196). (C–E) DYRK1A, c-MET and EGFR protein expression in PANC-1 xenografts (at 44 days after tumour cell inoculation) assessed by western blot (C), and immunohistochemistry (E); the protein levels quantified in immunoblots were plotted for correlation analysis (D) (n=12; Spearman’s correlation test). (F) qRT-PCR mRNA analysis (shCtrl, n=10; shDYRK1A.1, n=6; shDYRK1A.2, n=5; Mann-Whitney test). *p<0.05, **p<0.01, p***<0.001, ns, not significant.

We next analysed the expression of EGFR and c-MET in PANC-1 xenografts. Both EGFR and c-MET proteins were substantially reduced in DYRK1A knockdown tumours (figure 5C,E). Comparison of protein levels between the different samples shows positive correlation between protein levels of DYRK1A and those of c-MET and EGFR (figure 5D). No changes were detected in c-MET and EGFR transcripts (figure 5F), suggesting that DYRK1A was acting on EGFR and c-MET at the post-transcriptional level. These results agree with those observed in the genetic model with reduced expression of Dyrk1a, and therefore suggest that DYRK1A could be regulating c-MET protein accumulation.

To obtain more evidence of the relationship between DYRK1A and c-MET and EGFR, we conducted expression studies in paraffin-embedded human pancreatic tumours (figure 6A). Human tumour samples were stratified on the basis of high (H-score: 226.6±20) or low (H-score: 155.9±25.9) DYRK1A protein levels using mean immunohistochemistry intensity. c-MET, but not EGFR, staining positively correlated with DYRK1A content (figure 6B,C). c-MET and DYRK1A correlation at the mRNA level was studied in two independent cohorts of larger numbers of samples (TCGA and GSE62452 cohorts), but no significant correlation was found (supplementary figure S6A).

Figure 6

DYRK1A protein levels positively correlate with c-MET expression in human pancreatic tumours. (A) Representative images of immunohistochemical staining of DYRK1A, c-MET and epidermal growth factor receptor (EGFR) in low-grade pancreatic intraepithelial neoplasia (PanIN) and pancreatic ductal adenocarcinoma (PDAC) specimens. High magnification images are shown from low-grade PanIN lesions (scale bars, 10 µm and 100 µm). (B) Tumour samples (n=20) were categorised according to DYRK1A expression levels into high and low, according to the H-score, and rescored for EGFR and c-MET. The data show mean±SD of each category (Student’s t-test). (C) Correlation analysis using the H-scores of DYRK1A and c-MET (Spearman’s correlation test).

DYRK1A controls c-MET stability through SPRY2

To investigate whether DYRK1A was indeed modulating c-MET protein accumulation, PANC-1 cells were used as the model system. Indeed, reduction of DYRK1A protein amounts in knockdown experiments correlated with diminished c-MET protein levels but not with mRNA levels (figure 7A,B). This effect could be due to a reduced protein synthesis or increased degradation; however, silencing of DYRK1A led to a reduction in the half-life of c-MET, as assessed with the protein synthesis inhibitor cycloheximide (figure 7C and supplementary figure S7A), supporting a role for DYRK1A in controlling c-MET protein stability. Critically, treatment of cells with the DYRK inhibitor harmine also reduced c-MET protein levels (figure 7D), suggesting that DYRK1A’s catalytic activity is required for controlling c-MET accumulation.

Figure 7

DYRK1A regulates c-MET stability. (A,B) Steady state levels in shControl and shDYRK1A PANC-1 cells for c-MET protein (A, representative image and quantification; n=6) and c-MET mRNA (B, n=3) (Mann-Whitney test). (C) Graph showing c-MET protein decay in shDYRK1A PANC-1 cells after treatment with cycloheximide (n=4; two-way analysis of variance (ANOVA)). (D) Representative western blot analysis of c-MET accumulation in PANC-1 cells treated with harmine or vehicle (as control). The graph shows the corresponding quantitative analysis in which the c-MET levels at time=0 were arbitrarily set as 1 (n=4; one-way ANOVA). (E) Time course of c-MET protein levels in shControl and shDYRK1A PANC-1 cells in response to hepatocyte growth factor (HGF) stimulation. The graph shows the quantification of independent experiments in which the c-MET level at time=0 was set as 1 (n=5; one-way ANOVA). (F) HGF-induced c-MET accumulation in PANC-1 cells in the presence or absence of the proteasome inhibitor MG132. (G) DYRK1A re-expression rescues c-MET from degradation. Western blot analysis of c-MET in shDYRK1A.4 PANC-1 cells and in silenced cells expressing a sh4-resistant form of DYRK1A. *p>0.05, **P>0.01, ***p<0.001; ns, not significant. A cross-reacting band in A, D and E is indicated with an asterisk.

Accumulation of c-MET responds to HGF treatment as a result of stimulus-dependent endocytosis, intracellular trafficking and degradation.29 Addition of HGF resulted in a time-dependent regulation of c-MET protein levels, which was sensitive to the proteasome inhibitor MG132, indicating that HGF-dependent degradation was active in PANC-1 (figure 7F). Notably, c-MET degradation kinetics on HGF treatment was accelerated in shDYRK1A cells (figure 7E). The effect was DYRK1A-specific, since rescue of DYRK1A expression restored c-MET protein levels (figure 7G).

Upon DYRK1A downregulation in PANC-1 cells, an increase in the protein levels of its paralogue DYRK1B was observed (supplementary figure S7B), perhaps as a compensatory response. We cannot rule out that the harmine-dependent effect on c-MET accumulation was the result of also inhibiting DYRK1B, since the drug does not discriminate between the two DYRK family members.30 In fact, silencing of DYRK1B expression affected c-MET accumulation (supplementary figure S7C). To evaluate the role of DYRK1A in c-MET accumulation in the absence of DYRK1B, we assessed the behaviour of the receptor in the BxPC-3 cell line, in which DYRK1B expression was not detected (supplementary figure S2A,F). BxPC-3 cells responded to harmine treatment by reducing c-MET accumulation (supplementary figure S7D); likewise, silencing of DYRK1A in this cell line, either in normal growth conditions (supplementary figure S7E) or after stimulation with HGF (supplementary figure S7F), triggered reduced c-MET protein levels. The results thus support that DYRK1A mediates c-MET protein accumulation.

Various signalling pathways are activated on HGF binding to c-MET receptor, which trigger different cellular responses.29 Among them, activation of the mitogen-activated protein kinase signalling cascade, as well as its downstream target extracellular signal-regulated kinase (ERK), is involved in cell proliferation. Notably, DYRK1A knockdown cells displayed reduced phosphorylation of ERK in response to HGF compared with control cells (figure 8A), suggesting that the effect of DYRK1A in c-MET regulation impacts the activation of this pathway. Notably, knockdown of DYRK1A in MIA PaCa-2 cells, lacking endogenous c-MET, had no impact on cell proliferation, pointing to c-MET signalling being required for the DYRK1A proliferation effects (figure S7G,H).

Figure 8

DYRK1A modulates c-MET accumulation through SPRY2 phosphorylation. (A) Time course of extracellular signal-regulated kinase (ERK) phosphorylation on hepatocyte growth factor (HGF) stimulation in shControl and shDYRK1A PANC-1 cells. A representative western blot and the quantification of pERK1/2 levels relative to ERK2 are shown (n=3; Mann-Whitney test). (B) Time course of c-MET accumulation in shControl and shSPRY2 PANC-1 cells in response to HGF stimulation. (C) Western blot analysis of c-MET protein levels in shControl and shDYRK1A U2OS cells transfected with a phosphomimetic (T75D) or a non-phosphorylatable (T75A) SPRY2 variant (supplementary figure S8B for a similar experiment in HeLa cells). Quantification of c-MET is shown, relative to t=0. *p<0.05. Cross-reacting bands in B and C are indicated with asterisks.

Phosphorylation of c-MET at Y1003 has been shown to regulate the stability of the receptor.31 We reasoned that the reduced c-MET accumulation could be due to increased phosphorylation of this residue in conditions of DYRK1A silencing. However, no such increase was detected (supplementary figure S8A). DYRK1A-dependent modulation of EGFR turnover works through the RTK inhibitor SPRY2.32 SPRY2 positively regulates c-MET signalling in colon cancer and rhabdomyosarcoma cells by affecting c-MET mRNA and/or protein levels.33 34 However, overexpression of SPRY2 inhibits HGF-mediated signalling in leiomyosarcoma cells and hepatocellular carcinoma cells.35 36 Given these contradictory results, we first analysed the potential involvement of SPRY2 in the regulation of c-MET protein levels in pancreatic tumour cells. Silencing of SPRY2 in PANC-1 cells decreased the accumulation of c-MET following HGF stimulation (figure 8B), pointing to a requirement of SPRY2 for maintaining c-MET protein levels in this cell line. We have previously shown that DYRK1A phosphorylates SPRY2 at threonine residue 75.22 Overexpression of a SPRY2 T75 phosphomimetic mutant (T75D) partially reversed the c-MET reduction in protein levels in cells in which DYRK1A was silenced (figure 8C and supplementary figure S8B). In contrast, the expression of a non-phosphorylatable SPRY2-mutant (T75A) did not restore c-MET protein levels (figure 8C and supplementary figure S8B). These results suggest that DYRK1A phosphorylation of SPRY2 could be antagonising c-MET degradation.

Discussion

Our results show that DYRK1A is upregulated in PDAC tumour samples, which could be an event that promotes tumour formation through stabilisation of c-MET, thus increasing the signalling efficiency of this RTK. Increased expression of DYRK1A is observed in low-grade PanIN, in both human samples and the KPC mice, suggesting a potential role in early phases of PDAC.

We show that DYRK1A reduction led to a significant reduction in pancreatic cancer cell proliferation, migration and invasion. A previous report showed the effect of reducing the DYRK1A paralogue DYRK1B in clonogenic assays,13 but we show here that both kinases have proproliferative effects in PDAC cell lines. Consistent with this observation, DYRK1A silencing was highly effective in inhibiting tumour growth in human xenograft PDAC models. Notably, Dyrk1a monoallelic deletion reduced tumour progression and increased survival of Ela-myc transgenic mice. The cellular effects appear to be dependent on DYRK1A kinase activity and, more importantly, the in vivo effects on tumour growth were mimicked when the animals were treated with a Dyrk1 inhibitor. The xenograft data support a cell autonomous protumorigenic role of DYRK1A. However, in addition to decreased cell proliferation, other DYRK1A-dependent mechanisms may converge to ameliorate carcinogenesis in vivo. For instance, the kinase might help to reprogramme the neoplastic cells by modulating crosstalk signals in the tumour milieu, or through its proangiogenic potential.37

Our data show that inhibition of DYRK1A leads to reduced c-MET protein levels in the tumour context. In agreement with our in vitro results, we found a positive correlation between the expression of c-MET and DYRK1A in clinical tumour samples of PDAC. We have previously shown that DYRK1A regulates EGFR stability in adult neural stem cells,32 and this role has also been demonstrated in glioblastoma;16 moreover, we identified a role for DYRK1A in the stabilisation of VEGFR in primary endothelial cells.37 Our findings in PDAC cells indicate that DYRK1A increases c-MET accumulation by impairing its degradation, suggesting a common theme for DYRK1A as regulator of RTK accumulation.

The regulation of c-MET stability is a complex process in which different effectors might play a role depending on the cellular context.29 In the case of pancreatic tumour cells, the mechanisms regulating c-MET stability are poorly understood; therefore, the DYRK1A (or DYRK1B) roles in this process will need further investigation. However, our results support a role for DYRK1A acting upstream of SPRY2 to modulate c-MET targeting for degradation. SPRY proteins are negative regulators of RTKs activated pathways by targeting different elements within the RTK signalling cascades.38 In the case of c-MET, both positive and negative roles have been described for SPRY2, depending on the cellular context.33 34 36 Our results support a role of SPRY2 as a positive regulator of c-MET in PANC-1 cells; additionally, our data indicate that c-MET stabilisation might require DYRK1A-dependent SPRY2 phosphorylation, highlighting a regulatory mechanism similar to that of EGFR.32 DYRK1A is an active kinase in PANC-1 cells under normal growth conditions (supplementary figure S8C,D) and after HGF stimulation (supplementary figure S8E). The fact that we did not observe significant changes of DYRK1A-associated kinase activity following HGF activation (supplementary figure S8E) could suggest that the constitutive activity of DYRK1A is required to maintain c-MET protein levels; alternatively, regulation in response to HGF could be exerted at the level of substrate availability. Finally, as DYRK1A is a pleiotropic kinase, its modulatory role on c-MET stability could be just one of the DYRK1A-dependent activities that participate in its protumorigenic potential in PDAC. For instance, the hedgehog and Notch pathways are known to be important in PDAC,39 40 and DYRK1A acts as a positive regulator of these pathways.41 42

Overexpression of c-MET is a common alteration in PDAC and has been associated with tumour growth as well as resistance to chemotherapy.27 43 Targeting the HGF-c-MET pathway has been approached for the treatment of a variety of solid tumours with deregulated c-MET.44 However, tyrosine kinase inhibitors (TKI) of c-MET, monoclonal antibodies or small molecules designed to block the HGF/c-MET interaction have proven insufficient.45 As this failure has been suggested to be due to improper drug design, targeting c-MET for degradation might represent an effective approach. In this regard, the preliminary clinical activity obtained with the bivalent anti-c-MET antibody LY2875358 in a phase II trial might be related to its ability to induce internalisation and degradation of c-MET.46 c-MET-targeted therapy has been also tested in combination with EGFR TKIs.47 Of note, EGFR-targeted therapies have also been evaluated in PDAC.48 Our work shows that inhibition of DYRK1A resulted in reduced accumulation of both EGFR and c-MET proteins, and that DYRK1A and DYRK1B paralogues function similarly in this regard. These observations, together with the fact that both kinases are inhibited by the same type of molecules, prompt us to propose that using DYRK1A/DYRK1B as a target for therapy in PDAC could be considered as a combinatorial therapy targeting both EGFR and c-MET for degradation, thus leading to more effective anticancer therapies.

Acknowledgments

The authors thank Manuel Dominguez-Fraile for his help with mice genotyping. The group of SL thanks MEIC ’Centro de Excelencia Severo Ochoa' and the EMBL partnership for their support. JB is a FPI predoctoral fellow (BES-2014-069983) and XB-DR is a FPU fellow. The authors thank the Banc de Tumors-Biobank core facility of the Hospital-Clínic-IDIBAPS for technical help (work supported by the Xarxa de Bancs de Tumors de Catalunya - XBTC). The authors also acknowledge the CERCA Programme/Generalitat de Catalunya for their support. This work was developed at the Centro Esther Koplowitz (Barcelona, Spain).

References

View Abstract

Footnotes

  • SL and CF share senior authorship.

  • JL and JB contributed equally.

  • Contributors JL and JB designed and performed the experiments, analysed the data and gave input to the manuscript; MC analysed human histological samples; XB-DR performed computational analysis; EN-M performed experiments; MG provided RNA samples of patients with PDAC; ECV and MLA provided KPC and Dyrk1a +/- mice; SdlL and CF devised the project, coordinated the work, analysed data and wrote the manuscript. All authors approved the manuscript.

  • Funding This work was supported by grants from the Spanish Ministry of Economy, Industry and Competitiveness (MEIC) (BIO2014-57716-C2-2-R, BIO2017-89754-C2-2R to CF and BFU2016-76141-P to SL), and receives partial support from the Generalitat de Catalunya (SGR14/248 to CF and SGR14/674 to SL). CIBER de Enfermedades Raras and de Enfermedades Hepáticas y Digestivas are initiatives of the Instituto de Salud Carlos III (ISCIII). The group of CF is partially financed by the ISCIII (IIS10/00014), co-financed by Fondo Europeo de Desarrollo Regional (FEDER), and acknowledges the support of the COST Action BM1204 EUPancreas. MG acknowledges grant PI17/01003 from ISCIII.

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval The study was approved by the HCB Institutional Ethics Committee. All animal procedures met the guidelines of European Community Directive 86/609/EEC and were approved by the Ethical Committee (CEEA-University of Barcelona) and by the local authorities (Generalitat de Catalunya).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.