Casein kinase 1α has a non-redundant and dominant role within the CK1 family in melanoma progression

Human melanoma cell lines were cultured for this study in RPMI 1640 medium with 2 mM L-Glutamine and 10 % fetal bovine serum (FBS; Biochrom, Berlin, Germany), penicillin, and streptomycin. They were subcultured 1–2 times a week when they reached 80 % confluency using Trypsin/EDTA (0.05 %/0.02 %) for detachment [9, 12]. The melanoma cell lines Malme-3 M, MDAMB435, M14, UACC62, SKMel28 and A375 originated from the NCI60 cell panel of the National Cancer Institute (NCI-DCTD repository). The melanoma cell lines WM35, WM115, WM793, WM3734, WM266-4, WM1366, 1205 LU, and 451 LU were generously provided by M. Herlyn (Philadelphia, USA). SbCl2 and SKMel19 were provided by C. Garbe (Tübingen, Germany). SKMEL30 was obtained from the DSMZ (Braunschweig, Germany) and SKMel147 was a kind gift of M. Soengas (Madrid, Spain). Melanocytes, primary fibroblasts and keratinocytes were isolated from human foreskin as described previously [13‐15]. All of the cell lines used in our study were authenticated by sequence analysis of defined genes.

2.5 × 10⁵ melanoma cells in 6well cavities were transfected with 50 pmol siRNA using RNAiMAX (Invitrogen, Darmstadt, Germany) according to the manufacturers protocol. The following siRNAs were used: siCSNK1A1 sense gaauuugcgauguacuuaa-dTdT, siCSNK1A1 antisense uuaaguacaucgcaaauuc-dTdG; siCSNK1D sense ugaucagucgcaucgaaua-dTdT, siCSNK1D antisense uauucgaugcgacugauca-dTdT; siCSNK1E sense ccuccgaauucucaacaua-dTdT, siCSNK1E antisense uauguugagaauucggagg-dGdA; siNONSIL sense acaacauucauauagcugccccc, siNONSIL antisense gggggcagcuauaugaauguugu (all synthesized by biomers.net, Ulm, Germany)

Wild type CK1 isoform cDNA was amplified using the Human Multiple Tissue cDNA (MTC) Panel II (Clontech, Saint-Germain-en-Laye, France) and isoform specific primers. CK1 cDNAs were cloned into the inducible lentiviral vector PLVX-tight-PURO (Clontech) by using In-fusion-HD Liquid Kits (Clontech) according to the manufacturer’s protocol. Sanger-sequencing was performed for verification of the correct cloned cDNA. Lentiviral particles were produced in HEK293T cells using the second-generation packing and envelope plasmids pCMVΔR8.2 and pMD2.G. Cells were transduced with lentiviruses as described previously [16] and doxycycline inducible melanoma cells were generated according to the manufacturer’s instructions (Tet-on Advanced System, Clontech). For overexpression of CK1α the previously described adenovirus was used [9].

Small molecules were dissolved in DMSO and treatments were carried out using the indicated concentrations with vehicle controls. The following substances were used: Pyrvinium pamoate (Sigma, Taufkirchen, Germany), IC261 (Sigma), D4476 (Sigma), PF670462 (Sigma). Doxycycline hyclate (Applichem, Darmstadt, Germany) was dissolved in ddH₂O and used at the indicated concentrations.

For the analysis of proliferation and survival of melanoma cells, 2.5x10³ cells were seeded into 96-well plates and cultured with the indicated inhibitors for the indicated periods of time. After washing of the cells with PBS, 100 μg/ml 4-methylumbelliferyl heptanoate (Sigma, Taufkirchen, Germany) in PBS were added and incubated for 1 h at 37 °C. Microplates were measured in a fluorescence microplate reader (Berthold, Bad Wildbad, Germany) with Ex355/Em460 nm in sixtuplicates. Dose–response curves were generated using GraphPad Prism version 6 (GraphPad Prism Software Inc.).

2 x10⁵ melanoma cells per 6-well cavity were seeded and either transfected using siRNA or treated with 4 μg/ ml doxycycline to induce the overexpression of CK1δ and ε or transduced with the adenovirus (CK1α overexpression). Cells were cultured for 48 h before permeabilization and fixation of the cells in 70 % ice-cold ethanol for at least 1 h. Then they were re-suspended in PBS with 100 μg/ml RNAseA (Applichem, Darmstadt, Germany) and 50 μg/ml propidium iodide (Sigma, Taufkirchen, Germany) and stained for 30 min. FACS analysis for the detection of the distribution of the cells in the each cell cycle phase was performed with a LSRII FACS (BD, Heidelberg, Germany) using the FACSDiva software.

2.5 × 10³ SKMel19 cells were cultured on 1.5 % noble agar (Difco/BD, Heidelberg, Germany) coated 96well plates to form spheroids within 3 days. For overexpression of CK1 isoforms either 2 μg/ml doxycycline were added on the second day or the medium was supplemented with the adenovirus. After 3 days spheroids were embedded into 1 mg/ml collagen I (Corning/BD, Heidelberg, Germany) diluted in complete growth medium and cultured for four more days. In case of treatment inhibitors were added to the medium. Daily microphotographs were taken and the area of the spheroids was measured using ImageJ and normalized to the size at day 0 after collagen embedding for the evaluation of tumor cell invasion into the collagen matrix. After 4 days spheroids were stained using 1 μM calcein-AM (Life technologies, Darmstadt, Germany) and 100 ng/ml propidium iodide (Sigma, Taufkirchen, Germany) for fluorescence live-dead staining of the melanoma cells. Fluorescence was detected with an Axiovert fluorescence microscope (Zeiss, Jena, Germany). Mean fluorescence intensities of the red channel were used to determine relative cell death induction.

Total RNA was extracted from cells using the NucleoSpin RNA kit (Machery-Nagel, Dueren, Germany). Complementary DNA was made out of 1 μg total RNA using SuperScript II reverse Transcriptase (Invitrogen, Darmstadt, Germany) according to the manufacturer’s protocol. Quantitative real-time PCR (qRT-PCR) was performed with the SYBR green mix LightCycler 480 (Roche, Mannheim Germany). The relative expression levels of CK1 isoforms were determined using the ∆∆Ct-method method with ACTINB or 18S rRNA as reference genes. The primer sequences were as follows: CSNK1A1 forward 5’-aatgttaaagcagaaagcagcac-3’ and reverse 5’-tcctcaattcatgcttagaaacc-3’. CSNK1D forward 5’-acaacgtcatggtgatggag-3’ and reverse 5’-gaatgtattcgatgcgactgat-3’. CSNK1E forward 5’-tgagtatgaggctgcacagg-3’ and reverse 5’-tcaaatggcacacttgtctgt-3’. CSNK1G1 forward 5’-ctgtgaccgaacatttactttga-3’ and reverse 5’-tgcacgtattccattcgaga-3’. CSNK1G2 forward 5’-gaccttcacgctcaagacg-3’ and reverse 5’-ccggtagattaggctcttggt-3’. CSNK1G3 forward 5’-tgcaacaatccaaaaaccagt-3’ and reverse 5’-ctgcaaggtgagctctcaaa-3’. ACTINB forward 5’-ttgttacaggaagtcccttgcc-3’ and reverse 5’-atgctatcacctcccctgtgtg-3’. 18S rRNA forward 5’-cggctaccacatccaaggaa-3’ and reverse 5’-gctggaattaccgcggct-3’.

Protein lysates (30 μg) were subjected to SDS-PAGE and semi-dry blotting onto PVDF membranes (Roche, Mannheim, Germany). The antibodies used were as follows: anti-CK1α (Santa Cruz Biot., Heidelberg, Germany), anti-CK1δ (Santa Cruz Biot.), anti-CK1ε (Santa Cruz Biot), anti-p53 (Santa Cruz Biot), anti-p21 (Cell Signalling, Heidelberg, Germany), anti-β-catenin (Cell Signalling), anti p-S45-β-catenin (Cell Signalling) anti-β-actin (Cell Signalling). HRP conjugated secondary antibodies were used (Cell Signalling and Santa Cruz) and ECL substrates for chemoluminiscent detection. Densitometric semi-quantification was done by normalizing the band intensities of the target protein to the signal of β-actin with Scion Image.

2.5 × 10⁵ melanoma cells were seeded into 6well plates and transfected with 2 μg Super8xTOPFlash 16 h porst seeding using ScreenFectA (Genaxxon, Ulm, Germany) as recommended by the manufacturer. Twenty-four hours later cells were reseeded into 96 well cavities and the expression of isoforms was induced by the addition of doxycycline or of the adenovirus for 48 h. Then cells were lysed with 50 μl of passive lysis buffer (Promega, Mannheim, Germany) and luciferase activity was analyzed using D-luciferin as a substrate (Sigma) in a TriStar luminometer (Berthold, Bad Wildbad, Germany).

Nevi, primary and metastatic melanoma FFPE biopsies were sectioned, heat induced epitope retrieval (HIER) was performed using citrate buffer pH6 and the sections were stained using 1:100 rabbit anti-CK1α (Abcam ab 136052), 1:1000 mouse anti-CK1δ (Abcam ab85320) and 1:100 goat anti-CK1ε (Santa Cruz sc-6471). As secondary antibodies donkey anti-goat(Cy3), donkey anti-mouse(Cy2) and donkey anti-rabbit(Cy5) were used (all 1:250; JacksonImmunoResearch/Dianova, Hamburg, Germany) before staining the nuclei with 1 μg/ml DAPI (Sigma, Taufkirchen, Germany). Biopsies were microscopically analyzed using a confocal microscope system (Leica TCS SP2, Heidelberg, Germany) and the mean fluorescence intensity of representative cells was quantified using the Leica LCS software. For semi-quantification the mean fluorescent intensities of at least 30 cells per sample were background subtracted and presented as relative fluorescence units.

A 23mer peptide containing the exon 3 phosphorylation sites of β-catenin was synthesized as previously described [9] and the NH₂ terminus was labeled with biotin. Melanoma cells were lysed using passive lysis buffer (Promega, Mannheim, Germany), and 5 μg of the protein lysates were incubated in kinase buffer (Cell Signalling, Heidelberg, Germany) together with 10 μg of biotin-labeled peptide for 30 min at 37 °C in streptavidin-coated 96well plates (Life technologies, Darmstadt, Germany). Plates were washed with PBS-T and anti–phospho-Ser45-β-catenin antibody (Cell Signaling) was added (1:500). HRP-conjugated secondary antibody (Cell Signalling) was used to detect the phosphorylated substrate measuring TMB substrate (Cell Signalling) at 450 nm in a microplate reader (Berthold, Bad Wildbad, Germany).

Invasion was assayed using invasion chambers coated with or without Matrigel basement membrane matrix (BD Biocoat Matrigel invasion chambers, BD Biosciences, Heidelberg, Germany) as described previously [9, 16]. After incubation for 20 h at 37 °C the invaded cells were fixed and counted after cell staining with hematoxilin-eosin. The assays were performed in triplicates, six fields were counted per transwell filter and the invasion index was calculated according to the manufacturerer’s protocol.

The kinetics of cell migration was assayed using the xCELLigence Real-Time Cell Analyzer (RTCA DP; Roche). CIM-plate 16 wells used and 10,000cells were plated in each well using serum-free DMEM. The lower medium chamber contained DMEM with 10 % FCS. Cells were allowed to settle for 30 min at room temperature before being placed in the RTCA DP in a humidified incubator at 37 °C with 5 % CO₂. Data were recorded every 15 min for 24 h. Plotted curves represent the averages from three independent measurements.

We analyzed expression of the CK1- isoforms α, δ and ε on RNA and protein level in normal human melanocytes (NHM) and melanoma cell lines representing the different progression stages in melanoma from radial growth phase (RGP), vertical growth phase (VGP) and metastatic melanoma (MM) (Fig. 1a-c). We found a consistent downregulation of CK1α expression on RNA and protein level in RGP, VGP and metastatic melanoma cell lines compared to NHMs. NHMs expressed significantly more CK1δ RNA compared to the melanoma cell lines. However, CK1δ protein expression was variable without significant differences in the analyzed melanoma cell lines. CK1ε expression was low in all cell lines analyzed and could not be detected in NHMs on protein level (Fig. 1a-c). CK1 γ1, γ2 and γ3 RNA expression was almost not detectable in the cell lines analyzed (Additional file 1: Figure S1A). Therefore, we focused in the following experiments on the CK1 isoforms α, δ and ε.

Next, we analyzed RNA and protein expression of the CK1 isoforms α, δ and ε in vivo in tissue samples of benign nevi, primary melanomas and metastatic melanomas using real-time PCR and immunofluorescence analyses, respectively. RNA expression of all three CK1 isoforms did not differ significantly in the different tissue types (Fig. 1c). By trend, CK1α RNA levels were reduced in preparations of metastatic melanoma. In contrast, on protein level we found a significant downregulation of all three CK1- isoforms in metastatic melanomas compared to primary melanoma cells (Fig. 1d). In summary, we found in melanoma cell lines in vitro and in melanoma cells in vivo a consistent downregulation of CK1α RNA and protein expression in metastatic melanoma cells. Furthermore, we detected a downregulation of CK1δ and ε protein expression in metastatic melanoma cells in vivo compared to primary melanoma cells. This did not correlate with RNA expression and with the expression levels of melanoma cells in vitro.

So far it remains unknown whether the individual CK1 isoforms can regulate expression of the other isoforms in melanoma cells. Therefore, we downregulated expression of the CK1 isoforms α, δ or ε in the two human melanoma cell lines SbCl2 and SKMEL19 using isoform-specific siRNAs and analyzed RNA and protein expression of all three CK1 isoforms. As shown in Fig. 2a, downregulation of CK1α or CK1δ did not affect protein expression of the other isoforms in both cell lines. However, downregulation of CK1ε expression induced CK1δ expression most strongly in SKMEL19 cells (Fig. 2a). Combined inhibition of CK1α and CK1δ did only slightly affect CK1ε protein expression in SKMEL19 cells. However, downregulation of CK1α and CK1ε increased CK1δ protein expression, again most strongly in SKMEL19 cells. Downregulation of CK1δ and CK1ε had no effect on CK1α expression. These data suggest that CK1δ and ε regulate each other in a compensatory way and the expression is not or only mildly influenced by CK1α, whereas CK1α expression is independently regulated from CK1δ and ε.

To analyze whether overexpression of the specific isoforms resulted in similar effects we upregulated specifically CK1α expression by adenoviral gene transfer as previously reported [9] and CK1δ and CK1ε by a doxycycline-inducible lentiviral system in the two human melanoma cell lines SbCl2 and SKMEL19 (Fig. 2b). Overexpression of CK1α diminished only expression levels of CK1ε in SbCl2 and only at the highest induced expression level of CK1α. Induction of CK1δ reduced CK1ε protein levels in SKMel19 cells whereas elevated CK1ε levels were associated with lower CK1δ protein expression in SbCl2 cells (Fig. 2b). CK1α expression was not significantly affected by upregulation of the other CK1- isoforms. These data indicate that the δ and ε isoforms negatively regulate expression of each other. Analysis of RNA expression of the individual CK1 isoforms after induction of gene expression using real-time PCR indicated that overexpression of CK1α, CK1δ or CK1ε did not significantly influence RNA expression of the other CK1- isoforms (Fig. 2c). In summary, our data show that CK1δ and CK1ε negatively regulate expression of the respective other CK1 isoforms on a post-transcriptional level, whereas CK1α expression is not significantly affected by the other CK1- isoforms in melanoma cells.

Next, we looked for the functional effects of modulation of CK1- isoform specific gene expression on survival and proliferation of melanoma cells. First, we knocked down CK1α, CK1δ or CK1ε expression in SbCl2 and SKMEL19 melanoma cells using specific siRNAs (Fig. 2a). Ninety-six hours after transfection we analyzed survival and proliferation of the cells (Fig. 3a, Additional file 2: Figure S2A). In both cell lines the downregulation of CK1δ or CK1ε expression alone had no significant effect on cell growth or cell cycle. However, downregulation of CK1α expression retarded cell growth and increased the number of cell in the G1 phase of the cell cycle in SbCl2 melanoma cells, but not in SKMEL19 cells (Fig. 3a, Additional file 2: Figure S2A, B) confirming our previous study [9]. To further ascertain the effect of reduced CK1 activity on melanoma cell survival and proliferation we treated five different human melanoma cell lines with increasing doses of the CK1δ/CK1ε dominant inhibitors D4476 [19], PF670462 [20] or IC261 [21] and measured cell viability 72 h after treatment. As shown in Fig. 3b all three inhibitors did not significantly reduce melanoma cell viability. In a 3D spheroid culture model using collagen-embedded SKMEL19 spheroids similar results were obtained (Fig. 3c). At the highest concentration of IC261 a reduction in the size of the spheroids was observed which, however, was not accompanied with cell death induction (Fig. 3c). Only treatment of the cells with the CK1α activator pyrvinium resulted in propidium iodide positive dead cells (Fig. 3c). Also, overexpression of CK1δ or CK1ε in SbCl2 or SKMEL19 melanoma cells did not change melanoma cell viability and cell cycle (Fig. 3d, Additional file 2: Figure S2C). In contrast, activation of CK1α by pyrvinium [22] (Fig. 3b, c) or overexpression of CK1α in SbCl2 or SKMel19 melanoma cells (Fig. 3d) significantly reduced melanoma cell viability and induced apoptosis (Figs. 3b-d, Additional file 2: Figure S2C). These data indicate that CK1δ and CK1ε are not essential for melanoma cell survival and proliferation, whereas overexpression of CK1α reduces viability of melanoma cells. This suggests that CK1α is the most important CK1 isoform in melanoma cells with a non-redundant function in tumorigenesis.

In order to evaluate a further putative function of the CK1 isoforms in tumorigenesis - an increase in the migratory behavior of the tumor cells - we induced the expression of CK1α, δ and ε isoforms in SKMEL19 melanoma cells by doxycycline treatment and measured the migratory potential of the cells over time using the XCelligence system. Overexpression of CK1δ or ε in the melanoma cells led to no difference in the migratory behavior compared to the non-induced cells (Fig. 4a). However, overexpression of CK1α significantly decreased migration of the melanoma cells. 3D spheroid assays confirmed the results revealing no influence of the CK1- isoforms δ and ε on melanoma cell invasion of SKMEL19 cells into a collagen I matrix (Fig. 4b). CK1α overexpression significantly reduced the invasive growth within the monitored 4 days and again induced cell death. To further evaluate the effect of the CK1- isoforms on the invasive potential of melanoma cells we used an organotypic skin reconstruct using SbCL2 cells with siRNA mediated knockdown of the three CK1- isoforms which were seeded together with primary human keratinocytes as an epidermal layer. Since SbCL2 cells originate from an RGP melanoma they do not have the capacity to invade deep into the dermal part by breaking through the basal membrane which separates epidermal from dermal parts. Knockdown of CK1α resulted in a pro-invasive phenotype indicated by dermally invading melanoma cell nests as we showed before [9]. Knockdown of the other two CK1- isoforms δ or ε had no detectable effects on the growth characteristics in the skin reconstruct model (Fig. 4c). Our data indicate that CK1δ and ε do not affect survival and migration/invasion of melanoma cells in contrast to CK1α which seems to be the dominant active CK1- isoform in melanoma cells.

It is known that β-catenin is a substrate of CK1α, δ and ε [3]. Whereas phosphorylation of β-catenin at Ser45 by CK1α results in degradation of β-catenin, CK1 δ/ε are involved in the activation of the Wnt/β-catenin pathway by the phosphorylation of dishevelled (Dvl). We analyzed whether overexpression of the individual CK1- isoforms as described above affects expression and activity of β-catenin signaling. Interestingly, β-catenin total protein levels did not change 1–2 days after CK1- isoform specific overexpression (Fig. 5a). However, as expected phosphorylation of Ser45 of β-catenin was increased after overexpression of CK1α (Additional file 3: Figure S3A) and this directly correlated with the influence of CK1α levels on the capacity to phosphorylate Ser45 in melanoma cells in a kinase assay (Fig. 5b). Overexpression of CK1α in SKMEL19 enhanced the kinase activity causing Ser45 phosphorylation, whereas the respective knockdown in SbCl2 decreased this activity. The other CK1- isoforms δ and ε did not show significant impact on the phosphorylation of Ser45 of β-catenin (Fig. 5b).

In order to measure the general effect of CK1- isoforms on the canonical Wnt-signaling pathway we used a firefly reporter system (Super8xTOPFlash) and tested the luciferase activity in lysates of SKMEL19 cells after induction of CK1- isoform specific overexpression. As expected, CK1α overexpression decreased the endogenous signaling activity, whereas CK1δ and ε enhanced the canonical Wnt signaling (Fig. 5c). Doxycycline treatment alone as a negative control moderately induced the reporter, however to a much lesser extent as with CK1δ or ε overexpression. These results confirm an inhibitory effect on Wnt/β-catenin signaling of CK1α and an activating effect of CK1 δ/ε in melanoma cells.

In addition, CK1α, δ and ε are known to influence activity of p53 signaling by specific phosphorylation. Overexpression of CK1δ and ε increased the protein levels of p53 and its target p21 in SbCl2 and SKMEL19 melanoma cells. In contrast, overexpression of CK1α did not influence p53 and p21 expression in this analysis (Fig. 5a). This indicates that p53 signaling is predominantly activated by CK1 δ/ε and not by CK1α in melanoma cells. However, knockdown of CK1α increased p21 expression (Additional file 3: Figure S3B). This goes in line with previous findings that MDM2 is a target of CK1α and CK1 δ/ε can phosphorylate p53 at N-terminal activating phosphor-sites [23].