DYRK1B as therapeutic target in Hedgehog/GLI-dependent cancer cells with Smoothened inhibitor resistance

Abstract

A wide range of human malignancies displays aberrant activation of Hedgehog (HH)/GLI signaling, including cancers of the skin, brain, gastrointestinal tract and hematopoietic system. Targeting oncogenic HH/GLI signaling with small molecule inhibitors of the essential pathway effector Smoothened (SMO) has shown remarkable therapeutic effects in patients with advanced and metastatic basal cell carcinoma. However, acquired and de novo resistance to SMO inhibitors poses severe limitations to the use of SMO antagonists and urgently calls for the identification of novel targets and compounds.Here we report on the identification of the Dual-Specificity-Tyrosine-Phosphorylation-Regulated Kinase 1B (DYRK1B) as critical positive regulator of HH/GLI signaling downstream of SMO. Genetic and chemical inhibition of DYRK1B in human and mouse cancer cells resulted in marked repression of HH signaling and GLI1 expression, respectively. Importantly, DYRK1B inhibition profoundly impaired GLI1 expression in both SMO-inhibitor sensitive and resistant settings. We further introduce a novel small molecule DYRK1B inhibitor, DYRKi, with suitable pharmacologic properties to impair SMO-dependent and SMO-independent oncogenic GLI activity. The results support the use of DYRK1B antagonists for the treatment of HH/GLI-associated cancers where SMO inhibitors fail to demonstrate therapeutic efficacy.

Keywords: DYRK1B; GLI transcription factors; Hedgehog/GLI signaling; Smoothened drug resistance; basal cell carcinoma.

Figure 1. The DYRK1 inhibitor harmine blocks canonical HH/GLI signaling

A. Evolutionary distance of DYRK family members and mode of action of distinct DYRK members on GLI activation (DYRK1A) and GLI degradation (DYRK2). B. DAOY human medulloblastoma cells harbor a responsive canonical HH/GLI signaling system. Treatment with the SMO agonist SAG (100nM) results in activation of GLI1 expression that is quantitatively abolished in the presence of the clinically approved SMO inhibitor vismodegib (vismo) (0.5 μM). Treatment with recombinant sonic HH protein yielded comparable results (data not shown). C. qPCR analysis showing repression of GLI1 mRNA (left) and PTCH mRNA expression (right panel) in SAG-stimulated DAOY cells in response to vismodegib (0.5 μM) or harmine treatment (10 μM and 20 μM). D. Analysis of concentration-dependent inhibition of HH pathway activity and IC₅₀ calculation of 10.9 μM for the natural DYRK inhibitor harmine. E. Efficient inhibition of GLI1 protein expression in SAG-stimulated DAOY cells either treated with vismodegib (0.5 μM) or harmine (10 μM and 20 μM).

Figure 2. Genetic perturbation of DYRK1B interferes with canonical HH/GLI pathway activation

A.-B. qPCR analysis of GLI1 (A) or PTCH mRNA expression (B) in SAG-treated DAOY cells stably transduced with scrambled control shRNA (shcont), shRNA against DYRK1A (shD1A), or two shRNAs against DYRK1B (shD1B#1, shD1B#2). C. Western blot analysis of GLI1 protein expression in SAG-stimulated DAOY cells expressing the respective lentiviral shRNA constructs and a third shRNA against DYRK1B (shD1B#3). D. Western blot analysis of Ptch-deficient murine BCC cell lines showing abrogation of Gli1 expression by harmine treatment (left panel) (10 μM and 20 μM) and by shRNA against Dyrk1b (shD1b) but not by shRNA against Dyrk1a (shD1a) (right panel). Fine black lines indicate cropping of intermediate lanes from the same Western blots. ACTB/Actb: human/mouse beta actin loading control.

Figure 3. DYRK1B targeting inhibits SMO-dependent and SMO-independent activation of GLI

A. Illustration of HH/GLI signaling and SMO-targeting in PTCH- or SUFU-deficient cells. In PANC-1 pancreatic cancer cells, TGFb and RAS control GLI1 expression independent of SMO. In Ewing sarcoma cells (A673) the EWS-FLI1 oncoprotein directly activates GLI1 expression. B. shRNA-mediated depletion of SUFU renders DAOY cells resistant to SMO inhibition by vismodegib (vismo). Western blot analysis showing that stable expression of shRNA against SUFU (shSUFU) results in activation of GLI1 expression. Note that GLI1 expression in SUFU depleted cells is resistant to SMO inhibition by vismodegib. shcont: scrambled control shRNA. C. Western blot showing GLI1, GLI2 and GLI3 expression in wild-type and SUFU-deficient DAOY cells in response to DYRK1A/B knock-down. Note that RNAi against DYRK1B (shD1B) but not against DYRK1A (shD1A) strongly reduces GLI1 and moderately reduces GLI2 expression in SUFU-depleted (shSUFU) DAOY cells. GLI3 expression and processing are unaffected by DYRK1A and DYRK1B targeting. D. Harmine treatment (10 μM and 20 μM) inhibits Gli1 protein expression in both Ptch-deficient and Sufu-deficient mouse embryonic fibroblasts (MEF). E. Human pancreatic adenocarcinoma cells PANC-1 express detectable levels of GLI1 protein in response to TGFb/SMAD and RAS signaling [23]. GLI1 expression in PANC-1 cells is independent of SMO activity since vismodegib (vismo) treatment does not affect GLI1 protein levels (left panel). RNAi against GLI1 (shGLI1) and against DYRK1B (shD1B) but not against DYRK1A (shD1A) efficiently represses GLI1 protein expression. DYRK1B targeting does not affect non-HH/GLI effectors such as STAT5 (or STAT3 and CTNNB, data not shown). F. qPCR analysis of PANC-1 cells showing that inhibition of DYRK1B (shD1B#1, shD1B#2) but not of DYRK1A (shD1A) reduces expression of the GLI target BCL2. G. GLI1 expression in the Ewing sarcoma cell line A673 harboring the EWS-FLI1 oncogene is resistant to SMO inhibition by vismodegib (vismo) treatment (left panel). While shRNA against DYRK1A (shD1A) does not affect GLI1 expression in A673 cells, knock down of DYRK1B with two distinct shRNAs (shD1B#1, shD1B#2) decreases GLI1 expression (right panel). shGLI1 knockdown demonstrates specificity of the anti-GLI1 antibody used for detection of GLI1 in PANC-1 and A673 cells. ACTB/Actb: human/mouse beta actin loading control. Fine black lines indicate cropping of intermediate lanes from the same Western blots.

Figure 4. A novel DYRK1 inhibitor efficiently repressing SMO-dependent and SMO-independent GLI1 expression

A. Chemical structure of DYRKi, a novel DYRK1 inhibitor. B. Concentration-dependent inhibition of Hh/Gli signaling in murine Gli luciferase reporter cells by DYRKi resulting in an IC₅₀ of 3.7 μM. C. DYRKi efficiently blocks HH pathway activity in both SAG-stimulated wild-type human medulloblastoma cells (DAOY, WT MB +SAG) and SMO-inhibitor resistant, SUFU depleted medulloblastoma cells (ΔSUFU MB). D. shRNA mediated depletion of SUFU expression renders SAG-stimulated human medulloblastoma cells resistant to SMO inhibition by vismodegib (vismo). Data in C and D were calculated as a function of GLI1 mRNA expression in the respective samples. GLI1 mRNA expression was determined by qPCR. GLI1 mRNA levels of SAG-treated/solvent controls (wild-type DAOY) or solvent-only treated SUFU depleted DAOY cells were set to 100 percent. E. GLI1 protein expression in SAG-stimulated DAOY medulloblastoma cells treated with SMO-antagonists vismodegib (vismo, 0.5 μM), cyclopamine (cyc, 5 μM) or with DYRK1 inhibitors DYRKi (1 μM and 5 μM) or harmine (10 μM and 20 μM). F. Inhibition of GLI1 protein expression in SUFU depleted DAOY medulloblastoma (shSUFU) cells by DYRKi treatment (1 μM and 5 μM). Note that vismodegib (vismo, 0.5 μM) fails to reduce GLI1 expression. shcont: scrambled control shRNA. G. DYRKi treatment (1 μM and 5 μM) represses Gli1 protein expression in both Ptch-deficient and Sufu-deficient mouse embryonic fibroblasts isolated from Ptch- and Sufu knockout mice, respectively. Vismodegib (vismo) inhibits Gli1 expression in Ptch-deficient cells only. H. Treatment with the proteasome inhibitor bortezomib largely reverses the suppressive effect of DYRKi on GLI1 and GLI2 protein expression, supporting a model of post-translational regulation of GLI1 and GLI2 stability by DYRK1B. Fine black lines indicate cropping of intermediate lanes from the same Western blots. ACTB/Actb: human/mouse beta actin loading control.

Figure 5. DYRK1B targeting inhibits the malignant properties of GLI1-dependent human pancreatic cancer cells

A.-B. Growth of GLI1 expressing PANC-1 spheres in 3-D cultures. Sphere formation is resistant to SMO inhibition by vismodegib (vismo) and shRNA against DYRK1A. By contrast, harmine (10 μM and 25 μM), DYRKi (5 μM) and shRNA against DYRK1B (shD1B) efficiently prevent the formation of tumor-initiating spheres. Sphere formation shown in A) was quantitatively analyzed and the number of tumor-initiating (ti) spheres plotted in B). For all experiments, identical numbers of live cells were seeded into 3D matrix cultures. We noted that inhibition of DYRK1B does not simply induce cell death but prevents the formation of large spheres formed by highly clonogenic, putative tumor-initiating cells [16]. C.-D. Xenograft analysis of in vivo tumor growth of PANC-1 (n=7) in C. and highly metastatic GLI1-dependent L3.6pl pancreatic cancer cell lines (n=6) in D. shControl: cells lentivirally transduced with scramble control shRNA, shDYRK1B: cancer cells stably expressing shRNA against DYRK1B. E. Oral administration of DYRKi (100 mg/kg/d) significantly reduces in vivo tumor growth of GLI1-dependent pancreatic cancer cells (L3.6pl) (n=20). Control mice (n=8) received solvent only (control). * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001;