Background
Melanocytes are derived from the embryonic neural crest. During and after closure of the neural tube in the trunk of the embryo, the neural crest is induced by an intricate crosstalk between bone morphogenetic proteins (BMPs) and Wnt-signaling [
1‐
4], which could be reproduced in human induced pluripotent stem cells in vitro [
5]. In the course of this process, neural crest cells undergo morphological changes accompanied by an altered adhesion and migration capacity. A corresponding morphological transformation of chick embryonic lens epithelial cells was designated as epithelial-mesenchymal transition (EMT) [
6]. After induction, neural crest cells migrate along their designated medial and lateral pathways. Neural crest cells following the medial pathway form spinal ganglia and the autonomic ganglia of the sympathetic chain. The neural crest-derived progenitors of melanocytes (melanoblasts) migrate along the lateral pathway and colonize the epidermis [
7]. During invasion of the epidermal/dermal layer of the skin melanoma cells (arising from melanoblasts or melanocytes in the epidermis) morphologically pass through an EMT-like program and thus recapitulate their former neural crest cell specific migratory capacities.
In previous experiments, we observed that human SKMEL28 [
8,
9] and mouse B16-F1 melanoma cells [
10] spontaneously integrate into the neural crest and resume a physiological neural crest cell migration after transplantation into the neural tube of the chick embryo. In contrast to melanoma cells, neurospheres generated from adult mouse subventricular zone (SVZ) stem cells integrate into the neural crest and perform neural crest cell migration only after pre-treatment with bone morphogenetic protein-2 (BMP-2) [
11]. BMPs are constitutively expressed in melanoma cells [
12]. In line, we showed that physiological neural crest migration (after transplantation into the neural tube) and invasive growth (after transplantation into the optic cup of the chick embryo) of melanoma cells can be ablated by the BMP-antagonist noggin [
13,
14] and that vice-versa malignant invasion can be induced by pre-treatment with BMP-2 in non-transformed human melanocytes [
15] clearly demonstrating that embryonic neural crest signaling pathways play important roles in the invasive growth of melanoma cells in vivo.
In a second approach, we demonstrated that the embryonic transcription factor β-catenin is frequently activated during melanomagenesis via non-mutational alterations like reduced expression levels of casein kinase 1α resulting in protein stabilization of the central Wnt-signaling player β-catenin [
16]. Furthermore, we showed in vitro a pro-migratory and pro-invasive function of β-catenin in melanoma cells implanted into human organotypic skin reconstructs [
17] which seems context dependent due to controversial reports in the literature. In line with a pro-invasive function of neural crest signaling pathways, it was reported that leflunomide, an inhibitor of dihydroorotate dehydrogenase, leads to an abrogation of neural crest development in zebrafish and decreases melanoma growth in vitro and in mice. Leflunomide exerts these effects by an inhibition of neural crest development [
18], also suggesting a close connection between developmental pathways and melanomagenesis. Leflunomide is an agonist of the ligand-activated transcription factor aryl hydrocarbon receptor (AhR) and activation by leflunomide leads to a dysregulation of Wnt signaling not least because its function as ligand-dependent E3 ubiquitin ligase in the β-catenin degradation pathway [
19‐
21]. This was shown to have a tumor suppressing function in intestinal cancer [
22]. However, the peculiarity of the canonical Wnt-signaling in the context of melanoma has produced conflicting theses about the correlation of the expression levels of β-catenin and the progression of the disease [
23,
24], which shows that the role of β-catenin in melanoma metastasis remains largely unclear. This was nicely summarized by Weeraratna et al. [
25]. Some studies indicated that β-catenin suppresses invasion of melanoma cells and that loss of β-catenin predicts a poor survival rate in melanoma patients [
26‐
29]. Other studies showed that increased abundance or stabilization of β-catenin leads to increased melanoma metastasis, both in vitro and in vivo [
30‐
34]. However, Wnt ligands do play a role in melanoma cell invasion. Consequently, Wnt5a promoted melanoma cell invasion and polarized protein localization by depalmitoylation of the melanoma cell adhesion molecule (MCAM) at cysteine 590 [
35]. Interestingly, Wnt5a does not exlusively mediate its effects on melanoma cells via a non-canonical, protein kinase C (PKC) dependent Wnt signaling pathway [
36,
37], but also stimulates β-catenin transcriptional activity during Wnt5a-mediated melanoma invasion and metastasis [
33]. Lastly, it was demonstrated that in addition to driving melanomagenesis and invasion of melanoma cells, intrinsic β-catenin signaling prevents anti-tumor immunity, leading to T-cell exclusion and resistance to anti-PD-L1/anti-CTLA-4 monoclonal antibody therapy in melanoma mouse models [
38], which could have great clinical significance.
Going back to embryologic neural crest induction, we asked in this project whether the Wnt3a/β-catenin signaling pathway was also directly involved in melanoma cell adhesion/migration in vitro and in vivo in an embryonic micro-environment of the neural crest. To this end we used in vivo chick embyo models including the experimental growth of melanoma brain [
15,
39,
40] and liver metastasis [
41,
42] along with in vitro models.
Methods
Cell lines, cell culture, and generation of Wnt3a-conditioned medium
The following human metastatic melanoma cell lines were used in this study: SKMEL28, A375, BLM, SKMEL19 and. 451Lu. Melanoma cells were cultivated in HEPES-buffered, 2 mM L-Glutamin containing RPMI1640-medium (Gibco/ Life Technologies, Darmstadt, Germany), which was supplemented with 10% fetal calf serum (Biochrom, Berlin, Germany) and 100 U ml− 1 Pen-Strep (Gibco/ Life Technologies, Darmstadt, Germany). All cultures were maintained at 37 °C in a 95% air / 5% CO2 atmosphere at 85% humidity. For generation of Wnt3a-conditioned medium (Wnt3a-CM), NIH3T3 murine fibroblasts constitutively over-expressing Wnt3a (kindly gifted by Prof. Michael Schwarz, Institute of Toxicology, University of Tuebingen, Germany) were cultured in DMEM-medium (Life Technologies, Darmstadt, Germany) supplemented with 1% FCS for 48 h. NIH3T3 conditioned medium (3T3-CM) served as control.
Transcriptome analysis of SKMEL28 melanoma cells and of zebrafish embryo neural crest
mRNA expression analyses were performed in triplicates with SKMEL28 melanoma cells (3 experimental groups: untreated, Wnt3a pre-conditioned for 24 h, or 0.5 μM PKF115–584 (gifted by Novartis Oncology; Novartis Institutes for Biomedical Research, Cambridge, MA) pre-conditioned for 24 h) as described elsewhere [
43,
44].
siRNA transfections and lentiviral shRNA
Twenty nanometer siRNA was transfected into SKMEL28 melanoma cells using the riboxx-FECT (riboxx Life Sciences) reagent as recommended by the manufacturer. Briefly, 3 × 10
5 melanoma cells were seeded per 6 well cavity, and the next day cells were transfected using combined Opti-MEM:siRNA and Opti-MEM:riboxx-FECT (12 μl per cavity) mixtures. The incubation time before addition to the cells was 15 min and the final volume was 2 ml. Cells were further incubated for 24 h before the usage in cellular assays. Lentiviral particles were generated and used as previously described for the shRNA mediated knockdown of β-catenin in melanoma cell lines [
17].
Scratch assay
5 × 105 melanoma cells were seeded into 6well plates and grown until confluency followed by overnight serum starvation. Six hours before scratching the medium was exchanged with either a 1:1 mixture of Wnt3a-CM, 3T3-CM, or normal culture medium containing 0.5 μM PKF115–584 (Novartis Institutes for Biomedical Research, Cambridge, MA). Then a scratch was applied using a 200 μl standard pipette tip (Greiner Bio-One International, order No. 739290, Germany) and detached cells were removed by exchanging the medium once again with the corresponding treatment medium. Microphotographs were taken at 0, 12 and / or 24 h post scratching to measure the closed area using ImageJ (V 1.50b, NIH, USA). At least three biological replicates were measured in quadruplicates to analyze the closed area in % (mean +/− SD).
Generation of cell aggregates, determination of primary aggregate formation
SKMEL28 cell suspensions (1 ml) were transferred into sterile gas permeable biofoil bags (Biofoil, Heraeus-Kulzer, Hanau, Germany) as described previously [
13,
45]. Before sealing, either Wnt3a-conditioned medium or PKF115–584 (0.5 μM) was added into each bag. In the second set of experiments, the siRNA-transfected SKMEL28 cells were used without addition of drugs or conditioned medium. Bags were sealed and placed in open cylindrical plastic containers and were continuously kept in constant rotation (2 rpm) for 24 h using a roll mixer (Cellroll Integra Biosciences, Fernwald, Germany) in a cell incubator at 37 °C with 95% air and 5% CO
2. To quantify the primary melanoma cell aggregation, aggregates were next transferred into six-well-plates and photographed with a microscope (Olympus IX50) mounted camera (Olympus E330). For morphometric analysis, 800–1200 aggregates per treatment were measured using AxioVision software (Zeiss, Oberkochen, Germany). Results were statistically evaluated using One way ANOVA,
p-values of < 0.05 were considered as statistically significant.
Primary melanoma cell migration
For determination of cell migration, the melanoma aggregates were placed into six-well-chambers in culture media (without de novo addition of Wnt3a or 0.5 μM PKF115–584) and further incubated for 24 h at 37 °C in a humidified incubator with 5% CO2. The aggregates and their homogeneous cellular outgrowths were photographed as described above. For morphometric analysis, the area of the remaining aggregate and the total area covering the aggregate plus the outgrown/migrated cells (representing the migrated cells) were determined in 200–300 aggregates per treatment group using AxioVision software (Zeiss). The capacity for migration (cell migration index) was calculated by dividing the area of the cellular outgrowth by the size of the aggregate and statistically evaluated (One way ANOVA) with multiple comparison (Dunn’s multiple testing).
Luciferase reporter assay
For the assessment of Wnt/β-catenin signaling activity SKMEL28 melanoma cells were transfected with the Super8xTOPFlash plasmid as described previously [
17,
46]. Cells were treated for 24 h before measuring the luciferase activity which was normalized to cell viability.
Quantitative real-time PCR
RNA isolation from melanoma cells was performed using the NucleoSpin RNA Kit (Machery-Nagel, Dueren, Germany) followed by reverse transcription using Maxima reverse transcriptase (Thermo Fisher Scientific, Dreieich, Germany) as recommended by the manufacturers. Reverse transcribed cDNA was subjected to SYBR Green based real-time PCR as described previously [
46] using the following primers: INHBA forward 5′- agctcagacagctcttaccaca-3″, INHBA reverse 5′ ttttccttctcctcttcagca-3′, CYR61 forward 5′-aaggagctgggattcgatg-3′. CYR61 reverse 5′- aggctccattccaaaaacag-3′. ANGPTL4 forward 5′- acgatggctcagtggacttc-3′. ANGPTL4 reverse 5′- cgtgatgctatgcaccttctc-3′, FABP7 forward 5′- aagtctgttgttagcctgga-3′, FABP7 reverse 5′- agggtcataaccattttgc-3′. CT values were normalized to 18 S rRNA expression.
3D spheroid assay
SKMEL28 melanoma cell aggregates were formed overnight before embedding the spheroids into a collagen matrix with or without 3T3 / 3T3-Wnt3a fibroblasts (1.5 × 105 cells/ml) and cultured using RPMI with 10% FCS. Sprouting (invasion) of SKMEL28 cells was radially measured (n = 6) at 0 h and 48 h using ImageJ, and an invasion index calculated as ratio of the 48 h to the 0 h value.
Organotypic skin reconstructs
Organotypic tissue skin reconstructs (TSR) were generated as described previously [
40]. Briefly, human fibroblasts were seeded into a collagen type I (1.35 mg/ml) matrix before the addition of an epidermal layer consisting of HaCat keratinocytes and human melanoma cells. TSR were exposed to control medium (3T3-conditioned TSR medium) or Wnt3a-conditioned TSR culture medium for 14 days before fixation and parrafine embedding for immunohistochemical evaluation.
Preparation of eggs, transplantation procedures, processing of the chick embryos
White leghorn chicken eggs were obtained from A. C. Weiss GmbH&Co. KG (Kirchberg/Iller, Germany). Egg handling and transplantations of SKMEL28 melanoma cells into the neural tube (untreated (
n = 8), Wnt3a-preconditioned (
n = 10), 0.5 μM PKF115–584-preconditioned (
n = 12)), of BLM melanoma cells into the rhombencephalon (untreated (
n = 8), 0.5 μM PKF115–584-pretreated (
n = 8)), or of SKMEL28 melanoma cells into the rhombencephalon (control shRNA (
n = 5), β-catenin shRNA (
n = 5)) of stage 12/13 chick embryos according to Hamburger and Hamilton (HH) [
47] were performed as previously described [
15,
39]. For the shRNA experiment we used SKMEL28 cells with stable knockdown of β-catenin [
17]. Chick embryos were further incubated after melanoma cell transplantation for 48 h (neural tube) or 96 h (rhombencephalon), respectively.
For the chorioallantoic membrane (CAM) metastasis assay eggs were incubated for 7 days at 38 °C with 60% humidity. On day 7 of chick embryo development a small window was made in the shell under aseptic conditions. The window was resealed with adhesive tape and eggs were returned to the incubator for 24 h. On day 8, the CAM was gently abraded with a sterile cotton swab to provide access to the mesenchyme and BLM melanoma cell suspensions (5 × 10
5 cells) were seeded in 10 μl of PBS onto the upper CAM vasculature and further incubated at 37.5 °C for 7 days. On day 15 chick embryos were euthanized and CAM and livers removed for isolation of genomic DNA (gDNA). Melanoma cell derived human gDNA was detected by qPCR as described elsewhere [
42].
Immunohistochemistry and western blot
The following primary antibodies were used in this study: anti-HMB-45 (1:20, Dako/ Agilent, Hamburg, Germany), anti-β-catenin (#9562, 1:50, Cell Signaling Technology, Leiden, Netherlands) and anti-Ki67/MIB1 (1:100, Dako/ Agilent). Immunohistochemistry of a human melanoma tissue microarray (TMA) using an anti-β-catenin antibody (1:100, Cell Signaling #9562) was performed as described previously [
46]. For western blotting the following antibodies were used: anti-Phospho-Akt (Ser473) (#4060), anti-Akt (#9272), anti-PTEN (#9188), anti-beta-actin (#3700) (all Cell Signaling Technology) and anti-beta-catenin sc-7963 (Santa Cruz).
All work with the TMA and human material was approved by the local ethics committee (305/2017BO2).
Generation of Kaplan-Meier plots
Gene expression and clinical information were derived from The Cancer Genome Atlas (TCGA, skin melanoma dataset (n = 454)) (Network C.G.A., 2015). Patients were subclassified according to gene expression values (e.g. upper 11% and lower 11%). Five year survival rates were evaluated with the log-rank test. Multivariate Cox regression included gene expression, age, gender and pathologic stage. The survival curves were visualized by Kaplan-Meier plots.
Statistical analyses
Statistical analyses were performed with GraphPad Prism 6.0 using the following tests: Fisher’s exact test, One way ANOVA, Mann-Whitney. p-values < 0.05 were considered as statistically significant.
Discussion
In the present study, we demonstrate a novel role of Wnt3a and the β-catenin signaling pathway in neural crest migration and malignant invasion of human melanoma cells.
Current therapeutic strategies for the treatment of metastatic melanoma focus on two major approaches with proven clinical efficacy: (i) direct targeting of activated oncogenes in melanoma cells such as BRAF [
53] or (ii) indirect targeting of melanoma cells by T-cell stimulation with anti-CTLA4- or anti-PD-1-antibodies [
54,
55]. Although these therapies caused a paradigm shift and were able to improve the 3-years overall survival of patients diagnosed with metastatic melanoma between 2011 and 2014 to 23% [
56], both approaches bear major drawbacks, which are reflected by the limited duration of the initial clinical response. Only a subpopulation of melanomas harbors the crucial oncogenic BRAF-mutation, and even in mutated melanomas a therapy resistance rapidly develops [
57]. We have recently shown that β-catenin is one potent mediator of resistance towards BRAF inhibition [
46]. In line, high levels of ZEB1 expression (an EMT inducer) are associated with inherent resistance to MAPKi in BRAFV600-mutated cell lines and tumors [
58]. Likewise, only a half of the patients clinically responds to T-cell stimulation, which is at least partially due to the fact that cytotoxic CD8
+ T-cells only recognize major histocompatibility complex (MHC) class I (MHC-I)-expressing melanoma cells. However, the alteration of MHC-I expression together with an impaired response to interferons is a frequent event during cancer (and melanoma) progression, allowing cancer cells to evade the endogenous or therapeutic immunosurveillance [
59]. A second plausible explanation for resistance to the novel immunotherapies might be the tumor-intrinsic oncogenic signals such as active β-catenin signaling, that mediate T-cell exclusion at the site of the tumor and thus resistance to anti-PD-L1/anti-CTLA-4 therapy [
38,
60]. Such mechanisms might be reflected by the association of WNT3A expression and melanoma patient survival which we have elaborated in this project. Therefore, additional and fundamentally different therapeutic approaches are still desperately needed to improve therapies and finally overall- and long-term survival of advanced melanoma patients.
Our approach is to draw an analogy between embryonic growth and cancer growth. In particular, neural crest signaling pathways seem to be a promising target for the inhibition of melanoma cell invasion and metastasis [
14]. Therefore, in the current study we first addressed the spatial expression of β-catenin in primary human melanomas. Interestingly, we found that β-catenin was predominantly expressed in melanoma cells of the invasive front with a spindle-like morphology. Therefore, we hypothesized that β-catenin-inhibition could affect melanoma cell migration and invasion in the neural crest. In the embryo, emigration of neural crest cells from the neural tube is designated as EMT. EMT represents a complex change in cell morphology and migratory potential of embryonic cells and is induced in the embryo mainly by BMPs and Wnt-signaling [
1‐
4], and vice versa inhibited by their antagonists. EMT comprises two consecutive steps [
61,
62]: (i) the neural crest compartment is induced in the epithelium of the neural tube, which is morphologically characterized by the disintegration of the basal lamina in the region of the lateral roof plate. (ii) Neural crest cells are induced to start migration from the dorsal edges of the neural tube along their designated medial and lateral pathways. Hence, EMT (governing embryonic neural crest migration and possibly melanoma cell invasion in the patient) of melanoma cells as neural crest descendants should be analyzed in the neural crest environment.
To verify our analogy hypothesis, we therefore used our chick embryo model in two different experimental settings: First, we injected human melanoma cells into the lumen of the neural tube of stage 12/13 HH chick embryos to analyze their capacity for spontaneous neural crest migration. Before injection, the melanoma cells were pre-conditioned with either the agonist Wnt3a or with the β-catenin-inhibitor PKF115–584. Interestingly, the agonist and the antagonist had opposing impacts on melanoma cell behavior in the neural crest compartment: Wnt3a enhanced, and PKF115–584 abrogated the spontaneous neural crest migration of SKMEL28 cells in vivo, which is in line with the effects of the neural crest-inducer BMP-2 and its physiological antagonist noggin [
13]. The impact of Wnt-signaling on melanoma cell migration and EMT (reflected by changes in aggregation and migration) could be reproduced in vitro. This finding is challenged by several publications that rather show the opposite effects [
26‐
29]. However, a recent study clearly showed that canonical Wnt signaling is a pro-invasive factor in the context of active PI3K signaling pathway due to PTEN deficiency [
63]. This effect was found to be mediated by context dependent regulation of catabolic processes by β-catenin. Since in our cell culture models the PI3K signaling pathway is activated (high phosphorylation of Ser473 of AKT) we believe that our models reflect this situation. This is further substantiated by the bad impact of low PTEN expression levels on melanoma patient survival in the context of a high expression of WNT3A (whereas PTEN levels have no impact on survival for WNT3A low melanomas).
RNA expression analyses of melanoma cells pre-conditioned with either Wnt3a or PKF115–584, and of zebrafish embryo neural crest cells demonstrated overlapping genes. Using siRNA mediated down-regulation, we could functionally demonstrate an impact on EMT of the overlapping genes in melanoma cells in vitro, thus corroborating the RNA-expression results. In this respect, we were able to identify inhibin beta A as novel, β-catenin-regulated candidate gene that hinders melanoma cell invasion. Interestingly, by screening melanoma data from The Cancer Genome Atlas, we discovered a correlation of high expression of CYR61 or (by trend) INHBA with increased overall survival of melanoma patients, clearly underlining the clinical significance of our findings. This is in line with a previous report showing that Cyr61 expression in melanoma cells reduces tumor growth and metastasis [
64].
Since we also detected MITF and its target gene tyrosinase (TYR) to be regulated upon modification of canonical Wnt signaling in SKMEL28 melanoma cells it could be that MITF is at least partially involved in the pro-migratory effects in vitro and in vivo. This would support a recent publication, which found that MITF is required for melanoma cell migration and invasion independent of BRN2, meaning in MITF
high and MITF
low melanoma cells [
65]. Again, this is a topic full of conflicting data which is mirrored by the complexity of the MITF rheostat model [
66]. For example, Fane et al. could recently show that MITF and BRN2 are inversely correlated and BRN2 expression drives melanoma cell migration and invasion via the nuclear factor IB [
67]. Since we did not see strong effects on MITF gene expression in all cell lines tested, we do not assume MITF to play a leading part in the observed effects. TYR was additionally identified as a direct target of LEF1 dependent gene transcription [
68].
The question arose, if the physiological neural crest migration of melanoma cells was equal to malignant invasive migration, as encountered in advanced melanoma patients, and if such malignant invasion was also accessible for pharmacological manipulation through inhibition of endogenous Wnt-signaling. Thus, we injected untreated or PKF115–584 pre-conditioned human melanoma cells in the rhombencephalon of the chick embryo and analyzed malignant invasion after 4 days. As expected, the untreated melanoma cells formed tumor nodules in the roof plate and adjacent mesenchyme, with single cells and streets of cells invading the chick host tissues and blood vessels. Interestingly, only actively invading melanoma cells with a mesenchymal morphology had a prominent cytoplasmic β-catenin expression, and not the cells in the tumor nodule displaying a compact, epithelial-like morphology. This was in line with melanoma cells in primary melanomas displaying a high cytoplasmic β-catenin expression at the invasive front. To our surprise, the PKF115–584 pre-conditioned melanoma cells had undergone apoptosis, and in none of the embryos of this cohort a viable tumor nodule could be detected. This is particularly interesting since 48 h after injection into the neural tube, even PKF115–584 pre-conditioned melanoma cells were still positive for the proliferation marker MIB1, and suggests that inhibition of Wnt-signaling with PKF115–584 might lead to both complete inhibition of neural crest migration and invasion (after 48 h) and to a delayed induction of cell death (up to 96 h). This is in line with previous reports showing both anti-proliferative and anti-invasive effects of β-catenin-inhibition in melanoma cells [
16,
17,
30].